Towards the End of Hunger

Table of contents :
1. Introduction
Part I. CONCEPTS
2. From hunger to food security: a conceptual history
Part II. HISTORICAL TRENDS
3. Agricultural and food output
4. Land use and agricultural productivity
5. Agricultural trade
6. Food consumption patterns
7. Food access and nutrition
8. Conclusions about food and hunger trends
Part III. FUTURE PROSPECTS
9. Projections on food and agriculture
10.Impact of climate change on agriculture
11.Concluding remarks
TECHNICAL AND METHODOLOGICAL APPENDIX
12. Measuring historical trends
13. How to see the future: projection methods

Citation preview

T OWA R D S T H E E N D O F H U N G E R

© Hector Maletta, 2016 This edition: © Universidad del Pacífico Av. Salaverry 2020 Lima 11, Perú www.up.edu.pe

TOWARDS THE END OF HUNGER Hector Maletta 1st edition e-book: August 2016 Cover design: Icono Comunicadores ISBN e-book: 978-9972-57-363-7 E-book available at www.up.edu.pe/fondoeditorial

BUP Maletta, Hector. Towards the end of hunger / Hector Maletta. -- 1st edition e-book. -- Lima : Universidad del Pacífico, 2016. 493 p. 1. Hunger -- Research 2. Food supply 3. Agricultural productivity 4. Agricultural ecology I. Universidad del Pacífico (Lima) 363.8 (SCDD) Member of the Asociación Peruana de Editoriales Universitarias y de Escuelas Superiores (APESU) and of the Asociación de Editoriales Universitarias de América Latina y el Caribe (EULAC). The Universidad del Pacífico does not necessarily endorse the contents of the works that it publishes. Any total or partial reproduction of this text by any means without the permission of the Universidad del Pacífico is prohibited. All rights reserved under the law.

CONTENTS

1. Introduction.............................................................................................. 7 Part I. CONCEPTS 2. From hunger to food security: a conceptual history................................. 13 Part II. HISTORICAL TRENDS 3. Agricultural and food output................................................................... 43 4. Land use and agricultural productivity..................................................... 71 5. Agricultural trade................................................................................... 115 6. Food consumption patterns................................................................... 137 7. Food access and nutrition...................................................................... 169 8. Conclusions about food and hunger trends............................................ 189 Part III. FUTURE PROSPECTS 9. Projections on food and agriculture....................................................... 193 10. Impact of climate change on agriculture................................................ 215 11. Concluding remarks.............................................................................. 305 TECHNICAL AND METHODOLOGICAL APPENDIX 12. Measuring historical trends.................................................................... 319 13. How to see the future: projection methods............................................ 367 REFERENCES............................................................................................ 475

5

1. Introduction

Food is the most primary human need. Yet in spite of stunning technological achievements and economic development, humankind has not overcome hunger. More than 800 million people suffer undernourishment. When and how will this scourge be eradicated? Ever since Malthus (1798), concerns about hunger and food supply have pervaded the economic and social development literature, and in recent decades have become a major concern of international organisations and humanitarian donors. How to feed a rising world population is an often posed question, which is made more pressing by fears that climate change and higher food prices may reduce the food supply and restrict access to food. Many headlines and press reports leave the impression that the world’s starving masses are growing larger and hungrier, that the food situation is bad and worsening, that the world is increasingly unable to provide food for everyone, that a ‘food crisis’ is in the offing or is already occurring, and that future prospects are grimmer still. However, the facts (see box) tell a different and, on the whole, more encouraging story. These facts may change how many people understand the international food and hunger situation and should contribute to shifting the hunger agenda towards new priorities and challenges. • Hunger is diminishing. The world is close to attaining the Millennium Development Goal of halving the prevalence of hunger by 2015 relative to 1990-1992. Though the hungry are still many, their number is decreasing over time in absolute numbers and as a percentage of population. The food and hunger situation is improving, though the fight against hunger is far from over.

7

Introduction

• The prevalence of child malnutrition (stunting and underweight) is also decreasing. In its place child and adult overweight and obesity are rapidly increasing in both developed and developing countries. • Population growth has greatly decelerated, and is expected to keep decelerating in the future. • Current food output is more than enough to feed all humankind, and it is growing consistently faster than population. Per capita food supply is increasing, and its rate of increase is itself accelerating. • Availability of land is not a significant constraint for increasing food production. More than 95% of the growth in agricultural production since 1961 (including crops and livestock) came from increased production per hectare; less than 5% came from increased land area used for crops or pasture. Moreover, most of this small increase in farmland occurred in the first part of that period (1961-1980); in more recent decades, world food production kept growing while world farmland stalled or slightly decreased. • By the same token, crop production in particular has also been rapidly increasing, in total or per capita terms, while total cropland has been nearly stagnant since 1985, with some periods of slight decrease (until 2000) or slight increase (more recently). About 95% of crop output growth is explained by increased output per hectare of cropland. On the other hand, less than one half of the world’s Prime and Good land suitable for rain-fed cultivation is actually under crops and there is ample room for using more. • Projections of future agricultural growth to meet expected demand up to 2050 or 2080 envisage just a small net impact of climate change on agricultural production, as a result of a balance of negative and positive impacts in different regions, and intervening adaptations. Agricultural growth will entail only small increases in total farmland, or in land under crops: most of the growth would still come from extra output per hectare, even under hypotheses assuming that the rate of technical progress and productivity growth in agriculture would be significantly slower in the coming decades. • Even under pessimistic hypotheses, the extent of hunger is expected to keep decreasing throughout this century until it becomes non-significant over all regions of the world, even if technical progress in agriculture slows and there is little expansion of farmland, and even when expected population growth, impacts of climate change on agriculture, and use of food crops as biofuels are taken into consideration.

8

Hector Maletta

Hunger, like poverty, is still a pressing reality for many millions of people. The battle against hunger is still on-going. But past trends and future prospects are better than many people appear to think. Many such pessimistic beliefs are due to no more than the fact that bad news makes better headlines than good tidings. Importantly, mistakenly believing that hunger is overwhelming and increasing may create a feeling of powerlessness. Some of the false impressions about hunger arise not from actual figures or the scientific literature, but from press reports or TV documentaries that rest on inaccurate or inappropriate statistics, or on faulty concepts about what hunger is, how to measure its extent and severity, or what the future may hold. They also feed on the widespread penchant of people to expect the worst and probably also on the well-meaning efforts of many organisations that try to raise awareness of the issue of world hunger and to attract resources to the cause of promoting food security and eradicating hunger. However, arguing on the basis of false premises often backfires. Most of the available data and scientific assessments tell a more nuanced and less pessimistic story. In fact, humans are actually overcoming hunger, for the first time in their long evolutionary history. Understanding food, hunger, and food security issues thus requires careful discussion of the concepts involved, the methods used for their assessment, and the resulting figures. This book analyses the trends and prospects on world food production and consumption, and includes a discussion of concepts and methods commonly used. It is a book with a good many tables and charts. They are necessary, because the subject matter is complex, and the misperceptions of conventional wisdom can only be revealed through careful examination of the facts. Much still needs to be done in regard to world hunger. It is declining, but could decline faster if the right policies were enacted. This book, however, is not (mainly) about what needs to be done; it is not about solutions and policies for reducing hunger and improving food security. Here, we primarily concentrate on what has been happening and what will probably happen in the future, leaving detailed policy discussions for another occasion, lest this book becomes excessively long. Indeed, only a limited discussion of policy challenges is included. Awareness of trends and prospects is the best and inevitable first step to devising better policies, especially for those regions where hunger is more severe and expected to last longer. Hunger has been present for a long time in the history of humankind. This book, however, discusses only recent historical trends in key matters related to hunger (food production and consumption, food trade, access 9

Introduction

to food, malnutrition), mostly restricted to the last half century. A longerterm perspective can be found in Fogel (2004), for example, who studies the ‘escape from hunger’ since 1700 with prospects until 2100. A still more long-term view is taken by Oded Galor (2011), who studies the escape from ‘the Malthusian trap’ that afflicted humankind for millennia from the days of our distant prehistoric ancestors. Our own scope will be more limited: we analyse trends observed over the past half century, and prospects extending just a few more decades into the future. This puts limits on the time scope of our discussion. This book is also about trends and not episodic crises. It will not dwell on transient events such as the price surge in food commodities in the mid-1970s, the more recent price surge during the late 2000s, or particular episodes of famine. Nor will it examine the more distant past or the more distant future. Likewise, some specific issues are not discussed extensively, mainly to keep the text within a manageable length, and to focus instead on issues about which information is available at the world level and over a longer time span. Issues linked to excessive or qualitatively inadequate food (such as obesity, food waste, food safety, and the regrettable popularity of junk food) are discussed only sparingly. The book has three parts. The first discusses the concepts employed. The second part examines recent historical trends, demonstrating how food and hunger have evolved over the past half century to reach their current state. The third part explores future prospects and is based primarily on existing projections of future developments in agriculture and hunger and incorporates the expected effects of various key interacting factors: population growth, economic development, technical progress in agriculture, climate change, increasing use of crops for biofuels, and more. The last part of the book is a Technical and Methodological Appendix which provides a detailed discussion of many technical aspects related to the subjects discussed in the book. We urge readers to pay attention to these technical aspects, because many misconceptions about food and hunger arise from using faulty figures, misunderstanding key concepts or methods, applying defective methodologies, adopting wrong assumptions, or drawing unwarranted conclusions. To avoid these dangers, and whenever in doubt, the reader is strongly advised to refer to the detailed explanations supplied in the Technical and Methodological Appendix.

10

Part I. CONCEPTS

2. From hunger to food security: a conceptual history

2.1. A taste of hunger There is frequent talk in the press as well as in the development literature and academic publications about the extent and prospects of hunger in the world, especially in connection with natural disasters such as floods and droughts, countries and human groups afflicted by chronic poverty and/or violent conflict, or similar predicaments. Hunger is also a focus of attention in relation to challenges such as high food prices in international markets, the extent of food waste, or how much extra food will have to be produced in the future to eliminate existing hunger and to keep a growing population free of hunger; the issue is also frequently present in discussions of the likely impacts of climate change on agriculture and food. However, such frequent reference to hunger is often based on imprecise notions about what is actually meant. Hunger is a household word, familiar to everybody and thus easily communicable, but it has many layers and dimensions that warrant careful conceptual analysis. The technical notions underlying the problem of ‘hunger’ may be quite different from the popular notion of ‘hunger’. On a very elementary level, hunger is a subjective feeling of craving for food or a desire to eat, something everyone can experience, especially after some hours of fasting. This subjective feeling, however, is not necessarily correlated with nutritional requirements: obese people feel hungry every day, despite their ample reserves of most nutrients, and many people crave more food than they need. At the other extreme, starving people frequently lose appetite as a result of their condition, and cease to feel hunger. 13

From hunger to food security: a conceptual history

Humans have an inborn tendency to eat more food than they immediately need. This extra amount of appetite was selected for over several million years of evolution, as our ancestors survived in small bands of hunters and gatherers whose luck at finding food varied wildly from one day or week to the next. Individuals with a willingness to eat more than was immediately needed were able to accumulate bodily reserves of energy, protein, vitamins, and minerals during the (generally short and unpredictable) spells of abundance. This enhanced their chances of survival and reproduction over the (usually longer) periods of relative scarcity. Genes coding for extra appetite thus tended to be selected for through differential survival and reproduction, and became quite frequent in the human gene pool. In a world where food supplies are more reliable, these tendencies to crave more food than necessary may become superfluous and indeed deleterious for health and survival. Besides its primary meaning (a desire to eat) the word hunger has also been taken to be synonymous with insufficient intake of food over a period of time, even if those concerned may not actually feel hungry, and indeed not implying any investigation as to their feelings or cravings: a level of intake is usually judged to be insufficient in comparison with a certain external norm or standard established by doctors and nutritionists. Such insufficiency, in turn, may refer to the lack of any of several things. Food provides various ingredients necessary for life and the supply of any of them may be insufficient. Food provides fuel (e.g., carbohydrates or fats) used by body cells to generate energy (commonly measured in calories). Food also provides protein, i.e., chemical chains of amino-acids that are the building blocks of body tissues; once eaten, the protein in foods of animal or vegetal origin is broken down into its component amino-acids, which are then used by the body to synthesise its own human proteins (or eventually burnt for energy if not used for producing proteins). Food also supplies minerals and vitamins: the body cannot synthesise any vitamin other than vitamin D (the bodily production of which is activated by exposure to sunlight). All other vitamins, as well as all minerals, are normally taken from food. The body needs food to stay alive, to perform physical activity, to enable organs (like the brain or liver) to work properly, to recover from infections, and to build new tissues as required for growth in pregnant women, children, and adolescents. A proper diet should cover all these needs; an unbalanced diet may perhaps cover or exceed dietary energy needs but fail to provide enough of other nutrients. Food intake, moreover, is not everything; other conditions are required, both before and after food is eaten. Before food intake occurs, people must have access to food. In turn, before being accessed, the food must be produced, 14

Hector Maletta

and thus production and distribution of food and the ability to access it are important elements in understanding (and avoiding) ‘hunger’. And hunger does not end with food access and intake; after food is eaten it must be biologically processed and utilised inside the body, absorbing its nutrients and using them for the various purposes listed above. Problems hindering bodily food utilisation (e.g., intestinal infections) may offset the potential benefits of food intake. Lack of resources, poverty, or isolation may hinder food production or food access. Health and sanitation shortcomings may restrict biological utilisation of food by the body. Infectious diseases (such as diarrhoea) can reduce the body’s ability to assimilate food and to use it effectively; infections may also waste energy by producing and dissipating heat in the form of fever. Thus, hunger has a number of aspects that require more precise concepts and sub-concepts as well as their corresponding operational definitions and measures. Various concerns about food and hunger in the world have coalesced around the notion of food security, which covers all aspects of food described above. As the concept of food security has evolved, many discussions have taken place around some key questions: how much food is needed? Is food production sufficient for all? Will it continue to be sufficient in the future? How is it distributed among people? How much hunger is there? How much obesity? How are these conditions defined and measured? What is the likely future of hunger in the world? How is this likely to be affected by climate change? The way these questions have been addressed through the concept of food security has evolved over time as discussed below.

2.2. Food security as food self-sufficiency Many people identify food security with national food self-sufficiency. Accordingly, a nation has food security if it produces all the food required by its population. This notion is sometimes extended to smaller units; for instance, the self-sufficiency approach would assert that a rural household has food security if it produces enough food to cover its own needs. There was indeed a time when food security was regarded as a matter of selfsufficiency, but that approach has long been abandoned. Food security as a concept underwent a remarkable transformation during recent decades (Maxwell and Smith 1992; Frankenberger and McCaston 1998; Maxwell 2001; DEFRA 2006; CFS 2012). The concept emerged and gained worldwide relevance in the wake of the short-term food price surge in the mid-1970s 15

From hunger to food security: a conceptual history

following the 1973 oil crisis and amid fears of explosive population growth. This resulted in worldwide publicity for both the 1974 World Population Conference in Budapest, which addressed the issue of demographic growth, and the 1974 World Food Conference. The Food Conference discussed food security at the world level in terms of global food supply for a growing population. Food security was defined as ‘avai­l­ability at all times of adequate world supplies of basic food-stuffs […], to sustain a steady expansion of food consumption […] and to offset fluctuations in production and prices’ (UN 1975). The definition did not consider food security at the national or local levels, or the role of food trade in bringing food from one nation to another (or from one local area to another within a nation). It was also silent on the question of access to food by households and individuals, and on a number of other related issues such as food quality or food waste. The 1974 definition referred to supplies of food worldwide, with no explicit reference to individual nations. However, after adopting a definition of food security based on sufficient world supplies of food, a recommendation of the Conference and many documents derived from it was that to contribute to world food security, every nation should strive to achieve food selfsufficiency at the national level. The rationale behind this recommendation was apparently the trivial notion that if every nation were able to feed itself, world food sufficiency would also be achieved, and then food insecurity at the world level, as defined by the Conference, would not exist. The notion of worldwide sufficiency on which the 1974 definition was based was thus downscaled and applied to individual nations in the form of national selfsufficiency, i.e., positing as an ideal situation that each nation should produce all the food its citizens consume. The usual interpretation of this principle was that a country lacked food security if it had a food deficit, i.e., if it failed to be food self-sufficient. Food imports were the mark of food insecurity. Until the 1980s, the concept of food security continued to be centred on national self-sufficiency, especially in cereals, the main staple food of humankind. Foreign trade was not seen as a way of enhancing food security but as a sign of its absence - that a country had to import food was of itself an indication of limited food security, especially in poor countries. Food surpluses in some countries were not seen as a means of reducing the food insecurity of food-deficit countries; only domestically produced food counted against food insecurity. For countries without food self-sufficiency, their capacity to import food (e.g., with revenue from exports of non-food products such as minerals) was usually not examined. No difference existed in principle between food-deficit countries with different levels of per capita income or 16

Hector Maletta

non-food exports, e.g., between Mauritania and Saudi Arabia, or between El Salvador and Belgium. Only later was a specific category of ‘low-income food-deficit countries’ created in FAO statistics in an effort to focus attention on countries where lack of self-sufficiency originates in low income, or rather low export revenue, and not non-self-sufficiency in food production. On the other hand, only aggregate self-sufficiency at the national level received attention; domestic inequality in food consumption or in access to food among individuals or households, or different degrees of deficit or surplus among sub-national regions, were absent at that stage from the definition of food security. One common argument at the time was that dependence on foreign trade entailed exposure to capricious fluctuations in world markets, which could be avoided by relying on domestic production. This line of reasoning, of course, disregarded the instability created by fluctuations in domestic production (mainly due to climatic vagaries but also to macroeconomic imbalances, domestic inflation, distorted relative prices, domestic economic cycles, and other factors). Fluctuations in domestic output could indeed be offset or smoothed by foreign trade, which is unlikely to always fluctuate synchronously with local variations of climate, economic activity or inflation. This approach also assumed that local food prices could be successfully controlled and decoupled from world prices. Even if world prices surged, a self-sufficient country could allegedly keep domestic food prices low. This was a plausible proposition up to the 1970s and in many cases up to the 1980s, because price controls were widespread, trade was limited, government intervention in (and control of ) food trade was common, international financial flows were restricted, and exchange rates were officially fixed (at single or multiple rates) as per the 1944 Bretton Woods agreements, in force until 1973 but sustained in many developing countries until the 1980s and in some cases even longer. Developing country currencies were usually undervalued, and thus imported food was abnormally expensive compared to local foodstuffs. Thus, local and foreign prices were, in many cases, actually decoupled and local prices were kept under government control. In subsequent years and decades, this became increasingly difficult to achieve. The global economy underwent profound changes starting in the mid-1970s, including a shift towards liberalisation of trade and financial flows across national borders. This was not so much the result of a change in economic ideas as an inevitable consequence of the collapse of the Bretton Woods system; if currencies were no longer convertible into gold at a fixed US dollar price and therefore allowed to float, then freer international financial flows were a 17

From hunger to food security: a conceptual history

necessity. Ideas and policies were adjusted accordingly. At the same time, this led to acceleration of the world trade liberalisation process that started in 1948 with the General Agreement on Trade and Tariffs (GATT). Successive rounds of trade negotiations under GATT achieved gradual and incomplete but nonetheless very significant reductions in trade barriers (tariff and non-tariff). This led to the creation of the World Trade Organization (WTO) in 1996, which established tariff ceilings and general agreements on trade liberalisation. Although barriers to trade remained in place for some sectors, most notably for farm products, trade in general became more liberalised (including trade in food), a trend that has increased even more since the WTO was established. In many ways, national economies became ever more interconnected as regards trade, investment, technology and other aspects. This course of events, in addition to the evident inability of many developing countries to keep their public finances in order and to repay their foreign debt, led to structural reforms and economic liberalisation in both developed and developing economies. Again, this was less a result of a change in ideological winds than a necessity imposed by the new international economic reality. Ideological debate took place mainly on the question of how (and how fast) to adjust to this new economic situation, but it rested on the objective changes that had occurred and continued to occur in the world economy. The process of structural adjustment started with the establishment, shortly before 1980, of new lines of credit for this purpose extended by both the World Bank and the International Monetary Fund. The impact of these changes became clear during the Mexican debt crisis of 1982. It continued over the following years, especially during the 1990s. Older regimes of fixed exchange rates, determining wide differences between international and local prices (calculated at official rates of exchange) tended to be replaced by more realistic, and often floating, exchange rates. Realignment of currencies also meant that domestic prices became more closely aligned with international levels. The economic and financial systems of nations increasingly became internationalised as local activities and transactions implied global consequences, and domestic policies became increasingly conditioned by the global economic system. When parts of that global system faced a crisis, the repercussions spread rapidly over the entire globe, as was the case of the financial crisis that started in 2008 and caused the global slowdown known as The Great Recession. In the post-war era, developed countries (Europe, the United States, and Japan) had established farm protection policies that led to the systematic 18

Hector Maletta

production of surpluses, routinely purchased or subsidised by governments to protect farmers from ruinous drops in producer prices. This, however, tended to change after the 1992 reform of the European Common Agricultural Policy, and the establishment of the World Trade Organization - two events that triggered other similar developments worldwide. Other shortcomings of the concept of food security defined as national food self-sufficiency also surfaced. A key defect of this concept was that it was centred on food availability, neglecting matters of distribution and access: a country with sufficient food supplies, matching or exceeding its total food needs, may still have a large percentage of its population experiencing hunger or malnutrition. By concentrating on the aggregate (national or worldwide) level, a definition of food security centred on sufficient food supply ignores inequalities between zones, groups, households, and individuals. As noted above, another major shortcoming of the 1974 definition and its incarnation in national policies was an inconsistent treatment of food trade. Self-sufficiency was sought only at the national level; it was not ordinarily implied that the goal should also be that each province or locality in a country must be self-sufficient, implicitly allowing that some zones might have a food deficit, and implicitly entrusting domestic trade with the task of ensuring full coverage of food needs within countries. Thus, international food trade was rejected as a resource for food security, but no such exclusion existed for domestic trade. The latter was implicitly seen as manageable, while foreign trade was considered unpredictable and capricious. Ultimately, the decisive criterion was not self-sufficiency as such, but that governments keep political control over the food supply. If food was to come from abroad, it made the polity dependent on other nations, whereas domestic production and trade were (supposedly) under the aegis of the national government. This concept of food security was more about national sovereignty and national security than about ensuring that all people receive adequate food. A related aspect of this sort of double standard in regards to trade and food security is monetary in character. Domestic trade is conducted in local currencies, whereas foreign trade requires foreign monies. The former can be issued by a government, whereas the latter must come from exports, foreign investment, or financial transfers. Many developing countries by the 1970s and 1980s were caught in a chronic balance of payments quagmire, with stagnant exports and increasing imports (driven in part by a growing population, and also by domestic development policies based on subsidising local industries to achieve import substitution). 19

From hunger to food security: a conceptual history

This problem was often accentuated when countries incurred serious fiscal deficits and attempted to make up for their insufficient revenue (or excess expenditure) by issuing domestic currency and adopting loose monetary policies, thereby creating domestic inflation, increasing public indebtedness, and gradually reducing the purchasing power of their domestic currency to obtain internationally traded goods (such as food). Their capacity to import was chronically insufficient; their capacity to export was often insufficient to support the imports demanded by their growing population. Devalued currencies made imported commodities increasingly unaffordable. In many cases, subsidising the domestic price of imported food was a political necessity that gradually became fiscally unbearable. Reducing food imports (in fact any imports) was necessary due to balance of payments constraints. Food selfsufficiency at any cost thus seemed justifiable in the eyes of governments to avoid collapses of national economies, and the consequent political upheavals triggered by acute balance-of-payment crises, huge fiscal imbalances, and (often) runaway inflation. However, the root cause was not imports but the underlying distortions and imbalances of national economies. Only after economies gradually opened up and faced the necessity of economic reform was this gradually perceived. A major consequence of these combined developments was a growing awareness that international food flows (including ordinary trade as well as food donations) are a necessary component of food security. Facilitating food trade, as well as aligning exchange rates to make imported food affordable, increasingly appeared to be an important element to ensure adequate food supplies in food-importing countries. Food aid became a large (and quasipermanent) component of the food supply in many low-income countries, and the international food trade grew rapidly, both in absolute terms and as a share of total food output. Acceleration of growth in developing countries (especially in Asia) further increased food trade since the 1990s, and drove food prices up in the 2000s. Another conceptual problem with definitions of food security proposed in 1974 and thereafter was their neglect of export earnings, including those of agricultural origin, and other foreign revenue such as remittances. A country producing insufficient food and importing much of its food needs, but which is able to pay for its imported food with regular and abundant foreign revenue from remittances or exports will have no trouble fulfilling the food needs of its population. Nonetheless, such a country will appear to have an extreme degree of food insecurity according to the self-sufficiency definition. Export revenue may come from non-farm commodities such as oil, minerals, or 20

Hector Maletta

industrial goods, and from non-food or non-staple-food farm products such as jute, spices, tea, coffee, or tobacco. This neglect of exports logically implied that many rich countries, as well as many developing nations with abundant non-food exports or large remittance inflows, would be incongruously classed as food insecure at the national level just because they ‘depended’ on food imports. In actual fact, even if food output in those countries might be insufficient to cover domestic supply, such countries could have regular access to (imported) food insofar as they have a permanent source of foreign revenue. On the other hand, closed economies with insufficient financial capacity to import, such as North Korea, might suffer famine or food shortage precisely because they could not afford to import food when needed. Environmental aspects were also absent from the 1974 definition of food security, just as they were ab­sent from definitions of economic development at the time, except in works dealing with Malthusian views of resource constraints on population growth such as the ‘population bomb’ and ‘limits to growth’ debates sparked by Ehrlich (1968) and Meadows et al. (1972).1 Conflicts between the goals of ensuring food self-sufficiency and preserving the environment were not explicitly envisaged. In fact, some developing countries, in the name of the overriding goal of food self-sufficiency, strongly resisted calls to stop deforestation or to provide stronger environmental protection. ‘Food first’ often implied ‘the environment later’. Likewise, food quality was ignored in the early definitions of food security. Attention was centred on the quantity of food, and particularly on staple food, chiefly cereals. Concerns about diets that were poor in micro-nutrients were not paramount, while concerns about food safety and appropriateness to prevailing cultural values and food habits were largely absent from food security definitions. In a way, this was tolerable since the definition referred mostly to staple food, chiefly cereals, providing the bulk of food energy, and it was implicitly admitted that certain specific cereals were preferred in each country, like rice in China and maize in Mexico. However, this was not explicitly incorporated into the definition that identified food security with sufficient world supplies of food. A shortage of rice in an Asian country could be offset by an extra supply of wheat at the world level, even if rice-consuming populations were not likely Both books and their catastrophic forecasts were criticised for internal methodological shortcomings and lost most of their appeal some years afterwards when such disasters failed to materialise and observed trends diverged from their forecasts. The authors of both books recently revisited their work, mostly to defend it (Meadows et al. 1992 and 2004; Ehrlich and Ehrlich 2009). On the ‘population bomb’ issue see various (pro and con) articles in the Vol.1 No.3 (2009) of the Electronic Journal of Sustainable Development (http://geog.utm.utoronto.ca/desrochers/The_Population_Bomb.pdf). 1

21

From hunger to food security: a conceptual history

to substitute other cereals when facing a rice shortage. A possible insufficiency of vitamin A, leading to widespread night blindness, was not examined in the context of food security (there were only some concerns about protein in the 1970s and 1980s, leading to the incorporation of pulses and some dairy products in food aid packages). Consistent with the tendency to ignore food preferences, the various food aid policies implemented in the 1970s and 1980s provided cereals or other foodstuffs as dictated by excess supply at the source, regardless of local preferences. Wheat was dispatched as food aid to countries accustomed to the consumption of maize, thus contributing to aid-driven dietary change. Some of the strictest safety norms prevailing in donor countries were in some cases ignored or relaxed for donated food. A further problem with the early definitions of food security was their indifference towards economic behaviour and the operation of markets. Instead, there was a preference for a framework more akin to logistics and central planning. A shortage of food was defined as a deficit of supplies relative to usual or normative requirements and estimated mostly in a mechanical fashion, applying past trends or fixed coefficients, with not much room for adaptive behaviour, and no major role for relative prices and their effect on producer and consumer behaviour. This may be adequate in a command economy or for prisons or barracks, where food is rationed and centrally distributed, but not for a market economy. Finally, another problem was that early definitions did not involve risk, uncertainty or probability. Food security (or insecurity) was defined just as a factual situation of sufficient (or insufficient) food supply; the definition referred to objective facts of production, stocks, and use, without much concern given to preferences, expectations or beliefs, and little consideration of uncertainty, risk, or probability.

2.3. Conceptual shifts Since the early 1980s, the concept of food security evolved away from the concept of self-sufficiency. The first major blow to the early view of food security as a balance of aggregate supply and needs was Amartya Sen’s (1981) short book on famines, which showed that many famines occurred in situations of adequate or excess food supply. Shortly afterwards, a report was issued by the UN-sponsored Independent Commission on International Humanitarian Issues (ICIHI 1985) portraying famines as man-made disasters (in contrast to prevailing views attributing them chiefly to drought, plant disease, floods, and other natural disasters). 22

Hector Maletta

The centrality of individual (or household) access to food and the existence of acute malnutrition or famine in the presence of overall food balance or even food surplus had already been observed in a number of studies on nutrition. Sen’s contribution, however, gave the issue general and formal theoretical standing and made it practically impossible to continue discussing food security from a supply viewpoint without considering distribution and access. Some early foreshadowing of the upcoming conceptual change was evident during the 1980s. For instance, a FAO report in 1983 introduced food access (in addition to food supply) to the definition of food security: it required “ensuring that all people at all times have both physical and economic access to the basic food that they need” (FAO 1983). However, the focus was still on supply: ensuring access was mostly conceived as having enough supplies for everyone. Even this modest conceptual change, however, was not included in official terminology until much later. A number of conceptual trends then emerged in the evolution of thinking about food security, including: ŒŒ A shift in the unit of analysis from the world and the nation towards households and individuals. ŒŒ A gradual inclusion of foreign trade as a legitimate source of a country’s food supplies. ŒŒ A shift from food availability to food access and utilisation as central elements of food security. ŒŒ A stronger linkage of food security with economic and social development. ŒŒ A shift from ‘food only’ or ‘food first’ to a more inclusive ‘livelihood’ perspective. ŒŒ A tendency to consider not just objective situations of food shortage but also subjective perceptions of food security or insecurity. ŒŒ A tendency to consider market mechanisms and incentives, and the behaviour of firms and consumers as necessary components in the theory of food security. ŒŒ A widening of the concept to more explicitly include concerns about nutrition and food safety. This also includes complementing concerns 23

From hunger to food security: a conceptual history

about dietary energy (calories) and staple foods (cereals, tubers) with a growing consideration of micro-nutrients and ‘hidden hunger’. ŒŒ A growing linkage of food security with environmental concerns. ŒŒ A growing consideration of local food habits, the right to food, ‘food sovereignty’, and smallholder production. Some of these conceptual shifts are briefly discussed in the following sections.

2.3.1. From food supply to food access The most important conceptual change is undoubtedly the move from an emphasis on sufficient supply to an emphasis on individual access. What matters about hunger or food security is not the mere existence of food, but the possibility for people to acquire food and be able to consume it. This connects food security with poverty, income distribution, livelihoods, and the social and economic organisation of the food system, including food production as well as other aspects. This conceptual change is not only a change in emphasis, but the result of a realisation that the world has mostly overcome the ‘Malthusian’ condition in which humans have lived for millennia - a condition in which mere physical survival, the mere replacement of one generation by another, was perpetually threatened by hunger and early death (for illuminating analyses on the historical process of overcoming the ‘Malthusian’ era, see Fogel [2004] and the overarching theoretical framework of Galor [2011]). Once humans were able to produce more than enough for survival and life expectancy at birth could be extended almost threefold, from about 30 to over 80 years over the last three centuries, the problem was no longer centred on the supply of food (which is now sufficient to feed all humankind with a surplus) but on ensuring that everybody gets their share. This is mostly related to matters other than food production: it involves social and economic development, and with it, higher income, better skills, adequate health care, and a higher and more widespread capacity to acquire adequate food for everybody, every day.

2.3.2. From the nation to the household Consideration of inequality in food access produced a shift in emphasis from the aggregate level of the nation or the world to the micro level of individuals and households, as is usual when distribution is concerned. This made it

24

Hector Maletta

necessary to take urbanisation and markets into account. In the post-war years and, in many cases, up to the 1970s or later, most developing economies were seen as overwhelmingly rural, and rural populations in turn were seen as composed of subsistence farmers or pastoralists who relied little on markets for their food needs and had a limited capability to respond to market incentives. This view has been long discredited, as evidence to the contrary mounted. Untrue even in those times, it has been shown to be an even more inadequate description in recent decades. Subsistence farming families grew increasingly less able to live off their land, if they ever could, as many small holdings become subdivided through inheritance, the importance of markets increased in rural areas, and sources of rural income diversified. Rural dwellers increasingly grew cash crops and engaged in non-farm self-employment (in petty commerce and other trades) as well as farm and non-farm wage employment, often involving temporary migration to cities, other rural areas, or abroad. More permanent rural-urban migration intensified and, as a result, urban populations in developing countries increased very quickly during the second half of the 20th century. These processes meant that many developing nations became predominantly urban, and most of the rural population became market-dependent for most of their food needs (including profound changes in traditional diets). Migration also created a steady flow of remittances that became an ever larger part of Third World household budgets, including rural households. This growing reliance by peasants on cash income, mostly obtained off-farm, also was associated with an increase in per capita food consumption and a decrease in undernourishment and malnutrition in most countries, as we will see later in this book. The idealised notion of the self-sufficient peasant family tended to persist far longer than its real-life counterpart. One of the consequences of this biased view was that any sudden drop in subsistence agricultural production tended to be automatically interpreted as a food-access problem for rural families, which could only be solved by food aid since the families in question were supposed to have no other sources of income and to rely mostly on the food from their own farms. Commercial food supplies (from abroad or even from other regions of the country) were not regarded as a solution for peasants lacking self-sufficiency because peasants were supposed to lack other sources of income to buy food in the market. Adequate markets were also supposed not to exist in many places, or to be highly imperfect, dominated by greedy middlemen who would hoard the food and charge grossly distorted prices, starving the hungry farmers and their families. If the shortage was nationwide, importing food from abroad was also not seen as a legitimate arrangement, as it might create more ‘dependency’ and reduce food security (equated to 25

From hunger to food security: a conceptual history

food self-sufficiency at the national level). It would also use up the nation’s chronically scarce reserves of hard currency. Thus, donated foreign food aid distributed by the government (through non-market channels) to the affected people or to the food industry was thought to be the only solution – not only in the short term (e.g., after a crop failure due to severe drought), but also on a more permanent basis. Many countries started receiving a steady flow of regular food aid, the so-called ‘program food aid’ as distinct from ‘emergency food aid’. This strategy created disincentives for local producers, further distorted local diets, and generated long-term aid-dependency in beneficiary populations. Since the 1980s, the growing availability of survey information at the household level gradually cleared the fog and a better picture began to emerge of rural livelihoods, rural employment, rural income, food consumption patterns, seasonal migratory movements, and other essential data. Also, the rapid processes of urbanisation led analysts to shift the focus towards food security in urban areas, a problem much highlighted since the 1980s in studies on urban poverty and the ‘informal urban sector’. In many countries, the urban poor are more numerous than the rural poor, and many urban poor lack adequate food and nutrition. Also, in many countries a majority of subsistence farmers were shown to depend mostly on the market to acquire food, and to pursue diversified livelihoods relying on various sources of income. Thus, analyses of food security shifted emphasis from the national to the household level, defined indicators in terms of household food consumption, and also analysed intra-household differences in individual food consumption or nutritional status.

2.3.3. From food to livelihoods The notion of food security tends to isolate food needs from other needs, and agricultural production from other economic activities. However, in practice, people have a variety of simultaneous needs (food, water, shelter, clothing, health care, sanitation, education, and many more) that they must balance in terms of the effort and resources devoted to their satisfaction according to the ordering of preferences and budget constraints that guide the economic behaviour of individuals and households. Just as household needs are diverse, household livelihoods in both rural and urban areas include not only food production for subsistence, but also (and increasingly) other gainful activities and sources of income: cash crops, mining, manufacturing, transportation, construction, commerce and miscellaneous services. These activities are carried out by households through the use of a variety of assets: physical 26

Hector Maletta

(reproducible) means of production, natural resources, family labour, social relations and connections (social capital), and financial resources.2 Chambers and Conway (1992) produced one of the first formulations of the so-called sustainable livelihoods framework, which was widely used to define livelihood vulnerability in terms of the entire range of household needs and activities rather than focusing only on food. Frankenberg and McCaston (1998) was a key contribution to translating food security into the wider concept of livelihood security (see also Scoones [1998] and Swift and Hamilton [2001]). Rural livelihoods were no longer seen as being exclusively dependent on farming, but increasingly on diversified operations (Ellis 1998, 1999). This approach, which emphasises sustainable livelihoods, was readily complemented by the concept of sustainable development where environmental considerations were paramount. Thus, the concept of food security gradually incorporated human needs beyond food and other aspects of human activity aside from subsistence farming, and shifted attention towards environmental sustainability. Since the emphasis was on individuals, intra-household distribution of food also became an issue.3 They must use the proceeds not only to procure food but also to cover other needs. Livelihoods thus include multiple sources of income, and a diversified expenditure structure. Thus, the household (or family) came to be seen as an entity that uses resources to obtain revenue, and uses revenue to satisfy needs. Considering a family as a sort of ‘enterprise’, with assets that are mobilized to provide a means of living, is bound up with the concept of livelihoods, which gave rise to the ‘livelihood framework’ that has become dominant in the analysis of households and vulnerability in developing countries since the 1990s.4

Financial resources include money (cash or bank accounts) and other ways of storing purchasing power in more or less liquid forms (including traditional items like jewellery or livestock, or more formal ones like corporate stocks or government bonds). Access to additional credit (not actual indebtedness) is also a financial resource (indebtedness is a liability, not an asset). Access to remittances from emigrant household members and to other private or public transfers may be counted as revenue from social capital, broadly defined. 3 Though many of the concepts are different and incorporate elements from outside economics, the idea of the household as an economic unit using assets to produce goods, services and income in order to satisfy the needs of its members is closely related to the ‘economics of the family’ as sketchily proposed by some authors since the 1950s and formulated more systematically in the 1970s by Gary Becker (1976, 1981). 4 Studying and quantifying intra-household inequality in food access and nutrition poses many methodological difficulties. For an analysis of the issues involved see Haddad et al. (1997). Deaton (1997) and Lazear and Michael (1988) provide a more general review of intra-household distribution and its measurement through household surveys. 2

27

From hunger to food security: a conceptual history

2.3.4. Objective and subjective food security Up to the 1980s, food security was entirely centred on objective situations of food supply shortages or, to some extent, lack of access to food. Subjective feelings or beliefs, and reactions of economic agents to the food situation, were largely ignored. People classified as food secure or insecure may have had diverging opinions, but they were not asked. The classification of people as food secure or insecure was technocratic, not democratic. The emergence of the concept of subjective food security (or insecurity) marked another conceptual shift. A scale for measuring subjective food insecurity was introduced in the early 1990s, and then adopted as a regular module in the US Current Population Survey from 1995 (Radimer 1990, 2003; Radimer et al. 1990; Radimer et al. 1992; Bickel et al. 2000). This scale has been extended to other countries, both developed and developing (Frongillo et al. 1996; Frongillo 1999). In this approach, the household head or spouse is asked about his or her perceptions and expectations of the household’s ability to meet current and near-future food needs and actions taken (or intended to be taken shortly) in case of insufficient access to food, such as looking for extra income, reducing the size and number of meals, or curtailing the quality of food, etc. A number of such questions are combined into a scale that ranges from complete food security to severe food insecurity in subjective terms. This not only shifted the emphasis from objective to subjective indicators; it also shifted attention from the current or past situation to expectations or plans about the uncertain future. Thus, household risk assessment and management was introduced into the concept of food security. This differed from those forms of risk assessment conducted by technocrats in that it was undertaken by the potential victims of food insufficiency themselves. The severity of the risk is also indicated by the importance and subjective cost of remedial measures that households adopt or contemplate adopting as coping mechanisms (or strategies) to prevent or alleviate food insecurity (Maxwell 1996; Maxwell et al. 1999, 2008). Some of these mechanisms or strategies aim at increasing income or subsistence production, whereas others aim at reducing food intake or modifying intra-household distribution; some may be just short-term manoeuvres that soon run their course, while others may become permanent features of household behaviour. Some involve limited cost (e.g., doing without some ‘luxury’ foods) while others require substantial sacrifices (e.g., taking children out of school and sending them to work, entering ruinous indebtedness, or in some extreme cases, even selling a pre-adolescent daughter into forced marriage or sex slavery in order to reduce 28

Hector Maletta

household food needs and raise cash to feed the rest of the family). Some strategies are unsustainable in that they reduce household assets or future income (e.g., selling off the family livestock in order to buy food). The frequency and severity of (adopted or intended) coping strategies may be used as an indicator of the degree of objective or subjective food insecurity. Subjective food insecurity has thus become a frequent complement to more objective notions of food security. However, subjective food security can be misleading. Some people claiming to be food insecure are actually obese, or have sufficient access to food, or regularly eat much more than they need. On the other hand, some people in the poorest nations do not complain unless hunger is extreme simply because a degree of food insufficiency is habitual to them. More generally, different people feel food insecure at different levels of food access relative to their objective needs. However, an objective norm or standard still seems necessary.

2.3.5. Vulnerability and food insecurity The conceptual shifts from food to livelihoods and from objective to subjective food security (or insecurity) were also accompanied by increased prominence of the concept of vulnerability. The very notion of ‘security’ refers to a reliable expectation about the future. Such security would diminish if a threat to such expectations were perceived. From a subjective point of view, a person, household, or country feels ‘food secure’ when its present and future access to food is perceived as secured or certain. The chances of lacking food in the future are the very essence of vulnerability: the likelihood of experiencing hunger, or food insufficiency, at some point in the future. Since the 1990s, several authors using the ‘sustainable livelihoods’ approach tended to include vulnerability in the definition of food security (e.g., Maxwell 1996; Maxwell et al. 1999; Dilley and Boudreau 2001). This referred especially to the chances of having spells of food deprivation or scarcity, more than a chronic situation of insufficient access to food. Thus, for instance, the food supply of subsistence peasants who obtain their food from their farms in a semi-arid environment could be seen as vulnerable to drought. Likewise, people with informal and precarious livelihoods (street hawkers, casual labourers, and the like) are also likely to face periods of scarce income and therefore to be exposed to spells of reduced access to food. It is important to note that a person or household may be vulnerable to food shortage even if they currently have adequate access to food – the concept of vulnerability has an intrinsic inter-temporal dimension. Subjective perceptions of vulnerability 29

From hunger to food security: a conceptual history

also provide important clues about realities faced by households, and also constitute behavioural motivations to engage in preventive measures aimed at reducing the risk as well as reactive measures to reduce harm caused by a spell of poor income or food shortage. Another possible approach to this issue, based on objective information rather than subjective feelings, rests on the conceptual identification of vulnerability with instability or variance. If food supply or food access varies over time, a family obtaining just above the minimum amount of food may be deemed vulnerable in the sense that observed variance over time is an indicator of the chances for that family to fall beneath the minimum level. The same approach has been used (especially in a number of World Bank studies) to address vulnerability to poverty; households with incomes just above the poverty line may suffer variation of income over time, and these variations may put them below the poverty line at some point or another. The livelihoods approach encompasses a more elaborate view of vulnerability. Essential concepts in this regard are shocks and coping strategies. Any household with a certain habitual livelihood may suffer external shocks (a drought, a spell of unemployment, a surge in food prices) which may entail inadequate food access at least for a time; households must ‘cope’ with such emergencies by resorting to various ‘coping mechanisms’ or ‘coping strategies’, and may also have some kind of ‘insurance mechanism’ in place to make them resilient to such events. Coping strategies are many, and differ in their degree of ‘severity’ and sustainability (Maxwell et al. 1999). Severity is hard to define in this context, but the notion intuitively refers to the (pecuniary or other) costs of adopting a strategy. The frequency and severity of coping strategies may be used as an indicator of the seriousness people attribute to their predicament concerning security in food access. Food security vulnerability also has been assessed in relation to household assets, and to subjective perceptions of food insecurity. In the sustainable livelihoods framework (Scoones 1998), paucity of assets is the main factor determining the resilience or vulnerability of a livelihood. These assets can be broken down into five categories: reproducible physical assets, access to natural resources, financial capital, human capital, and social capital. All household activities (productive or reproductive in the broad sense) involve mobilising and using the assets. Preserving or losing assets is a key factor of continued resilience or increased vulnerability; this applies to all kinds of risks, including food insecurity as well as other kinds of harmful deprivation. Through its link to vulnerability and livelihoods, food security is linked to risk, uncertainty, 30

Hector Maletta

emergencies, and coping strategies that households may adopt to prevent or cope with such circumstances. Thus, the evolving concept of food security, formulated nowadays in terms of livelihoods, includes an implicit or explicit regard for risk reduction. The risk in question is not the same for everybody; the widespread (though seldom clearly defined) notion of vulnerability, which is essential to the sustainable livelihoods framework, is rooted in the idea that some individuals or households are more vulnerable than others, i.e., more likely to suffer the effects of adverse situations in the future, or less likely to adapt to such adversities, because of their paucity of assets, lack of resilience, or higher exposure to external shocks. Vulnerability is mainly a function of a household’s characteristics, chiefly its various assets (physical, natural, financial, human, and social) and its capability to use them in order to satisfy needs and respond to shocks. Assets are not static endowments; households may get richer or poorer and can gain or lose resilience as they accumulate, lose, or liquidate assets, and as their capabilities are enhanced or diminished. Their likelihood of withstanding future shocks, especially in the long run, will depend not on their current but on their future assets and capabilities, which are also subject to uncertainty and may be influenced by today’s decisions.

2.4. The World Food Summit definition The 1980s and early 1990s witnessed a string of financial crises and economic and financial adjustment processes in developing economies, as countries tried to cope with a changing international environment. During the same period, international agencies and donors started giving more attention to the social costs of structural adjustment, including influential studies such as Cornia et al. (1987) and the book by Amartya Sen and Jean Drèze (1989), both sponsored by United Nations agencies. In this context, emphasis shifted to poverty reduction as the main way to reduce food insecurity and hunger. A major contribution to this shift of emphasis was the World Bank’s 1990 International Development Report devoted to poverty (WB 1990), in which the bank introduced its universal poverty lines of one and two USD a day per capita (in PPP USD at 1985 prices, later updated). In this context, FAO organised a World Food Summit which convened in Rome in 1996. The Summit redefined food security and its new definition was ratified (with minor variants) by later such summits in 2002, 2005, and 2009. Food security was re-defined entirely in terms of access and entirely in terms of individuals. 31

From hunger to food security: a conceptual history

The official definition in the 1996 Summit’s Plan of Action (WFS 1996) was as follows: Food security exists when all people, at all times, have physical and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life.

This definition became the one that is universally used; the only change was the more recent addition of ‘social access to food’ to take into account food deprivation caused by social custom or other noneconomic reasons such as ethnic or gender discrimination (for instance, the fact that in some cultures, women routinely receive less than their fair share of the food accessed by the household). With this modification, the definition requires that all people at all times have physical, social, and economic access to food, leaving the rest of the definition untouched. This was the version sanctioned by the World Food Summit of 2009 (WFS 2009, footnote 1). Thus redefined, food security is not primarily concerned with aggregate balance between global or national production and demand, but is centred instead on individual access to food. This definition required a shift from the country to the individual level and from an emphasis on availability to one on access. Of course, availability is a necessary condition for access, but it is not of itself a sufficient condition. Access also requires the capacity to acquire food at the individual or household level and is thus intimately related to poverty. Availability, in turn, does not equate to domestic production since food may be made available through importation. Trade has become a component of food security. Along with this radical change in perspective, this definition includes a number of other innovations. Besides incorporating the notion that food should be quantitatively sufficient and accessible, the definition also includes food safety and nutritional needs and recognises food preferences, therefore giving high priority to the qualitative and subjective aspects of food. The reference to an active and healthy life means that the normative level of food consumption is not to be restricted to the bare minimum required for survival but should be enough to sustain physical activity and health throughout life. This may also include the food required to sustain a certain level of voluntary physical activity by individuals who do not do heavy physical work in order to avoid the dangers of sedentary lifestyles. Economic, social, political, and

32

Hector Maletta

environmental factors are also emphasised. The 1996 Summit declaration, for instance, stated: Poverty is a major cause of food insecurity, and sustainable progress in poverty eradication is critical to improve access to food. Conflict, terrorism, corruption and environmental degradation also contribute significantly to food insecurity. Increased food production, including staple food, must be undertaken. This should happen within the framework of sustainable management of natural resources, elimination of unsustainable patterns of consumption and production, particularly in industrialized countries, and early stabilisation of the world population.

The notion that food security implies access to food ‘at all times’ entails an inter-temporal view of food security, requiring not only stability of food supplies over time but also ensuring people’s future ability to access food (e.g., by having a steady source of income enabling them to acquire food). Since we are uncertain about the future, the concept of food security ‘at all times’ implies an evaluation of the probability of having access to food at different times in the future; thus, an uncertainty and risk component was (rather obliquely) introduced into the concept of food security, although that element has not as yet been fully incorporated into standard operational indicators. All this makes evaluating long term food security especially difficult. As a matter of fact, most analyses of food security confine themselves to the present, the very recent past or the immediate future. This is a severe limitation for analyses on the impact of long term processes (such as climate change or economic growth) on long term food security. On the other hand, short-term food security concerns are strongly motivated by events such as economic downturns, wars, and natural disasters, rather than long-standing structural factors and high prevalence of chronic lack of adequate food access. Last but not least, as noted above, the importance of trade in addressing local food shortages was more explicitly underlined in the 1990s. The 1996 Rome Declaration emphatically stated that: We agree that trade is a key element in achieving food security. We agree to pursue food trade and overall trade policies that will encourage our producers and consumers to utilise available resources in an economically sound and sustainable manner. [...] We will strive to ensure that food, agricultural trade and overall trade policies are conducive to fostering food security for all through a fair and market-oriented world trade system. 33

From hunger to food security: a conceptual history

The notion of trade as ‘a key element in achieving food security’ was in stark contrast with the traditional tendency to regard food imports as a sign of food insecurity. Even before the 1996 Summit, an early indication of the shift away from a definition based on self-sufficiency was the fact that commercial food imports were recognised in practical assessments of food security as a legitimate part of domestic food availability; only the ‘uncovered gaps’ (i.e., food needs not covered by local food output and also above the country’s financial capacity to import) were considered as an indication of a food shortage at the national level. This can be seen for instance in the methodology used by FAO’s Global Information and Early Warning System (GIEWS) to foresee and assess emergency food aid needs.5 Ex ante food security prospects assess availability of food from domestic production in addition to the ‘normal’ or expected level of commercial food imports, and consider the remaining gap that could not possibly be covered by the country’s capacity to import (or to subsidize imported food) as ‘food aid needs’. The second World Food Summit, held in 2002, reaffirmed ‘that trade is a key element in achieving world food security.’ The final Declaration of the High-Level Conference on Food Security held in Rome in June 2008, in the midst of an acute food price spike in international markets, also committed to efforts ‘to help farmers, particularly small-scale producers, increase production and integrate with local, regional, and international markets’ and urged ‘the international community to continue its efforts in liberalising international trade in agriculture by reducing trade barriers and market distorting policies.’ The 2009 World Summit once again reaffirmed the same views. There seems to be a consensus that a more liberalised international market for agricultural commodities would help improve food access worldwide and reduce world food insecurity, especially in a context of high international food prices. Such a market would also help poor countries to raise export

The GIEWS (http://www.fao.org/giews/english/index.htm) was established after the 1974 World Food Conference. GIEWS routinely conducts Crop and Food Supply Assessment Missions (CFSAM) to estimate food aid needs in countries facing a present or impending food emergency (CFSAM methodology: http://www.fao.org/docrep/011/i0515e/i0515e00.htm). CFSAMs are the objective basis for decisions by food aid donors, and also provide guidance for the UN World Food Programme (WFP) which is in charge of food aid delivery and targeting. WFP participates in CFSAMs to assess the numbers and location of the most vulnerable groups in need of food aid; donors also take part as observers. 5

34

Hector Maletta

revenue and accelerate development by enabling access for their products to markets in developed countries, some of which are still heavily protected.6

2.5. Dimensions of food security Even before the 1996 Summit, analyses of food security had gradually expanded their focus away from the sole consideration of (aggregate) availability to include other dimensions. A common formulation is to distinguish between four dimensions or ‘pillars’ of food security, implicit in the Summit definition: ŒŒ Availability of food - from production, trade, and existing stocks, minus non-food uses. ŒŒ Access to food (physical and principally economic access). ŒŒ Biological utilisation of food in the human body - affected by infections. ŒŒ Stability of food availability, access, and utilisation. These dimensions have often been examined separately and on an unequal footing, instead of being regarded as essential components. Sometimes availability is the dominant concern, sometimes food utilisation (nutrition) is the focus, sometimes access. Some analyses focus on chronic or habitual conditions, others on seasonal fluctuations or occasional emergencies. In fact, all the dimensions are essential. Emphasis has shifted since the 1990s towards priority of access, thus moving the focus away from national or regional aggregates towards households and individuals. The 1996 Summit definition responds to the perceived centrality of access. Availability is thus seen as a necessary (albeit not sufficient) condition for access, and access is also required to be stable and reliable. Moreover, while shifting emphasis from collective aggregates (such as nations) down to individuals, access comes to embrace biological utilisation as well; a person has actual ‘access’ to food or to food’s nutritional components (energy, Historically, the pursuit of self-sufficiency has not been successful. Self-sufficiency policies in Maoist China and North Korea failed to ensure adequate food supply and produced several bouts of famine. Recent food export bans (wheat in Argentina, maize in Bolivia, and similar policies in Russia), adopted in the wake of food price surges in the late 2000s, caused declines in domestic output and increases in imports, and did not enhance or cheapen domestic consumption. See Maletta (2013a) for the Bolivia case as compared with Peru’s more relaxed market-oriented policy (Maletta 2013b). 6

35

From hunger to food security: a conceptual history

protein, vitamins, and minerals) only to the extent that infections and disease do not hamper their biological utilisation in the body. As a matter of fact, this explicit inclusion of the body’s ability to process and use food shifts the emphasis even further down the chain, from individuals to their organs, tissues, cells, and metabolic processes. Food security requires not only that food be reliably available and (physically and economically) accessible to households, but also that individuals within households have effective access to food and are able (on a continuous basis) to biologically utilise it within their bodies. Thus, food security may decrease due to outbreaks of disease, even in the absence of any problem with food availability and access. Access by individuals and not only by households also draws attention to the intrahousehold distribution of food - a common issue in developing countries (see Haddad et al. [1997] for a wide-ranging examination of intra-household allocation of food and other resources). Access to adequate food by all people at all times - the definition of food security - is a prerequisite to achieving certain desirable outcomes. The main outcome of a food security condition is sound nutritional status on the part of the individuals making up a population. However, indicators of food security per se, concerned chiefly with food availability and access, are often produced separately from indicators on nutritional status. In fact, different world organisations are charged with maintaining the databases that are relevant to each issue (FAO for food security, WHO for nutrition). There is, in fact, a high correlation between the two families of indicators, but the databases have a different structure - resulting in non-comparable coverage, different frequency of the data, and generally differing availability of information depending on the problem at hand. Some of the most pressing problems, such as consumption of micro-nutrients (vitamins and minerals) are not currently included in either group of indicators. Indicators of child malnutrition are not routinely collected on an annual basis but in sample surveys (albeit not in all countries) every four or more years, in contrast with indicators of food availability and access produced on an annual or seasonal basis in most countries. Thus, besides the complex shifts in ideas that have completely transformed the concepts used to assess hunger - from national self-sufficiency in staple food to actual measurement of individual intake or subjective feelings of food deprivation - there is also widespread heterogeneity as regards methods for collecting and processing the relevant information. Hunger measurement remains elusive to a certain extent, especially for the less visible forms of hunger such as micronutrient deficiency or intra-household access inequality, 36

Hector Maletta

which are seldom, if ever, measured in a systematic way. Traditional indicators related to calories may no longer be the most significant ones in a world where other concerns take precedence, such as micronutrients, food quality, diet imbalance, saturated fats, or food waste. Most of these areas are still not subject to systematic collection of routine statistical data around the world. Most people will recognise extreme cases of hunger, such as those that occur during famines or in exceptional circumstances such as war or severe natural disasters, but the hidden forms of food inadequacy are more difficult to gauge. In addition to the need to ensure better conceptual definitions, there is the challenge of devising better measurement methods and more systematic collection of the right data. Not all dimensions of hunger, then, are regularly monitored on a worldwide basis. The main indicators are the prevalence of undernourishment or inadequate food access (FAO) and anthropometric measures of nutritional status (measured in health and nutrition surveys assembled by WHO). Undernourishment is defined as habitual access to an amount of dietary energy that is below the minimum amount compatible with good health. Major anthropometric indicators are based on height and weight, mainly those calculated for children under five and centred on failure to attain adequate height and weight for their age, as well as those calculating overweight and obesity among people of all ages. The major database on food production and trade around the world is FAOSTAT, the statistical system maintained by FAO, by means of which the prevalence of undernourishment is estimated. The WHO, for its part, maintains a database of anthropometric measurements of nutritional status and other related indicators. We will make extensive use of both these sources. There are no regular worldwide statistics on the short term (seasonal) stability of food consumption. Also, since undernourishment indicators are based on the supply of dietary energy but no regularly maintained worldwide database exists on the supply of micronutrients, the latter can only be gauged indirectly on the basis of consumption of nutrient-rich foods, such as fruit or vegetables.

2.5.1. Food security and nutrition In parallel with conceptual developments regarding food security, there emerged a framework for conceptually organising the efforts to fight malnutrition. This framework is related not only to food, but also to sanitation, health care, and other issues. In fact, once food intake has taken place, the biological utilisation of food within the body may be hampered by bad health. 37

From hunger to food security: a conceptual history

Thus, for instance, gastro-intestinal infections may cause diarrhoea and the consequent loss of water and nutrients; any infection (gastro-intestinal or not) may cause unwanted dissipation of dietary energy in the form of heat (fever). Unsafe water supply is a central cause of infection, and dehydration is a major consequence of diarrhoea, often resulting in death. Proper feeding of infants (especially through maternal lactation) provides not only a safe and adequate supply of nutrients but also ensures good health, especially in areas without adequate water supply. Thus, adequate feeding patterns, water supply, sanitation, hygiene, health care, and other similar components are at the forefront of concerns regarding nutrition, in addition to adequate access to food. Indicators of nutrition also differ from those of food availability and access. They are mostly centred on the anthropometric outcome of good nutrition, especially in children; the most widely used ones are weight and height, whereby actual measurements of children or adults are compared to standards based on healthy and well-nourished individuals; WHO (1995, 2007) are currently the main anthropometric standards. Other nutritional indicators look for signs of micro-nutrient deficit; this includes measuring iron in blood samples, or calcium in bones, and may also include looking for clinical and epidemiological evidence of related diseases such as scurvy (due to lack of vitamin C), night blindness (vitamin A deficiency), or kwashiorkor (massive energy-protein deficiency). Clinical and biochemical indicators are complex and expensive, and are seldom applied on a large scale. However, some demographic and health surveys take blood samples from surveyed individuals, especially mothers and young children, most frequently to look for evidence of anaemia (i.e., iron deficiency). Some diseases are extreme outcomes only observed in situations of famine. The first conceptual formulation linking food security with these issues, beyond general concerns about nutrition, was the development of the concept of ‘nutrition security’ during the 1990s (see a historical account in CFS [2012]). While ‘food security’ was mainly promoted by FAO, ‘nutrition security’ was a concept initially raised by other international bodies such as the World Health Organization (WHO) and UNICEF. This does not mean that these organisations worked separately; in fact, FAO and the Committee for World Food Security have incorporated the concept of nutrition security or some variant thereof. FAO proposed the following formulation:

38

Hector Maletta

Nutrition Security is achieved when secure access to an appropriately nutritious diet is coupled with a sanitary environment, adequate health services and care, to ensure a healthy and active life for all household members (FAO 2011c:10).

This was expanded upon by the Committee for World Food Security (CFS 2012), which discussed several terminological variants. While the concept of food security is centred on access to food by households and people, nutrition security, according to the CFS, focuses on actual consumption and biological utilisation of food and nutrients by individuals. On the concept of nutrition security, the Committee stated that: ‘It combines having access to adequate food that fully satisfies nutritional needs with non-food factors that enable a person to metabolise their food and use the nutrients to support growth and maintenance of the body and to carry out basic life functions’ (CFS 2012:6). The multi-agency Committee ultimately opted for the combined term ‘food and nutrition security’ to emphasise both access and utilisation, and recommended an encompassing definition: Food and nutrition security exists when all people at all times have physical, social and economic access to food, which is safe and consumed in sufficient quantity and quality to meet their dietary needs and food preferences, and is supported by an environment of adequate sanitation, health services and care, allowing for a healthy and active life. (CFS 2012: 8)

This definition includes actual consumption of food and not just access to food, and emphasises the requirement of adequate conditions regarding sanitation and health. Behind these terminological variations, and aside from the concern for conceptual precision, lies the need for a common language that recognises and embodies the roles and functions of the various international agencies involved in the matter, allowing them to engage in the fight against hunger with their own particular emphases and nuances. In practice, the concept of food security is the one that is more widely used, with the understanding that it includes all aspects concerning nutrition (i.e., utilisation) and not just availability and access, as well as the stability of all these dimensions over time. In the following chapters, we will analyse trends and prospects regarding food availability, food access, and malnutrition, with some references to stability. The analysis is centred on figures for major world regions, and only occasionally downscaling to specific countries. Past trends (chs. 3-8) refer mainly to the past half century, i.e., from the early 1960s to the early 2010s. Most data are taken from international databases on agricultural production and nutrition, based on national statistics and those compiled by UN specialised agencies 39

From hunger to food security: a conceptual history

such as FAO and WHO. Future prospects (reviewed in chs. 9-11) are based on projections prepared by broad-based teams at international agencies (e.g., FAO), research institutions (like IIASA or IFPRI), or (for climate change) the IPCC. Many aspects of our subject require careful attention to data and methods, which are often disregarded or only cursorily considered by authors and readers alike. In fact, many otherwise excellent efforts at analysing the problems of food and hunger suffer from a failure to consider and adequately address certain obscure but crucial methodological issues. Details on data sources as well as various methodological and technical aspects are discussed in the Methodological and Technical Appendix (chs. 12-13). We strongly recommend paying due attention to the technical details, since many misunderstandings about food and hunger are rooted in the finer points of definitions, indicators, and other technical matters.

40

Part II. HISTORICAL TRENDS

3. Agricultural and food output

3.1. Data and regions This chapter examines trends in food production, on the world scale and for major world regions. The main source of information on production, land use, trade, and consumption is FAO’s FAOSTAT database (http://faostat.fao. org), which consolidates national agricultural statistics in a harmonised way (see Ch. 12 for technical details).7 While the analysis here is mostly on the world and regional levels (see ‘Regions’ box), data for specific countries (and also more detailed information about specific agricultural products) are readily available in the FAOSTAT database.

FAOSTAT-based analyses in this book (chs. 3-6) mostly refer to the half century from 1961 to 2011, though some figures and tables also include data for more recent years, such as 2012 or 2013, made available up to April 2015.

7

43

Agricultural and food output

Regions FAOSTAT data on production, trade, and consumption are reclassified into five major world regions: • Africa (including North Africa as well as Sub-Saharan Africa) • Asia (including Oceania/Pacific and excluding Asian countries that were part of the former USSR) • Latin America and the Caribbean (or LAC) • North America (USA and Canada) • Europe (including parts of the former USSR technically located in central and western Asia) Putting the Asian parts of the former USSR within the ‘Europe’ region is necessary for intertemporal consistency since the entire USSR was included in ‘Europe’ up to its dissolution in 1991. Some data from other sources come with their own regional subdivisions, which may group countries in a somewhat different way. See Section 12.1 for more details.

3.2. Aggregation and measurement of world food output FAO agricultural statistics provide data on the output of food and non-food agricultural products. Food production refers only to agricultural products that are used (at least in part) for food purposes (e.g., maize, meat, apples), and thus excludes non-food farm products (e.g., wool or tobacco).8 An analysis is required of how the amount and growth of agricultural production, whether food or non-food, is assessed. It is not enough to simply add up various products by weight, and to express agricultural or food output in metric tons.9 For instance, one cannot add apples and oranges, or worse still, chicken and watermelons, vegetable oil and potatoes, milk and tomatoes, and so on with hundreds of heterogeneous products. Nor is it sufficient to Food products include all farm products used for food, even if part of their output is actually used for other purposes (e.g., as animal fodder). Stimulants (coffee, tea) are not classified as food products because they do not contribute significant amounts of dietary energy or nutrients. Gross production refers to entire farm output; net production excludes the amounts used within the farming sector as seed or feed. Actual food supply in a particular territory equals domestic production, minus net exports, minus domestic stock change, minus domestic non-food uses like seed, feed, non-food industries, and minus post-production losses along the processing and marketing chain. See Ch. 12 for details. 9 The terms ‘metric ton’ and ‘tonne’ are used interchangeably throughout this book. Their respective abbreviations are ‘MT’ and ‘t’. 8

44

Hector Maletta

add up food products by the dietary energy they can generate, i.e., in calories or joules. Food provides not just energy but also protein, vitamins, and minerals, and there is no simple way to reduce all these nutritional ingredients to a single figure. Some foods produce ‘empty’ calories devoid of any other nutrient (e.g., refined sugar), some are almost pure fat (butter or vegetable oil), and some foods rich in vitamins and minerals provide small or negligible amounts of energy (e.g., broccoli or lettuce). A common way to aggregate heterogeneous goods is to add them up by price. The economic value of a product is a synthesis of the various factors involved in its production and use: the cost of producing it, its objective qualities related to demand, and the various influences of taste, culture, quality, custom, and fashion that happen to increase or decrease its worth in the eyes of individuals. However, as in other problems of economic aggregation, the simple addition of monetary amounts may be misleading: prices change over time and differ from one place to another, and exchange rates used to convert currencies often result in acute misalignment in price levels between countries. As a result of these factors, the real size of agricultural or food production, when valued at market prices and market rates of exchange, cannot be reliably compared over time or across borders. As in other economic aggregations, a measure of real output must be adjusted for the factors mentioned, ensuring that differences in value reflect actual differences in physical quantity, and not inflation over time, price differences across countries, or the varying purchasing power of various currencies. A good measure of overall food output and its growth over time should reflect the real value of food production, tallied at constant and uniform prices (constant over time, uniform across countries), correcting for inflation, and accounting for inter-country differences in relative prices and misalignments in the purchasing power of money. Such a measure would be a reliable index of the physical quantum of production, irrespective of differences or fluctuations in prices or currencies. There is a further reason for choosing economic rather than physical measures: the main purpose of assessing agricultural and food output in the present context is concern about food security. Since food security is defined in terms of access to food, the economic value of food is the relevant measure to assess real food production and trade. At every income level, people choose a particular combination of food items for their food basket - a choice conditioned by income, food preferences, and food prices. To compare these choices over time and across borders, adjustments are needed for inflation, relative prices, and purchasing power. A similar approach is usually followed for internationally comparable measurements of income and poverty. 45

Agricultural and food output

This kind of measurement has been available since 1961 in FAOSTAT estimates of the value of agricultural production at constant world-average producer prices expressed in USD of uniform purchasing power by means of agricultural purchasing power parity (PPP) conversion rates. Output of various crops and livestock is aggregated in value terms, using the output-weighted world average of domestic producer prices for a certain base period. These prices, originally expressed in local currencies, are converted into a common currency (the US dollar) at purchasing power parity (PPP) conversion rates. Such prices and PPP rates are computed by the so-called Geary-Khamis approach. Methodological details are discussed in Ch. 12, Section 12.2. This procedure corrects for (a) inflation over time, (b) price differences in agricultural products across countries, and (c) misalignments of exchange rates causing variability over time and across countries in the purchasing power of money relative to agricultural commodities. It is thus a quantity index, a measure of real output level and real output growth (see details in Rao 1993 and FAOSTAT metadata). At the time of writing, FAOSTAT figures for value of production since 1961 are based on the 2004-2006 world average of producer prices. Choice of base period, however, does not significantly alter conclusions about trends, as shown in Section 12.2.4. It is important to keep in mind that FAO’s estimates of the value of agricultural or food production refer only to products from crops and livestock, excluding fishery production. Statistics on physical production of fish and seafood are available in FAOSTAT, but not their value at PPP-adjusted producer prices. Valuation at export prices (expressed in USD at market exchange rates) are also available, but not for the entire half century analysed here. Fishery production, however, represents only a small proportion as compared to agricultural production. At current prices and market exchange rates, the 2010 world value of production from fishery capture and aquaculture was 217 billion USD (FAO 2012a: 51), or about 6% of the world gross output of crops and livestock, worth 3,309 billion USD in the same year (also at market exchange rates); fish and seafood supply was just 34 kilocalories/ person/day, or 1.3% of total dietary energy at the world level (FAOSTAT Food Balance Sheets). Fishery physical output, on the other hand, has been growing in parallel to farm food products (see box ‘What about fish?’ in Section 3.3).

46

Hector Maletta

3.3. Growth of world food production World food output has grown strongly and steadily in recent decades, clearly ahead of population, both at the world level and in every major region, which has ensured steadily increasing food availability per person. This includes basic staple foods such as cereals and also every other category of food (meat, dairy products, fruit, vegetables, and more) and also total real food output. According to FAO estimates, output has been rising in real terms since the beginning of its series. The real net value of agricultural and food production (at constant and uniform 2004-2006 world-average producer prices in PPP USD) grew steadily from 1961 (Figure 1).10 Except for some very minor oscillations, almost every year has been an all-time high.

Billion PPP USD at 2004-06 prices

Figure 1. Net value of crop and livestock output (food and non-food products), 1961-2013, in billion USD, valued at world-average producer prices of 20042006, converted into USD at agricultural PPP conversion rates. 2500 2000 1500 1000 500 0 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001 2005 2009 2013 Food

Non-food

Source: FAOSTAT (retrieved April 2015).

To avoid duplication, net output refers to output leaving the agricultural sector of each country. It excludes the amounts of crop and livestock products used for feed and seed within the agricultural sector of the same country. However, gross output grew approximately at the same rate. Net output represents about 90% of gross agricultural and food outputs; the net/gross ratio for agricultural production has been rising in recent decades, from 89% in 1961 to 92% in 2013. For food products, the increase was similar, from 88% to 92%, indicating in both cases a slightly diminishing share of feed and seed in the gross value of production. Food products represent about 95% of all agricultural value of production. This percentage has been gradually (albeit slowly) increasing, from about 94% in 1961 to over 96% in 2013.

10

47

Agricultural and food output

The growth rate of all agricultural production, as well as food production alone, was about 2.5% per year (mean annual growth in the period from 1961-1963 to 2009-2011, using three-year averages for a more reliable estimate). The growth of food production was slightly faster than the growth of all agriculture since non-food production grew significantly less and more irregularly, at a yearly rate of 1.66%. The average annual growth rates of all agricultural production were similar in all intervening decades, as shown in Table 1. Thus, world food production maintained a near-stable rate of growth for half a century, around 2.5% per year, with only slight variation from one decade to the next, fluctuating within a range from 2.3% to 2.8%. Table 1. Annual growth rates of net value of agricultural production, food and non-food products, 1961-1963 to 2009-2011 1961-1963 1961-1963 1969-1971 1979-1981 1989-1991 1999-1901 to 2009-2011 to 1969-1971 to 1979-1981 to 1989-1991 to 1999-2001 to 2009-2011 Agriculture

2.48%

2.63%

2.35%

2.39%

2.24%

2.80%

Food

2.52%

2.68%

2.40%

2.40%

2.33%

2.81%

Non-food

1.66%

1.78%

1.42%

2.46%

0.26%

2.41%

Net production valued at world-average producer prices, converted into USD at agricultural PPP conversion rates. Net production excludes products used as seed or feed. See Section 12.2. Source: FAOSTAT.

During this half century (1961-2011), world food production more than trebled in real terms, while population merely doubled; consequently, per capita food output grew by 50% (Table 2 and Figure 2). Thus, there has been positive overall growth in food production, both total and per capita. Since the 1970s, moreover, growth rates of total and per capita output have tended to accelerate over time.

48

Hector Maletta

Table 2. Growth of world total and per capita net real output of food products (valued at constant 2004-2006 world average producer prices in PPP USD), 1961-1963 to 2009-2011. Index numbers (1961-1963=100)

Annual growth rate (%) Farms' Per capita net food Population food output prod.

Period

Farms' net food output

Population

Per capita food prod.

Period

1961-63

100

100

100

1961-63 to 2009-11

2.52%

1.66%

0.82%

1969-71

124

117

106

1961-63 to 1969-71

2.68%

2.04%

0.63%

1979-81

157

142

111

1969-71 to 1979-81

2.40%

1.89%

0.42%

1989-91

199

169

118

1979-81 to 1989-91

2.40%

1.80%

0.63%

1999-01

250

195

128

1989-91 to 1999-01

2.33%

1.42%

0.85%

2009-11

330

223

148

1999-01 to 2009-11

2.81%

1.31%

1.44%

Includes food products from crops and livestock (fisheries excluded). Net food production is total gross value of food products minus food products used as seed or feed. See Technical and Methodological Appendix. Source: FAOSTAT.

PPP Dollars per capita (2004-06 prices)

Figure 2. Per capita world real net food output (crops and livestock), at 20042006 world-average producer prices in PPP USD, 1961-2013. $320

$280

$240

$200

$160 1961

1971

1981

1991

2001

2011

Source: FAOSTAT (retrieved April 2015).

Over the long term, the annual growth rate in real per capita food output averaged 0.82% per year, but with a tendency toward speedier growth in recent decades, as shown in Table 2. Per capita output accelerated from 0.42% per year in the 1970s to a yearly 1.44% in the 2000s. This acceleration of growth in per capita food output was mostly due to slower demographic growth rates, which gradually defused the ‘population bomb’ greatly feared in the 1960s and 1970s (e.g., Ehrlich 1968). 49

Agricultural and food output

Population growth rates (average per decade) peaked at 2.04% in the 1960s but steadily decreased to 1.2% in the 2010s, and are expected to keep on decreasing. This was the principal factor of acceleration in per capita food output growth, though total output also grew faster in the 2000s than in previous decades. What about fish? Fish is a component of food supply, but it is not included in FAOSTAT estimates of the value of agricultural production at PPP producer prices. Thus, it cannot be added to the remaining agricultural or food output in terms of value. Fish and seafood are, however, reported in physical terms in FAOSTAT’s fishery section, and also in food balance sheets, commodity balances, and undernourishment statistics. Therefore, we include fish and seafood in our analysis of food consumption (Ch. 6). Their dollar value is reported in FAOSTAT trade statistics and in FAO fishery yearbooks. Fish and seafood make up a small part of the global food supply: at the world level (as of 2011) these products supply just 34 daily kilocalories per capita (1.18% of all dietary energy), and 6% of total protein. These amounts are not large enough to alter output trends based on crops and livestock. Fish and seafood production tonnage increased from an index of 100 in 1961-1963 to 351 in 2009-2011, growing at a yearly 2.65% (Table 3). Per capita fish and seafood output actually decreased in the 1970s, but has been growing in more recent decades. Overall, from 19611963 to 2009-2011, per capita fish output tonnage increased by nearly 60%. The per capita growth rate for fish and seafood tonnage was 0.95% for the whole half century, but fisheries total and per capita growth rates peaked in the 1980s and have decelerated in recent decades.

Table 3. Growth of total and per capita physical output of fish and seafood (tonnes), 1961-1963 to 2009-2011. Index numbers (1961-1963=100) Period

Fish/seafood Population production

Annual growth rate (%)

Per capita production

Period

Fish/seafood Population production

Per capita production

1961-63

100

100

100.0

1961-63 to 2009-11

2.65%

1.66%

0.95%

1969-71

151

117

128

1961-63 to 1969-71

5.29%

2.04%

2.58%

1979-81

172

142

122

1969-71 to 1979-81

1.34%

1.89%

-0.61%

1989-91

235

169

139

1979-81 to 1989-91

3.14%

1.80%

1.36%

1999-01

295

195

151

1989-91 to 1999-01

2.31%

1.42%

0.86%

2009-11

351

223

158

1999-01 to 2009-11

1.76%

1.31%

0.45%

Source: FAOSTAT.

50

Hector Maletta

3.4. Growth in regional food output Growth in food production was uneven across regions (Table 4). As indicated above, the world’s food output grew at 2.52% per year from 1961-1963 to 2009-2011. Long-term annual growth was faster than average in regions dominated by developing and emerging countries such as Asia (3.50%), Latin America (3.24%), and Africa (2.92%). It was much lower than average in developed regions like North America (1.80%) and Europe (0.80%). Table 4. Annual growth rates of net real food production from 1961-1963 to 2009-2011, by regions, at constant 2004-2006 world average PPP producer prices. World Africa Asia LAC N. America Europe

1961-1963 to 2009-2011 2.52% 2.92% 3.50% 3.24% 1.80% 0.80%

1961-1963 to 1969-1971 2.68% 2.94% 3.11% 3.37% 2.18% 2.23%

1969-1971 to 1979-1981 2.40% 1.58% 3.00% 3.28% 2.63% 1.66%

1979-1981 to 1989-1991 2.40% 3.14% 3.92% 2.51% 0.81% 0.90%

1989-1991 to 1999-2001 2.33% 3.09% 3.89% 3.36% 2.10% -1.18%

1999-2001 to 2009-2011 2.81% 3.85% 3.52% 3.73% 1.35% 0.70%

Source: FAOSTAT. See notes to Table 2 and the Regions box in Section §3.1.

A remarkable feature of these data concerns Africa - its food output has been growing rapidly in recent decades. This output increased by 3.85% per year in the last decade considered, 1999-2001 to 2009-2011 - faster than in Asia (3.52%) and Latin America (3.73%). African annual food output growth had slowed down from 2.94% in the 1960s to 1.58% in the 1970s, but grew faster than 3% per year over the following three decades, and faster in the 2000s relative to the 1980s and 1990s. In contrast, food output in Asia has slowed down somewhat in recent years, growing faster in the 1980s (at 3.92%) and 1990s (at 3.89%) than it did in the 2000s (though its rate of growth in that decade was a still strong 3.52% per year). Asian economies have grown vigorously in all sectors, including agriculture, but their rapid initial growth rates are becoming more moderate in food and other sectors while African growth rates in this sector are speeding up. Very similar to food output is the record of total agricultural output growth since 1961. Total net agricultural output (also including non-food products such as tobacco, coffee, or vegetal fibres, and valued in the same manner) grew at 2.48% per year from 1961-1963 to 2009-2011 (Table 5). The relative ranking of the regions as regards growth rates is similar to the ranking observed with food output, with Africa, LAC, and Asia at the top, and North America and Europe at the bottom. This refers again to net output, excluding farm output used as feed or seed within the agricultural sector. 51

Agricultural and food output

Table 5. Annual growth rates of real net agricultural output, 1961-2011. World Africa Asia LAC N. America Europe

1961-1963 to 2009-2011 2.48% 2.83% 3.47% 3.06% 1.70% 0.78%

1961-1963 to 1969-1971 2.63% 2.99% 3.18% 2.96% 1.87% 2.24%

1969-1971 to 1979-1981 2.35% 1.42% 2.93% 3.14% 2.58% 1.66%

1979-1981 to 1989-1991 2.39% 3.06% 3.93% 2.35% 0.79% 0.88%

1989-1991 to 1999-2001 2.24% 3.00% 3.72% 3.15% 2.06% -1.23%

1999-2001 to 2009-2011 2.80% 3.71% 3.52% 3.69% 1.27% 0.67%

Source: FAOSTAT.

The geographical distribution of total agricultural output, or food output in particular, changed significantly during these fifty years. Asia went from producing 32% of agricultural output in 1961-1963 to nearly 52% in 20092011; Africa and Latin America also increased their shares, albeit by a smaller proportion, while North America and Europe both reduced their contribution to world output, from a joint share of 50% in 1961-1963 to 26.3% in 20092011. Changes in the geographical distribution of food output were similar (Table 6). Table 6. Regional shares of world net value of agricultural and food production, 1961-2011.

World Africa Asia LAC N. America Europe

Net agricultural production (%) 1961-1963 1989-1991 2009-2011 100.0% 100.0% 100.0% 7.5% 7.5% 8.8% 32.7% 41.9% 51.9% 10.0% 11.0% 13.1% 15.5% 12.7% 10.8% 34.5% 26.9% 15.5%

Net food production (%) 1961-1963 1989-1991 2009-2011 100.0% 100.0% 100.0% 7.4% 7.5% 8.9% 32.4% 41.2% 51.3% 9.3% 10.8% 13.1% 15.3% 12.9% 10.9% 35.7% 27.7% 15.9%

Source: FAOSTAT (value of production). Net (agricultural or food) production excludes agricultural output used as seed or feed. Food production excludes agricultural non-food products such as jute, cotton lint, wool, tobacco, and others. Output is valued at constant 2004-2006 world average producer prices in PPP USD.

As mentioned before, population growth rates decelerated in recent decades in every region. World population growth slowed down from a yearly 2.04% in the 1960s to 1.2% per year in the 2010s. The regional differences in demographic rates are shown in Table 7.11

The rates in this table correspond to the 2010 revision of the UN Population Estimates, which are the ones used by FAOSTAT at the end of the period covered by this analysis. Figures have slightly changed in more recent revisions of the UN estimates, as new census data are added to the database; these slight changes pertain mostly to the period after 2000. See technical details on population estimates and projections in sections 12.2.6 and 13.2.1.

11

52

Hector Maletta

Table 7. Annual growth rates of population by region and period, 1961-2011.   World Africa Asia LAC N. America Europe

1961-1963 to 2009-2011 1.68% 2.66% 1.83% 2.00% 1.06% 0.47%

1961-1963 to 1969-1971 2.04% 2.56% 2.37% 2.66% 1.18% 0.90%

1969-1971 to 1979-1981 1.89% 2.70% 2.15% 2.39% 0.97% 0.65%

1979-1981 to 1989-1991 1.80% 2.79% 2.00% 2.03% 1.03% 0.52%

1989-1991 to 1999-2001 1.42% 2.52% 1.49% 1.69% 1.11% 0.14%

1999-2001 to 2009-2011 1.31% 2.72% 1.26% 1.37% 1.03% 0.25%

Source: FAOSTAT, based on the 2010 Revision of UN Population Estimates.

Population growth rates in the developed world increased during their industrialisation but have been decreasing since the middle or late 1800s. In the developing world, the demographic transition is still on-going; population growth rates are generally declining, after peaking at different periods depending on level of development. In Asia and Latin America, these rates peaked earlier, and have been falling since the 1960s. In Africa, the transition happened later, and is apparently still on-going - population growth increased from the 1960s to the 1980s, peaking at 2.79% per year in the 1980s; it fell in the 1990s but rose again in the 2000s, though it did not return to its earlier rates. Per capita food output is growing faster in predominantly developing regions, i.e., Asia, Africa, and LAC, than in developed areas such as Europe and North America (Table 8). For the whole period from 1961-1963 to 2009-2011, it grew by 0.82% worldwide, led by Asia (1.64%) and Latin America (1.22%). The overall performance of Africa in per capita food output growth has been poorer (0.25% per year over the entire half century) due to its rapid population growth, but the worst periods for Africa were the decades before 1990 – especially the 1970s, when per capita food output on that continent actually decreased. It started growing again at 0.34% per year in the 1980s, accelerating to 0.55% in the 1990s, and 1.10% in the 2000s. Table 8. Annual growth rates of real per capita net food output by world region, 1961-1963 to 2009-2011.   World Africa Asia LAC N. America Europe

1961-1963 to 2009-2011 0.82% 0.25% 1.64% 1.22% 0.73% 0.33%

1961-1963 to 1969-1971 0.63% 0.37% 0.73% 0.69% 1.00% 1.32%

1969-1971 to 1979-1981 0.51% -1.09% 0.83% 0.87% 1.65% 1.01%

1979-1981 to 1989-1991 0.59% 0.34% 1.88% 0.47% -0.22% 0.38%

1989-1991 to 1999-2001 0.89% 0.55% 2.37% 1.64% 0.97% -1.31%

1999-2001 to 2009-2011 1.44% 1.10% 2.23% 2.33% 0.31% 0.45%

Source: Based on FAOSTAT data on value of net food production, valued at constant 2004-2006 producer prices converted into USD at agricultural PPP conversion rates.

53

Agricultural and food output

Per capita output grew more slowly, and at generally decreasing rates, in Europe and North America. North America (the United States and Canada) increased its per capita food output at 0.73% per year over the half century, perceptibly more than Europe (0.33%), but below the world average. In the case of Europe, per capita food output actually fell in the 1990s, mostly due to the collapse of the centrally-planned economies, but its performance in the 2000s was better. Per capita food output in North America suffered a transient decline in the 1980s, from which it later recovered. Agricultural or food outputs and the growth thereof are not, however, direct indicators of food security, since (as seen above) food security is not defined by food production but by food access. Food security (for a country as well as for an individual or a household) does not require self-sufficiency in the production of food, but people being able to access food. What matters for improving food security is (1) sustained growth of per capita world food output to ensure sufficient supply, and (2) improved and more extended access to food. Access includes physical access (mostly depending on trade) and economic access (mostly depending on the level and distribution of income). The preceding figures illustrate the growth of food supply: per capita world food output has been growing, in total and per capita terms, and its growth has accelerated in recent decades. Economic access to food, and levels of effective food consumption, will be examined later (in chs. 6 and 7) after reviewing food production and trade.

3.5. Composition of agricultural output The above discussion about growth in real food output must be complemented by an analysis of the composition of agricultural output (including food and non-food products), which has also changed over time as its various components grew at different speeds. After a general review of output composition in this section, we will then take a closer look at the main staple food, i.e. cereals, and some other important food products. The growth rates of the various groups of products differ greatly (Table 9). Over the entire half century, the fastest growing groups were nuts, oil crops, and vegetables, increasing at about 3.2%-3.4% per year, whereas pulses (1.13%) and starchy roots (1.65%) were the slowest. Cereals and sugars grew somewhat slower than all agriculture, whereas fruit and meat grew somewhat faster than average. The faster growth rate of nuts, oil crops, and vegetables accelerated in the most recent quarter century (1986-2011) relative to the previous period (1961-1986), as did the growth rate of fruit, dairy and eggs, 54

Hector Maletta

pulses, and tubers. On the other hand, the growth rates of cereals and meats, as well as of eggs and dairy products, decelerated. Cereals increased their share of total value of production from 1961 to 1986, but then shrunk back to their original share, about 15.5%. Oil crops, vegetables, meats, and other edible products (especially cocoa, coffee, and tea) increased their share, while the shares of starchy roots (i.e., potatoes and other tubers), pulses, and dairy products (grouped here with eggs) diminished. Table 9. Growth rates and composition of agricultural net real output, by product group, 1961-2011.

All agric. products Cereals Starchy roots Sugar crops, honey Pulses Nuts Oilseeds Vegetables Fruits Meats Dairy and eggs Other, edible* Other, inedible**

Annual rate of growth 1961-2011 1961-1986 1986-2011 2.45% 2.51% 2.40% 2.44% 3.08% 1.79% 1.65% 1.01% 2.29% 2.39% 2.79% 2.01% 1.13% 0.76% 1.49% 3.19% 1.94% 4.46% 3.40% 3.12% 3.69% 3.25% 2.58% 3.93% 2.48% 2.41% 2.55% 2.60% 2.99% 2.21% 1.91% 1.97% 1.86% 2.52% 1.88% 3.16% 1.71% 1.78% 1.65%

Index 2011 (1961=100) 336 333 226 327 175 481 533 495 340 361 258 347 234

Percent share of total 1961 1986 2011 100.0% 100.0% 100.0% 15.6% 18.0% 15.5% 6.4% 4.4% 4.3% 3.4% 3.6% 3.3% 2.1% 1.4% 1.1% 0.8% 0.7% 1.1% 5.0% 5.8% 7.9% 8.1% 8.2% 11.9% 11.9% 11.6% 12.1% 22.8% 25.6% 24.4% 17.8% 15.6% 13.7% 1.5% 1.3% 1.5% 4.5% 3.8% 3.2%

(*) Other, edible: cocoa, coffee, tea, maté, other infusions, hops, species. Coffee and tea are classified as non-food products in FAOSTAT. (**) Other, inedible: vegetal fibres (cotton lint, jute, etc.), pyrethrum, wool, tobacco, rubber, gums, beeswax, silk-worm cocoons, other. Net real output = net value of agricultural production (excluding amounts used as seed or feed) valued at 2004-2006 world average producer prices, converted into USD at agricultural PPP conversion rates. Source: FAOSTAT.

These differential rates of growth and corresponding changes in percentage shares of total agricultural value of production are consistent with trends in demand - the world population is increasing its consumption of vegetable lipids and prefers a more diversified diet with more fruit and vegetables. Perhaps contrary to conventional wisdom, meat output grew only slightly above average, and most of that growth was in the former period. In 19862011 meat output grew at 2.21%, more slowly than agricultural production as a whole (2.40%) and at half the growth rate of nuts (4.46%) and vegetables (3.93%). The rate of growth of meat production was 2.99% per year in 19611986 but decelerated to 2.21% in 1986-2011. Among meats, the fastest growing was poultry, while bovine meat tended to lag behind - cattle meat 55

Agricultural and food output

grew at 1.65%, pork at 2.98%, chicken at 5.09% over the entire half century. But every one of these three kinds of meat decelerated their growth in 19862011 relative to the previous quarter century. These trends in net real value of production are mirrored, of course, by trends in consumption, as will be shown later. While beef production and consumption has been growing relatively slowly (more slowly than population in recent decades, implying a decline in per capita terms), chicken is on a rapid worldwide multi-decadal rise, whereas pork has been growing at an intermediate pace (above the rate of population growth). It is not possible, for reasons of space, to analyse in detail the output of every product or product group that could be analysed by total and per capita output, and (in the case of crops) in terms of harvested land and physical yields. We shall examine growth in their consumption in Ch. 6, especially in Section 6.2. However, the next section (3.6) examines the output of the main staple food, cereals, in more detail.

3.6. Growth in cereal production Cereals have been the basic foodstuffs of humankind since the Neolithic agricultural revolution. They provide a substantial fraction of human dietary energy intake - a fraction that is higher in less developed countries. Some cereals are also a major component of livestock feed. Thus, cereal output growth (total and per capita) is a key issue as regards world food supplies. In fact, most international food aid consists of cereals, and cereal availability is the main concern of international bodies when they discuss international food security and hunger emergencies. For example, the reports from the UN Committee on Food Security (e.g., CFS 2011a, 2011b, 2013) and the monthly FAO reports on the World Food Situation are cen­tred on the availability of cereals (see: http://www.fao.org/worldfoodsituation/csdb/en/).

3.6.1. Total cereal output and yields World cereal output has been steadily growing, from 805 million metric tons in 1961 to 2,533 million metric tons in 2013. Growth continues - cereal harvests in 2014 were slightly higher than the previous year (based on the FAO estimate as of April 2015). From 1961 to 2013, cereal output grew at an annual rate of 2.22% (using 3-year averages, it grew at a yearly 2.11% from 1961-1963 to 2011-2013).

56

Hector Maletta

World cereal output undergoes frequent but small fluctuations, mostly due to weather variations in the various producer countries; local fluctuations are greater, but at the world level, the net oscillations are inconsiderable. There were some relatively stagnant periods (e.g., in the late 1990s) and periods of faster growth (e.g., 2002-2014). In recent years, and besides the usual small fluctuations, output growth has not been significantly affected by the world recession that started in 2008 or the food commodity price surges in 20072008 and 2010-2011 (Figure 3). Preliminary data indicate that output in 2014 was at an all-time high, as had previously been the case in 2013, 2011, 2008, 2007, and many earlier years. Cereal output is definitely on a secular rising trend. Figure 3. World production of cereals (million tonnes), 1961-2014. 3000

Million tonnes

2500 2000 1500 1000 500 0 1961

1971

1981

1991

2001

2011

Sources: FAOSTAT (retrieved April 2015) up to 2013; for 2014, the rate of growth for 2013-2014 from the FAO World Food Situation (WFS) report issued April 2015 was applied to the final FAOSTAT estimate for 2013. Figures in FAO WFS reports (http://www.fao.org/worldfoodsituation/csdb/en/) are preliminary. Rice production counted in milledequivalent terms.

The cumulative growth rate of cereal output between 1961-1963 and 20092011 was a yearly 2.11%, and 2.13% from 1961-1963 to 2012-2014. This was achieved mostly through increased cereal yields per harvested hectare (Figure 4).

57

Agricultural and food output

Figure 4. World average cereal yield (tonnes per harvested hectare), 1961-2013.

Tonnes per harvested hectare

4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0

1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001 2005 2009 2013

Source: FAOSTAT (retrieved April 2015). Rice measured in milled equivalent terms.

About 95% of the growth in cereal output tonnage from 1961 to 2013 is attributable to increased cereal yield per unit of harvested area, and only 5% to increased harvested area. There was just a small increase in the harvested area of cereals between these years, from 648 to 722 million hectares at a very small rate of 0.21% per annum, while output grew at 2.24%. In fact, the harvested cereal area was stagnant; in 2010-2013, it was lower than in some years during the 1970s and 1980s. In contrast, yields have shown a strong and steady upwards trend. The average yield went from 1.35 t/Ha in 1961 to 3.85 t/Ha in 2013. Regional cereal yields in 1961 were mostly around one tonne per hectare, except in North America where they were above 2 t/Ha. Cereal yields in North America grew to over 6.6 t/Ha by 2013. Africa was the slowest growing region, but still doubled its yields from 0.81 MT/Ha in 1961 to 1.62 MT/Ha in 2013. The other regions (Europe, Asia, and Latin America) trebled their yields, from barely above 1 to about 4 t/Ha. Table 10 and Figure 5 illustrate these changes. Table 10. Cereal yields by region, 1961-2013 (tonnes per harvested hectare). Region

1961

1971

1981

1991

2001

2011

2013

World Africa Asia LAC N. America Europe

1.35 0.81 1.21 1.27 2.20 1.38

1.89 0.99 1.66 1.49 3.38 2.05

2.25 1.24 2.11 2.00 3.84 2.05

2.69 1.24 2.78 2.18 4.02 2.71

3.13 1.29 3.21 3.00 5.05 3.19

3.67 1.49 3.85 3.78 6.19 3.64

3.85 1.62 3.97 4.19 6.67 3.70

Source: FAOSTAT (retrieved April 2015). Rice expressed in milled-equivalent terms.

58

Annual growth rate 1961-2013 2.03% 1.34% 2.31% 2.32% 2.15% 1.91%

Hector Maletta

Figure 5. Cereal yields by region, 1961-2013 (tonnes per harvested hectare) 8.00 7.00

tonne /Ha

6.00 5.00 4.00 3.00 2.00 1.00 0.00 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001 2005 2009 2013 World Africa Asia LAC North America Europe Source: FAOSTAT (retrieved April 2015).

Between the high yields of North America, well above the rest, and the much lower yields of Africa, other regions are clustered around the world average. Note that the great U.S. drought of 2012 caused a large fall in yields in North America, but even these diminished yields were higher than yields attained in any year before 2004. North American cereal yields in every year from 2004 to 2013, including the poor 2006 and 2012 crops, were in fact higher than all yields before 2004. To sum up, cereal output is growing strongly and steadily, and almost all growth comes from higher yields with very little from expansion of cereal harvested area. There were, of course, countries or regions where the harvested area increased and others where it decreased; the above figures refer to the world total. Likewise, yield increases varied across countries and across different cereal crops. Annual yield variability is smoothed when all cereals and countries are considered as a whole; more variability exists in particular areas and for particular crops due to localised variability in weather conditions. The composition of world cereal output has not changed much over time. The tonnage shares of the various cereals register only small changes (Table 11): wheat increased its share from about 28% to about 30%; other cereals 59

Agricultural and food output

(barley, sorghum, millet, and others) decreased from 29% to 27%; while maize stayed at around 24-25% and rice at around 18% for the entire half century. Table 11. Percentage composition of world cereal output tonnage, 1961-1963 to 2009-2011. 1961-1963

1969-1971

1979-1981

1989-1991

1999-2001

2009-2011

Total

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

Maize

24.9%

25.4%

24.6%

24.4%

25.4%

25.1%

Rice (milled equiv.)

18.2%

18.3%

18.9%

18.4%

18.0%

17.9%

Wheat

27.9%

28.9%

28.4%

29.7%

29.3%

30.0%

Other cereals

29.0%

27.4%

28.1%

27.5%

27.3%

27.1%

Source: FAOSTAT.

Table 12. Annual growth rate of world cereals output tonnage, 1961-63 to 2009-11. 1961-63 to 2009-11

1961-63 to 1969-71

1969-71 to 1979-81

1979-81 to 1989-99

1989-91 to 1999-01

1999-01 to 2009-11

Total

2.11%

3.60%

2.59%

1.85%

0.85%

1.96%

Maize

2.96%

3.80%

4.04%

1.43%

2.24%

3.49%

Rice (milled equiv.)

2.40%

3.87%

2.49%

2.77%

1.56%

1.61%

Wheat

2.23%

4.01%

3.11%

2.48%

0.50%

1.45%

Other cereals

0.32%

2.83%

0.48%

0.68%

-1.70%

-0.14%

Source: FAOSTAT.

In regard to cereal output growth on the world scale (Table 12), the decadal rates generally decelerated from the 1960s to the 1990s, but increased again in the 2000s for all cereals taken as a whole, and also for each major cereal. In the 2000s, rice and wheat grew at rates of 1.61% and 1.45%, respectively - well below the growth rate of maize (3.49%). This was probably related to the increased demand for maize for use in ethanol fuels. However, the recent growth rate of maize is in fact slower than its rates in the 1960s (3.80%) and 1970s (4.04%). Minor cereals (chiefly barley, sorghum, and millet) began to slow down in the 1970s and 1980s (relative to the 1960s) and have been actually decreasing in output in the 1990s and 2000s. Regional cereal output. The production of cereals has increased in all regions over the last half century, albeit at different rates (Table 13). 60

Hector Maletta

Table 13. World cereal output tonnage by region: percent share and growth rate, 1961-1963 to 2009-2011. Percent share of regions in world cereal tonnage  

1961-1963

1969-1971

1979-1981

1989-1991

1999-2001

2009-2011

World

100.0%

100.0%

100.0%

100.0%

100.0%

100.0%

Africa

5.9%

5.3%

4.8%

5.5%

5.7%

6.8%

Asia

34.2%

35.0%

37.3%

41.6%

44.6%

45.1%

LAC

5.6%

5.9%

6.0%

5.6%

7.1%

7.7%

N. America

22.9%

21.8%

23.7%

19.8%

20.2%

19.5%

Europe

31.4%

31.9%

28.1%

27.5%

22.3%

21.0%

Annual regional growth rates of cereal tonnage (% per year)   World

1961-1963 to 2009-2011

1961-1963 to 1969-1971

1969-1971 to 1979-1981

1979-1981 to 1989-1991

1989-1991 to 1999-2001

1999-2001 to 2009-2011

2.11%

3.60%

2.59%

1.85%

0.85%

1.96%

Africa

2.40%

2.32%

1.51%

3.21%

1.32%

3.67%

Asia

2.70%

3.93%

3.24%

2.95%

1.56%

2.07%

LAC

2.79%

4.35%

2.80%

1.09%

3.25%

2.78%

N. America

1.77%

2.94%

3.46%

0.03%

1.06%

1.61%

Europe

1.25%

3.79%

1.30%

1.64%

-1.24%

1.31%

Source: FAOSTAT. Rice expressed in milled equivalent terms.

Over the whole half century (from 1961-1963 to 2009-2011), the annual rate of growth was higher in Latin America (2.79%), Asia (2.70%), and Africa (2.40%), well above the world average (2.11%). North America (1.77%) and Europe (1.25%) experienced much slower growth (marked in Europe by a decline in the 1990s in connection with the general collapse of centrally-planned economies). For all regions (and for most countries), the cereal output of recent years has been the highest on record, as also has been the case at the world scale. Although cereal output in Europe and North America has been growing at a slower pace relative to the world average, those two regions still produce 40% of the world’s output (down from 54% in the early 1960s); Asia has been contributing 45% in recent years (up from 34% in 1961-1963), with the balance distributed between Africa and Latin America. It is remarkable that Africa (in spite of being the slowest growing region as regards yields) nonetheless nearly doubled its cereal yield and tripled its tonnage from 19611963, and increased its share of the world’s output from 4.8% in 1979-1981 to 6.8% in 2009-2011. 61

Agricultural and food output

3.6.2. Per capita cereal output As seen above, world cereal production has been growing at an average annual rate of 2.11%. However, per capita cereal output remained remarkably stable over recent decades (Figure 6). During the 1960s and early 1970s it increased from 250 to 320 kilograms/person/year (kgpyr). It stagnated thereafter, fluctuating between 300 and 340 kgpyr, including a slightly decreasing trend in 1986-2002 and a slightly rising trend thereafter which continued at least up to 2014. The high per capita outputs of 2013-2014 were only slightly above the levels already attained in 1984-1986. Figure 6. World per capita cereal output, 1961-2014 (kg/person/year). 400 350 kg/person/year

300 250 200 150 100 50 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Source: FAOSTAT and FAO World Food Situation website (both retrieved April 2015). Data for 2014 are preliminary. Rice expressed in milled equivalent terms.

Why should per capita cereal output (and effective demand) seemingly stagnate in a world of increasing per capita production of food, increasing cereal yields, and overall increasing levels of income? Indeed, why should this be the case when total and per capita food consumption have also increased in terms of dietary energy and other nutrients, as FAOSTAT data demonstrate? The key reason is that increasing incomes entail changes in the composition of food demand; it is this shift in demand that drives production of specific foods. As their incomes grow, people limit their intake of cereals and increase their intake of other foods, such as meat, milk, vegetable oil, fruits or vegetables. This is partially offset by population growth and also by increasing use of cereals for purposes other than food. The net outcome is a relatively stable per capita cereal output. This output is absorbed mainly by food demand for 62

Hector Maletta

cereals, but there are also other uses of cereals: some go into the production of animal feed or biofuels, some get wasted, and some are used in other industries (e.g., beer or cosmetics). This chapter is limited to food production; demand for food products will be discussed later (Ch. 6). As noted, there is a slightly decreasing tendency in per capita consumption of cereals as food, offset by an increase in other uses of cereals.

3.7. Production of other selected crops In this section, we briefly review the trends of some other major agricultural products in terms of total output, area, and yield, primarily on the world scale. The purpose here is not to examine the whole list of agricultural products but rather only some specific crops that are important food items and whose production may have grown at different rates in the past half century, possibly influenced by economic development and changes in prevailing diets. We concentrate on roots and tubers such as potatoes and cassava; pulses such as beans or lentils; oil crops and vegetable oil; and all fruits and vegetables, which are the main source of micronutrients – vitamins and minerals – in the human diet.

3.7.1. Roots and tubers Production of roots and tubers has been increasing since 1961, though on the whole it increased far less than population, which is to say that per capita output decreased. By 2013, the world was producing 836 million tonnes of tubers, including 368 million tonnes of potatoes, 276 of cassava, 110 of sweet potatoes, and 80 million tonnes of other tubers (yams, taro, yautia, and others) as shown in Figure 7.

63

Agricultural and food output

Figure 7. World output of roots and tubers, 1961-2013. 900 800

Million tonnes

700 600 500 400 300 200 100 0 1961

1971 Potatoes

1981

1991

Cassava

2001

Sweet potatoes

2011

Other

Source: FAOSTAT, retrieved April 2015.

As shown in Figure 8, the yields of these products grew relatively slowly in comparison with other agricultural products: just 56% for the whole group. The average yields of sweet potatoes were rather erratic - they doubled from 1961 to the early 1980s, stagnated until 2000, and tended to decline thereafter. Potato and cassava yield growth has been steadier, though not spectacular. Figure 8. World average yields of roots and tubers (tonnes/hectare), 1961-2013. 25

Tonnes/Hectare

20 15 10 5 0

1961

1971 Potatoes

1981 Sweet potatoes

Source: FAOSTAT, retrieved April 2015. Total includes other tubers.

64

1991

Cassava

2001

Total

2011

Hector Maletta

3.7.2. Oil crops and vegetable oil From 1961 to 2013, the production of oil crops expanded by a factor of 6.5, at an annual rate of 3.67%, in an area of land that only increased at half that rate (1.82% per year), as shown in Figure 9, while yields grew from 1.34 to 3.43 tonnes per harvested hectare (Figure 10). More than half the world production in 2013 was made of only two oil crops: soybeans and oil palm, with 276 and 266 million tonnes respectively; production of the next two largest crops - rapeseed and seed cotton - was only 73 million tonnes each. At the same time, the production of vegetable oil increased even faster, expanding by a factor of 9.3 and growing at an annual rate of 4.38% (Figure 11). This extraordinary growth in the production of vegetable oil for human consumption (accompanied by the production of oilcakes for animal feed, and biodiesel, all from the same oil crops) also implies strong growth in per capita terms since, in the meantime, population grew at a mere 1.6% per year. This contrasts with the relatively stagnant per capita output of cereals; the difference mainly expresses significant changes in food consumption patterns, as will be reviewed later. Figure 9. Oil crops worldwide: area harvested and production, 1961-2013.

Million tonnes or hectares

1200 1000 800 600 400 200 0

1961

1966

1971

1976

1981

1986

Harvested area (Ha)

65

1991

1996

2001

Production (t)

2006

2011

Agricultural and food output

Figure 10. Average yield of oil crops (tonnes per hectare), 1961-2013. 4.00 3.50

Tonnes / Ha

3.00 2.50 2.00 1.50 1.00 0.50 0.00

1961

1966

1971

1976

1981

1986

1991

1996

2001

2006

2011

Figure 11. World production of vegetable oil, 1961-2013 (in millions of tonnes). 180 160

Million tonnes

140 120 100 80 60 40 20 0

1961

1971

1981

1991

2001

2011

3.7.3. Pulses Dry leguminous grains (pulses) like beans or lentils are very important food components, especially in traditional diets with limited intake of animal protein. Their production has increased in recent decades, especially since the 1980s, and so has their yield. The harvested area increased far less, and only did so in very recent years. However, pulses increased less than other crops (from 40 to 75 million tonnes) and their yield only grew from 0.7 to

66

Hector Maletta

0.9 metric tons per hectare. Output per capita, as a result, actually decreased (from about 13 to about 10 kg/person/year) in the period from 1961 to 2013, as people shifted consumption to other sources of protein like meat (chiefly poultry, fish, and pork) as well as eggs and milk.

80

90.0

70

80.0

60

70.0 60.0

50

50.0

40

40.0

30

30.0

20

20.0

10

10.0

0

Area (million Ha)

Production (million tonnes)

Figure 12. World harvested area and production of pulses, 1961-2013.

0.0 1961

1971

1981

1991

Production (tonnes)

2001

2011

Area harvested (Ha)

3.7.4. Fruits and vegetables Production of fruits and vegetables as a whole grew from 400 to 1,800 million metric tons from 1961 to 2013. Yields grew from about 8 to about 15 metric tons per hectare. These figures are influenced not only by changes in the area and yield of each individual product, but also by changes in the product mix; however, the trend is similar for each of the most significant sub-groups of fruits and vegetables.

67

Agricultural and food output

2000

16.0

1800

15.0

1600

14.0

1400

13.0

1200

12.0

1000

11.0

800

10.0

600 400

9.0

200

8.0

0 1961

Yield (tonnes /Ha)

Production (millions of MT)

Figure 13. World production and yield of fruits and vegetables, 1961-2013.

7.0 1971

1981

1991

Production (left axis)

2001

2011

Yield (right axis)

3.8. Conclusions on agricultural and food output growth World agricultural output, and more specifically the output of primary food products (from crops, livestock, and fisheries), has been growing consistently ahead of population during the past half century (i.e., since 1961, the start of the world-scale series maintained by FAO). While population doubled, production trebled, increasing per capita production by about 50% over the period considered. Growth in total output (aggregated in money terms at constant and uniform prices of equal purchasing power) kept a more or less constant pace (around 2.5% per year), but since population growth is decelerating, growth of per capita output has been faster in recent times as compared with the earlier part of the past half-century. The Great Recession that hit the world economy starting in 2008 and the rise in agricultural prices that peaked in two successive surges in 2007-2008 and 2010 (before subsiding in 2011-2015) made no dent in the rising trend of food production, which continued unabated. Growth of output is faster in the periphery of the world economy, i.e., in Asia, Africa, and Latin America, where it is growing at rates well above 3% per year. The rates are much slower in North America and Europe. The share of Europe and North America in total world output has declined: more than

68

Hector Maletta

half the world output now comes from Asia, and the shares of Africa and (most especially) Latin America have increased as well. Per capita production of specific foods grew at different rates as patterns of utilisation changed over time. Thus, per capita production of staple food, like cereals, has remained relatively stagnant since the mid-1970s, with a slight decreasing trend from the mid-1980s until 2000-2001, and a slightly rising trend thereafter. Faster growing products include oilseeds and vegetable oils, fruit and vegetables, and some animal products led by poultry and pork. Production of beef has grown more slowly and remains relatively stagnant, as does cereal production, when assessed in per capita terms. These different patterns of growth (as will be seen later when our focus shifts to consumption) respond mainly to changes in food demand, led by differential income growth in various parts of the world. Changes in demand include changes in the composition of human diets (with increasing consumption of animal products and non-staple crop products), faster growth of food consumption in the periphery (due to faster demographic and economic growth), and increasing use of crops for non-food purposes such as animal feed and (incipiently) biofuels. Food consumption, however, will be discussed later (Ch. 6). Before that, we will turn our attention to land use (Ch. 4) and international trade in agricultural products (Ch. 5).

69

4. Land use and agricultural productivity

The land question. Agriculture and food production, as we have seen, more than trebled in fifty years, and per capita production increased by half, even though this was the period of the fastest population growth in the history of humankind. How was this remarkably successful outcome achieved? Is humankind perhaps straining and rapidly exhausting all available land for this enormous increase in agricultural production required by unprecedented population growth? How much growth came from using extra land and how much from increases in productivity? Are humans in danger of running out of land to feed a growing world population? How much land is left untapped? Ever since Malthus, one major and frequent concern about the food situation and prospects for the future has been the limited amount of land available for food production. As the population increases, it is expected, in principle, that farmland will need to expand, possibly into less fertile areas or onto land intended for other purposes, not suitable for agriculture, or meriting protection (e.g., forests and other wildlife areas). It is thus rather encouraging to discover the following facts: (a) Land used for agricultural production (including crops or livestock) is not increasing; it expanded rather slowly from 1961 to around 1991, and did not increase much thereafter: global farmland is stagnant, and in fact has been slightly decreasing. (b) Nearly all growth in agricultural production in the last half century (including crops and livestock) was not due to the use of extra farmland, but rather to increasing production per unit of farmland. Increased productivity accounts for more than 95% of total agricultural growth, 71

Land use and agricultural productivity

whereas increased agricultural land area accounts only for the remaining residual of about 4-5%. (c) The same is true for crops alone: land used for crops (both permanent and temporary) has not increased significantly in recent decades, and most growth in crop production (also about 95%) has come from increased production per hectare; and finally: (d) There exists a large area of land that is suitable for crops but not currently used for this purpose. This land may potentially be used for crops, even without expanding irrigation or encroaching onto forests. It is unused but usable land, i.e., land suitable for crops that is not forested, not built-up, and not otherwise protected. In fact, more than half of all land suitable for rain-fed crops is as yet not used for crops. In addition, in some arid or semiarid regions, the expansion of the area under irrigation may make unsuitable land suitable, just by providing water – however, this possibility is not considered in this analysis. Support for these assertions is provided in the next sections.

4.1. Major uses of agricultural land Land is, of course, a major factor in agricultural production. Agricultural land (or farmland for short) is any land used for growing crops or raising livestock. Land used for crops (cropland for short) is the sum of all arable land (used for temporary crops) plus land with permanent crops. A third category is land with permanent meadows and pastures (grassland or pasture for short), where the meadows and pastures may be either naturally grown or cultivated. As per FAOSTAT definitions, ‘permanent crops’ do not include permanent cultivated meadows and pastures. Crop production can easily be associated with specific areas of land, namely the area where each crop was grown. A crop’s yield is its output per hectare. Livestock ‘yields per hectare’, on the other hand, are harder to define and calculate - even if many animals live off permanent pastures, others are fed crop products such as barley or soybean cakes. Livestock cannot really be associated with any specific type or area of land, since their fodder may come from permanent meadows and pastures, crop products, crop residues, and other sources (domestic or foreign). It is in fact possible for a farm to raise livestock without having any farmland by purchasing fodder on the market or letting animals graze on public or communal pastures. 72

Hector Maletta

The main measures relating crops to land are yield, usually defined as the physical output (in metric tons) per harvested hectare; and cropping intensity (mean number of harvests per year, per hectare of cropland). To understand the latter concept, it is necessary to learn how harvests are counted since this is done differently for temporary and permanent crops. In the case of arable land (devoted to temporary crops), a given plot may be harvested more than once per year if two or more temporary crops are successively grown on it. In such cases, each harvest counts towards total harvested area, thus one physical hectare of cropland where two temporary crops are grown and harvested within the same year is equivalent to two hectares of harvested area. Most arable land is harvested only once per year, but some areas undergo double or triple cropping, while other areas may not be harvested at all in a given year due to crop failure after sowing, drought, temporary fallow, rotation with pasture, or some other reason. Thus, a single hectare of arable land may translate into zero, one, or more harvested hectares in a given year. In the case of permanent crops, two successive crops in one year are impossible. In fact, a hectare under permanent crops, if it is harvested at all, is counted as just one harvested hectare per year, even if product gathering occurs during various or extended periods of the year as is the case, for instance, with tea or coffee fields. However, a given hectare of land with permanent crops may not be harvested at all during a given year; the main reason is that, at any given time, some permanent crops are still in the growing stage and are not yet yielding any product (e.g., young fruit trees). Some may not be harvested for other reasons (e.g., because some pest or disease destroyed the product before it could be harvested). As a result, the cropping intensity of permanent crops cannot be greater than one and is usually less. Thus, any given hectare of cropland may not be harvested, be harvested once, or (in the case of temporary crops) be harvested more than once per year. Average cropping intensity may be less or more than 100%. In fact, world cropping intensity is less than 100%, though it increased from 70% to 85% between 1961 and 2011. In some specific regions, it may be greater than 100% (e.g., by 2011, the harvested area of Asia was equivalent to 109% of Asian cropland, due to extensive double cropping). In this chapter, we analyse the growth of total farm production (products from crops and livestock) in relation to farmland (cropland and pastures), and growth in crop production in relation to cropland. At constant 73

Land use and agricultural productivity

productivity, agricultural output may only grow due to expansion of the land area used for farming; any observed excess growth should be regarded as due to increased farmland productivity, i.e., additional value of production per hectare, at constant and uniform prices. Likewise, crop output growth may be explained by expansion of cropland or by increased production per hectare of cropland. Specific factors influencing agricultural productivity include cropping intensity, changes in the crop mix, irrigation, improved seeds, and technical changes of various sorts. Below, we discuss (1) the role of land use and land productivity in agricultural output growth; (2) availability of crop-suitable land around the world, and (3) expected demand for additional agricultural land.

4.2. Changes in agricultural land use Total estimated farmland (used for temporary or permanent crops or for pasture) from 1961 to 2011 is charted in Figure 14. The main uses of farmland are shown at Table 14, which also reports the area of farmland equipped for irrigation. The most striking feature of this data is that total farmland (i.e., land used for agricultural production around the world) has remained remarkably stable in recent decades, in spite of rapid growth in agricultural production. Farmland increased marginally from 4.46 billion hectares (BHa) in 1961 to 4.92 BHa in 1993. This is an increase of just 10%, or 0.3% per year over 32 years - a period during which world farm output increased by 128%, and world population by 82%. After 1993, farmland actually ceased to expand; it remained stagnant for a decade, with a very small increase from 4.92 to 4.94 BHa from 1993 to 2003, and later hovered around 4.90 BHa in 2007-2011.

74

Hector Maletta

Figure 14. World farmland (crops and pastures), 1961-2011 (billion hectares). 5.00

Bilion hectares

4.90 4.80 4.70 4.60 4.50 4.40 4.30 4.20

1961

1971

1981

1991

2001

2011

Source: FAOSTAT.

Table 14. World agricultural land by land use, 1961-2011 (thousand hectares). Land uses A=B+E Agricultural land (farmland) B=C+D Land used for crops (cropland) Arable land C (used for temporary crops) D Land with permanent crops Permanent meadows and E pastures F Land equipped for irrigation cropland equipped for G=F/B % irrigation H=B-F Rain-fed cropland

1961 1971 1981 1991 2001 2011 4,459,881 4,580,126 4,667,083 4,838,953 4,935,491 4,911,631 1,370,571 1,425,084 1,454,866 1,521,515 1,516,944 1,552,977 1,282,133 1,328,867 1,353,995 1,402,818 1,382,624 1,396,279 88,437

96,217

100,871

118,697

134,321

153,938

3,089,311 3,155,042 3,212,217 3,317,438 3,418,546 3,358,655 160,994

187,442

226,264

260,426

292,550

318,297

11.7%

13.2%

15.6%

17.1%

19.3%

20.5%

1,209,577 1,237,642 1,228,602 1,261,089 1,224,394 1,234,680

‘Land with permanent crops’ excludes land with permanent cultivated meadows or pastures. Permanent meadows and pastures may be natural or cultivated. Some land ‘equipped for irrigation’ may not be currently irrigated. Source: FAOSTAT.

Despite stagnant use of farmland in the years since 1993, world population kept growing, albeit at declining rates, and agricultural production increased steadily and at accelerating rates; from 1993 to 2011, farm output grew by 54% and world population increased by 23%, but farmland did not rise. By 2011, farmland area worldwide was 4.91 BHa, the same as in the early 1990s and just 10% above its 1961 level. World farmland per capita was halved from 1.44 Ha in 1961 to 0.70 Ha in 2011, while in the same period, per capita output of food products from crops and livestock grew by 48% (Table 1). Without productivity growth, feeding the current world population at the current per capita levels would have required more than twice the land area actually used for agriculture around the world. 75

Land use and agricultural productivity

Cropland, i.e., arable land in addition to land under permanent crops (excluding permanent cultivated meadows and pastures), slowly expanded over a quarter century, from 1.37 BHa in 1961 to nearly 1.51 BHa in 1986; this was followed by some years of very slow growth until it peaked at nearly 1.53 BHa in 1993, and then remained nearly stagnant (1.51-1.53 BHa) until the early 2010s (Figure 15). Figure 15. World cropland: Sum of arable land and land with permanent crops (excluding permanent cultivated meadows and pastures), 1961-2011 (billion hectares). 1.60

Billion hectares

1.55 1.50 1.45 1.40 1.35 1.30 1.25

1961

1971

1981

1991

2001

2011

Source: FAOSTAT.

What appears to be a weakly revived trend of cropland expansion did occur in the most recent years (from a low of 1.515 BHa in 2002 to 1.55 BHa in 2011, with a temporary reversal in 2007). Preliminary data for 2012 (not shown in the figure) indicate a further increase to 1.57 BHa, but it is not yet clear whether this foreshadows further expansion in the future or is just a temporary fluctuation in the stagnant long-term trend prevailing since the late 1980s. Even with this recent small increase, the broad long-term picture is still that world crop production took place for a quarter century in the context of a near-stagnant cropland area and nevertheless achieved significant growth in crop output (increasing 80% in the quarter century from 1986 to 2011). Land equipped for irrigation accounts for just a fraction of total cropland, albeit an increasing fraction, expanding from 11.7% in 1961 to 20.5% in 2011. Thus, irrigated land area practically doubled in the past half century, increasing at an annual rate of 1.37%. Because of this relatively fast expansion

76

Hector Maletta

in irrigation, rain-fed cropland has been even more stagnant than total cropland: by 2011, it was practically at the same level it was in 1971. The remaining farmland, i.e., grassland, or permanent (natural or cultivated) meadows and pastures, grew until the early 1990s, stagnated during that decade, and actually decreased after 2000 (Figure 16). Figure 16. World land with permanent meadows and pastures (natural or cultivated), 1961-2011 (billion hectares). 3.50

Bilion of hectares

3.40 3.30 3.20 3.10 3.00 2.90

1961

1971

1981

1991

2001

2011

Source: FAOSTAT.

Permanent meadows and pastures covered 3.1 BHa in 1961. They slowly but steadily expanded during three decades to slightly over 3.4 BHa in the early 1990s, then remained at that level until the early 2000s before declining from 2001 to 3.35-3.36 BHa in 2009-2011. Both the increase and subsequent decrease represent a minor proportion of existing pasture land - an increase of about 10% from 1961 to the 1990s, and a decline of about 2% in the 2000s. These changes, especially those after 1990, are perceptible in Figure 16 only because the vertical scale of the chart has been deliberately limited to make them visible. Thus, the decline in total agricultural land discussed above is entirely due to a reduction in the area covered by permanent meadows and pastures; cropland has been relatively stable for a quarter century, though it recently showed a small measure of growth. In short: cropland is nearly stagnant or growing very little, and pasture land is declining slightly, while crop and livestock output is increasing. However, this differs somewhat across regions (Table 15 and Table 16).

77

Land use and agricultural productivity

Table 15. Agricultural land, cropland, and grassland by region, 1961-2011 (thousand hectares). Total agricultural land (farmland) 1961

1971

1981

1991

2001

2011

World

4,459,881

4,580,126

4,667,083

4,838,953

4,935,491

4,911,631

Africa

1,057,226

1,073,150

1,076,738

1,106,690

1,129,230

1,169,696

Asia

1,541,304

1,611,011

1,661,995

1,774,121

1,839,837

1,764,094

LAC

561,451

614,682

654,711

690,044

709,992

741,021

N. America

517,573

502,200

494,291

494,941

482,686

474,098

Europe

782,328

779,084

779,349

773,158

773,747

762,722

Total land used for crops (cropland)* 1961

1971

1981

1991

2001

2011

World

1,370,571

1,425,084

1,454,866

1,521,515

1,516,944

1,552,977

Africa

168,909

183,714

190,665

205,518

224,369

258,303

Asia

472,070

493,738

507,594

557,860

563,938

567,718

LAC

102,996

127,973

142,859

152,066

162,932

188,047

N. America

235,302

243,472

241,349

239,571

230,216

210,660

Europe

391,294

376,188

372,399

366,501

335,490

328,248

Permanent meadows and pastures (grassland)** 1961

1971

1981

1991

2001

2011

World

3,089,311

3,155,042

3,212,217

3,317,438

3,418,546

3,358,655

Africa

888,317

889,436

886,073

901,172

904,861

911,393

Asia

1,069,234

1,117,273

1,154,401

1,216,261

1,275,899

1,196,376

LAC

458,455

486,709

511,852

537,978

547,060

552,974

N. America

282,271

258,728

252,942

255,370

252,470

263,438

Europe

391,034

402,896

406,949

406,657

438,257

434,473

(*) Arable land + land with permanent crops, not including cultivated meadows and pastures. (**) Including both naturally growing and cultivated meadows and pastures. Source: FAOSTAT

78

Hector Maletta

Table 16. Total farmland, cropland and grassland: Annual growth rates (1961-2011) and percentage distribution in 1961 and 2011. Annual rates of growth

% of world total

Total agricultural land (farmland) 1961-2011 1961-1971 1971-1981 1981-1991 1991-2001 2001-2011

1961

2011

World

0.20%

0.33%

0.19%

0.36%

0.20%

-0.05%

100.00%

100.00%

Africa

0.21%

0.19%

0.03%

0.27%

0.20%

0.35%

23.71%

23.81%

Asia

0.28%

0.55%

0.31%

0.65%

0.36%

-0.42%

34.56%

35.92%

LAC

0.58%

1.14%

0.63%

0.53%

0.29%

0.43%

12.59%

15.09%

N. America

-0.18%

-0.38%

-0.16%

0.01%

-0.25%

-0.18%

11.61%

9.65%

Europe

-0.05%

-0.05%

0.00%

-0.08%

0.01%

-0.14%

17.54%

15.53%

1961

2011

Total land used for crops (cropland) 1961-2011 1961-1971 1971-1981 1981-1991 1991-2001 2001-2011 World

0.26%

0.49%

0.21%

0.45%

-0.03%

0.24%

100.00%

100.00%

Africa

0.89%

1.06%

0.37%

0.75%

0.88%

1.42%

12.32%

16.63%

Asia

0.39%

0.56%

0.28%

0.95%

0.11%

0.07%

34.44%

36.56%

LAC

1.26%

2.75%

1.11%

0.63%

0.69%

1.44%

7.51%

12.11%

N. America

-0.23%

0.43%

-0.09%

-0.07%

-0.40%

-0.88%

17.17%

13.56%

Europe

-0.37%

-0.49%

-0.10%

-0.16%

-0.88%

-0.22%

28.55%

21.14%

1961

2011

Permanent meadows and pastures (grassland) 1961-2011 1961-1971 1971-1981 1981-1991 1991-2001 2001-2011 World

0.17%

0.26%

0.18%

0.32%

0.30%

-0.18%

100.00%

100.00%

Africa

0.05%

0.02%

-0.04%

0.17%

0.04%

0.07%

28.75%

27.14%

Asia

0.23%

0.55%

0.33%

0.52%

0.48%

-0.64%

34.61%

35.62%

LAC

0.39%

0.75%

0.50%

0.50%

0.17%

0.11%

14.84%

16.46%

-0.14%

-1.08%

-0.23%

0.10%

-0.11%

0.43%

9.14%

7.84%

0.22%

0.37%

0.10%

-0.01%

0.75%

-0.09%

12.66%

12.94%

N. America Europe

Source: FAOSTAT.

Over the past half century total farmland growth was positive in Africa, Asia, and Latin America, but negative in North America and Europe. Total farmland was stagnant in the 2000s as a result of reductions in Europe, North America, and Asia, while Africa and Latin America kept expanding albeit slowly. This was the result of various regional changes in both cropland and grassland. 79

Land use and agricultural productivity

Africa and Asia together contain about 60% of all farmland and 53% of cropland, while another 15% of farmland and 11% of cropland is in Latin America (Table 16). In the last half century both farmland and cropland grew faster than the world average in Latin America, Africa, and Asia, and both slightly decreased in Europe and North America. In 2001-2011, farmland only expanded (slightly) in Latin America (at 0.36% per year) and Africa (at 0.35%). In the same period cropland expanded in LAC and Africa (at 1.42% and 1.44% per year, respectively), had a negligible (0.07%) growth rate in Asia, and significantly decreased in Europe and North America where it had already declined in every decade since 1971. This brief examination of agricultural land use shows that total farmland grew very slowly since 1961, peaked in the early 1990s, and changed very little ever since (with a slight decline in the 2000s). The vigorous worldwide growth that more than trebled agricultural production since the 1960s therefore has not been accompanied by a proportionate expansion in the use of land for crops or livestock, and this contrast has been even more stark since the turn of the 21st century. As mentioned before, this stagnant use of land occurred while world population grew significantly and there was an even faster increase in farm output. This growth did not require much extra farmland; it came mostly from extra output per hectare, as discussed below in more detail. Cropland, as well as farmland, expanded only about 10% between 1961 and the mid-1980s, and then remained largely stagnant but did start to grow slightly again in the late 2000s. Among major regions, Africa and Latin America showed more significant growth of cropland in the 2000s (at 1.39% and 1.31% per year, respectively). Cropland has been stagnant lately in Asia, while it continues its gradual long-term reduction in Europe and North America. Grasslands, for their part, have decreased globally in recent years, but this decrease is confined to Asia; in other regions, grassland is stable or still expanding, albeit very slightly. Data on land use and farm production in all regions of the globe during the last half century confirm something that has been known since the Industrial Revolution - growth in agricultural production (and food output) depends more on increasing land productivity than on farmland expansion. About 95% of cumulative growth in 1961-2011 is explained by increased output per hectare, and about 5% by addition of extra land, both for crops alone and for the whole farm sector (comprising crops and livestock). This process of increasing productivity per unit of land is discussed next. 80

Hector Maletta

4.3. Land use and land productivity Agricultural output (crops and livestock) measured in economic terms may be seen as the product of total farmland multiplied by the average value of production per hectare of farmland, or farmland productivity. Likewise, crop production may be regarded as the product of two analogous factors: the area of cropland, and the average output per hectare of cropland (i.e., cropland productivity).

4.3.1. Farmland expansion and agricultural output growth As we have seen, farmland has increased very little (on the world scale) since 1961, expanding by just 10% in half a century, with almost all the increase occurring up to 1993, whereas agricultural output more than trebled in the same period. Thus, the growth of agricultural output should be explained mostly by increases in productivity. Farmland productivity, in relation to total agricultural production, is here defined as the ratio of farm output value (crops and livestock products valued at constant and uniform prices) to total farmland (i.e., the sum of arable land, land with permanent crops, and permanent meadows and pastures). If farmland productivity were constant, agricultural output would grow in proportion to the growth of farmland. Any excess growth over and above the growth allowed by expansion of farmland is attributable to growth in farmland productivity. This is illustrated in Figure 17, which charts the cumulative growth of farm output since 1961, i.e., the additional level of output, over and above the output attained in that initial year, split into the portion attributable to additional farmland (keeping productivity constant), and the portion attributable to higher farmland productivity (more output per hectare of farmland).

81

Land use and agricultural productivity

Billion PPP USD, 2004-06 prices

Figure 17. Contribution of extra farmland and increased farmland productivity to growth since 1961 to world agricultural output, including crops and livestock (billions of PPP USD at 2004-2006 prices) 1800 1600 1400 1200 1000 800 600 400 200 0 1961

1971

1981 Farmland

1991 Farmland productivity

2001

2011

Source: FAOSTAT.

In Figure 17, the thin, darker area along the bottom represents growth attributable to additional farmland added since 1961, i.e., the output that would have been attained if output per hectare had remained constant. The lighter (and much larger) area above it reflects growth derived from higher output per hectare. Additional farmland contributed 4.3% of total cumulative agricultural growth, while farmland productivity growth contributed the remaining 95.7%. Expansion of the agricultural frontier, i.e., the use of more land for agricultural production, therefore played only a minor role in the growth of world agricultural output from 1961 to 2011. Moreover, the limited expansion of farmland since 1961 occurred mostly in the first three decades of the period considered. Farmland grew worldwide only until the early 1990s and then flattened, with a small decrease in the 2000s. In contrast, output grew throughout this period. This pattern of increasing land productivity accompanied by a stagnant use of land for agriculture is present in all world regions (Table 17). Real output per hectare grew in all regions, but chiefly in Asia, Africa, and Latin America. World farm output per hectare has been growing at more than 2% per year for half a century, and so far shows no sign of slowing down. In fact, it is accelerating: in the latest decade considered here (1999-2001 to 2009-2011), it has been growing at 2.79% per year, well above the half-century average (2.26%) and above all the preceding periods considered. Acceleration relative to the 1980s or 1990s is perceptible in all regions, and especially in Africa, Asia, and Latin America. 82

Hector Maletta

To sum up, farmland expansion, at constant productivity, explains just 4.3% of total agricultural growth in 1961-2011. The remainder (95.7%) is explained by increased output per hectare of farmland. In recent decades, furthermore, additional land has not been added on a global scale, and has thus contributed nothing to the notable increase in world agricultural production since about 1990. Table 17. Annual growth rate of net agricultural output (crops and livestock) per hectare of farmland, 1961-2011. 1961-1963 to 2009-2011

1961-1963 to 1969-1971

1969-1971 to 1979-1981

1979-1981 to 1989-1991

1989-1991 to 1999-2001

1999-2001 to 2009-2011

World

2.26%

2.35%

2.15%

2.03%

2.02%

2.79%

Africa

2.58%

2.85%

1.33%

2.83%

2.81%

3.12%

Asia

3.16%

2.72%

2.60%

3.25%

3.32%

3.84%

LAC

2.49%

2.05%

2.43%

1.84%

2.77%

3.29%

N. America

1.89%

2.14%

2.76%

0.78%

2.32%

1.51%

Europe

0.85%

2.25%

1.69%

0.92%

-1.22%

0.92%

 

Farm output = Value of net crop and livestock production (excluding products used as feed or seed) at world average 2004-2006 producer prices, converted into international dollars at AgPPP conversion rates. Farmland = Agricultural land = Arable land, land with permanent crops, and permanent meadows and pastures. Source: FAOSTAT.

4.3.2. Cropland expansion and crop production growth The previous section deals with total agricultural output (crops and livestock) and total agricultural land (arable, permanent crops, and permanent meadows and pastures). The same general conclusion as reached above may hold if the analysis is limited to crop output and cropland. Additional cropland at constant cropland productivity contributed just a small fraction (5.7%) of cumulative crop output growth from 1961 (Figure 18), while 94.3% of gross crop output growth reflects increased cropland productivity, i.e., more output per cropland hectare.12

This estimate is based on gross crop output, including amounts used as animal feed (and also amounts used as seed, but these are very minor compared to feed). The use of gross output is sensible here, because what is sought is total crop production. However, the conclusions would be nearly identical if they were based on net crop output.

12

83

Land use and agricultural productivity

Table 18. Growth rate of gross crop output value per hectare of cropland, 1961-2011, at constant 2004-2006 prices. 1961-1963 to 2009-2011

1961-1963 to 1969-1971

1969-1971 to 1979-1981

1979-1981 to 1989-1991

1989-1991 to 1999-2001

1999-2001 to 2009-2011

World

2.13%

2.30%

1.97%

1.74%

2.30%

2.39%

Africa

1.83%

2.34%

0.53%

2.61%

2.19%

1.61%

Asia

2.76%

2.59%

2.49%

2.40%

3.23%

3.08%

LAC

1.69%

0.98%

1.65%

1.69%

1.92%

2.06%

N. America

2.01%

1.74%

3.52%

0.54%

2.29%

1.96%

Europe

1.27%

2.62%

1.01%

0.67%

1.14%

1.17%

 

Gross crop output = value of production of all crops (food or non-food, temporary or permanent, including amounts used for feed or seed) at world average 2004-2006 producer prices, converted into dollars at AgPPP conversion rates. Cropland = Arable land and land with permanent crops (not including permanent cultivated meadows and pastures). Source: FAOSTAT.

Figure 18. Contribution of additional cropland and higher gross crop output per cropland hectare to world gross crop output growth, 1961 to 2011, in billions of AgPPP dollars at 2004-2006 prices. Billion USD at 2004-06 PPP prices

$1,200 $1,000 $800 $600 $400 $200 $0 1961

1971

1981 Cropland

1991

2001

2011

Cropland productivity

Source: FAOSTAT.

As shown in Table 18, world crop output per hectare grew at a yearly rate of 2.13% from 1961-1963 to 2009-2011. Growth rates, however, first decelerated from the 1960s to the 1980s, and then accelerated again after 1990. In the 2000s, the rate was a yearly 2.39%, above the half-century average rate of 2.13%. Asia’s crop output per hectare grew at 2.76% per year in 1961-2011, faster than any other region over the same half century; its 84

Hector Maletta

growth rate slowed down from the 1960s to the 1980s and accelerated again in the two latest decades, as also occurred in Latin America. The growth rates of output per hectare (either total agricultural output or just crops) are due not only to higher physical yields for each particular crop or livestock, but also to more efficient allocation of resources and knowledge, increased cropping intensity (e.g., two or more crops per year on the same piece of land), shifts to more valuable products, changes in the geographical distribution of production across countries and continents, new technology (mechanical, chemical, biological, managerial), and a catch-up process leading to more extended and intensive use of already available technologies, all resulting in rapidly growing output per hectare. Farmers across the world, including those in Asia, Africa, and Latin America, are changing how they farm in a number of ways: altering their product mix; using more inputs like fertiliser and improved seeds; expanding irrigation or improving its efficiency; using short-cycle crops to cultivate two or more crops per year on the same piece of land; adopting non-tillage methods of production for extensive crops, thus reducing soil erosion and saving on machinery and fuel; and trying to catch up with growing market demand, existing technology, and available opportunities for improvement. Since not everyone is at the production efficiency frontier, catching up with available technologies is an on-going process. Large differences in productivity exist across regions and countries (and also across sub-national regions of the same country and even among farmers in the same zone), and there are considerable differences in the extent to which modern inputs and technologies are used. This suggests that the process of catching up has a long way to go. As a result of this process, whereby existing knowledge is gradually diffused among farmers, the growth of land productivity worldwide may be expected to continue to be, for a long time, above the rate of expansion of the technological possibilities frontier - which is itself advancing, albeit at a somewhat slower pace. To sum up, just about 5% of total growth in agricultural production since 1961 may be attributed to expansion of the agricultural frontier, i.e., to an increase in agricultural land use. This is true of total farm output (crops and livestock) and of crops alone. Moreover, most of that small increase in farmland and cropland occurred in the first decades of that period; since 1990, cropland remained stable and farmland has slightly contracted.

85

Land use and agricultural productivity

4.3.3. Components of crop output growth Expansion of cropland area, at constant cropland productivity, directly explains (as we have seen above) just 5.8% of all the cumulative growth in crop output (which was 230% in fifty years); the other 94% of growth in crop output reflected higher productivity per hectare of cropland. The latter in turn is the joint result of changes in cropping intensity, crop yields, and crop mix. Available data allows not only growth between cropland and its productivity to be apportioned; cropland productivity itself may be analysed into its component factors, chiefly the level of cropping intensity (area harvested per year as a fraction of total cropland), the crop mix (in terms of the shares of total harvested land allocated to the various crops), and the physical yields of the various crops (metric tons per harvested hectare). According to FAOSTAT definitions, a particular piece of cropland may be used for temporary or permanent crops. In the case of temporary crops, it is often the case that two or more crops are successively grown on a particular hectare during a particular year; each temporary crop harvested in a particular hectare counts towards total harvested area. Some hectares, of course, may be not harvested at all due to temporary fallow, crop failure, or other reasons. In the case of permanent crops, part of them may be young growing plants not yet yielding any product (and thus not harvested). For those that are productive, only one harvest is counted per year, even if the actual gathering of products is distributed throughout the year. The net result is that total harvested area may be larger or smaller than total cropland; in practice, it is smaller at the world level but larger in some regions (notably Asia), indicating differing levels of average cropping intensity. Cropping intensity is equivalent to total harvested area divided by total cropland. Some hectares of cropland may be in fallow, or suffering crop failures, or holding permanent crops that are still at the growing stage, and (as a result of any of these possibilities) not harvested at all. Most cropland hectares are harvested only once a year, but some support two or more successive temporary crops per year. As a net outcome, cropland area and harvested area usually differ. Total harvested area was 71% of cropland area in 1961, but this cropping intensity ratio increased to 85% by 2011 (Figure 19). This is also the case in almost all regions (Table 19). The only region showing a (slight) decline in cropping intensity is Europe, where it oscillated at around 66-69% in 1961-1981 but fell to 60-63% during the 1990s and remained at this level in the 2000s.

86

Hector Maletta

Figure 19. World cropping intensity (harvested area of temporary and permanent crops as a percentage of cropland), 1961-2011. 90% 85% 80% 75% 70% 65% 1961

1971

1981

1991

2001

2011

Source: FAOSTAT.

Figure 20. Cropping intensity by region, 1961-2011: total harvested area of temporary or permanent crops, as a percentage of total cropland. 120% 110% 100% 90% 80% 70% 60% 50% 40% 30%

1961

1971 World

Africa

1981

1991

Asia

LAC

Source: FAOSTAT.

87

2001 N. America

2011 Europe

Land use and agricultural productivity

Table 19. Cropping intensity by region, 1961-2011.  

1961

1971

1981

1991

2001

2011

World

70.9%

73.0%

77.1%

75.9%

78.2%

85.2%

Africa

61.6%

69.9%

63.8%

77.1%

82.8%

88.7%

Asia

90.4%

92.1%

97.3%

94.1%

97.7%

108.2%

LAC

70.3%

71.8%

75.4%

73.6%

74.9%

79.7%

N. America

45.9%

46.9%

59.2%

52.1%

54.2%

58.1%

Europe

66.5%

66.6%

68.6%

64.2%

60.3%

63.2%

Cropping intensity = harvested area of all temporary and permanent crops (except cultivated pastures) as a percentage of total cropland (which includes ‘arable land’ and ‘land with permanent crops’). In the case of multiple successive temporary crops grown on the same land during the same year, each harvest counts towards total harvested area. Source: FAOSTAT.

Cropping intensity is highest in Asia, increasing from 90% in 1961 to 108% in 2011; it is lowest in North America, where it nonetheless increased from 46% to 58% during the same period. It grew from 61% to 89% in Africa and from 70% to 80% in LAC, both closer to the world average than Asia, North America, and (most recently) Europe. Increasing cropping intensity meant that a 13% increase in cropland (from 1961 to 2011) translated into a 36% increase in harvested land. Increased cropping intensity directly explains about 9% of total crop output growth during that period, over and above the 5.8% directly explained by cropland expansion. The rest of crop output growth is explained by increased productivity per harvested hectare, which reflects a combination of changes in the yields of individual crops, and changes in the crop mix, i.e., in the allocation of harvested area among the various crops. Changes in the yields of individual crops, at constant crop mix, explain a substantial portion of growth in the productivity of crops per harvested hectare. Yields of specific crops are usually measured as metric tons per harvested hectare. To obtain an index of the general increase in yields, the various crops should be added with adequate weights. Adding up the tonnage of heterogeneous products (such as oranges, cauliflowers, watermelons, maize, or blueberries) makes little sense. Adding in terms representing specific qualities like calories or vitamins likewise makes little sense, since food provides various nutrients (carbohydrates, protein, vitamins, minerals) and there is no obvious way to add them up into a single figure - products with high energy values may lack other nutrients (e.g., vegetable oil or sugar, the typical sources of ‘empty calories’) while others provide specific micronutrients with little 88

Hector Maletta

energy content (e.g., broccoli). Like other heterogeneous sets of products, the most usual weights for a quantity index of crop production per hectare are prices, in this case the constant and uniform PPP producer prices used by FAO to calculate the value of agricultural output. A quantity index reflects changing quantities weighted by constant prices, just as a price index reflects changing prices weighted by constant quantities. Quantity indexes may be expressed as a relative index number (typically set at 100 in the base year) or as a monetary value. For convenience, we use the latter option, but the monetary values are simply indicators of real change over time. The value of output per harvested area, measured in PPP USD at 2004-2006 prices, grew from 479 USD in 1961 to 1,169 USD in 2011, an increase of 689 USD per hectare. These figures are the net result of two processes: changes in the various individual yields and changes in the crop mix (the set of shares of total harvested area occupied by the various crops). Keeping the crop mix constant measures the pure change in yields, with no interference for changing allocation of land among crops. The growth of physical yields at constant crop mix explains 68% of that increase of harvested land productivity (value of output per harvested hectare); the 2011 average productivity at the 1961 crop mix would have been 950 USD per harvested hectare, an increase of 470 USD above the 1961 value. This means the direct impact of changes in the crop mix explains nearly a third of the growth in productivity per harvest. Figure 21 shows the course followed by average productivity both at the actual crop mix of every year (solid line) and with the crop mix fixed at its 1961 state (dotted line). The two evolved close to each other up to the early 1980s, but then started to diverge. The gap between the two lines reflects the change in productivity due to changes in the crop mix. Changes in the crop mix invariably tended to increase productivity, indicating that changes in the allocation of cropland tended to favour crops with higher value per hectare, i.e., with higher yield, higher price, or both.

89

Land use and agricultural productivity

Figure 21. Real growth since 1961 in average value of crop production per harvested hectare, 1961-2011, in PPP USD at 2004-2006 prices, weighting crops by either their actual or their base-year share of harvested land. $800 $700 $600 $500 $400 $300 $200 $100 $0

1961

1971

1981

1991

At actual crop mix

2001

2011

At 1961 crop mix

Source: FAOSTAT.

The two main conclusions that can be drawn from this rapid examination are that the addition of extra cropland played just a very minor role in crop output growth, and that changes in cropping intensity and in the crop mix were important complements of increased physical yields as the main drivers of crop output growth. Changes in the crop mix are mostly driven by forces outside agriculture mainly changes in demand for various products, prompted by changes in consumer preferences, relative prices, or income growth. Changes in yields are chiefly the result of development and adoption of technical innovations of various sorts (better seeds, improved plant protection, new machinery, better farming practices, and the like). These technical changes may have their origins in agriculture itself or in industries providing inputs for agricultural activity (e.g., seeds and machinery). The net effect of these factors is growth in productivity. The growth of output may thus be seen as the combined result of using additional resources (land, labour, machinery, seeds, etc.) and achieving increased productivity of such resources. The latter seems to have been the dominant driver in recent times, though the addition of extra inputs was predominant in the earlier years of the past half century. It is to this matter that we now turn.

90

Hector Maletta

4.3.4. Factor productivity and agricultural growth The history of agricultural land and output in the latest half century, as in preceding periods following the Industrial Revolution, shows that farm production and thus food production depend more on innovation and efficiency than on the area of land used by farms. In fact, they depend more on innovation and efficiency than on the growth of all production inputs including land, infrastructure, equipment, labour and materials, which grew slowly or even decreased due to input-saving technologies (Fuglie 2012; Fuglie et al. 2012). These studies show that total factor productivity is the driving force behind agricultural growth, and its growth is accelerating, thus offsetting stagnant land use and a declining rate of growth in the use of the various inputs. Starting in 1975, FAOSTAT incorporated a series on capital investment in agriculture. In addition, the International Labour Organisation has been providing estimates of agricultural employment since 1980. The latter series is used by von Cramon-Taubadel et al. (2011) to estimate the contribution of land, labour, and capital to the growth of agricultural production. Fuglie (2012) achieved wider coverage in terms of countries and periods, and resorted to complementary sources in order to overcome some of the limitations in the FAOSTAT capital investment series; one of these limitations is that the figures tend to ignore variation over time and space in the quality of capital goods and agricultural inputs, since a common unit value for all items of the same kind is used (e.g., all tractors are given a single unit value, regardless of horsepower and other characteristics). As a result, improvements in the quality of equipment or inputs are not distinguishable from improvements in the efficiency of farmers when using such equipment or inputs - both are merged into agriculture’s total factor productivity. Stephan von Cramon-Taubadel et al. (2011: 307) estimate that from 1975 to 2007, agriculture’s total factor productivity (also including increased productivity in the production of inputs) increased at a yearly rate of 1.7% on the world scale. The rate was lower but still significant in Sub-Saharan Africa (0.9%) and Latin America (1.0%). It was higher in China (2.1%) and the composite North America/Pacific area where TFP growth was estimated at 2.0%. The rest of Asia (minus China) grew – according to the study – at a yearly 1.4%, the same as Europe, while the estimate for the Near East and North Africa (NENA) region was 1.5%, and 1.7% for all transition economies.

91

Land use and agricultural productivity

At the world level, von Cramon-Taubadel et al. estimate that 1.0 percentage points in the TFP growth rate were due to technical change and 0.7 points to improved efficiency in the use of existing technology.13 This partition shows significant differences across regions; in the NENA region almost all improvement was in efficiency (1.4 out of 1.5) while technical change was only responsible for the remaining 0.1 point. The opposite occurred in all other regions considered, where technical change grew faster than efficiency (Table 20). Table 20. Estimated changes in TFP and its components by region, 1975 to 2007.  

Efficiency change

Technical change

TFP change

NENA

1.4%

0.1%

1.5%

SSA

0.3%

0.6%

0.9%

NAP

-0.7%

2.7%

2.0%

LAC

0.3%

0.7%

1.0%

Asia excl. China

-0.9%

2.3%

1.4%

China

0.9%

1.3%

2.1%

Europe

0.3%

1.1%

1.4%

Transition

0.7%

1.0%

1.7%

World

0.7%

1.0%

1.7%

Source: von Cramon-Taubadel et al. 2011:308, in FAO 2011a.

Estimates by Fuglie (2012) are more complete in their coverage and employ other supplementary sources of data concerning capital goods and materials used in agriculture, and the cost share of each factor. The period of analysis is also longer (1961-2009). The estimates of the TFP growth rate are somewhat lower than those found by von Cramon-Taubadel et al. (2011), though still quite significant (Table 21). Growth rates are based on regressing each variable against time (i.e., estimating the β coefficient in equations such as ln(xt) =α+βt, where t is time and x may be output, TFP or another variable).

Data correspond to the most complete variant of the authors’ estimates (last row of their table on p. 307).

13

92

Hector Maletta

Table 21. Output, input and TFP growth rates for world agriculture. Period 1961-2009

Gross output

Total inputs

Total factor productivity

Output per worker

Output per hectare

2.23%

1.28%

0.95%

1.19%

2.00%

1961-1970

2.74%

2.55%

0.18%

1.13%

2.45%

1971-1980

2.30%

1.70%

0.60%

1.58%

2.09%

1981-1990

2.12%

1.50%

0.62%

0.62%

1.75%

1991-2000

2.21%

0.55%

1.65%

2.00%

2.16%

2001-2009

2.49%

0.65%

1.84%

2.80%

2.64%

1971-1990

2.25%

1.53%

0.72%

1.11%

1.97%

1991-2009

2.29%

0.70%

1.59%

1.97%

2.27%

Source: Fuglie (2012), Table 16.3.

The importance of TFP has increased over time, in parallel to the decrease in the relative importance of extensive growth in the use of inputs (land, labour, and capital). This is true for the various decades as well as for two long periods: 1971-1990 and 1991-2009. Between these two periods, the growth rate of TFP doubled, while the growth rate of input use was halved. Analysis by regions shows disparities between developed and developing countries considered as a whole. Fuglie (2012:307) provides results for many regions and sub-regions across the world, broadly consistent with the overall picture presented. The figures for the main groups of countries (developed, developing, and transition economies) are shown in Table 22. Use of inputs actually diminished during recent decades in developed and transition countries, and expanded at slower rates in developing countries. The contraction in the use of inputs in developed countries was more than offset by accelerating TFP growth, resulting in positive growth in output. In transition countries, the reduction in inputs (mostly related to the collapse of centrally-planned economies around 1990) was not compensated by TFP during the 1990s, causing output to decrease. However, positive growth resumed in the 2000s as the decrease in inputs tended to be much slower and productivity growth restarted. In developing countries, in contrast, a steady output growth rate of around 3% per year was dominated by input application growth in the earlier periods, and by TFP in the later ones. Increasing TFP indicates that the potential product per hectare of agricultural land is rapidly increasing. Output (total and per hectare) is certainly growing well ahead of population growth, but it could grow even faster. If it does not, this will be primarily related to demand. 93

Land use and agricultural productivity

Table 22. Agricultural growth and TFP by groups of countries, 1961-2009.  

1961-1970

1971-1980

1981-1990

1991-2000

2001-2009

2.74%

2.30%

2.12%

2.21%

2.49%

Output growth rate World Developed

2.05%

1.93%

0.72%

1.37%

0.58%

Developing

3.15%

2.97%

3.43%

3.64%

3.34%

Transition

3.27%

1.32%

0.85%

-3.51%

1.96%

Input growth rate World

2.56%

1.70%

1.50%

0.56%

0.65%

Developed

1.06%

0.29%

-0.64%

-0.86%

-1.86%

Developing

2.46%

2.04%

2.31%

1.42%

1.13%

Transition

2.70%

1.43%

0.27%

-4.29%

-0.32%

World

0.18%

0.60%

0.62%

1.65%

1.84%

Developed

0.99%

1.64%

1.36%

2.23%

2.44%

TFP growth rate

Developing

0.69%

0.93%

1.12%

2.22%

2.21%

Transition

0.57%

-0.11%

0.58%

0.78%

2.28%

Source: Fuglie 2012, Table 16.4.

In many projections of future agricultural growth, it is assumed that the rate of growth in productivity will eventually slow down, an assumption presumably due to a precautionary principle. In fact, we will see below (in Section 9.2.4) that major past projections of agricultural output have systematically understated the growth in productivity. Their expectations about technical progress were too low; instead of decelerating, growth in total factor productivity is accelerating and shows no sign of relenting. The relatively slower growth in some regions has more to do with the withdrawal of resources from agriculture than with any slowdown in technical progress or less improvement in efficiency.

4.4. Availability of land suitable for rain-fed crops Farmland and cropland both seem to have plateaued around 1990, with little or no growth in recent years, while production and productivity soared. Was the overall stagnation in land use a consequence of land scarcity? Have humans already used up nearly all land potentially usable for crops and livestock? How much land is actually left that is suitable and available for agricultural production, and how much land is expected to be put into agricultural use in the coming decades? These questions are discussed next. 94

Hector Maletta

4.4.1. Basic concepts Only part of the land area around the world is suitable for crop growth. Sandy deserts and frozen tundra are in principle not suitable for that purpose. Land suitability depends mostly on four factors: soil, climate, crop choice, and the cropping practices of farmers. Climate, in turn, affects agriculture mostly through temperature and precipitation. Since human activity and ingenuity is involved, evaluating the agricultural suitability of land must be conditional on certain production practices and techniques. For instance, arid land or even desert may be made suitable for crops through irrigation, conveying water to the crops from a surface source (river, lake, spring), or extracting water from underground aquifers. If irrigation water is itself scarce, watersaving irrigation techniques could be used, such as drip irrigation, enabling farmers to irrigate crops in the desert with very small amounts of water (as practiced, for instance, in Israel’s Negev desert and elsewhere). Similarly, a cold climate (within certain limits) may be offset by greenhouses. Lands that are not suitable for certain crops may be suitable for others; bananas or olives do not grow easily in Canada, but wheat does – albeit only certain varieties adapted to the Canadian climate. Thus, determining how much land is ‘suitable’ must be defined in relation to some specific set of crops (or crop varieties) and cropping techniques. In this book, we deal with land that is suitable for rain-fed crops and a specific set of major food crops. Land actually used for crops is only part of all the land suitable for growing crops that exists around the globe. It may also include some land that is marginal or not suitable but is nonetheless cropped. One important question is how much suitable land is still available, should it be needed for future crop growth. Land that is potentially suitable but is not available includes suitable land already in use for crops and suitable land that is used for other purposes (built-up areas, and areas that are forested or otherwise strictly protected). The latter could still be cropped, but only if the forest is cleared or protected areas are left without protection. However, calculations about available suitable land usually exclude not only land already under crops but also land that is devoted to those other uses (buildings, roads, forests, or protected areas). Of course, crops may also be grown on non-suitable land. There are poor people who farm very marginal lands, obtaining a very meagre output. Crops can also be grown on desert land, albeit at higher costs and using technology such as computer-driven localised drip irrigation. Also, some previously unsuitable land might be rendered suitable through human ingenuity, e.g., supplying extra water to arid areas through new or improved irrigation works. 95

Land use and agricultural productivity

The present discussion, however, refers to land suitable for rain-fed crops that is not yet used and is available in the sense of not being built-up, forested, or otherwise strictly protected. Suitability for rain-fed crops requires sufficient rainfall (adequately distributed throughout the year), as well as adequate temperatures during the crop cycle, and adequate soil characteristics. Soil quality, temperatures, the amount and seasonal distribution of rainfall, crop choice, and cropping technology jointly determine the expected yield. Suitability is estimated in relation to a wide set of crops (and crop varieties), understanding that a piece of land is deemed ‘suitable for rain-fed crops’ if at least some of the crops (or some of their varieties) can be grown on it.

4.4.2. Land classification: suitability and availability FAO’s World Soil Map and other soil-related data resources (http://www.fao. org/soils-portal/en/), as well as its Global Agro-Ecological Zones (GAEZ) programme which combines soil and climate data (http://www.fao.org/nr/ gaez/en/), have been used by FAO and IIASA to identify land areas across the world with different current uses and various degrees of crop suitability, assessed in relation to various crops if grown under rain-fed conditions and assuming the use of certain kinds of technology. Each type of land may suffer from various constraints as re­gards rain-fed crops: temperature and rainfall regimes, soil quality, slope, and so on. Table 23 shows the standard land classes identified by FAO. Suitability classes are defined according to the attainable yields of a wide set of major crops, as a percentage of the maximum yield attainable in a total absence of soil and climate constraints. Very few lands are able to produce at the maximum rain-fed yield predicted by agricultural scientists under a no-constraints assumption. In practice, even on the best lands, the yields actually obtained by farmers are below the theoretical maximum. Thus, assigning a piece of land to the highest land class (Very Suitable) does not require the maximum yield to be obtained: this class includes all lands where yields are above 80% of the unconstrained standard. The ‘Suitable’ category has yields of between 60% and 80%, while land with yields between 40% and 60% of the maximum are still classed as ‘Moderately Suitable’. These three categories, comprising all lands with yields above 40% of the potential, are deemed ‘Suitable’ and also referred to as ‘Prime and Good Land’. The rest is deemed ‘Marginal’, though some of it is ‘Marginally Suitable’, or ‘Very Marginally Suitable’.

96

Hector Maletta

Table 23. Land classes according to suitability for rain-fed crops, and world area by class. Broader class

Broad class

Class

Class name

Yield, as % of reference

Total land surface Not constrained (reference) Prime Suitable

Marginal

Good

Marginal

VS

World area (Millions Ha) 13295

100%

Very Suitable

80-100%

1315

Suitable

60-80%

2187

MS

Moderately Suitable

40-60%

993

mS

Marginally Suitable

20-40%

1111

vmS

Very Marginally Suitable

5-20%

1627

NS

Not Suitable

< 5%

6061

S

Source: Alexandratos and Bruinsma 2012:104. Crop groups considered include cereals; roots and tubers; sugar crops; pulses; and oil-bearing crops.

These degrees of suitability for rain-fed crops (defined by the achievable yield as a percentage of the technically maximum potential) may vary according to the choice of technology level. Land that is not suitable under traditional technology may be rendered suitable with more advanced techniques of production. Assessments of land classes are based on a ‘reference technology’. The basic data for suitability of land is the comprehensive FAO analysis of the world Global Agro-Ecological Zones (GAEZ) database. Its latest update at time of writing is reported by Fischer et al. (2012). Here, land classification is based on the ‘mixed-technology’ level, as explained next, and refers to a wide set of important crops. The GAEZ approach first determines the climatic and edaphic characteristics of each zone, and then estimates the expected yield under rain-fed conditions for a variety of crops, expressed as a percentage of the yield expected under no constraints and assuming a certain technological level. Four technological levels are considered: low, intermediate, high, and mixed. Estimates presented here are for ‘mixed’ technology, which assumes that high technology is used on the best classes of land, intermediate technology on land of intermediate quality, and low technology on low-grade or marginal land (Fischer et al. 2012:56). The analysis is based on a very detailed grid for the entire land area of the planet. Levels of agricultural technology were defined as follows (Fischer et al 2012:56):

97

Land use and agricultural productivity

Low-level inputs/traditional management. Under the low input, traditional management assumption, the farming system is largely subsistence-based and not necessarily market-oriented. Production is based on the use of traditional cultivars (if improved cultivars are used, they are treated in the same way as local cultivars), labour intensive techniques, and no application of nutrients, no use of chemicals for pest and disease control, and minimum conservation measures. Intermediate-level inputs/improved management. Under the intermediate input, improved management assumption, the farming system is partly market-oriented. Production for subsistence plus commercial sale is a management objective. Production is based on improved varieties, on manual labour with hand tools and/or animal traction and some mechanisation. It is medium labour intensive, uses some fertiliser application and chemical pest, disease and weed control, adequate fallows and some conservation measures. High-level inputs/advanced management. Under the high input, advanced management assumption, the farming system is mainly market-oriented. Commercial production is a management objective. Production is based on improved high yielding varieties, is fully mechanised with low labour intensity, and uses optimum applications of nutrients and chemical pest, disease, and weed control.

4.4.3. Land suitability, availability and use Besides suitability, it is necessary to consider whether a piece of land is actually available for cultivation: some of it may already be cultivated; some may be covered by forests; some may be otherwise protected (e.g., non-forested national parks); and some may be covered by buildings or roads. The balance after subtracting all these categories represents land suitable for rain-fed crops that is also available for cropping. All the land area around the globe (13.3 billion hectares) has thus been classified according to its crop suitability and current use. Table 24 provides the resulting distribution of global land.

98

Hector Maletta

Table 24. Land suitability classes for rain-fed crops, by actual use (world total, millions hectares).

A

Prime and good land

Land uses

Total land

Total

VS

S

All land uses

13295

4495

1315

2187

Marginal land MS

Total

mS

vmS

NS

993

8800

1111

1627

6061

B

Under forest

3736

1601

453

854

293

2135

342

530

1263

C

Built-up land area

152

116

41

61

14

37

12

10

15

D

Other land, strictly protected a

638

107

30

50

27

530

39

59

432

 

2671

791

1222

659

1746

718

1028

 

1559

1260

442

616

201

299

120

104

75

Rain-fed

1283

1063

381

516

166

220

93

84

43

Irrigated

276

197

61

100

35

79

27

20

32

 

1412

349

606

458

1521

598

923

 

with some E=A-B-C-D Land rain-fed potential F=G+H G H I=E-F

Land used for crops (1999-2001) b

Net balance c

(a) Includes only protected land that is not presently used for crops, built-up, or under forest (e.g., wildlife reserves). (b) Arable land and land with permanent crops, excluding cultivated permanent meadows and pastures. (c) Includes suitable and land that is marginally or very marginally suitable, excluding the non-suitable (NS) class. Adapted from Alexandratos and Bruinsma (2012:104) (rows reordered; headings slightly modified). Totals may not add up due to rounding.

It turns out (Table 24, line A) that 4495 MHa worldwide are very suitable or suitable (i.e., Prime or Good land) for rain-fed cultivation, including only the best three land classes: 1315 MHa in the VS class, 2187 MHa in the S class, and 993 MHa in the MS class. Marginally suitable lands (mS and vmS) are excluded from this calculation. Deducting all areas under forest, built-up, or otherwise strictly protected (e.g., non-forest national reserve parks) results in a net balance of 2672 million Ha of Prime and Good land for rain-fed cultivation. Part of these lands (1260 MHa) were devoted to crops in the baseline period (1999-2001) leaving a net balance of 1412 MHa of Prime and Good land that was not yet used for growing crops, was neither under forest nor built-up, and was not otherwise strictly protected. As cropland worldwide has not changed much since 2001 (Table 14 and Figure 15), this assessment is still substantially applicable in the 2010s. The usable but yet not cropped Prime and Good land area (1412 MHa) is larger than the Prime and Good land that is already effectively cropped (1260 MHa as of 1999-2001, with little change afterwards). Thus, more than half of all crop-suitable and usable land is not yet used for crops. Cropland, therefore, could potentially be doubled using only Prime and Good land, without encroaching onto forests, protected land, built-up areas, or marginal 99

Land use and agricultural productivity

land, and not considering any expansion of irrigation on otherwise marginal land. This balance of land suitable for rain-fed crops but not currently cropped is mostly covered by grasses or other non-forest vegetation such as shrubs. However, it may include some cultivated permanent meadows and pastures, since many countries fail to make a clear distinction between natural and cultivated permanent meadows and pastures. About 300 MHa of marginal land classes are currently used for crops (line F in Table 24), i.e., 19% of all land used for such purposes. Some 70% of crops growing on marginal land are rain-fed and the rest is irrigated. The share of irrigated crops is about 20% in marginally suitable (mS) and very marginally suitable (vmS) land, and nearly 44% on not suitable (NS) land where rain-fed yields are below 5% of the constraint-free potential. It may be inferred from these figures that most crops grown on marginal lands are rain-fed and that average yields on marginal lands must be very low. Some valuable crops (e.g., fruit or vegetables) may be grown on marginal land if irrigation is available, but most crops grown on marginal land are likely to be staple crops (e.g., coarse cereals or starchy roots) with low yields and a relatively low price per metric ton. On the other hand, about 80% of the land currently used for crops is Prime or Good land. However, since Prime and Good land (otherwise called Suitable land) is more productive, is used on ave­ra­ge with better technology compared with marginal land, and such land is devoted on average to more valuable crops, it can be inferred that significantly more than 80% of the world’s real crop output (probably over 95%) comes from Prime and Good land. The rightmost column of Table 25 shows where the available Suitable land is located. About 90% of the net balance of 1412 MHa of usable but unused Prime and Good land is in three major country groups: Sub-Saharan Africa (451 MHa), Latin America (363 MHa), and developed countries (447 MHa). It is clear from these figures that there is not much Suitable land left in Asia, especially in South Asia where the net balance is just 13 MHa (possible expansion: 30 for overweight and BMI≥30 for obesity (see WHO recommendations at http://www.who.int/mediacentre/ factsheets/fs311/en/). While nutritional deficiency indicators (like stunting and underweight) have been diminishing, obesity and overweight are increasing (de Onís et al. 2010). This process is occurring in all regions and almost all countries, developing and developed, for both sexes and all ages including small children, albeit its intensity is lower in Asia and higher in Latin America. Under-five overweight (Table 43 and Figure 60) is particularly worrying, since it foretells further future increases in adolescent and adult overweight and obesity. Table 43. Prevalence of overweight (including obesity) in children under five (WHO database).  

1990

1995

2000

2005

2010

2013

2015

Africa

4.5%

4.8%

5.2%

5.5%

6.0%

6.3%

6.5%

Asia

4.1%

4.0%

4.1%

4.3%

4.8%

5.1%

5.3%

LAC

6.5%

6.6%

6.7%

6.9%

7.0%

7.1%

7.1%

All developing

4.4%

4.5%

4.6%

4.9%

5.4%

5.6%

5.8%

Global

5.0%

5.1%

5.2%

5.6%

6.0%

6.3%

6.5%

Source: WHO Global Health Observatory Data Repository http://apps.who.int/gho/data/node.main.ngest?lang=en (accessed April 2015). Surveys up to 2013; figures for 2015 are projections.

The worldwide prevalence of overweight and obesity among preschool children has increased from 5.0% in 1990 to 6.3% in 2013 (projected to reach 6.5% in 2015). From 1990 to 2015, it is estimated to have grown from 4.4% to 5.8% in developing countries (and implicitly estimated to have increased from about 9.5% to about 12.5% in developed ones). It has been rapidly increasing in Africa (from 4.5% to 6.5%), in Asia (from 4.1% to 5.3%), and in Latin America and the Caribbean (6.5% to 7.1%). The tendency to higher prevalence of child overweight is thus not confined to rich countries: it is growing in poorer ones as well.

185

Food access and nutrition

Figure 60. Prevalence of overweight in children under five (1990-2013 and projection to 2015). 8% 7% 6% 5% 4% 3% 2% 1% 0% 1985

1990 Africa

1995

2000

2005

Asia

LAC

2010

2015

2020

All developing

Source: WHO (see note to table 43)

Adolescent and adult overweight and obesity are also on the rise, especially in North America and Eastern Europe, where more than half the population is overweight. In those regions, obesity afflicts on average more than a fifth of the population and over a third in some countries, especially affecting women. About 12% of the world population over age 20 was obese in 2008 (Body Mass Index ≥ 30), including 10% of men and 14% of women. Prevalence of adult obesity is related to income: it affected (in 2008) 3.8% of adults in Low Income countries, 6.6% in Lower Middle Income countries, 24% in Upper Middle Income countries, and 21.6% in High Income countries; in all cases prevalence was higher among women (WHO 2011:113). These figures for adults refer to obesity proper (BMI≥30), but prevalence is at least twice as great for the more inclusive concept of overweight (BMI≥25). Recent data (all after 2000) on prevalence of adult overweight in some selected countries were: 66.9% in the United States, 66.5% in Germany, 62.7% in New Zealand, 61% in the UK, 59% in Canada, 49% in France and Australia, 44% in Italy, 40.6% in Brazil, 59.7% in Chile, 55% in Peru, and 42% in Iran (WHO 2011:104-13; see also WHO Global Database on Body Mass Index, http://apps.who.int/bmi/index.jsp). Intriguingly, however, Asian trends are different. Overweight prevalence in the two most populous countries is much lower: China 18.9% and India 4.5%. Among developed countries, Japan also has a relatively low score, with only 23% overweight. Since this discrepancy between East and West extends to other less populous Asian countries, both

186

Hector Maletta

developing and developed, it is arguable that some specific Asian factors (cultural, genetic or both) may tend to produce less overweight and obesity on that continent, even in countries experiencing a rapid process of economic growth. The global study by Ng et al. (2014) found that between 1980 and 2013, and among adults, the prevalence of overweight (Body Mass Index ≥25) increased from 28.8% to 36.9% in men and from 29.8% to 38.0% in women. Obesity (BMI≥30) increased from about 11% to about 14% in women, and from about 7% to about 9% in men. The study found, however, that the increase of adult obesity in developed countries has slowed since 2006. To sum up, evidence shows a general decline in the prevalence of nutritional deficiency in children as indicated by anthropometric measures (stunting, wasting, and underweight) and a general increase in overweight and obesity among children and adults. The latter conditions attain very high levels in developed countries or, more generally, at medium and high levels of per capita income. The prevalence of stunting, underweight, and wasting in Africa, although declining, is still high and decreasing more slowly. The slower reduction in Africa is not due to insufficient growth of food production (which in growing faster in Africa than the world average) but to poor health and sanitation, widespread poverty, and (in some countries) persisting violence endangering lives and livelihoods.

187

8. Conclusions about food and hunger trends

In the half century since 1961, food output has consistently grown faster than population, thus increasing world food supply per person. Indeed, this growth occurred in all world regions, but was faster in the developing world. Food output increased more quickly in Asia, followed by Latin America and Africa. The more developed regions (North America and Europe) grew at a much slower rate. Most of the growth in agricultural production comes from increased land productivity. The total land area used for agriculture increased (albeit very slowly) up to 1990 but has been stagnant or decreasing in the latest two decades. Land used for grazing livestock (including natural and cultivated meadows) actually decreased, while total cropland (arable or with permanent crops) was stagnant. Over 95% of the farm output growth accumulated since 1961 (and nearly all such growth after 1990) came from increased output per hectare, both for total agricultural production (crops and livestock) and for the case of crops. Trade in agricultural and food products has increased much faster than production - a manifestation of a more inter-dependent world food system. Whereas food output trebled, food trade grew by a factor of 8.5 in real terms. All regions have significantly increased agricultural exports and imports. North America and Latin America are net exporters of food and, more generally, agricultural products. Asia, Africa, and Europe, in contrast, are net importers, although net importation into Europe has been steadily diminishing. In practice, only Africa and Asia are strong net importers of food and agricultural products. However, these regions devote a diminishing proportion of their total export revenue to the importation of food (about 4% in Asia and about 10% in Africa, as an average for 2007-2011). Volatility in international commodity prices was reflected in the nominal value of both 189

Conclusions about food and hunger trends

exports and imports of agricultural products, but neither Asia nor Africa altered the growing trend in their real food imports (or total and per capita food consumption) on account of price volatility. Per capita dietary energy supply has steadily increased and undernourishment has consistently decreased over time in all regions and on the world scale. Consumption of all major food items has increased, especially non-staple foods such as fruit, vegetables, and many foods of animal origin, while per capita food consumption of cereals (the main staple foods) is declining. Concurrently, key indicators of malnutrition (stunting, wasting, and underweight in children under five) are also improving consistently in all parts of the world, though figures are still high in some regions (chiefly Sub-Saharan Africa). The opposite problems (overweight and obesity) are rapidly increasing, even in children under five and also in other age groups.

190

Part III. FUTURE PROSPECTS

9. Projections on food and agriculture

9.1. Visions of the future Having reviewed historical trends in food supply, food consumption, access to food and malnutrition, we now turn our attention to the future of food and hunger. We are interested in the likely behaviour of the main variables defining the food situation: population, food production and trade, growth of income, and the resulting implications for the prevalence of hunger during the 21st century. Our primary time horizon is 2050, extended to 2080 or 2100 in some specific cases when such projections are available. Throughout, we will be using existing projections of population and food production and available studies and models simulating the future as an integrated interplay of various intervening factors, concentrating on large-scale studies undertaken by large research groups under the aegis of international organisations. We will also discuss the assumptions and possible biases or shortcomings of these projections and simulations to ascertain whether they may be too optimistic or too pessimistic, i.e., overstating or understating the likely future reduction of world hunger. Foreseeing the agricultural and food situation in the future involves projecting the food supply and demand, and the future prevalence of undernourishment and other indicators of food insecurity. Food supply depends on land use and land productivity. The food demand depends chiefly on population and income. Two other important factors should be taken into account: the expected impact of climate change on farm production and the growing use of food crops for producing biofuels.

193

Projections on food and agriculture

Thus, we will review projections of future agricultural production and the expected impact of climate change on agriculture, as well as the possible impact of biofuels on food availability, assessing implications for food security focused on dietary energy supplies, undernourishment, and malnutrition. Existing projections of future developments in agriculture and food are based on a series of studies carried out by various research groups and sponsored by international organisations such as FAO, the UN Population Division, the International Food Policy Institute, the Intergovernmental Panel on Climate Change, and others. Indicators of prospects for the future include projections of agricultural production encompassing the expected impact of climate change, and projections of demand for agricultural products for various purposes, such as food, biofuels, fodder and other uses. These projections are not ungrounded prophecies nor simple extrapolations from the past (see ‘Prophecies, extrapolations, predictions and projections’ in Section 13.1). Agriculture and food are part of a complex adaptive system, involving many factors intertwined in myriad ways. These factors include soils, climate, plant and animal physiology, the technicalities of growing crops and conducting animal husbandry, the economic aspects of agriculture, and other socioeconomic factors such as the level and distribution of income, and population growth. These aspects are interrelated, and cannot be projected independently of each other; a change in any factor triggers changes in other aspects, and these cause further changes in the original factor, leading it to change again. Because of this systemic connectedness, projecting the future development of a complex system cannot be done piecemeal, adjusting one factor at a time under a ceteris paribus clause; it cannot be assumed that a factor will change in a certain way while all other factors remain constant or change in their own independent way. This cannot be done for two different reasons: first, because the system is a complex interconnected one; second, because in the long term, all variables change, as do their inter-relationships and cannot possibly be ‘kept constant’. Projections and predictions. The concept of ‘projection’ has a specific meaning in this context. Predictions are unconditional statements about things to come, based on theory (like the prediction that a solar eclipse will take place on a certain date). Projections, on the other hand, are conditional statements linked to a specific scenario regarding how the future may unfold. Such scenarios contain assumptions about contingent events that may well not occur at all, or may occur in a different way. Projections simply make explicit the implications of a given scenario; however, the question of which particular scenario will eventually materialise is not part of the projection. See Section 13.1 for more details. 194

Hector Maletta

As projections of agriculture and food depend on a very complex network of factors involving natural resources, climate, social and economic institutions, technological developments and more, projections do not only use assumptions, but are also subject to great uncertainties about the precise way in which scenarios will unfold. Assumptions, as well as theories and methods underlying the links between assumptions and outcomes, have a most important bearing on the conclusions. Some seemingly innocuous assumption about demographic growth or the choice of some particular crop varieties to test crop response to climate change may have huge implications when incorporated into a projection. Since the results of projections are ordinarily more visible than the subtle and often crucial assumptions on which these results depend, the reader is urged to consider the methodological underpinnings of available projections, and to take the latter with a grain of salt. In this book, we do not produce any original projections of our own; we use projections that are already available, which have been prepared with great care by large teams of people drawing on extensive computational and information resources, under the auspices of important international organisations.

9.2. FAO projections of world agriculture FAO has been publishing projections of agricultural growth and food demand, using an iterative multivariate model, periodically updated and extended from the original projection to the year 2000 (FAO 1981; Alexandratos 1988) through more recent projections that extended the outlook to 2010, 2015, and 2030-2050 (FAO 1995, 2003a, 2006, 2011; Alexandratos and Bruinsma 2012 - hereinafter referred to as AB 2012).21 These FAO studies contain the most complete set of agricultural projections for the coming decades. FAO projections combine detailed information at the country and product levels regarding natural resources, past agricultural production, past and projected technological change, agricultural markets, and expected economic development in terms of income, trade, and other aspects. They are based on highly disaggregated data about each product, country, and agro-ecological zone accumulated by FAO in the course of its work in all corners of the globe. FAO’s efforts to foresee the future of agriculture and food are based on prospects of population and income, and on detailed agro-ecological mapping, Alexandratos (2011) provides an interesting review of problems involved in updating the 2006 FAO projections, but does not provide a new projection of agricultural output, which is presented instead in AB 2012.

21

195

Projections on food and agriculture

making extensive use of the FAOSTAT database and the accumulated store of in-house FAO information on every country in the world and all crop and livestock types. The general approach of FAO projections is to start with projections of demand, determined essentially by population, income, and estimated income elasticities and Engel coefficients for the various agricultural products. Demand projections are prepared for each individual product in each individual country. Agricultural production is estimated for each country and each particular crop or livestock product, based on a wide array of expertise and information on each country and product, including data on natural resources (land, soils, water, and climate), macroeconomic conditions, agricultural technology, and international trade. Projected production does not represent the maximum production attainable, nor does it represent any desirable outcome. It is the amount required to match projected demand, within constraints imposed by land, climate, inputs, and other relevant factors. Domestic demand for each product is projected to be covered by local production and possibly imports; projected domestic production may require imports to meet domestic demand or may be in excess of domestic demand, thus generating exportable surpluses. FAO’s approach to projecting the future of food and agriculture is not a formal mathematical model that generates outcomes automatically; rather, it is a heuristic endeavour based on a huge store of in-house information and certain external projections of population and income. The future course of demand and supply for each particular item is projected first at the country level, using country-specific and product-specific expertise; then all the partial results for specific countries and for specific products are reconciled at the world level, in an iterative way, in order to ensure that world supply and world demand match each other for all products. In this heuristic approach, FAO’s projections differ from other exercises that simply estimate the coefficients of a set of equations, which apply to all countries, and automatically determine the outcome. FAO projections reports do not explicitly take into account the possible effects of climate change, although it is possible that such effects have been incorporated in a rather general manner. Also (except in recent versions), FAO projections do not consider the prospective impact of using crops for biofuel production. FAO has separately worked with IIASA, the International Institute for Applied Systems Analysis (Laxenburg, Austria) to produce specific modifications of its projections incorporating the impact of climate change and biofuels (see, for instance, Fischer 2011). The climate-related aspects of FAO projections will be analysed in Ch. 10, along with other projections of 196

Hector Maletta

climate change impact on agriculture. In this chapter, we will focus on FAO’s projection up to 2050, primarily using the 2006 interim report and its more recent update (AB 2012).

9.2.1. FAO projections to 2050: population and income assumptions FAO projections to 2050 use the 2008 Revision of UN population estimates and projections and GDP projections prepared by the World Bank, both reproduced here in tables 44 - 46, employing the regional breakdown used in AB (2012). Table 44. Population projections used for FAO agricultural projections, 1970-2050 (million people).  

1970

2000

2006

2015

2030

2050

World

3688

6115

6592

7302

8309

9150

Developing countries

2597

4778

5218

5879

6839

7671

Sub-Saharan Africa

270

625

730

912

1245

1686

Near East/North Africa

181

387

432

504

615

726

Latin America/Caribbean

282

515

556

611

682

721

South Asia East Asia Developed countries

708

1375

1520

1729

2016

2242

1147

1857

1957

2096

2247

2255

1079

1318

1351

1396

1437

1439

Source: AB (2012:30), based on UN projections (2008 Revision).

Table 45. Annual population growth rates underlying FAO agricultural projections, 1970-2050.  

19702000

20002006

20062015

20152030

20062030

20002030

20302050

20062050

20002050

World

1.70% 1.26% 1.14% 0.86% 0.97% 1.03% 0.48% 0.75% 0.81%

Developing countries

2.05% 1.48% 1.33% 1.01% 1.13% 1.20% 0.58% 0.88% 0.95%

Sub-Saharan Africa

2.84% 2.62% 2.50% 2.10% 2.25% 2.32% 1.53% 1.92% 2.00%

Near East/North Africa

2.57% 1.85% 1.73% 1.34% 1.48% 1.55% 0.83% 1.19% 1.27%

Latin America/Caribbean

2.03% 1.28% 1.05% 0.74% 0.85% 0.94% 0.28% 0.59% 0.68%

South Asia

2.24% 1.68% 1.44% 1.03% 1.18% 1.28% 0.53% 0.89% 0.98%

East Asia Developed countries

1.62% 0.88% 0.77% 0.46% 0.58% 0.64% 0.02% 0.32% 0.39% 0.67% 0.41% 0.36% 0.19% 0.26% 0.29% 0.01% 0.14% 0.18%

Source: AB (2012:30), rates for some periods have been added, based on previous Table 44.

197

Projections on food and agriculture

Table 46. Assumed rates of economic growth in FAO projections of agriculture, 2006-2050. Region

Total GDP

Per capita GDP

2006-2030

2006-2050

2006-2030

2006-2050

World

2.47%

2.11%

1.49%

1.36%

Developing countries

4.47%

3.58%

3.30%

2.67%

Sub-Saharan Africa

4.64%

4.17%

2.34%

2.20%

Near East/North Africa

3.54%

2.92%

2.03%

1.72%

Latin America and Caribbean

2.45%

2.09%

1.58%

1.49%

South Asia

4.90%

4.05%

3.67%

3.14%

East Asia

5.51%

4.18%

4.90%

3.85%

Other developing countries

4.43%

3.49%

3.65%

2.98%

1.56%

1.34%

1.30%

1.20%

Developed countries

Source: AB (2012:36), projections based on figures provided to FAO by the World Bank’s De­velopment Prospects Group. The 2006 baseline GDP used by FAO is actually the 2005-2007 average.

9.2.2. FAO projections to 2050: The 2006 Interim Report The most recent versions of FAO’s projections of world agriculture, as mentioned before, extend to 2050. Preliminary projections with this time horizon were published in FAO 2006. A more recent update is AB (2012). We will primarily use the latter but, first, we will briefly review the former. The 2006 interim report entitled ’World Agriculture: Towards 2030/2050’ was based on FAOSTAT information up to the turn of the century and took 2000 (represented by the 1999-2001 average) as its baseline period. Table 47 shows the baseline historical rates of growth and FAO (2006) projected rates for various parts of the world.

198

Hector Maletta

Table 47. Growth rates of agricultural production, 1961-2001, projected to 2030 and 2050 (FAO 2006). Historical annual growth rates to 2001

FAO projection

1961-2001 1971-2001 1981-2001 1991-2001 2000-2030 2030-2050 2000-2050 World

2.3%

2.2%

2.1%

2.3%

1.5%

0.9%

1.3%

Developing countries

3.4%

3.6%

3.7%

3.8%

1.9%

1.1%

1.6%

Sub-Saharan Africa

2.3%

2.5%

3.1%

3.0%

2.7%

1.9%

2.4%

NENA

3.0%

3.0%

2.8%

2.4%

2.2%

1.4%

1.9%

LAC

2.9%

2.9%

2.8%

3.2%

2.2%

1.2%

1.8%

South Asia

2.9%

3.2%

3.2%

2.8%

2.1%

1.3%

1.8%

East Asiab

4.2%

4.5%

4.6%

4.8%

1.7%

0.6%

1.3%

- China

4.6%

4.5%

5.1%

5.1%

2.0%

0.6%

1.4%

- Rest of East Asia

3.3%

3.2%

2.8%

2.4%

1.9%

1.1%

1.6%

Developed countriesc

1.5%

1.2%

1.1%

1.4%

0.6%

0.3%

0.5%

Transition countries

0.3%

-0.8%

-2.1%

-3.1%

0.5%

0.2%

0.4%

a

d

a. Near East and North Africa. b. Developing countries in East and South East Asia, as well as in Oceania. c. Excludes all transition countries. d. Former centrally planned economies in Europe and Central Asia. Source: FAO (2006), Table 3.1. Rates for 2000-2050 are based on rates for 2000-2030 and 2030-2050. The baseline (indicated as 2000 in the table) is actually the 19992001 mean. FAO (2006) equates output with the value of agricultural production in PPP dollars at world average producer prices of 1989-1991 (FAO 2006:31). Past growth rates for China (not reported in the source table) are from current FAOSTAT figures computed in the same way, albeit at 2004-2006 prices. Rates for China after 2000, not provided in FAO (2006), are based on FAOSTAT figures for net agricultural production in 1999-2001 and on rates projected in FAO (2006) for East Asia (total) and for East Asia excluding China.

According to these FAO projections, the growth of agricultural output in the first half of the 21st century is very likely to decelerate relative to the second half of the previous century, with a world growth rate of just 1.3% per year from 2000 to 2050 and only 0.9% per year in 2030-2050. This amount is enough to cover the growth in projected demand originating in the assumptions regarding growth in population and per capita income. The envisaged reduction in growth rates is not the result of an inability to produce more, but of a slowdown in demand caused chiefly by slower population growth and the low elasticity of food demand relative to income. FAO output projections are the product of an integrated model starting with projections of demand and imposing global ex-post balance of supply and demand. Demand projections, in turn, depend on expected growth of population and income. Population growth is projected to decelerate according to UN projections and this is the decisive factor in decelerating demand growth. Income projections were rather conservative and demand for agricultural products tends to be 199

Projections on food and agriculture

income-inelastic, determining a less-than-proportional impact of income growth on the growth of demand for agricultural products. Projections of demand under an expected deceleration of population growth and rising per capita income imply not only an increase in overall food consumption but also changes in its composition (stagnant demand for staple foods and increasing demand for other food items). The latter effect appears to be more important than the former. As dietary energy supply reaches high levels (typically above 3100 kcpd) it tends to stabilise, but at higher income levels, diets tend to include more protein, vitamins, and minerals (mainly from animal products, fruits, and vegetables), and also more fats and oils, reducing the relative share of cereals in dietary energy and their share in the total value of agricultural production. This has already occurred in the past in high- and medium-income countries and regions, and is expected to continue as emerging countries reach higher levels of income. Growth in staple food consumption will occur primarily at the bottom of the income scale, as the poorest countries increase their meagre living standards, but this will not reverse the overall decline in the importance of staple food. These projections thus foresee a slowdown in global agricultural production growth from 2000 to 2030 and 2050, relative to previous periods. This is only natural; in the later decades of the 20th century (1961-2000), total demand grew by more than 200% as did production, whereas demand is expected (in FAO 2006) to increase by just 86% in the first half of the 21st century. The annual rate of output growth worldwide was above 2% in the decades prior to 2000, but is expected to be about 1.5% in 2000-2030, and 0.9% in 2030-2050, implying that production in 2050 will be 86% greater than in 2000, corresponding to expected growth in demand. The deceleration is projected to affect all regions, but is strongly influenced by FAO’s projection for East Asia and most especially for China. In the late 20th century, China’s farm production was growing at rates above 5% per year but it is (implicitly) projected by FAO (2006) to grow at about 2% in 2000-2030, and at about 0.6% in 2030-2050. China’s growth rate for (net) agricultural production has indeed slowed down (to 4.02% per year) in the period from 1999-2001 to 2011 but, as yet, it has not shown as marked a deceleration as is suggested by the projection; given its growth to 2011, it should grow at a mere 1.1% per year in 2011-2030 to reach the level projected for 2030 in FAO (2006), which is possible, but not very likely. On balance, it seems the projection of farm output in FAO (2006) is rather conservative, though not altogether unrealistic.

200

Hector Maletta

9.2.3. FAO projections to 2050: The 2012 Update In 2012, FAO published an updated version of its projections to 2050 in a report by Alexandratos and Bruinsma (AB 2012). These updated projections to 2050 use 2006, represented by the average of 2005-2007, as the baseline period, unlike FAO (2006) which used 1999-2001. The overall growth rate of agricultural production from 2005-2007 to 2050, projected in AB (2012), is about 1.1% per year, implying a total increase of about 60% in 44 years. This growth rate is considerably below historical rates, chiefly as a consequence of the projected slowdown in global demand due to decelerating population growth and also due to relatively modest hypotheses regarding future economic growth. It is also below the growth envisaged in the interim report (FAO 2006).The lower growth of demand envisaged in AB (2012) as compared to FAO (2006) is not due to lower projected levels of population or income; in fact AB (2012) projections are based on the same UN 2008 Revision of population projections, and the same GDP growth hypotheses from the World Bank’s Development Prospects Group as used in FAO (2006). The difference emerges from updated production data and also from multiple adjustments in the implicit country-specific and product-specific assumptions about future technical change, land use, and income elasticities of de­mand, i.e., the factors that drive future demand. Projected demand and updated baseline data, in turn, drive projected production. In line with previous versions of FAO projections, the authors state: [T]he projected increases are those required to match the projected demand as we think it may develop, not what is required to feed the projected world population or to meet some other normative target. Our projection is not a normative one (AB 2012:7).

Projections of agricultural production. Projected rates of growth in AB (2012), compared with historical data, are provided in Table 48. Following the pattern of FAO (2006), it is assumed that future rates will decelerate very quickly as compared with the immediately preceding period. They differ little from those in FAO (2006), but tend generally to be slightly lower; for instance, the growth rate foreseen at the world level for 2030-2050 is lowered from 0.9% (Table 47) to 0.8% (Table 48).

201

Projections on food and agriculture

Table 48. Historical and projected growth rates of agricultural production (AB 2012).   World Developing countries Idem, excl. China

1970-2007

1980-2007

1990-2007

2005/20072030

2030-2050

2005/20072050

2.1%

2.1%

2.2%

1.3%

0.8%

1.1%

3.5%

3.5%

3.4%

1.6%

0.9%

1.3%

2.9%

2.9%

2.9%

1.8%

1.2%

1.5%

Sub-Saharan Africa

2.7%

3.2%

3.1%

2.5%

2.1%

2.3%

NENA

3.0%

2.8%

2.6%

1.6%

1.2%

1.4%

LAC

2.9%

2.9%

3.5%

1.7%

0.8%

1.3%

South Asia

3.0%

2.9%

2.5%

1.9%

1.3%

1.6%

East Asia Idem, excl. China Developed countries

4.2%

4.2%

4.1%

1.3%

0.5%

0.9%

3.1%

2.7%

2.7%

1.5%

0.9%

1.3%

0.6%

0.2%

0.3%

0.7%

0.3%

0.5%

Source: AB (2012:63). Rates are based on the net value of crop and livestock production at world-average 2004-2006 producer prices, converted into USD at PPP conversion rates. East Asia includes developing countries in East Asia, Southeast Asia, and Oceania. NENA = Near East & North Africa.

These growth rates lead to the growth in total and per capita agricultural output shown in Figure 61 (total output), and Figure 62 (per capita). In these figures, the historical annual series for 1961-2013 (latest FAOSTAT data available at the time of writing) are added for illustrative purposes. Projected absolute values for 2030 and 2050 are based on FAOSTAT data for 20052007 and the AB (2012) projected growth rates from 2005-2007 to 2030 and from 2030 to 2050. Figure 61. FAO: Historical (FAOSTAT) and projected (AB 2012) world agricultural output (billions PPP USD at 2004-2006 prices).

Billion $ppp at 2004-06 prices

$3500 $3000 $2500 $2000 $1500 $1000 $500 $0 1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 2060

202

Hector Maletta

Figure 62. FAO: Historical (FAOSTAT) and projected (AB 2012) world per capita agricultural output, (PPP USD per person, at 2004-2006 prices). $400

PPP USD per capita

$350 $300 $250 $200 $150 $100 $50 $0 1940

1960

1980

2000

2020

2040

2060

FAO projections from 2005-2007 to 2030 and 2050 imply a deceleration of growth rates of total output relative to rates observed in the recent past. There are reasonable grounds for FAO’s projection of decelerated growth (see AB 2012) as there were in FAO (2006). Those grounds do not refer to problems meeting future demand, but to slower growth of demand due to slower demographic growth and expected deceleration of economic growth in emerging countries, most notably in China. The projections are, therefore, conservative. Even so, per capita production (in PPP USD) is expected to grow from $295 in 2005-2007 to $340 in 2050. Over that period agricultural production would rise by 60% and per capita output by 15%. Note that actual growth after 2006 has been faster than the average growth projected for the 2006-2030 period, for both total and per capita output; as a result, the actual per capita output of 2013 was already above the projected level of 2030. The underlying growth projections for the output of specific major foods are also conservative. Cereal production is estimated to have been 2068 million metric tons (MMT) in 2005-2007, and is projected to expand to 2,720 MMT by 2030, and 3,009 MMT by 2050. The implied growth rates are 1.2% per year from 2005-2007 to 2030, and 0.9% per year from 2030 to 2050. Growth rates for China and East Asia have been further reduced in this update relative to FAO (2006). For the comparable period 2030-2050, FAO (2006) foresaw growth rates of 1.7% for all East Asia, and 1.9% for East Asia excluding China; in AB (2012) those rates fell to 0.5% and 0.9% per year respectively. This makes AB (2012) projections even more conservative than those in FAO (2006). 203

Projections on food and agriculture

Projections of agricultural trade. FAO projections of agricultural production and demand include estimates of agricultural trade, estimated mostly in net terms, as a difference between projected production and projected demand for each country. The general conclusion of this exercise is that developing countries will continue their tendency (observed since the 1960s) towards decreasing net agricultural exports and developing a (relatively small) net deficit in agricultural trade. The agricultural trade balance of developing countries, as estimated in AB (2012), was positive in the 1960s, with a surplus of about 10 billion USD. This surplus decreased during the 1970s and practically reached zero in the 1980s. It then entered negative territory, reaching a deficit of about 10 billion USD from the mid-1990s to the mid-2000s. This is in accordance with our estimate of the agricultural trade balance (Section 5.4). FAO projections for the future in AB (2012) imply a continuation of this trend, reaching a yearly deficit of about 45 billion USD by 2050. This is not that much in relative terms: it would represent just a small fraction (around 1%) of the entire agricultural output of developing countries (now producing over 50% of the total and tending to increase their share in the future) and a much smaller fraction of developing countries’ future total GDP. It would also be a small fraction of their total trade volume. There is thus no implication that increasing net food imports would create a significant problem for the economy of developing countries as a whole (though they may cause problems for particular countries). It should be noted that these imports are projected in terms of effective demand and not based on normative considerations; in other words, the model foresees that under the adopted projections for population and income, developing countries will have the purchasing power and foreign revenue required to afford such imports. It follows that developing countries would make up for the deficit by means of non-farm export revenue and flows of foreign income and capital (otherwise the model would not foresee the existence of effective demand for such an amount of farm imports). Since developing and developed countries together comprise the entire world, the increasing deficit projected for the former would be reflected in an equal surplus for the latter. In other words, from the start of the 21st century, the rich countries would be net providers of agricultural products to the poorer parts of the world. However, as we have seen in Section 5.4, these imbalances constitute a very small proportion of total trade and include the effect of counting exports at FOB prices and imports at CIF prices. 204

Hector Maletta

If a country has a negative (or positive) trade balance in one particular product or in some particular category of goods (e.g., agricultural products) this is of no special significance since countries also export and import other commodities (oil, gas, metals, diamonds, industrial products, services). In addition, there are also regular flows of financial resources going to and from each country, including flows of income (e.g., remittances from immigrants to relatives in their home country) and flows of capital (e.g., direct foreign investment). Effective demand for imports, or more precisely, demand for the foreign currency required for imports, is funded by all such flows. As noted above, the FAO projection model is not a normative one. It is based on projections of effective demand, not on needs. If a country is projected to import a certain amount of food, it is not because it is judged to need such additional food but because its projected population is deemed to want those goods and to have sufficient income to afford their purchase. Projected farm trade is derived from a comparison between effective domestic demand (i.e., domestic purchasing power expected to be allocated to the purchase of farm products) and projected domestic supply of farm products, for a certain projected level of per capita income. At the same time, the existence of a flow of imports indicates that an equivalent flow of exports is created somewhere else. World exports and world imports are essentially the same (except for small statistical discrepancies). It is important to cover all agricultural trade, and not only food products, because many developing nations export non-staple farm products and import staple (and other) food. As shown previously (Table 27), cereal trade is but a small fraction of agricultural trade (about 7% as of 2010) and a much lower proportion of total trade. Major farm commodities usually studied in this regard (cereals, meats, and vegetable oils) make up merely 25% of total agricultural exports. Non-food farm exports (including many tropical products coming from developing countries) make up a large and increasing share of total trade (about 29% in 2010). The real question about future agricultural trade is not about major commodities such as cereals, sugar, or oil crops, or whether developing countries will or will not be self-sufficient in staple food, but about the development of trade in agricultural and other products, either farm and non-farm, either food or non-food, and about growth in income and the increased capacity of developing countries to export all sorts of goods and services and to attract income and capital to sustain their growth and provide their people access to food.

205

Projections on food and agriculture

In fact, FAO projections imply that developing countries will be able to afford importing an increasing amount of farm products. FAO projections are obtained from an integrated assessment model based on projections of effective demand. Therefore, if the model foresees that developing countries will have a (small) negative agricultural trade balance, this implies that the model foresees that these countries will have the financial capacity to pay for it, either with non-farm exports or through other financial flows. If, as projected by FAO, developing countries as a whole will indeed have a net deficit in agricultural trade, however small, this implies that they will necessarily have the income required to afford it, either from trade in non-farm products (oil, gas, and minerals, as well as manufactured goods and traded services, much of which are being relocated to the periphery of the world economy) or from other flows of income and capital accruing to developing countries. This result is necessarily implied by FAO’s results on farm trade, albeit not explicitly discussed in the reports. Projections of land use. The envisaged growth of agricultural production in AB (2012) implies only a small expansion in the area of arable land. In this projection, growth in arable land (i.e., land used for annual crops) would explain only 10% of total projected crop growth in the world; increased cropping intensity (less land in fallow and/or more land in multiple cropping) would explain another 10%, while increased output value per harvested hectare would explain the remaining 80%. This is in line with the percent contributions of these growth factors observed in the past, not only in our previous study of trends, but also in AB (2012): they estimate that (from 1961 to 2007) arable land contributed 14% of observed crop output growth, cropping intensity 9%, and output per harvested hectare the remaining 77% (AB 2012:98).22 Since the availability of land suitable for rain-fed crop production greatly exceeds its actual use for crops, as was already seen in Section 4.4, the envisaged increases in the use of arable land are quite modest. Total arable land in use in 2005-2007, which was estimated at 1592 MHa, would just marginally expand to 1647 MHa in 2030 and to 1661 MHa in 2050 (AB 2012:109). The rate of expansion in arable land, at both the world and regional levels, would be slower than the already small expansion observed during the past half century, although in line with data for the 1990s and 2000s, which indicated stagnant arable land area. For cereals, their projection implies that in developing countries, the harvested area of wheat would increase just 1% from 2005-2007 to 2050, while the harvested rice area is projected to decrease by 9%, and the area of Land with permanent crops, as well as grassland, was not considered in this estimate, which was limited to arable land, i.e. land devoted to temporary crops.

22

206

Hector Maletta

maize is projected to expand by a stunning 41% (AB 2012:99), all within the modest aforementioned total increase in arable land use. Projections of dietary energy supply. Projections of demand, output, trade, and population allow AB (2012) to make projections of future supplies of dietary energy and future levels of undernourishment. World dietary energy supply is projected to increase from 2772 kilocalories/person/day (kcpd) in 2005-2007 to 3070 in 2050, i.e., an increase of about 11%. By that time, even the two regions with the lowest supplies would have more than enough calories to meet their needs even allowing for unequal access: Sub-Saharan Africa would have 2740 kcpd and South Asia 2820 kcpd. Both figures are about 23% higher than corresponding estimates for 2005-2007, and are also roughly equivalent to the 2005-2007 world average (2772 kcpd), as shown in Table 49. All other regions would have over 3000 kcpd. Table 49. Historical and projected supplies of dietary energy: Kilocalories/person/day, 1969-1971 to 2050.  

1969-1971 1979-1981 1989-1991 1990-1992 2005-2007

2015

2030

2050

World

2373

2497

2634

2627

2772

2860

2960

3070

Developing countries

2055

2236

2429

2433

2619

2740

2860

3000

Sub-Saharan Africa

2031

2021

2051

2068

2238

2360

2530

2740

NENA

2355

2804

3003

2983

3007

3070

3130

3200

LAC

2442

2674

2664

2672

2898

2990

3090

3200

South Asia

2072

2024

2254

2250

2293

2420

2590

2820

East Asia

1907

2216

2487

2497

2850

3000

3130

3220

3138

3223

3288

3257

3360

3390

3430

3490

Developed countries Source: AB (2012:23).

Projections of undernourishment. According to projections included in AB (2012), world undernourishment in 2050 would be reduced to non-significant values (below 5%) in almost all regions including South Asia and would be just slightly above that threshold in Sub-Saharan Africa (7.1%). This SSA projection would be below the levels estimated in 2005-2007 for Latin America or the NENA region and close to the 5% level below which undernourishment is not statistically significant (see Table 50). Since the AB (2012) report was prepared in 2011, its authors were unable to take into account the more recent revision of FAO undernourishment statistics, reflected in FAO-SOFI (2012) and subsequent SOFI issues. Since this revision 207

Projections on food and agriculture

tended to show a better situation than previous estimates for the years after 1990, it is likely that future updates of FAO projections would suggest further improvements from now to 2050, relative to those in AB (2012). Table 50. Historical (unadjusted) and projected prevalence of undernourishment in developing countries, 1990-1992 to 2050 (percentage of population).  

1990-1992

2005-2007

2015

2030

2050

Developing countries

19.7%

15.9%

Sub-Saharan Africa

33.6%

27.6%

11.7%

7.9%

4.1%

21.4%

14.5%

7.1%

6.0%

7.4%

6.0%

4.7%

3.4%

LAC

12.2%

8.5%

6.3%

4.1%

2.5%

South Asia

21.5%

21.8%

16.1%

10.5%

4.2%

East Asia

19.2%

11.0%

6.8%

4.2%

2.8%

NENA

Source: AB (2012:26). East Asia includes also Southeast Asia and Oceania. Figures for 1990-1992 and 2005-2007 are from FAO-SOFI (2008), before the more recent methodological revision of undernourishment introduced in FAO-SOFI (2012), which reduced the estimates for the period since 1990. Projections of future undernourishment are influenced by those baseline figures.

In summary, FAO forecasts an increase of about 60% in demand for agricultural products from 2005-2007 to 2050, and considers that such an increase could be met by agricultural production without using much additional arable land (which is nonetheless amply available) and with land productivity increasing less than in the recent historical record. Under these projections, per capita dietary energy supplies would significantly increase between the base period and 2050 in all regions, and the prevalence of undernourishment would be non-significant at the world level and would reach non-significant levels (or be close to them) in all regions of the world. FAO projections do not make explicit the possible impact of climate change, which we will examine separately below. Before doing so, it is worth asking whether FAO projections are indeed correct or whether they exhibit some positive or negative bias, either towards excessive optimism or excessive pessimism.

9.2.4. Are FAO projections likely to be right? Both the latest round of FAO projections of food output (FAO 2006) and its update (AB 2012) use approximately the same methodology employed for previous rounds (FAO 1981, 1995, 2003a). After so many years, some preliminary assessment can be made of these previous FAO projections to ascertain whether they show any systematic tendency to overstate or understate

208

Hector Maletta

the actual development of agricultural production. The general conclusion of this assessment is that they are generally very conservative, and routinely fall short of actual growth. FAO (1981) is not so easy to evaluate, because it follows a rather different (and more normative) projection methodology as compared to later calculations. Projections since FAO (1995) contain estimates of the future value of agricultural production in PPP USD (albeit with different base years, but this does not greatly affect the growth rates). FAO (1995) growth projection, from the 1988-1990 average (centred on 1989) to 2010, foresaw output growing at a yearly 1.8% and population at 1.56%, whereas actual data show that in that period, farm output actually grew at 2.4% per year and population at 1.34%. Thus, production grew faster than projected and population grew more slowly. As a result of such differences in output and population growth, per capita agricultural output (which had been foreseen to grow at a yearly rate of 0.23%) in fact grew 4.7 times faster, at 1.08% per year. Per capita agricultural output in 2010 was actually 12.7% higher than projected in FAO (1995). Similarly, FAO (2003a) presented a projection of the value of agricultural production from the observed 1997-1999 average (centred on 1998) to 2015. Population was presumed to grow at 1.2%, and production at 1.6% per year. In fact, population did indeed grow at 1.2% from 1997-1999 to 2013, and the expected rate from 1997-1999 to 2015 (as per UN latest population projections) is also 1.2%, as projected in FAO (2003a). However, agricultural production grew faster than projected: from 1997-1999 to 2013 it increased at an annual rate of 2.5% per year, well above the yearly rate of 1.6% projected by FAO from the same baseline period to 2015. Consequently, the 2013 per capita output was 15.7% higher than the (interpolated) 2013 value resulting from the FAO (2003a) projection to 2015. For the period from 1999-2001 to 2030, FAO (2006:4, 16) projected an output growth rate of 1.5% per year, with population growth at 1.0%; equivalent to about 0.5% yearly growth per capita, well below the rates of 2006-2013. AB (2012:29, 95) project agriculture to grow at 1.3% and population at 0.94% from 2005-2007 to 2030, implying a per capita growth rate of 0.33%, also below the growth observed since 2007. The 2013 per capita output in PPP USD ($324) was already well above the level of $318 foreseen for 2030 in AB (2012), as shown in Figure 63.

209

Projections on food and agriculture

Figure 63. Real per capita world net agricultural output since 1989, in PPP USD at world average producer prices of 2004-2006: actual value (1989-2013, based on FAOSTAT data retrieved April 2015) compared to exponential growth projected in FAO 1995 (1988-1990 to 2010), FAO 2003a (1997-1999 to 2015), FAO 2006 (1999-2001 to 2030), and AB 2012 (2005-2007 to 2030). $325 $305 $285 $265 $245 $225

1985

1990

1995

2000

Actual FAO (2003) projection to 2015 AB (2012) projection to 2030

2005

2010

2015

2020

FAO (1995) projection to 2010 FAO (2006) projection to 2030

Thus, FAO projections have systematically understated future growth in per capita agricultural production. Since they systematically understate future growth, FAO projections are thus regarded as usually being too pessimistic. FAO output projections, as discussed before, are not meant to represent potential or maximum production, nor are they intended to reflect normative concerns about the food needed to adequately feed the world. They are rather an estimate of production derived from a projection of effective demand, dictated in turn by expected growth of income and population, under constraints imposed by natural resources and by expected improvements in productivity. Projected demand, then, is not a normative target of adequate demand, but simply the expected level of achievable production in relation to effective demand for farm products, derived from expected growth in population and income, and background estimates of demand elasticities and Engel coefficients, combined with restrictions due to available natural resources and expected changes in yields, irrigation, cropping intensity, and other technological factors. They do not attempt to assess what ought to happen but to foretell what is likely to happen. As a matter of fact, the projected level of production (calculated to meet future effective demand) is well below potential production, however defined. 210

Hector Maletta

Production could be much greater, either by using more land or achieving more production per hectare (through wider diffusion of existing technology and by expansion of the productivity frontier if technology keeps improving as it has been doing in the past). In theory, of course, the production required to meet projected demand might conceivably be larger than the maximum potential production given available resources and technology, but this has not been the case, at least not at the world level. The world at large is considered to have enough resources to sustain the production required to meet future demand, even if demand projections were much more challenging than those estimated by FAO. Available land suitable for crops (as shown in Section 4.4) far exceeds the area of land currently under crops; it is also larger than the area that is expected to be put under crops in the coming decades. FAO estimates of land availability are based on careful consideration of agro-ecological zones defined by soil and climate, and purported future output is projected to take place without exceeding (in fact using only a fraction of ) existing natural resources. Humankind is using only about one-half of the available land that is suitable for rain-fed cultivation and FAO projections imply only a small expansion in the use of such land for rain-fed crops, and a small increase in irrigated land. In fact, extra land or extra irrigation are introduced in FAO projections only when projected growth in productivity (output per hectare), which is supposed to slow down, fails to meet projected demand and local resources and costs offer the chance of producing the extra goods needed domestically instead of importing them from abroad. The projected growth rates of land productivity are indeed conservative, implying a strong deceleration relative to previous decades. Access to food is closely correlated with per capita GDP and is bound to improve since per capita GDP is expected to increase, but the GDP growth assumptions used by FAO are also conservative in comparison with other estimates. Nikos Alexandratos and Jelle Bruinsma note that: The GDP assumptions adopted in this study were kindly made available by the Development Prospects Group of the World Bank. This is one of the most conservative scenarios among those available for several countries. (AB 2012:2)

The projections adopted envisage that global GDP will grow by a factor of 2.5 in the 44-year period between 2005-2007 (centred on 2006) and 2050, at an annual rate of 2.11% (AB 2012:36, their Table 2.4). As the authors explain, other projections from the World Bank foresee world GDP expanding by a 211

Projections on food and agriculture

factor of 3.6 in the same period, at a yearly rate of 2.88%; the French research centre CEPII envisages a factor of 3.3; Price-Waterhouse-Coopers forecasts factors between 3.3 and 4.1; and the various IPCC SRES scenarios used in the Third and Fourth IPCC assessments envision GDP expanding by factors between 2.2 and 7.1 from 2000 to 2050, at annual rates between of 1.58% and 4.00% (these are the GDP growth rates required to produce the greenhouse gas emissions in the various scenarios). On average, these other projections are significantly less conservative than the one chosen by Alexandratos and Bruinsma (AB 2012:34-35). A less pessimistic GDP projection would translate into better access to food and even less undernourishment, at the target dates of 2030 and 2050. Higher GDP may also generally produce faster growth in agricultural productivity, as investment in agricultural R&D and its application by farmers are also associated with higher levels of national output and income. In short, AB (2012), just like previous FAO projections, shows a perceptible tendency to understate future growth in agricultural (and food) production. Such biases in projected agricultural output would, of course, have an important impact on estimates of future changes in food supplies and on the prevalence of undernourishment. According to the preceding review of past projections, FAO output estimates for 2050 are likely to keep underestimating future demand for agricultural products and consequently future food output, possibly by a significant margin. The main source for this tendency is the very conservative nature of the assumptions adopted for future GDP growth, which (along with population) determine future demand. This is because production in FAO projections is primarily driven by demand and the underestimation of performance means that projections of demand have been (and are likely to keep being) too conservative. FAO projections show that with such modest projections of demand, sufficient farm production is attainable. If demand growth is corrected upwards, the logical question would be whether this faster growth of demand could be met by production. The answer seems to be affirmative. FAO assumptions on future growth of income and demand, and corresponding growth in production, imply an expansion of global demand by about 60%, which is met by projected agricultural production with a very small increase in arable land use, and with land productivity increasing at a slower pace than in the past. This suggests that in the likely event of a larger increase in future demand, further growth in agricultural production would be feasible. It would require combining an additional area under cultivation with some additional growth in output per hectare (at rates that still would be well below those observed 212

Hector Maletta

until now). Some calculations on the impact of climate change illustrate this feasibility; for instance, some of the projections by the IIASA research team on land use and climate change, which we shall review below, are based on scenarios of climate change assuming rapid population growth and contemporaneous increases in per capita GDP. These calculations show, as we shall see, that future production would meet future demand (projected up to the 2080s) even in scenarios of very rapid population growth cum significant growth of per capita income. Even in these cases, production is projected to meet demand without using much additional agricultural land and without imposing any significant additional stress on irrigation water supply. This will be considered in the next chapter of this book which deals with the impacts of climate change; it is mentioned here to underline the point that FAO’s projections do not suggest any difficulty in meeting future demand, even if demand grows higher than expected and even if FAO’s assumptions about future productivity growth continue to be too conservative, as they have invariably been in the past. FAO projections to 2050 did not consider possible effects of future climate change on agriculture. This might be justified since many evaluations of such change calculate that - even if global temperature rises by about 3°C during the century - temperature increases up to 2050 would not be more than 1-2°C, and that such an increase might not be harmful for agriculture. World climate, however, is expected to change over the entire century and beyond, and future agricultural production may be affected. How, and by how much, is the question, and it is not easy to answer.

213

10. Impact of climate change on agriculture

10.1. Concepts and methods A change in climatic conditions, chiefly in temperatures and precipitation, will undoubtedly have an impact on agriculture as it will on other aspects of life on Earth. But understanding and estimating such an impact is not straightforward. In this case, we specifically refer to climate change caused (at least in part) by emissions of greenhouse gases (GHG), mainly carbon dioxide (CO2). Climate is defined as the normal or average meteorological conditions in a certain place, and its normal range of variability. Such conditions include the level and the range of variation of temperatures and precipitation during the various seasons of a typical year. Normal or average climatic conditions (including the degree of inter-annual variability and the expected frequency and intensity of extreme events) are expected to change during the 21st century as a net result of both human actions (GHG emissions) and long-term natural trends. International climate projections have been prepared by the Intergovernmental Panel on Climate Change. The IPCC has published several assessment reports since the early 1990s; its latest issue (the Fifth Assessment Report) was published in 2013-2014. However, most of the available studies on the possible impact of climate change on agriculture have been based on previous projections, chiefly on those in the Third (2001) and Fourth (2007) IPCC assessments. Projections of climate change are based on several scenarios, defined by various possible paths of socioeconomic development in which higher or lower GHG emissions are expected to occur. These scenarios and their effects on the climate are quantified by means of various climate models, based on different approaches and using different assumptions. 215

Impact of climate change on agriculture

As noted before (Section 9.1) and explained at more length in Section 13.1, projections are not predictions: projections are conditioned by the choice of scenarios, models, and model assumptions. Climate models produce results at the planetary scale and for specific geographical areas. Both global and regional projections are affected by uncertainties connected with the choice of scenarios, models, and model parameters; uncertainty is higher, however, when models are used for projecting the future climate in particular regions or locations. The causal pathways through which GHG-induced climate change affects the biophysical side of agriculture are chiefly twofold: first, changes in temperature and precipitation may affect the growth of plants; second, higher atmospheric concentrations of carbon dioxide may also affect plant growth and water needs. General biophysical effect of climate change on plants and animals: Plants of any species growing in specific locations (including crops and vegetal wildlife) consist of varieties of each species that are adapted to each particular environment. Both natural selection and human intervention may determine which particular species and varieties grow at each location. Their yield fluctuates from year to year due to fluctuations in local weather, but their long-term local viability may be compromised if local climatic conditions significantly deviate from the range that is typical for each particular environment. The same is true for animals, which ultimately depend on plant food and water supply within a given ecosystem. There are usually many varieties and cultivars of each plant species, including most crops, and many breeds of each kind of animal; different varieties of the same species (and different cultivars of each variety) grow and thrive under different climates, albeit with different yields. Plant growth includes total plant biomass and also the growth of specific parts such as seeds, leaves or fruits; the effects of an environmental change could differ from one plant organ to another. In general terms, a significant alteration of the local climate in a particular area may expose a plant variety to climatic conditions that are not the most conducive to its growth. By making an area wetter or drier, hotter or colder, climate change may lead to a reduction in the growth of some plant varieties in a particular area, while other varieties or cultivars (that possibly did not grow there before) may thrive as a result. Specific biophysical effects on plants and animals derived from GHG-induced climate change: Projected anthropogenic climate change is not any climate change, but one resulting from GHG emissions (chiefly 216

Hector Maletta

CO2). Thus, in addition to the usual effects of climate change, there would be some specific effects resulting from higher amounts of greenhouse gases in the atmosphere. Higher atmospheric CO2 concentrations induce higher rates of photosynthesis and thus higher plant growth and yield (especially in C3 plants such as wheat) and a reduction in crop water needs (especially in C4 plants such as maize). C4 crops may also show increased photosynthesis and higher yield in the presence of elevated CO2, but only if growing under limited water availability, as more abundant carbon dioxide reduces water needs in C4 plants and thus permits extra growth. Otherwise, for C4 crops not constrained by lack of water, the effect of extra CO2 would mostly be a reduction in water needs, with only marginal or small increases in yields. Effects of climate change on farming practices and farm production. Besides its biophysical effects on plants and animals generally, the specific impacts of climate change on agriculture, and moreover on food production, also involve effects on the human side of agriculture. The impact of climate change on crops and livestock cannot be assessed without taking into account the human response to climate change. Crops and their yields, as well as livestock and its products, are not forms of wildlife but the outcomes of an adaptive human activity, and understanding the future of agriculture is impossible without ascertaining what shape this human activity will take. As climate changes, so too does farming but not because all individual farmers correctly sense (or anticipate) changes, rationally modifying their behaviour and making all the required technical choices; routine and ignorance as well as resource poverty lead one to expect nothing of the sort, especially in the least developed parts of the globe. Humans are fallible and prone to error. But humans are an adaptive species and a farming population would not persist in doing exactly the same thing while external conditions gradually change over several generations, and the population itself also changes through generational replacement and social or spatial mobility. It seems obvious that agriculture is a human activity and that humans are an adaptive species, but some assessments of the possible impact of climate change on agriculture are nevertheless prone to ignore the human side of agriculture. Thus, one possible approach to evaluating the effects of climate change on agriculture is a naive use of crop models that make use of mathematical relations (observed today or in the past) between environmental conditions (soil, temperatures or rainfall), crop characteristics, and crop management. The most traditional or conventional way is to take one such crop model, in which all variables have been calibrated to existing values (based on some 217

Impact of climate change on agriculture

specific variety of the crop at some specific place), and artificially simulate an alteration in some of the parameters (typically temperature or water availability), leaving the rest constant. This approach may be appropriate for short-term changes in a certain environmental variable (e.g., the effect of a relatively drier season on the yields of a rain-fed crop), but it is of course inappropriate for long-run projections because of the interdependencies between the variables. These interdependencies are crucial when long-term projections are involved; in the long term the ceteris are never paribus. The effect of marginally altering one variable while keeping the rest constant cannot be extrapolated to major changes extended over many years; and all the more so when the effect was originally estimated using an oversimplified tool, such as an econometric model composed of linear equations. Farmers at each location grow specific crops, and specific varieties and cultivars of each crop, according to prevailing long-term climate and soils. They also use specific crop management practices (type of tillage, depth of tilling, date of planting, form of cultivation, fertiliser application, plant protection, water management, soil conservation, and so on) also adapted to long-term local conditions. Since changes in the climate are gradual and a great many farmers are involved, trial and error would produce a multitude of good and bad results. The various outcomes would encourage the propagation of some practices and discourage or eliminate others. They would possibly drive out farmers who persisted with bad practices. At the very least, bad decisions would lead the respective farms and farmers to reduce their share in total production and their weight in average productivity. In the relatively long term (two or more generations), as is the case with projected climate change, even farm ownership, farm boundaries, and type of farm management would change. Eventually, the accumulation of individual decisions and market pressures would drive the average farmer and the average unit of land in each agro-ecological zone towards forms of agricultural production that are more suited to the existing environment, just as is the case now (and has been in the past). This does not imply perfect adaptation to the best solution available; there would be, as there is now, a gap between the productivity of an average hectare and the productivity that is potentially achievable at that location, and there would be a distribution of actual productivities around the actually achieved average; but actual average productivity can be expected to move in somewhat adaptive ways and probably in parallel with the displacement of the technological frontier (on the role of productivity in agricultural growth see Fuglie et al. 2012; Fuglie 2012; Baldos and Hertel 2013, 2014).

218

Hector Maletta

Climate change is a slowly unfolding change in the average historical conditions that guide current decisions made by farmers, and determines changes in farmer behaviour as it also determines change in farm structure and ownership. Over the time span envisaged in current estimates of climate change, several generations of farmers would manage each piece of land over several decades. There would also be changes in land ownership, type of farm organisation and management, farm-size structure, markets, technological frontier, farmer education, and much more. None of this can be ignored if one exogenously introduces the hypothetical climate of, say, 2050 or 2080 into the crop model. Farmers of the future will face a new climate (that will have come about gradually in fits and starts) equipped with the tools and information of the future, not with those of today. Gradual changes in climate (i.e., gradual changes in the average long-term conditions of the climatic environment) are accompanied by (usually larger) inter-annual fluctuations in weather. Thus, inter-annual ‘noise’ is much more visible than any underlying long-term trend. The biophysical effects of longterm climate change are thus not likely to be detectable from one year to the next, or even over a few years, on any local or regional scale. The human response to gradual climate change is a farming-system level outcome, obtained as a net effect of a myriad of individual decisions. Such decisions are informed (albeit imperfectly) by recent historical experience and (to some degree) by expert knowledge and advice, as perceived (and acted upon) by farmers who have a variable degree of cleverness and a variable ability or willingness to change their habits. At each turn, some may make a correct move while others do the opposite, as is usually the case when humans face uncertain conditions with limited information. Systemic change would also unfold in a gradual manner, and would emerge as a change in the mix of farm characteristics and practices, and in the mix of products (crop and livestock species and varieties) prevailing in each particular zone. Changes in the farming system, then, are expected to occur in a gradual and probably rather haphazard way if climate change gradually affects actual and expected agricultural production conditions over several generations. These farming changes do not solely depend on the behavioural changes of individual farmers. They should be seen as a population-level outcome at the scale of each agro-ecological zone, achieved through many changes at the level of individual farmers, but also by gradual replacement of individual farmers and gradual reorganisation of farm structure and management, in a process extending over periods of time that are relatively long by human standards. Such changes involve not only individuals changing their ways, 219

Impact of climate change on agriculture

but transformations in agrarian structure, land ownership, farm sizes, farm organisation (e.g., family farming vs. corporate agriculture), level of education and knowledge of farmers and farm managers, accessibility of technical information, development of markets for products and inputs, and so on. All these changes would be mediated by changing market conditions at the same location or elsewhere, either related or unrelated to climate change, and would happen even in the absence of any deliberate public policy intervention regarding agriculture or the impact of climate change on agriculture, and even in the absence of conscious deliberation or awareness of farmers about long-term climate change. It would thus be inappropriate to assess the potential impact of climate change on agriculture in the absence of human adaptation. However, impact analysis of that type, with a ceteris paribus clause that leaves adaptation aside, is quite common. The apparent conceptual justification for such an approach is the IPCC general conceptual framework on climate change impacts (IPCC 2007d, Glossary). That framework distinguishes potential impacts, not considering adaptation, and residual impacts once adaptations are accounted for. This distinction is useful for policy makers to assess the costs and benefits of possible or facultative adaptation of policy responses to natural events that require no human intervention to occur (e.g., the advisability of building additional coastal defences against a projected rise in sea level). But this is not the case with agriculture: the very process of farming involves an adaptive human activity, which therefore cannot be simply ignored or ‘kept constant’. Assessing impacts of climate change on agriculture requires assumptions about the future behaviour of farmers, which cannot sensibly be modelled as constant. Assessments cannot leave aside spontaneous farmer adaptations, because agriculture itself is an adaptive human activity. Such spontaneous farmer adaptations need not be optimum, as farmers are not infallible, nor can all of them be expected to be at the cutting edge of technology. Even new technology is not necessarily expected to offset any negative effects of climate change. In practice, even the most efficient groups of farmers usually get yields that are below crop trials conducted by agricultural scientists; and scientists at any given time can attain only limited improvements. This will still be the case in the future. This adaptive process does not require farmers to be clairvoyant, foreseeing all climate changes and unfailingly anticipating the future by timely adoption of the correct adaptation. Even in the unlikely hypothesis that all individual farmers universally fail to adapt during their lifetimes, the farming sector of each zone will adapt (to some degree) at the aggregate level, albeit involving the gradual 220

Hector Maletta

replacement of farmers and (to some extent) the reorganisation of farms via inheritance, land sales, and changes in land ownership or tenancy. Some ways of farming will be discredited and disused, while other ways will be replicated more widely. Farmer replacement is also important. Since climate change will unfold over several generations, it is evident that farmers in, say, 2050 or 2100 will not be the same individuals who ran farms in 2000 or 2010; some may be their grandchildren, while others might be complete strangers or impersonal corporations that happened to buy or otherwise acquire the land previously tilled by today’s farmers; these future new farmers may possibly produce an increasing share of total output and thus become more influential in total or per hectare agricultural output. The continuous, spontaneous, haphazard, and imperfect adaptation of farmers and other agents to the on-going conditions of markets and climate manifests itself as a statistical average evolving over many years amid ceaseless fluctuations and myriad trials and errors. Adaptation is never uniform, nor is it perfect. For any technology that is optimum for the prevailing conditions at a certain location, the average farm would be expected to operate with a certain degree of efficiency (which would generally be below the optimum). It is most likely that the future average degree of adaptedness would be similar to the degree observed in today’s farms, i.e., the average observed gap between scientific trials (or leading farms) and average farm outcomes. This gap is likely to narrow as prevailing forms of farm management gradually incorporate more modern techniques of production as has occurred in the past, but the gap would hardly disappear because science is always likely to be one step ahead, and also because farm use of technology is unlikely to be universal and perfect. Even advanced commercial farmers are known to be habitually below the performance of scientific crop trials. And individual farmers usually obtain a wide range of actual results distributed about the mean, with individual values depending on many idiosyncratic factors. Changes in the adaptedness of each farm and the degree of efficiency achieved on each piece of land would contribute to the average adaptedness of farming in each particular zone; they would also cause some farms to be more productive than others and thus determine the distribution of total output among various types of farmers and farms. It is to be expected that farming practices exhibiting lower adaptedness will either disappear, or else these farmers will gradually reduce their share of total output.

221

Impact of climate change on agriculture

Adaptation of agriculture to processes that are slow in comparison to a human lifetime is also a form of evolutionary change, though one that is much faster than anatomical changes during the evolution of biological species. It does not require individual farmers consciously deciding to change their way of farming on account of perceived or expected changes in the long-term local climate. It is best conceived of as a population/ecosystem/farming-system process whereby the prevailing distribution of farming activities in a region gradually evolves over time. This includes adaptation to environmental processes such as climate change or to socioeconomic changes such as population growth, technical progress, or changing market conditions. This adaptation may be voluntary (e.g., a farmer adopting new seeds) or due to forced circumstance (e.g., a bankrupt farm sold to a more entrepreneurial farmer). Just like any other form of adaptation to a changing (natural or social) environment, agricultural adaptation is selective. The process is based on the presence of variation among zones, farms and farmers, producing different outcomes that in turn cause some variants to become more frequent or more infrequent in the population and to increase or decrease their share of total agricultural output. Unlike biological evolution, where variants propagate only through biological reproduction, socioeconomic variants also are propagated by imitation, learning, and diffusion, and by the gradual marginalisation of less adaptive units (thus increasing average systemic adaptedness). Agricultural adaptation, understood as gradual selective change in a farming system within a given geographical area, or on the world scale, does not imply perfection or optimality; adapting is not optimising. Omniscient economic agents who know everything about everything and unerringly choose the best option exist only as simplifying assumptions in neoclassical economics textbooks. Real people are ‘creatures of a lesser god’, universally imperfect, that act on uncertain terrain and make fallible choices, often different one from the other. At any given time, some farmers are being driven off the land or see their output share reduced as a result of poor decisions in the past or an unexpected exogenous blow, while others are prospering or at least managing to survive and thrive in the farming business, thanks to their abilities or just plain good luck. Methods of assessment of expected impact of climate change on agriculture. The best approaches for assessing the probable impact of climate change on agriculture include allowances for technical change and for farming adaptation, especially the spontaneous adaptations that evolve in a farming system through micro-decisions of farmers and other economic agents. The 222

Hector Maletta

most usual ways to incorporate these factors are through so-called Ricardian models and Integrated Assessments. Ricardian models are econometric schemes whereby current farm revenue (or land value) across various climatic zones is estimated as a function of climate variables, controlling for other variables related to farm structure, soils, and (in some cases) macroeconomic environment. The models do not usually specify particular crops or particular kinds of livestock, but only the general characteristics of farms (such as size, type of ownership or management, land quality, and others). This approach quantifies the effect of a given change in climatic conditions (such as a 3°C increase in average temperature or a 20% drop in precipitation during a given season of the year), keeping other variables constant but allowing farmers to choose other crops or varieties or to otherwise adapt their farming practices. Some of these models allow for farmers introducing or improving irrigation while others take the extent of irrigation as a given, not subject to farmer choice. Some other approaches specify and quantify different equations for rain-fed and irrigated farms. Some use data on individual farms while others are based on aggregate data at the district level. The resulting spatial differences in farm revenue or land value, as observed today across different climatic zones, are used as an indicator of future changes resulting from gradual changes in the climate at every location. This approach accounts for adaptation but fails to account for the impact of increased CO2, since all the data are taken considering current carbon dioxide concentrations. Incorporation of irrigation in the models is also usually unsatisfactory or non-existent (for instance, some models assume that farmers can simply choose to use irrigation, not taking into account the actual availability of surface or underground water, nor the prior development of off-farm irrigation works such as dams and primary canals which would greatly influence individual farmer decisions). Integrated Assessment models combine several sub-models: crop models, agro-ecological zoning, climate change models downscaled to local conditions, socio-economic models at national and international levels, soil maps, crop experiments under varying CO2 concentrations, scenarios and assumptions about future technological change and economic growth, population projections, and other elements. These models allow for various possible assumptions regarding land use, crop mix, crop yields, water management, socio-economic environment, food demand, and other aspects. In fact, FAO projections of agricultural output, reviewed above, are themselves based on an integrated assessment approach, albeit not fully integrated with climate change models. 223

Impact of climate change on agriculture

Section 13.4 in the Technical and Methodological Appendix provides a more detailed discussion of the concepts and methods available for assessing the impacts of climate change on agriculture. Any such assessment should consider the various impact pathways; the analysis should be centred on ‘net’ impacts after taking spontaneous farmer adaptations into account; and the preferred approaches should be at least Ricardian models or Integrated Assessments, and particularly the latter. Estimated impacts should be calculated in relation to the output expected at a future date, thus incorporating projected changes in technology and land use; and they should cover the entire range of food or agricultural production, and not only staple foodstuffs such as cereals. The results should also be published with sufficient detail regarding substantive and methodological aspects and include estimates of the worldwide impact on agricultural output and prevalence of hunger. These requirements are only fulfilled by integrated assessment studies and most especially by those produced under the joint aegis of FAO and IIASA. We will now review these in more detail, as well as looking at some other models such as IFPRI’s integrated assessment models and Ricardian studies (especially those conducted by Robert Mendelsohn). First, however, we will review what the IPCC reports have to say on this matter.

10.2. The IPCC on agricultural impacts of climate change Assessments of agricultural impacts in IPCC reports have been rather general and qualitative, reviewing literature on various aspects of the issue, but not offering an estimate of the actual importance of the impacts of climate change on the future supply and demand of food. Also, many assessments (including many cited or reproduced in IPCC reports) refer only to potential impacts ‘not considering adaptation’, an approach that is wholly inadequate for the case of agriculture, as already explained. For this reason, we will briefly review recent IPCC impact assessments and then go on to present other quantitative estimates of impact produced by studies undertaken by academic researchers and by specialised centres or agencies such as FAO, IIASA, and IFPRI. These quantitative estimates were prepared using either the Ricardian or Integrated Assessment approaches. Still, it is important to review IPCC assessments, however incomplete, because they present the IPCC’S general conceptual framework as applied to this subject. All IPCC reports are available at http:// www.ipcc.ch/. The methods, definitions, and approach used by the IPCC to evaluate climate change and its impacts on agriculture are summarised in Section 13.4 in the Appendix. Here, we concentrate on two IPCC reports: the Fourth (2007) and the Fifth (2013-2014).

224

Hector Maletta

10.2.1. The IPCC AR4 report (2007) The IPCC Fourth Assessment Report (AR4) includes global and regional assessments of the possible impacts of climate change on agriculture (see IPCC 2007b - the report of Working Group II on Impacts and Vulnerability). It tends to confirm the conclusions of the 2001 Third Assessment (TAR): TAR indicated that impacts on food systems at the global scale might be small overall in the first half of the 21st century, but progressively negative after that. Importantly, crop production in (mainly low latitude) developing countries would suffer more, and earlier, than in (mainly mid- to high-latitude) developed countries, due to a combination of adverse agro-climatic, socio-economic and technological conditions. […] Many studies since the TAR have confirmed […] potentially large negative impacts in developing regions, but only small changes in developed regions, which causes the globally aggregated impacts on world food production to be small.23 (IPCC 2007b: 283-284)

This general assessment is quite important because it indicates that the overall impact is expected to be small at the world scale, with more positive prospects for developed countries (or, more precisely, in temperate regions as opposed to tropical ones). New information acquired after the 2001 TAR, as reported in the Fourth Assessment (IPCC 2007b:284-285), partly modified and amplified, but did not contradict these conclusions. Here are the main points, together with some comments: ŒŒ More demand for irrigation water; even after considering the increased water efficiency of some major crops (e.g., maize) under CO2 fertilisation, global demand for irrigation water would increase by 20% in 2080 compared with 1990 (Fischer et al. 2007), with more marked increases in tropical zones because of elevated evaporation. These estimates refer to plant water needs and do not take into account any increase in irrigation efficiency (more extensive use of pressurised irrigation as opposed to flood irrigation, etc.) which may fulfil plants’ increased water needs.

This passage cites the works of Fischer et al. (2002) and (2005); Parry (2004); and Parry et al. (2005). Other related studies, e.g. Fischer et al. (2007), are cited elsewhere in the same chapter. However, the AR4 report does not present (beyond a general citation of the papers) detailed results from such comprehensive integrated assessments, especially those by Fischer and associates at IIASA, which include quantitative projections of land use, water use, production, consumption, and undernourishment for the world and major regions. 23

225

Impact of climate change on agriculture

Comment: Average efficiency worldwide is now about 50% (Fischer et al. 2007). A very moderate increase in efficiency over the century plus the expected increase in irrigated areas would easily meet the increase in demand. Irrigated areas expanded at 1.77% per year in 1961-2000 according to FAO (2008e) and FAOSTAT/AQUASTAT data. Even a smaller rate of expansion in the 21st century, without much improvement in efficiency, would offset the expected increase in demand for irrigation water. On the other hand, a moderate increase in efficiency is also to be expected (Maletta and Maletta 2011:312-326).

ŒŒ Confirming a TAR finding, AR4 concluded that ‘including the effects of trade lowers regional and global impacts’ (IPCC 2007b:284). Comment: Of course food trade tends to offset local discrepancies between supply and demand; it is an expanding reality and must be included in any meaningful projection of food availability and access as per current definitions of food security. The role of trade is not only to convey food from one place to another; the existence of trade in farm products and inputs also enhances technical change and productivity as well as promoting higher growth. Existing integrated assessment studies, such as those carried out by IIASA and IFPRI, conclude that international food trade would increase as a result of climate change. The World Food Summits since 1996 have emphasised that trade is an essential component of food security.

ŒŒ Relative to TAR, AR4 found there are more possibilities of extreme events affecting agriculture. Comment: The IPCC (2010) report dealing with the possible effect of climate change on the frequency or intensity of extreme events tends to moderate the confidence placed in projections of increases in extreme events. Both reports (2007 and 2010) and also IPCC (2013a) insist on the difficulty of attributing extreme events to general or anthropogenic climate change. Most assessments related to this issue are reported to show a high degree of uncertainty regarding the probability of increasing or decreasing frequency or intensity of extreme events as a result of projected climate change. Projected changes in the average impact of extreme events are, in any case, incorporated into long-term climate projections, and into projections for agriculture which normally refer to decadal averages.

226

Hector Maletta

ŒŒ A review of the studies on the impact of climate change on crop yields confirms that ‘in mid- to high-latitude areas, moderate to medium local increases in temperatures (1°-3°C), across a range of CO2 concentrations and rainfall changes, can have small beneficial impacts on the main cereal crops. Further warming may have increasingly negative impacts, but this conclusion has medium to low confidence; in low latitudes, these simulations indicate that even moderate temperature increases are likely to have negative yield impacts for major cereal crops […] For temperatures increases [of ] more than 3°C, average impacts are stressful to all crops assessed and to all regions (medium to low confidence)’. (IPCC 2007b:285) Comment: Some of the results refer to studies in which farmer adaptation is not considered. This is problematic. Most crop species and varieties grown today in any particular environment and the corresponding techniques of production are adapted to that environment; any given crop would surely be stressed if grown in a different climate without changing the chosen variety or certain farming practices. Agriculture is itself an adaptive activity, adapted to the climate prevailing in each location, and therefore a projection ‘without adaptation’ is not a meaningful or useful concept. In fact, the above statement specifically refers to Figure 5.2 of the report (p.286) that provides results from many studies with and without adaptation, with the corresponding curvilinear regression lines. Most of the negative results correspond to simulation of impacts ‘without adaptation’. For studies where adaptation is considered (as befits the case of agriculture), the conclusions are not so discouraging; with the exception of wheat at low latitudes, the residual impact of climate change is null or positive for all crops considered (maize, wheat, and rice at mid-to-high latitudes, and maize and rice at low latitudes), even for simulations assuming +5°C, if only certain minimal adaptations are considered: changes in planting date, changes in the chosen cultivar, and shifts from rain-fed to irrigated conditions where water availability allows it. This does not include other possible adaptations (e.g., change in the mode of irrigation, improved efficiency within each mode of irrigation, changes in land use, shifting localisation of crops, or changes in fertiliser use, in the treatment of pests and diseases, and most especially in product mix, among others). If some of these other adaptations were considered the outcomes would be even more positive. The exception of wheat is for rather obvious reasons. At low latitudes, wheat is only grown by traditional farmers in a few specific locations (e.g., in the Sahel) with old cultivars especially adapted to Equatorial climates, at low yields. Other cereals such as millet, triticale, or sorghum are more suitable for such hot and dry conditions and are more abundantly cultivated. The finding that a temperature rise of more than 2°C would make wheat even less suitable at low latitudes is not surprising, since the crop is typically grown at higher latitudes and lower temperatures.

227

Impact of climate change on agriculture

The IPCC AR4 report also states that the above effects refer only to changes in temperature, not considering rainfall or elevated CO2, although ‘the studies summarized […] also indicate that precipitation changes (and associated changes in precipitation/evaporation ratios), as well as CO2 concentrations, may critically shape crop yield responses, over and above the temperature signal’ (p. 285). This is very important, because many of the models do not consider changes in rainfall (which may increase globally and in some areas as temperature rises, though diminishing in other areas). These studies, as mentioned before, also exclude other ways of increasing water availability (mode of irrigation or changes in irrigation efficiency). Almost none of the studies include the impact of elevated atmospheric carbon on yields or on crop water demand. These considerations lead us to regard the conclusions as unduly pessimistic and technically incorrect (global warming must entail changes in precipitation and CO2 effect on plants).

ŒŒ For sown fodder mixtures, AR4 confirmed the TAR finding that elevated CO2 increases the growth of legume pastures (such as alfalfa); AR4 extends these conclusions to temperate semi-natural grasslands, based on multiple free-air CO2 enrichment (FACE) experiments (p.287). Comment: These beneficial effects of elevated CO2 on natural grasslands and cultivated pastures also affect other crops. A large number of studies confirm these impacts on various crops. The IPCC (2007b) report states (p.276): Recent re-analyses of FACE studies indicate that, at 550 ppm atmospheric CO2 concentrations, yields [relative to current concentrations] increase under unstressed conditions by 10-25% for C3 crops, and by 0-10% for C4 crops (medium confidence), consistent with previous TAR estimates (medium confidence). Crop model simulations under elevated CO2 are consistent with these ranges (high confidence).24

These yield increases do not correspond to a doubling of CO2 concentrations from preindustrial levels (i.e., from 275 to 550 ppm), but to shifts from current concentrations (about 370-390 ppm at the time of the studies were done) to 550 ppm. The effect of doubling the concentration would presumably be greater, albeit with diminishing returns after a certain point.

24

228

Hector Maletta

Supporting data on this issue demonstrate a rather high positive impact of elevated CO2 on the yields of C3 crops like wheat and soybeans (midpoint +17.5%), and more moderate impacts (midpoint +5%) on C4 crops like maize, both relative to current CO2 concentrations at the time the studies were carried out (typically about 370-390 ppm). These CO2 impacts are always on currently grown varieties at each location and do not take into consideration any possibility of shifting (at each location) to more suitable varieties or crops in the face of gradual climate change. The final net increase in yields, resulting from both elevated CO2 and more adequate varieties, may be larger. Water savings are also crucial; expected decreases of rainfall (according to IPCC 2007) in maize-growing regions like Mexico and parts of the U.S. Midwest and Southwest would benefit from maize water savings at higher CO2 levels. Given the differential effect of CO2 on different crops, farmers are also likely to shift land use toward crops and crop varieties depending on a less negative or more positive net impact of climate change. These changes in crop mix and geographical distribution are not considered in the literature on CO2 impacts cited by the IPCC, which almost always refer to single crop species and current varieties at particular locations.

ŒŒ The IPCC (2007b) report (p.298) dismisses the idea that large numbers of wild plant species in European grasslands may be endangered by changes in temperature and rainfall: ‘such empirical model predictions’, the report states, ‘have low confidence as they do not capture the complex interactions with management factors (e.g., grazing, cutting and fertiliser supply’). Comment: This should also be true for other regions outside Europe, and also for crops. In addition, it illustrates the extraordinary importance of human action in determining the impact of climate change, even for wild vegetation such as natural grasslands, and much more so for crops and animal husbandry.

ŒŒ Some studies on the impacts of climate change (cited by the IPCC report) suggest a possible decline in the protein content of grasses due to higher temperatures. However, as the same IPCC passage (p.298) states, ‘an increase in the legume content of swards may nevertheless compensate for the decline in protein content of the non-fixing plants’.

229

Impact of climate change on agriculture

Comment: This finding is quite important for livestock activities based on natural pasture. It chiefly affects marginal grazing land with low nitrogen content; a warmer climate will preferentially encourage the growth of leguminous grasses that capture atmospheric nitrogen, enriching the soil, contributing to the growth of other grasses, and providing more protein for livestock. Recent satellite-based studies of global green cover in semi-arid regions confirm this conclusion since they show a general increase in green cover linked to increased CO2 concentrations in the atmosphere (Donohue et al. 2013). A simulation using the Hadley Centre climate model for equilibrium, in which the preindustrial concentration of CO2 was doubled, found an expected net global increase of 57% in net primary production of plants on the world scale (Hemming et al. 2013). This expected increase results from a gross increase of 75% in photosynthetic capacity and a 21% decrease due to plant responses to changes in the climate. It tends to support the idea that CO2 effects may be larger than climate effects (e.g., temperature and rainfall) when it comes to the impact of anthropogenic climate change on vegetation of all kinds (natural or cultivated). The Hemming results are subject to possible restrictions related to availability of non-CO2 nutrients for such growth, especially nitrogen, although experiments conducted by Melillo et al. (2011) suggest that soil warming releases extra nitrogen, thus greatly reducing this constraint. The general ‘greening’ of the planet due to increasing vegetation biomass has been further confirmed in a more recent paper by Liu et al. (2015).

ŒŒ The same passage of IPCC (2007b) also finds that although the plant diversity of natural grasslands may decrease with elevated CO2 and nitrogen deposition (because of the aforementioned optimum growth of legumes under higher CO2, causing relative expansion of legumes at the expense of other grasses), natural-grassland plant diversity increases with elevated precipitation, and (most importantly) shows no significant effect from warming. Comment: Fears that global warming will reduce biodiversity (at least in grasslands, and probably elsewhere), according to these IPCC findings should probably be viewed as overstated.

ŒŒ Increased temperature tends to reduce animal reproduction rates and milk production. The example provided in the report is that high-yield Friesian milk cows decrease their yield by one-third or more if transported to the tropics. However, even the expected impact of such sudden changes in climate is somewhat limited: ‘Production-response models for growing confined swine and beef cattle, and milk-producing dairy cattle, based 230

Hector Maletta

on projected climate outputs from GCM [Global Circulation Model] scenarios, have been developed by Frank et al. (2001). Across the entire U.S., the percentage decrease in confined swine, beef and dairy milk production for the 2050 scenario averaged 1.2%, 2.0% and 2.2%, respectively, using the CGC (version 1) model and 0.9%, 0.7% and 2.1%, respectively, using the HadCM2 model’ (IPCC 2007b:287). Comment: It is very important to note that these estimated decreases refer to current animal breeds if suddenly exposed to a different climate (like a Friesian cow transported by airplane from the Netherlands to the tropics). Therefore, they do not include the effect of gradual adaptation and (natural and artificial) selection of breeds to suit the gradual change to new climatic conditions occurring over a century, i.e., many generations of cattle affected by natural selection, selective breeding, breed crossing, geographical displacement of breeds, or genetic engineering. For instance: starting with Friesian cows, it took a few decades in the late 19th and early 20th centuries for Argentine breeders to get comparably high milk yields in ‘Hollandese-Argentine’ cows adapted to the conditions of Argentine prairies, implying differences in temperature and rainfall (relative to the Netherlands) even greater than those expected in the 21st century (relative to the 20th) under projected climate change. Similar results have occurred in Southern Africa, Brazil, Australia, and New Zealand, often involving the crossing of European (Bos Taurus) with Indian (Bos Indicus) breeds. It is also worth again recalling that any estimated percentage of reduction or increase due to climate change, besides allowing (in this case) for livestock adaptation, is to be applied to the expected production at the relevant future date. For instance, the aforementioned decreases of 0.7% to 2.1% in U.S. pork, beef, and milk, estimated by Frank et al. (2001) and cited by IPCC (2007b), are not only very small, but (if predicated for the climate of 2100) would apply to the level of US livestock production and productivity in 2100, probably much higher than the levels existing at the time of the study.

In addition to these IPCC findings on crop and livestock prospects under climate change, it is also important to note again that from the viewpoint of food security, it is not essential that the production of each food item continue at the same (total or per capita) rate at the very same locations where it occurs today; what is needed is sufficient worldwide output and adequate access to food, including physical access (through trade) and economic access (through income). If crops or livestock at some specific locations, even allowing for farm adaptation, should face an inevitable decline in production, there would also be increases in production in other places (chiefly at mid to high latitudes), which would partially or totally offset the former. In fact, all

231

Impact of climate change on agriculture

estimates foresee that growth in food trade will continue, with internationally traded food representing an increasing share of total food availability; and all these estimates also foresee increases in food consumption and declines in hunger, in all regions, even in the case of estimates based on extremely pessimistic assumptions. AR4 does not make an overall projection of future agricultural production with and without climate change in order to assess the net impact of climate change; this kind of assessment was indeed already available when the AR4 report was written (including Ricardian and integrated assessment models, that will be reviewed in sections 10.3 and 10.4) but these available studies are generally not discussed (and most are not even cited) by the IPCC Working Group II report of 2007.

10.2.2. The IPCC AR5 report (2013-2014) Available evaluations of the world-level impact of climate change on future agricultural production and food security are based on the climate change scenarios used in some of the first four IPCC reports. Few, if any, have as yet used the scenarios and projections of the IPCC Fifth Assessment Report or AR5 (IPCC 2013, 2014a and 2014b).25 IPCC (2014a) is the AR5 report on impacts, adaptation, and vulnerability. In relation to agriculture and food security, it summarizes a number of findings accumulated after the AR4 IPCC report of 2007 was issued and is better organised than the 2007 report. Unfortunately, however, the many factors at play are mostly analysed in a piecemeal manner; the report does not include any quantitative overall estimate of the (potential or residual) impact of climate change on world agricultural production or food access, nor an evaluation of the resulting world food situation during the 21st century once climate change and other intervening factors are taken into account. Most of the statements refer to one particular aspect (e.g., impact of higher temperatures on the yield of wheat, or the effect of increased atmospheric ozone on crops). However, the report does not arrive at or even aim to provide an overall estimate of all the factors at play in a projection of future agriculture. These include future income and One of the few examples using models and emission scenarios that were used in AR5 is the on-going programme of studies on agricultural impacts known as AgMIP, which we will briefly review in Section §10.4.3. The ‘representative concentration pathways’ (RCP) used in AR5 are described in IPCC 2008 and briefly characterised in the Technical and Methodological Appendix (see the closing paragraphs of Section 13.4.1).

25

232

Hector Maletta

population, projected climate change, projected or expected technical change, expected displacement of agro-ecological zones, expected changes in product mix due to changes in the structure of demand, and expected spontaneous adaptation of agricultural production systems to changes in climate. All these and more must be taken into account when projecting multidimensional changes in a complex adaptive system like agriculture. In addition to this lack of an overall prospect for agriculture under projected climate change, AR5, just like its predecessor reports, is mainly interested in measuring the differential impact of climate change if other things remain constant, not in foreseeing the future level of agricultural production. This is an important distinction; we regard climate change as one of the many factors affecting future production, along with changes in technology, land use, capital investment, demand size and composition, and others. Even if changes in the climate happen to have a significant effect, they should be combined with all the other factors to evaluate the course and ultimate pattern of agriculture and hunger during the 21st century. The 2014 IPCC report on climate change impacts on agriculture and food security is superior in many ways to the one issued in 2007. Its scope is wider, including not only agricultural production but food systems and food security; it improves its coverage of some issues that were treated more cursorily in AR4, such as atmospheric CO2 fertilisation, and includes some new topics such as the effects of atmospheric ozone on crops. It deals somewhat more extensively with integrated assessments, although persisting in using a piecemeal approach which does not produce any comprehensive projection of future agricultural production and food security at a planetary or regional level that is a joint function of climate change, technical progress, economic development, and other relevant factors. Because of this piecemeal approach, it is difficult to provide a summary account. Here, we will only underscore and comment on some specific points that are relevant to the purpose of this book. Relative to yields, IPCC AR5 summarises the evidence that has emerged since AR4 indicating that the effects on crops would be negative in the absence of adaptation: For the major crops (wheat, rice, and maize) in tropical and temperate regions, climate change without adaptation will negatively impact production for local temperature increases of 2°C or more above late-20th-century levels,

233

Impact of climate change on agriculture

although individual locations may benefit (medium confidence) (IPCC 2014b: 487) (emphasis ours).

This conclusion is interesting in several respects. First, it is stated with medium confidence, meaning there were problems with the robustness of evidence and insufficient consensus or agreement among experts and models; see IPCC (2014d:37) for the terminology used by the IPCC to indicate the degree of uncertainty. Second, the statement refers only to the three major cereals, the per capita demand for which has been stagnant overall during the latest decades, with a declining trend as income increases. Third, the statement refers to yields without adaptation, under significant changes in climate at each location, which is a quite unrealistic proviso; farmers at each location in 2100 would not keep growing the same crop using the same type of seeds their great-grandparents planted in 2000, and with the same farming techniques, disregarding the changes in climatic (and possibly many other) conditions during the intervening century. Fourth, and perhaps more importantly, the statement refers to yields of a current crop variety or cultivar at a specific location when the local climatic conditions are simulated to be different. As each crop model is based on varieties and cultivars suited to each location and climate and establishes its potential yield at current climate, current seed, current practices, current soil, etc., it is rather to be expected that any significant deviation in any of those factors would cause a decrease in yields. Such models typically represent a previous process of adaptation whereby the crop, the crop variety or cultivar, and the farming techniques were adapted to the prevailing climate and soil, and the current availability of agronomic technology. Gradually evolving climate change would favour another set of optimum decisions regarding crop, crop variety and cultivars, cropping techniques, and other aspects; for instance, using a variety of the same crop better adapted to a warmer climate could possibly offset the hypothetical increase in temperature, and changes in other factors (many yet to emerge in the future) would make the offsetting more likely. The overall impact reflected in the conclusion cited above is composed of opposite effects that occur in different places: regional analyses ‘show crop production to be consistently and negatively affected by climate change in the future in low-latitude countries, while climate change may have positive or negative effects in northern latitudes (high confidence)’. (IPCC 2014 c.b.:488). This still refers to climate change impacts without adaptation. In addition, no indication is provided regarding the relative importance of 234

Hector Maletta

lower and higher latitudes in the (present and future) production of cereals, nor the relative importance of cereals in overall crop production. And most especially, the report acknowledges that adaptation is the essential factor for food production: The relationship between climate change and food production depends to a large degree on when and which adaptation actions are taken (IPCC 2014b:494).

Given this, and considering the inadequacy of modelling farm production ‘without adaptation’, it is rather difficult to explain the salience given to theoretical effects modelled ‘without adaptation’ and the lack of salience given to effects that do include plausible adaptations, especially the spontaneous adaptations likely to be adopted by today’s farmers or (more probably) their successors several decades into the future. However, a collection of studies (not always strictly comparable), reflected in Figure 7-4 in IPCC (2014b:498), shows that a minimal set of adaptations greatly moderates or even reverses the effects of climate change on yields. In temperate regions, the effect with adaptation is positive (i.e., yields would increase with climate change, once farmers minimally adapt to the new conditions, even under the extreme hypothesis of warming up to +5°C). In tropical regions, the expected effect of climate change with adaptation would also be positive for rice and still negative but to a lesser degree in the case of wheat.26 In other words, if farming practices minimally adapt, rice in tropical regions would increase its yields under higher temperatures and the decrease in wheat yield would be lower than expected. The fifth impacts report of the IPCC (2014b) presents more abundant evidence on the positive effect on crops (and other plants) of elevated CO2 concentrations in the atmosphere. Moreover, it is stressed that change in cultivars may greatly enhance this effect, for instance in IPCC (2014b:499):

The studies included in the cited IPCC figure are quite heterogeneous in assumptions and methods, and this (according to the note to the figure) may explain certain strange results, e.g., the prospect for maize in tropical regions being worse with adaptation than without it. Controlling for other factors (such as CO2 concentration or absolute initial temperature at each location) may explain the anomaly; other less visible problems may exist in the rest of the figure, casting doubt on its usefulness. 26

235

Impact of climate change on agriculture

There is observational evidence, new since AR4, that the response of crops to CO2 is genotype-specific (Ziska et al. 2012). For example, yield enhancement at 200 ppm additional CO2 ranged from 3 to 36% among rice cultivars (Hasegawa et al., 2013).

Most crop models are based on a specific variety or cultivar, ignoring the wide variation among varieties and cultivars of any specific crop and that farmers are likely to adjust the cultivar and variety to the prevailing climatic conditions, especially if climate change takes place over many decades and affects several generations of farmers. The report also introduces recent studies showing that increased atmospheric ozone (O3) near the surface of the planet may partially offset the positive effects of additional CO2 (IPCC 2014b:499). There is, however, no global estimate of net impacts on agriculture of CO2 and O3 combined, or of possible moderation effects due to wider future diffusion of other industrial technologies such as renewable energy, cleaner industrial processes, electric cars, and the like. Even if findings about yields of cereals reported in AR5 are mostly negative, its findings on other crops are not so negative. One outstanding example is tubers, especially those grown in tropical climates such as cassava (manioc). The AR5 report on impacts (IPCC 2014b:499) highlights several studies on cassava, a major staple food especially in Africa and Latin America: Cassava (also known as manioc) is an important source of food for many people in Africa and Latin America and recent studies suggest [medium evidence, medium agreement] that future climate should benefit its productivity as this crop is characterised by elevated optimum temperature for photosynthesis and growth, and a positive response to CO2 increases (El-Sharkawy, 2012; Jarvis et al., 2012; Rosenthal and Ort, 2012).

Another source of benefits for crops under projected climate change come from early flowering and early maturity, as noted in the same passage of the IPCC report IPCC (2014b:499): Earlier flowering and maturity have been observed (robust evidence, high agreement) worldwide in grapes (Duchêne et al., 2010; García-Mozo et al., 2010; Jorquera-Fontena and Orrego-Verdugo, 2010; Sadras and Petrie, 2011; Webb et al., 2011), apples (Fujisawa and Koyabashi, 2010; Grab and Craparo, 2011), and other perennial horticultural crops (Glenn et al., 2013).

236

Hector Maletta

Changes in land suitability and use are recognised by AR5 (IPCC 2014b:504) as a major factor for determining future agricultural production, especially favouring temperate regions: As noted in the AR4, changes in land use, for example, adjusting the location of crop production, are a potential adaptation response to climate change. Studies since the AR4 have confirmed that high-latitude locations will, in general, become more suitable for crops.

The AR5 report on impacts and adaptation affords some attention to so-called ‘integrated assessments of climate change impacts on agriculture’ (see their Section 7.4), but the models and impacts involved are partial, mostly referring only to yields, and take note of a set of experiments not involving adaptation as well as other studies with a range of incremental adaptations (not well identified: see their Figure 7.7). These approaches, even if limited to a few crops, and concentrated only on yields, show more encouraging results than the conventional use of crop models mentioned earlier in this review. For instance, according to the aforementioned AR5 (2014b) Figure 7.7, the combined effect of climate change and higher CO2 in the atmosphere is estimated as a net reduction in yields of 0% to ‑2% per decade with a median of about ‑0.5% (equivalent to a median of -5% per century). Projected impacts ‘without adaptation’ show median effects between about ‑0.8% per decade in temperate regions and ‑1.5% in tropical ones (respectively ‑8% and ‑15% per century). It is important to recall that these projected yield reductions are modelled (not observed) effects on crops, based on crop models that take into account varieties used today and adjusted to present climate and edaphic conditions. The results were obtained by running the models under modified temperatures and CO2 concentrations but without taking into account future changes in technology, population, food preferences, land use, changes in crop mix, and sundry other factors. Moreover, these effects were only estimated in reference to yields per hectare in places where the chosen crop varieties are presently cultivated, and with adaption involving only minimal adaptive measures such as changing the date of planting but without changes in the extent and location of the areas planted, the farming practices applied, the crops grown, or other adaptive changes that would quite plausibly occur during the course of the century. The use of so called ‘integrated assessments’ in the IPCC’s (2014b) chapter on agriculture and food is thus quite elementary: it is limited to yields and is 237

Impact of climate change on agriculture

called ‘integrated’ only because it takes into account both the effects of climate and CO2, ignoring other factors such as changes in technology, demography, suitable or planted area, displacement of agro-ecological zones, extent and efficiency of irrigation, growth of per capita income, and so on. It does not take into account several existing integrated assessments that do consider all (or most of ) these factors, such as the various studies undertaken by IIASA and FAO. There is indeed a recognition of ‘increasing evidence that farmers in some regions are already adapting to observed climate changes, in particular altering cultivation and sowing times, crop cultivars and species, and marketing arrangements’ (IPCC 2014a:514). One simple incremental spontaneous adaptation (especially for rain-fed crops depending on the start of rains) is changing planting dates due to a changing climate. In this regard, the report states that: ‘Aggregated across studies, changing planting dates may increase yields by a median of 3 to 17%’. In addition, there is also recognition that ‘use of short duration cultivars could be desirable so as to reduce exposure to end-of-season droughts and high-temperature events [...] optimisation of crop varieties and planting schedules appear to be effective adaptations, increasing yields by up to 23% compared with current management when aggregated across studies’. The range of increases estimated to be caused by these factors (3% to 17% from changing planting dates and up to 23% by optimising crop varieties and planting schedules) are enough to offset the estimated biophysical effects of climate change on yields, and this does not take into account other factors such as new seeds, changes in area planted, displacement of agro-ecological zones, better farming practices, and many others. More generally, the chapter of the report discussed here evaluates a range of studies and presents a total figure showing that, on average, adaptation to climate change would increase yields by about 16-23% more than if no adaptation took place (measured as the difference between adapted and non-adapted simulations, estimated from their figures 7-8 (IPCC 2014b: 516). Since most expected decreases in yields without adaptation were below 20% (except some hypotheses of extreme climate change over +5°C in tropical areas, where yields decreased more in a hypothetical world where farmers planted their crops for decades without any adaptation), it can easily be concluded that the overall effect would be small or nil, and possibly positive, at most plausible future temperature increases, as is in fact stated in both AR4 and AR5.

238

Hector Maletta

All in all, however, the AR5 report, as was the case of its predecessors, takes a very simplistic view of adaptation and tends to understate its effects; indeed, the chapter on food and agriculture recognises this bias: It is notable that most of the above adaptations [...] are essentially either incremental changes to existing agricultural systems or are systemic changes that integrate new aspects into current systems. Few could be considered to be transformative changes. Consequently, the potential adaptation benefits could be understated (IPCC 2014b:516) (emphasis ours).

On the other hand, the distinction between incremental and transformative adaptations, as we have noted before, is rather fuzzy: evolutionary processes often attain transformational effects through incremental adjustments. In this case, ‘transformative changes’ seem to refer to the effect of major interventions outside individual farms, such as building major irrigation works or an overhaul of the production, processing, and marketing process. These kinds of measures are also ordinarily undertaken - as demonstrated, for example, in the rapid increase in areas equipped for irrigation. But transformative changes also occur in an incremental fashion, as is the case of changes in technology, from discovery to development and then gradual diffusion and further improvement by trial and error. This is also true in other similar macroprocesses reviewed (in relation to the past half century) in Part II of this book (changes in crop mix, improvements in water management efficiency, and so on). Whether transformative, incremental or both, all adaptive changes that are tried out, and then selected and propagated, would have an impact in reducing any harm caused by climate change or through exploiting the opportunities it offers. These impacts are insufficiently considered in the AR5 report on adaptation.

10.3. Integrated assessments: IIASA models Integrated assessments (see technical details in Section 13.5) combine various types of information and assumptions embedded in crop models, climate models, soil maps, agro-ecological zoning, socioeconomic models and projections, etc., and use various scenarios to project the possible behaviour of a complex adaptive system, such as climate change, and its impact on a human-driven activity such as agriculture. This section and the next review some of these studies using an integrated assessment approach. The studies primarily include the IMPACT model developed by the International Food Policy Research Institute (IFPRI) and the results of the Land Use Change group at IIASA, working in close alliance with FAO. 239

Impact of climate change on agriculture

In this section, we review the IIASA approach, especially the analysis of Günther Fischer (2011) in which the baseline FAO projections to 2050 are modulated to highlight the projected impacts of climate change and biofuels. We also refer to some previous reports from the IIASA team. This set of studies has the virtue of addressing the question of climate change impacts from the point of view of food security at the world level (and for major regions), and thus producing estimates of future food production, food consumption, and undernourishment that incorporate (in addition to other factors already considered in FAO [2006] and AB [2012]) the specific impact of climate change on total agricultural production, per capita food consumption, and the prevalence of undernourishment, which are the central concerns of this book.

10.3.1. Impact of climate change on land area suitable for crops FAO has estimated the land area with various levels of suitability for rain-fed crops, from VS (very suitable) to NS (not suitable), as seen in Ch. 4. The ‘suitable’ classes (VS, S, and MS), not including marginal or marginally suitable land, are grouped as Prime and Good land (Section 4.4). Most crops in the world are grown on about half of all existing Prime and Good land, in addition to some crops grown on marginal land. Half of all usable Prime and Good land is still available for cultivation, not including land that is covered by forests, built up, or otherwise strictly protected. The amount of land suitable for agriculture will be affected by climate change, which will modify the crop-suitability of land by altering current temperature and precipitation regimes. In particular, some lands will become cultivable or more suitable for cultivation, especially in medium and medium-high latitudes, while some other lands (chiefly in tropical and subtropical zones) may become less suitable based on climate projections for 2050 and beyond. It is to be noted that to construct these simulations, certain hypotheses must be adopted about the technology to be used, from simple or traditional cultivation with a low level of inputs (no mechanisation, little or no fertiliser, etc.) to high-input technologies. Some land that is only marginally suitable using traditional farming technology may be more suitable under a better system of production (e.g., applying soil and water conservation techniques, or using more fertiliser). The Land Use programme at IIASA, in collaboration with FAO, has produced some projections including the likely effects of climate change on the ex­tent of land suitable for rain-fed cultivation.

240

Hector Maletta

Fischer et al. (2002b) presents the results of some simulations of changes in land suitability under various hypotheses of (local) climate change, i.e., not resulting from general climate models but simply assumed to occur in each region. Only three crops were taken into consideration: rain-fed rice, wheat, and maize. Therefore, the resulting areas suitable for crops do not coincide with those discussed in Section 4.4, which is based on a broader set of crops. Three scenarios are presented by these authors regarding changes in local temperatures: +1°C, +2°C and +3°C, all of which are assumed to occur in all seasons. These thermal scenarios are also combined with varying increases in local precipitation (0%, 5%, or 10%), again for all seasons. As shown in Table 51, in these scenarios the area suitable for rain-fed cereal cultivation will increase worldwide, though the effect is greater for temperature increases of 1°C and 2°C, and smaller for 3°C. These three levels of warming (not considering precipitation) will increase the world area suitable for rain-fed cereals (classes VS, S, and MS) by 3.9%, 3.1% and 1.1% respectively. These effects are moderately amplified to 4-6% when precipitation is allowed to increase by 5% or 10%. Table 51. Sensitivity to local climate of usable land area suitable for rain-fed cereal cultivation. Local climate change hypotheses

Rainfall

0% (ref )

Temperature 0°C (ref )

0%

0%

0%

+5%

+5%

+10%

+10%

+1°C

+2°C

+3°C

+1°C

+2°C

+2°C

+3°C 1429.4

Million Ha Developing

1607.9

1587.0

1519.5

1429.4

1574.1

1511.4

1513.0

Developed

744.9

827.6

856.6

876.7

855.1

887.2

910.3

933.4

2444.6

2425.7

2378.7

2470.4

2461.0

2491.6

2454.0

Usable prime 2352.8 and good land World (VS, S, and Percentage change MS classes) Developing 0.0%

-1.3%

-5.5%

-11.1%

-2.1%

-6.0%

-5.9%

-11.1%

Developed

0.0%

+11.1%

+15.0%

+17.7%

+14.8%

+19.1%

+22.2%

+25.3%

World

0.0%

+3.9%

+3.1%

+1.1%

+5.0%

+4.6%

+5.9%

+4.3%

Upper panel, first data column: Reference or baseline area in Fischer et al. (2002b:87, 6th data column) (VS+S+MS land, at average 1961-1990 climate, based on land potential for growing rain-fed rice, wheat and maize with mixed levels of inputs, excluding areas that were built-up, forested, or otherwise protected by 1994-96). Figures in millions of hectares in other columns are based on the percentages in the lower panel, taken from Fischer et al. (2002b:102).

On the world level, the effect is positive (more land will be suitable), but the result will vary by region and by type of country. In particular, suitable land will shrink in developing countries (predominantly tropical) and expand in developed countries (located mostly in temperate or colder zones). However, the worldwide effects are relatively small, with the total area suitable for crops varying from a baseline of 2352.8 million Ha to various levels ranging from 2378.7 to 2491.6 MHa. 241

Impact of climate change on agriculture

These effects, however, correspond to assumed changes in local precipitation and temperature, not to any projection model of local or global climate change, uncertain though it may be. On the other hand, climate change projections do imply a worldwide increase in precipitation, but some areas will experience reduced precipitation, including some that are already subject to water-stress (e.g., Southern Africa or Northern Mexico). Unfortunately, the authors do not explore the effect of a decrease in precipitation. Changes in precipitation, on the other hand, will affect the classification of a piece of land as suitable or marginal only when rainfall changes pass the critical point at which MS land becomes marginal (or vice versa) for the crops (and technologies) considered. Precipitation will increase worldwide under climate change; only some areas of the world will be affected by reduced precipitation, and, in turn, only in some of these areas, will the change pass the critical point. These simulations by Fischer et al. include more detailed results by region. The effects in some regions are more substantial; of those where suitable land is increased, and only when considering a temperature increase of 3°C, North America’s suitable land will expand by 20% to 28% depending on the hypotheses adopted about precipitation change (0% to +10%); Russia’s suitable land will grow by 40% to 50%; most sub-regions of Europe will also expand in various degrees between 0.6% to 21% (except Western Europe where suitable land will decrease by -2.9% to -4.3%). In Central Asia (another region where low temperatures may be a major constraint for rain-fed cereals) the areas of Prime and Good land will increase between 6.6% and 51% depending on precipitation, while East Asia will expand by around 15-16% under all hypotheses; a more significant relative increase will occur in West Asia, where prime and good land will expand between 65% and 106%. Major areas where suitable land will diminish are South America (-19.8% to -22.7%), South Asia (-3.0% to -13.6%), and Southeast Asia (-11.7% to -21.2%). Much of the effect depends on the pre-existing climate of each region. For instance, in relatively dry areas, increased precipitation expands the suitable land area (relative to less change or no change in precipitation), but in some humid areas, simulated increases in precipitation may actually cause a reduction in suitable land, through increased probability of flooding. Such is the case of South Asia: under +2°C warming, the suitable area reduces in size by 3% if rainfall increases by 5%, but shrinks by as much as -8.7% if rainfall increases by 10%. A similar effect pattern is projected for rain-fed cereals in Southeast Asia.

242

Hector Maletta

Besides these simulations based on hypothetical increases of temperature and rainfall, Fischer et al. (2002b) also include a projection of changes in land suitable for major food and fibre crops, based on IPCC climate change projections downscaled to various regions. Climate change is projected according to the A1F1 ‘worst case’ scenario (IPCC 2000, 2007a), in which fossil fuel use as well as CO2 emissions and global warming intensify during this century. This scenario was quantified using one of the major Global Circulation Models available, developed by the Hadley Centre in the UK, and projected for the decade starting in 2080. According to that model, total land area suitable for rain-fed cultivation of major food and non-food crops will slightly increase by 2% worldwide. Total land suitable for rain-fed cultivation in the reference climate of 1961-1990 was 2598 MHa, increasing to 2649 MHa in the 2080s under the chosen climate projection scenario of intensified fossil fuel use. The small overall increase in land with potential for rain-fed cultivation (some 51 million Ha) is the net result of increases in some regions and decreases in others, typically expanding in developed and temperate countries, and falling in developing and tropical ones, generally by relatively small proportions in both directions. All of the above refers to the stock of suitable (and usable) land, whether used or not used for crops. The total area of suitable land effectively under rain-fed cultivation with annual or permanent crops for the baseline period (19941996), as estimated by Fischer et al. (2002b) on the basis of FAO data, was about 1245 MHa, in addition to 260 MHa of irrigated land. Of this total area of 1505 MHa of land used for crops, some corresponds to crops grown on marginal land (about 300 MHa according to data in Section 4.4), leaving about 1200 MHa of crops grown on suitable land. However, cultivated land will increase only marginally according to this IIASA simulation; as in the past, most future growth in output will come from increased land productivity rather than expansion of cultivated area, even if the IIASA simulation involves a lower rate of growth of productivity throughout the 21st century (as compared with recent decades). It may be surmised, based on past experience, that land productivity in turn will respond to a combination of increased application of other resources (labour, equipment, infrastructures including irrigation, and inputs such as fertiliser or plant protection), and growth of Total Factor Productivity; according to Fuglie (2012), the latter is by far the dominant factor and will reflect the combined effect of changes in the yields of each crop, changes in crop mix, and changes in cropping intensity.

243

Impact of climate change on agriculture

This IIASA projection of suitable land does not take into account these factors, nor does it consider any expansion in irrigated areas, some of which may occur in areas not currently suitable for rain-fed cultivation. Indeed, it only refers to the area that is (or will be in the future) suitable for rain-fed cultivation. Under these very conservative hypotheses, as shown in Table 52, land suitable for rain-fed crops is expected to expand significantly in middle and high latitudes: by 87 million hectares (MHa) in North America (+33% relative to the baseline figure), by 91 MHa in Russia (+56%), by 12 MHa in East Asia (+8%), and by 8 MHa in Central Asia (+64%). Reductions will occur in other regions: 54 MHa in South America (-11%), 20 MHa in Oceania (-17%), 18 MHa in Southern Africa (-47%), another 18 MHa in Eastern Africa (-5%), 15 MHa in Central Africa (-6%), and 14 MHa in Northern Africa (-78%). Table 52. Land suitable for rain-fed cultivation of major food and fibre crops, under baseline (1961-1990) and projected 2080s climates for SRES A1F1 scenario quantified using the Hadley Centre model. Region North America Eastern Europe Northern Europe Southern Europe Western Europe Russian Federation C. America and Caribbean South America Oceania and Polynesia Eastern Africa Central Africa Northern Africa Southern Africa Western Africa West Asia Southeast Asia South Asia East Asia Central Asia Developed countries Developing countries World

Land classed as Prime and Good for rain-fed crops (MHa) Reference climate Projected climate Change % change (1961-90) (2080s) 267 354 87 33% 111 112 1 1% 36 39 3 8% 42 45 3 7% 57 56 -1 -2% 163 254 91 56% 44 43 -1 -2% 473 419 -54 -11% 121 101 -20 -17% 333 315 -18 -5% 243 228 -15 -6% 18 4 -14 -78% 38 20 -18 -47% 169 162 -7 -4% 29 31 2 7% 115 107 -8 -7% 178 177 -1 -1% 147 159 12 8% 14 23 9 64% 798 960 162 20% 1800 1689 -111 -6% 2598 2649 51 2%

Source: Fischer et al. (2002a:74, 6th and 8th data columns). Prime and Good land areas comprise land classes VS, S and MS, excluding those presently forested, built-up, or otherwise strictly protected. These excluded lands totalled 681 MHa in 1961-1990 and 688 MHa in the 2080s. Marginal land, even if presently cultivated, is also excluded. Areas slightly differ from those in Table 51 due to considerations of suitability for other crops besides cereals.

244

Hector Maletta

According to these projections, developing countries will lose 111 MHa (6%) of rain-fed suitable land (Africa as a whole will lose 72 MHa, representing 39% of the continent’s baseline suitable land availa­bi­li­ty, and two thirds of the total loss in developing countries). Overall, on the other hand, developed countries will add 162 MHa (20%) to their rain-fed suitable area, for a total net gain of 51 MHa (+2%). These projections are, of course, dependent on the chosen model and scenario, and the quality of the regional downscaling of global circulation models such as the Hadley Centre model; but the major conclusions seem to be robust to the choice of scenario and model, such as the expansion of suitable land in northern latitudes (North America, Russia, sizable parts of Europe, Central Asia) and reductions in tropical areas, chiefly in Africa. Within large regions, it may be that some sub-regions are gaining ground while others are losing it; for instance, the temperate southern parts of South America will likely increase their suitable land whilst the projected losses in that subcontinent will be concentrated in its tropical latitudes. Changes in the area suitable for crops, as simulated in Fischer et al. (2002a:72-83 and 2002b:67-77), indicate a globally small though positive net effect of climate change in the area suitable for rain-fed cultivation, with significant differences in regional impacts. The global effect will be positive in currently developed countries, and negative in those that are still developing. However, only half of all land suitable for crops is actually used and this will still be true, according to all projections, after the projected effects of climate change on land suitability. Land scarcity does not seem to be a problem on a global scale, nor will it be accentuated by climate change. Since only one half of suitable land is currently under crops, the loss of some suitable land (e.g., in Africa) does not imply that actual suitable land used for crops will necessarily be reduced. According to IIASA projections, the expected increase in the actual use of cropland is very moderate and will withstand the projected reduction of suitable areas without causing a reduction in the area actually cultivated, even under the restrictions imposed in that study on future technological developments. Besides, changes in the use of land - even if the projections indicate a reduction in the expected growth of crop production (which is doubtful given the pace of technological progress and the projected growth in demand) will still be offset by increases in other regions and continuing expansion of international food trade (which in the past grew almost three times faster

245

Impact of climate change on agriculture

than food production: in 1961-2011 food output grew by a factor of 3.3 but food trade by a factor of 8.5).

10.3.2. Potential rain-fed cereal output with constant technology Changes in the quantity of land in each suitability class are not the whole story: within existing suitable land there will also be changes in potential production. Some lands (even with­in the broad categories of Prime and Good land) will reduce or increase their potential cereal yields. Furthermore, the abovementioned estimates of changes in potential production generally ignore technological change and adaptation of farm production - essential elements for any attempt at envisaging the future prospects of agriculture in the face of a changing climate. These exercises in impact assessment without taking account much in the way of adaptation are, however, useful for some purposes, as shown in this section. Fischer et al. (2002a and 2002b). As explained before, ideas about the potential impacts of climate change, without considering adaptation, are not really applicable to the case of agriculture, because agriculture is itself a form of adaptation of human agency to deal with prevailing environmental conditions and projecting its future shape must factor in the actions likely to be taken by farmers. However, as a purely theoretical exercise, restricted to a single group of products (rain-fed cereals), the IIASA group has produced some projections for 2080 that are worth reviewing. Fischer et al. (2002b) report on the results of combining climatic and crop models, limited to rain-fed cereals (excluding irrigated rice, other irrigated cereals, all other crops, and all livestock), and referring only to land suitable for cereals. The amount of land suitable for cereals may change because climate change may alter the boundaries of agro-ecological zones as temperature and precipitation patterns change over time, as seen in Section 10.3.1. Also, the potential yield of rain-fed cereals in a given location (under a given fixed technology) may change due to expected local changes in temperature and precipitation. ‘Potential’ production of cereals in Fischer et al. (2002b) is based on the technology used in the base period, i.e., cereal varieties and cultivars used and cultivation techniques prevailing at each particular location at the time of the study (mid or late 1990s). The model works in a recursive manner, year by year, on each of 2.2 million grid-cells across the whole land area of the planet (except Antarctica) and simulates farmers’ decisions (each year) on 246

Hector Maletta

the specific crop, variety, and cultivar they want to grow based on climate and market conditions, choosing from a relatively large set of possibilities. For the published simulation, a fixed technology for each location was used (in this case, a high-input technology, not necessarily the one actually used on average in each area).The comparison was between the potential outputs that could be achieved in the baseline period versus the potential outputs that could be expected in 2080, under the same technology. Farmers are assumed to choose, during each period, the most efficient mix among 83 established cultivars and varieties of a selection of cereal crops (wheat, barley, rye, rice, grain maize, sorghum, millet, and setaria), given estimated current climate and prices, and to use a technology based on a high level of inputs for rain-fed cultivation, according to a large number of crop models used for IIASA-FAO simulations. Climate changes were projected, as we have already seen, under the A1F1 ‘worst case’ scenario (IPCC 2000, 2007a), in which fossil fuel use as well as CO2 emissions and global warming are expected to intensify during the current century. The main conclusion is that the worldwide potential production of rain-fed cereals in 2080 will be significantly higher than the respective potential in 1961-1990 due to global warming, even if the exercise uses a fixed set of cereal species, varieties and cultivars, grown under the same agricultural method. Potential production of rain-fed cereals on suitable lands (with constant high-input technology and allowing farmers to maximize production at each period by choosing the optimum mix of the given cultivars) is expected to increase by 11% worldwide (from the reference climate of 19611990 to the projected climate of 2080), with major expected percentage increases in Russia (91%), North America (28%), Northern Europe (18%), and Central Asia (48%); and major relative reductions in Northern Africa (-76%) and Southern Africa (-41%). The production potential will increase by 24% in developed countries, and by 4% in developing ones, giving an average global increase of 11%. The main factor for the increase in potential output is the net addition of new crop-suitable land due to climate change. If the simulation is restricted to suitable land ‘currently cultivated’ (i.e., in 1994-1996 for simulations carried out by Fischer et al. 2002a), the cereal production potential under the same assumptions will be slightly higher (+1%) in the climate of 2050 than in the reference climate of 1961-1990, and slightly lower (-3%) under the projected climate of 2080 (Fischer et al. 2002a:65); both percentage changes are well within margins of uncertainty and are not to be regarded as significant. Production potentials under constant technology, then, will vary only slightly, or not at all, on land already cultivated at the base 247

Impact of climate change on agriculture

period. They are instead expected to increase due to the opening of new lands as a result of climate change (chiefly in temperate zones). The assumption of constant technology, adopted for these simulations, is only illustrative and not to be construed as a realistic projection of future methods of agricultural production. In fact, this entire exercise is mostly theoretical or of academic interest, for various reasons: (a) It considers only rain-fed cereals, ignoring irrigation, other crops, and

livestock.

(b) The simulation considers only a fixed set of cultivars available around 1995. However, cultivars are constantly being improved to suit different environmental conditions: some varieties and cultivars grown in the 1990s, kept constant in this Fischer simulation, have already been superseded at the time of writing (two decades later), and many will be antiquated by 2080, as agronomic research and bioengineering provide new varieties, and cultivars and seeds are adapted from other (e.g., drier or warmer) regions. (c) The technology level adopted (high-input) is not the one actually used by all farmers. Even under fixed technical knowledge, output may improve through wider and more efficient use of that knowledge (e.g., if the highinput technology as developed by 1995 is more widely adopted). (d) Other adaptations by farmers (besides choosing the most suitable crop for each climate and price structure) do normally take place, such as using forms of cultivation that improve water retention in soils (e.g., contour ploughing), applying more fertiliser, improving plant protection, etc. This is true even of that most basic scheme of food production, rain-fed cereal cultivation, which humans have practiced for at least 10,000 years in varying climates. The simulation, for instance, does not consider increases or decreases in double or triple cropping, a topic considered separately by Fischer et al. (2002a:75-83). However, the exercise in potential impact conducted by Fischer et al. shows that under the worst conditions (extreme climate change, minimal or no adaptation, no technical change, no irrigation, etc.), the potential production of rain-fed cereals will significantly increase worldwide - not as much as necessary to meet world demand in 2080, but enough to make it evident that the potential impact of climate change alone, even if the concept is not

248

Hector Maletta

strictly applicable to agriculture, will mean an increase, not a decrease, in the potential worldwide production of rain-fed cereals. Once other factors are considered (irrigation, new seeds, improved techniques, and the like), prospects are still better, as shown by the same team of researchers in their integrated assessment of impacts of climate change. Fischer (2011). Simulations with updated data and models are presented in Fischer (2011), reflecting IIASA model runs carried out in 2009. However, the analysis in this case is restricted to land already in cultivation. Under this restriction, the impact of climate change alone will be nil or slightly positive (as in Fischer et al. 2002a). Farmers in this simulation are still restricted to a fixed technology, and are allowed not only to choose crops and cultivars, but also the date of planting. Moreover, the exercise in this case includes two variants for the choice of cultivars; in one, farmers can only choose among local varieties and cultivars already in use in the same geographical area; and in the other variant, farmers are allowed to use best-adapted plant types such as those available elsewhere and adapted to higher temperatures. Change in rain-fed cereal production potential in the 2050s, relative to the reference climate, and considering only land presently cultivated with rain-fed cereals, will be 0% if farmers are limited to the use of their current local varieties, and +3% if current varieties from other zones are also allowed for (Fischer 2011:108, Table 3.11). Under the same hypotheses, however, the +3% impact estimated for the 2050s will be reduced to 0% in the 2080s, using the Hadley climate model with the A2 scenario (Fischer 2011:109, Table 3.12). This zero impact under the Hadley model for all cereals in the 2080s is the net result of a negative impact for wheat (-12%) and a positive one for maize (+7%). Results based on the CSIRO climate model are similar: -10% in potential wheat output and +9% for maize, in addition to an estimated +16% for sorghum (Fischer 2011: 110). The reported CSIRO runs do not include a total or average figure for all rain-fed cereals. All these results are obtained with models including CO2 fertilisation effects. Fischer also provides results without CO2 effects, but there is nothing to be gained in reproducing them here: increased CO2 availability certainly does affect photosynthesis and water needs and is therefore an integral and unavoidable component of the biophysical impact of climate change on crops or indeed vegetation in general. To sum up: these IIASA results suggest that climate change will cause a significant increase in the global area of land suitable for rain-fed cultivation, which includes land that can be also irrigated, and will determine an increase (or at worst, would not cause a reduction) in the production potential of 249

Impact of climate change on agriculture

rain-fed cereals, even without considering any technological progress between a baseline around 2000 and the target periods (2050s or 2080s). The global net increase in production potential found under such unrealistic conditions will be due mostly to expansion in suitable land and improvements in land suitability, concentrated in medium and high latitudes, chiefly in the Northern Hemisphere. Suitable land will diminish at some tropical latitudes, especially in areas such as Southern Africa, Northern Africa, Northern Mexico and parts of the US Southwest, due to reduced local precipitation. In any case, the suitable land area that is actually used for crops is limited, leaving a wide margin for worldwide expansion of cultivation if the need arises. Besides the new and improved lands at medium and high latitudes, there are (and there will continue to be) suitable unused (but usable) lands, non-covered by forests or otherwise protected, mostly located in Sub-Saharan Africa and tropical South America (AB 2012). Land scarcity, therefore, will not be a global problem for expansion of crops, either with or without climate change. If anything, there will be regional effects: suitable land may shrink in some areas, and expand in others.

10.3.3. Impact of climate change on agricultural GDP and hunger FAO projections of future agricultural production include an estimation of agricultural GDP as well as estimates of future prevalence of undernourishment and dietary energy supply. The impact of climate change on such aspects has been addressed by some Ricardian models (e.g., Mendelsohn 2000) and has also been included in integrated assessments. We shall review two such analyses, both originating at IIASA - one published in 2002 and the other in 2011. Fischer et al. (2002a). We have already cited Fischer et al. (2002a) for its estimates of climate change impacts on agricultural production potentials and land suitability. Another important result of that study refers to the impact of climate change on agricultural GDP or value added, and on the prevalence of hunger or undernourishment. These results are based on a combination of models (climatic, agronomic, and socioeconomic) in an integrated assessment with various assumptions. The starting point is the GDP of 1990. The published results of Fischer et al. (2002a) do not make explicit the absolute level of total and agricultural GDP by region, which was used as a baseline, nor are projections provided for the future up to 2080, with or without climate change, by region and climate scenario. However, some information is provided in the paper and a table is published

250

Hector Maletta

showing the relative impact of climate change on the expected level of future agricultural GDP. IIASA’s Basic Linked System (BLS) produces estimates of GDP by sector (at base-year prices – in this case 1990 prices in USD, the same baseline used by Mendelsohn 2000) for each of 34 countries or geographical areas included in the BLS. These areas were then aggregated into 10 broad regions, and macroeconomic variables such as GDP were calibrated to four SRES scenarios (A1F1, A2, B1, and B2) and two supplementary scenarios with a reduction of 45% in economic growth rates, A2-S and B2-S.27 This process generates endogenous growth rates for each region under each scenario, for both total and sectoral GDP. The growth rates of the world’s agricultural GDP resulting from the model vary between 1.20% and 1.54% in the different SRES scenarios. Several projections were prepared, using, as explained above, various climate models. The relative impact of climate change on agricultural GDP by 2080 (relative to a reference projection without climate change), as estimated with integrated assessment models by Fischer et al. (2002a), are based on a baseline projection of total and agricultural GDP. At the world level, the 1990 level of world agricultural GDP used by Fischer was 1077 billion USD (at 1990 prices), as reported in another paper by the same team (Tubiello and Fischer 2007:1040). Agricultural GDP in 2080 without considering climate change is expected to be between 2.9 trillion USD (B1) and 3.6 trillion USD (A2). Relative to the baseline, the above figures imply that world agricultural GDP is endogenously estimated to grow at annual rates varying between 1.12% (B1) and 1.37% (A2), in line with the exogenous (and conservative) growth rate employed by Mendelsohn (2000) (1.24%). These rates of growth are well below the rates observed from 1961 to 1990, which hovered around 2.5%. This is all the more notable when it is taken into account that in the two and a half decades since 1990, world agriculture has already grown at a faster rate (over 2.5%), and hence, if it is to reach the values obtained by Fischer et al. (2002a), it should grow much more slowly in the decades remaining until 2080. Therefore, the growth rates of agricultural GDP projected by Fischer et al. 2002 can be regarded, just like those of FAO (2006), AB (2012) and Mendelsohn (2000), as quite conservative. Growth of total GDP in Fischer et al. (2002a) is expected to proceed at the annual rates assumed in SRES scenarios, varying from 1.58% Fischer’s modified scenarios with lower GDP but the same climate change are not internally consistent: if economic growth is reduced by 45%, global warming would also be reduced, since future GHG emissions are a function of projected GDP and its assumed energy intensity. However, those modified scenarios are not used in the present context.

27

251

Impact of climate change on agriculture

to 4.00%. Fischer and colleagues project agricultural GDP under various SRES scenarios, and using various (not strictly comparable) climate models. The only climate model with variants of the main four SRES scenarios was the Hadley Centre HadCM3 model, with the results shown in Table 53. Table 53. Percentage impact of climate change on the agricultural GDP of 2080, relative to baseline projection without climate change, under various SRES scenarios. Projection to 2080 (HadCM3 model)  

A1FI

A2

B2

B1

World

-1.5%

-0.9%

-0.4%

-0.5%

Developed

-0.5%

0.2%

-0.7%

1.1%

7.5%

3.1%

2.7%

7.7%

Europe

-14.7%

-18.0%

-16.9%

-8.1%

Former USSR

-4.9%

-0.5%

1.9%

-0.9%

N. America

-1.9%

-1.2%

-0.3%

-0.9%

-4.9%

-3.7%

-1.6%

-7.0%

LAC

3.7%

1.4%

1.8%

4.1%

Southeast Asia

-3.7%

-5.0%

-3.5%

-0.8%

Centrally planned Asia

-6.4%

-2.1%

-0.8%

-2.0%

-4.3%

-4.2%

-2.8%

-1.1%

Developing Africa

Asia (total) Source: Fischer et al. (2002a:109).

According to this projection, the effect of climate change on the agricultural GDP of 2080 will be negative but small. In the worst-case scenarios (A1F1 and A2) world farm output will be barely 1.5% and 0.9% below a reference projection without climate change. In more benign scenarios (B2 and B1), the world level impact will be even smaller (-0.4% and -0.5% respectively). The effects vary by region. The most notable effects would be in Europe, where farm GDP would be reduced by a percentage between 8% and 18% depending on scenarios (curiously, the greatest reductions would be in the two scenarios based on a more fragmented world with lower economic development and more concern about the environment, i.e., A2 and B2). These impacts are somewhat lower under intensified use of fossil fuels (A1F1) and even lower with less intense use (B1). Other climate models like CSIRO or NCAR agree with the Hadley model on the limited magnitude of the effects, but vary in the details, which suggests these projections are highly uncertain if downscaled to particular regions. Under these conditions, and assuming that the population grows according to the IPCC SRES scenarios, including the explosive population growth underlying the demographically unrealistic A2 scenario, the percentage of the population

252

Hector Maletta

at risk of hunger (conceptually equivalent to FAO’s undernourishment) declines dramatically in all scenarios, with only limited impact from climate change. Table 54 shows the reference projections at the world scale, from 1990 to 2080, with and without the impact of climate change.28 Future prevalence of undernourishment in the absence of climate change will be mostly insignificant, at about 1-2% of the population, well below the 5% significance threshold (i.e., indistinguishable from zero). Even in the unlikely scenario of high population growth and lower economic development in A2, undernourishment will fall to 5.62% - just above the significance level of 5%. Table 54. Projected world undernourishment rates with and without climate change.  Scenarios Population (million) A1 B1 A2 B2 Undernourished population (million) A1 B1 A2 B2* Prevalence of undernourishment A1 B1 A2 B2

1990 Baseline

No CC

2080 With CC

5194 5194 5194 5194

8086 8086 13656 10135

8086 8086 13656 10135

824 824 824 824

108 91 768 233

136 99 890 245

+28 +9 +122 +12

15.86% 15.86% 15.86% 15.86%

1.34% 1.13% 5.62% 2.30%

1.68% 1.22% 6.52% 2.42%

+0.34 % points +0.09 % points +0.90 % points +0.10 % points

Difference due to CC

Source: Fischer et al. (2002a). Projection without climate change based on population in Table 4.1 (p.94), and persons at risk of hunger in Table 4.6 (p.100). Projection with climate change as provided in the text on pp. 112-115. Baseline figures for undernourishment in 1990 (based on FAO-SOFI 2001) do not reflect current versions (FAO-SOFI 2012). CC=Climate change. (*) Absolute figure with CC: estimated from 5% increase mentioned on p.115.29

Figures in Table 54 reflecting impacts of climate change up to the year 2080 refer to simulations with the Hadley Centre model used by Fischer (2002a). Figures that differ only slightly emerge from other models, such as CSIRO. 29 The relevant text is as follows: ‘Concerning people at risk of hunger [in the baseline projections without climate change], there is a substantial improvement observed in scenario B2, from more than 800 million currently to an estimated 350 million people in 2050 and down further to about 230 million in 2080. [...]. For both the HadCM3 and CSIRO climate projections, the estimated number of people at risk of hunger in the 2080s increases by about 5%’ (Fischer et al. 2002a: 114-115, emphasis ours). Curiously enough, Fischer et al. report their projections of undernourishment in a most incomplete manner; projections by region are reported only for reference scenarios without climate change (p. 100). For projections with climate change, only a few figures are offered as text comments, mostly limited to the world level as is the case of the passage cited here. 28

253

Impact of climate change on agriculture

Climate change will have only a small marginal effect upon these projections: according to Fischer (2002a), it will cause prevalence to increase by between 0.09 and 0.90 percentage points, the latter only under the unrealistic A2 population figures. In other words, save for a sudden explosion in demographic rates between now and 2080, undernourishment will practically disappear; even under the improbable population growth of A2, it will decline from the baseline (15.86%) to only 6.5%. The case of scenario A2, unlikely though it may be, actually proves that the world will feed an improbably high population by 2080, almost as well as it would feed the smaller populations envisaged in other scenarios. As the figures reveal, the higher projections of undernourishment for A2 do not result from projections of lower rates of agricultural production but from projections of an improbably high rate of demographic growth. Agricultural projections developed under the A2 climate scenario show a very high increase in output - the biggest of all the scenarios considered, and all of it within suitable agroecological zones not encroaching on forests - but nonetheless involving a considerable decrease in undernourishment by 2080 relative to the baseline year (1990), in spite of the high population growth. This higher output is mainly caused by higher food demand due to the high population of A2. The outcome for A2, then, shows that the world of 2100 could feed over 15 billion people with higher incomes than today, while reducing undernourishment from 15.86% to 6.5% and requiring only a small expansion of cultivated land. In all scenarios, the net impact of climate change on undernourishment is very low: climate change will imply prevalence increases of less than 1%; such small impacts will be practically irrelevant in view of the low level of undernourishment in all scenario projections, and because those undernourishment rates, with or without climate change, are mostly not statistically significant, or at worst, in A2, on the borderline of significance (the threshold of 5% for significance is due to inherent natural variability in energy needs and also to uncertainties and imprecision in the definition and estimation of undernourishment: see FAO 2008). It is worth noting that the expected percentage prevalence reported in Table 54 has been calculated for this text, but does not appear in Fischer et al. (2002a or 2002b): they report only the number of people at risk of hunger, which is misleading because the population is expected to increase in all scenarios, and thus the numbers of undernourished are bound to increase even if the situation fails to deteriorate or climate change does not occur. As 254

Hector Maletta

explained in Section 12.8, percentages (and not absolute numbers) are the correct indicator. In fact, the numbers of the non-undernourished are projected to increase far more than people at risk of hunger, but such absolute figures do not mean much. Fischer (2011) projections of agricultural GDP. In the context of more recent FAO projections of agricultural production to 2050, Fischer (2011) has produced a series of projections to 2050 and 2080, making explicit in FAO’s model the expected impacts of climate change and biofuel production, with updated information relating to his work in previous years. The background projections are based on UN Population projections (2008 Revision) and a projection of GDP growth used for FAO projections. GDP in Fischer (2011) is valued at 1990 prices. According to these baseline projections, the world population in 2050 will be 9.105 billion people and global GDP will be 98,014 billion USD at 1990 prices. Population growth from 2010 to 2050 is projected to increase at 0.72% per year, with a trend toward population stabilisation after 2050. Total GDP, aggregated in PPP terms and according to the baseline projections, is assumed to grow in 2000-2050 at an annual rate of 2.48%, implying that world per capita income will increase at about 1.76% per year over the same period of 40 years. This rate is quite conservative in view of the performance of the world economy in recent decades, even including the slump that started in 2008 and was not expected or considered in the baseline projections. However, this conservative rate is in line with other (also conservative) projections such as those by FAO or the OECD. As a baseline, Fischer (2011) provides a projection of population and agricultural GDP to 2030, 2050, and 2080, for the whole world and major regions, not including climate change or biofuels. His projection to 2050 was based on FAO (2011a) agricultural and population projections (population projected as per the UN 2008 Revision). A projected population for 2080 was also adopted, but not explicitly published by Fischer. We have estimated it based on Fischer’s figures for 2050 and demographic growth rates from 2050 to 2080 as per the UN 2010 Revision (the 2008 Revision did not provide projections beyond 2050). We cannot be sure that these population projections are the same as those used by Fischer for 2080, but they cannot be very different. As regards GDP, Fischer (2011) obtained GDP and agricultural value-added projections for 2080, but he failed to make them explicit, just as he failed to publish his population estimates for 2080. For illustrative purposes, GDP figures for 2080 have been projected here on the basis of trends, assuming that annual growth rates of GDP and agricultural GDP in 2050-2080 will slow down (relative to 2030-2050) by the same proportion 255

Impact of climate change on agriculture

as they are expected to slow down between 2000-2030 and 2030-2050 in Fischer’s baseline projection. For undernourished people in the reference FAO projection, Fischer reports (Fischer 2011:115) that the worldwide baseline number will be 760 million in 2030, 530 million in 2050, and 150 million in 2080. We use these figures to calculate the expected prevalence of undernourishment on those future dates, in the absence of climate change. These baseline projections ignore the possible effects of climate change and assume that no expansion occurs in the use of farmland or crops to produce biofuels. We had to use this convoluted procedure to establish the no-climate-change baseline, because the corresponding baseline figures were not explicitly published in Fischer (2011). Fischer then superimposed the effects of climate change on these baseline projections by means of two well-known climate models (Hadley Centre’s HadCM3 and Australia’s CSIRO), both including CO2 fertilisation (Fischer also provides estimates without considering CO2 fertilization; these are not reproduced here because they are of purely academic interest: the CO2 effect must be taken into account). Fischer uses, for these projections, the A2 scenario of GHG emissions as well as its climate change outcomes, coupled with the updated population projection provided by the UN 2008 Revision; this serves to avoid the exaggerated demographic hypothesis of the original A2 scenario. It is, however, regrettable that unlike Fischer et al. (2002a), Fischer (2011) does not explore the impact of other climate change scenarios, since the socioeconomic assumptions of A2 regarding the future (a fragmented world with slow economic development and no convergence between developing and developed income levels) seem somewhat unrealistic in view of existing trends in globalisation, and robust evidence that the periphery of the world economy is more dynamic than developed countries. With regard to the use of crops to make biofuels, Fischer formulates several scenarios, depending on the percentage replacement of fossil fuels by biofuels and the proportion of first and second generation biofuels. A major feature of these projections is that only liquid fuels for transport are considered. This means that no consideration is given to scenarios of increasing use of solid or gaseous biomass-based biofuels, especially for power generation and heating. The two biofuel scenarios with the most severe consequences in terms of land required and food production forsaken in favour of making biofuels are the so-called TAR-V1 and TAR-V3 scenarios, and especially the former (see box below).

256

Hector Maletta

Two scenarios for future use of liquid biofuels for transport

TAR-V1

Assumes transport energy demand as projected by IEA in the WEO 2008 reference scenario. Assumes that mandatory, voluntary, or indicative targets for biofuel use announced by major developed and developing countries will be implemented by 2020, resulting in approximately twice the biofuel consumption projected in WEO 2008. Second-generation conversion technologies become commercially available after 2015; deployment of second generation biofuels is gradual, reaching, by 2050, about 66% of transport biofuels in the U.S., 50% in other OECD economies, 40% in Russia, China and India, and 20% in other developing countries.

TAR-V3

Assumes transport energy demand as projected by IEA in the WEO 2008 reference scenario. Assumes that mandatory, voluntary, or indicative targets for biofuel use announced by major developed and developing countries will be implemented by 2020. Accelerated development of second-generation conversion technologies permits rapid deployment; 33% and 50% of biofuel use in developed countries from second-generation in 2020 and 2030, respectively.

Source: Fischer (2011:126, Table 3.26). WEO 2008: IEA World Energy Outlook. Only liquid biofuels for transport are considered, excluding (among others) solid crop-based biofuels for power generation or heating.

Fisher (2011) projections of undernourishment. Fischer (2011) produced a series of projections of agriculture, GDP, and undernourishment to 2050 and 2080, incorporating the expected impacts of climate change and the production of biofuels into the FAO model, with updated information relating to his work in previous years (Fischer et al. 2002, 2005; Tubiello et al. 2007). The worldwide results are shown in his estimates of undernourishment, summarised in Figure 64 and Table 55. Figure 64. Worldwide projected prevalence of undernourishment, 2000-2080, including effects of climate change and increasing use of biofuels. Based on FAO agriculture and population projections, A2 GHG emissions scenario, Hadley and CSIRO climate models, and the two ‘worst-case’ biofuel use scenarios (Fischer 2011). 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0%

2000

2030

2050

2080

Baseline FAO

CC Hadley

CC CSIRO

CC Hadley+TAR-V3

CC CSIRO+TAR-V3

CC Hadley+TAR-V1

CC CSIRO+TAR-V1

257

Impact of climate change on agriculture

Table 55. Projected worldwide undernourishment (2000-2080) in Fischer (2011). 1

2

3

4

5

6

7

8

9

Baseline projection (FAO) Population (million) GDP (billion 1990 USD)* Agricultural GDP (billion 1990 USD)* Undernourishment (million) Undernourishment (% prevalence) Impact of climate change (Hadley) Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence) Impact of climate change (CSIRO) Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence) Average of 2 and 3 Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence) CC Hadley + biofuels (TAR-V3)*** Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence) CC CSIRO + biofuels (TAR-V3)*** Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence) CC Hadley + biofuels (TAR-V1)*** Change in undernourishment (million) Undernourishment (million) Undernourishment(% prevalence) CC CSIRO + biofuels (TAR-V1)*** Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence) Average of scenarios 5 to 7*** Change in undernourishment (million) Undernourishment (million) Undernourishment (% prevalence)

2000

2030

2050

2080**

6,047 27,136 1,259 833 13.8%

8,231 62,165 1,836 760 9.2%

9,105 98,014 2,192 530 5.8%

9,752 119,686 2,349 150 1.5%

0 833 13.8%

1 761 9.2%

-3 527 5.8%

28 178 1.8%

0 833 13.8%

24 784 9.5%

4 534 5.9%

19 169 1.7%

0 833 13.8%

13 773 9.4%

1 531 5.8%

24 174 1.8%

0 833 13.8%

82 842 10.2%

39 569 6.2%

40 190 1.9%

0 833 13.8%

108 868 10.5%

46 576 6.3%

32 182 1.9%

0 833 13.8%

148 908 11.0%

99 629 6.9%

55 205 2.1%

0 833 13.8%

176 936 11.4%

104 634 7.0%

48 198 2.0%

0 833 13.8%

129 889 10.8%

72 602 6.6%

44 194 2.0%

(*) At 1990 prices, billion PPP USD. Conversion factor to 2005 PPP USD: 1.395. (**) Population 2000-2050: Fischer (2011, Table 3.1); population 2080: estimated from the figure for 2050, and the 2050-2080 growth rate envisaged in the 2010 Revision of UN Population Projections. Total and agricultural GDP estimated to grow at decelerating rates in 2050-2080, as per projected trends up to 2050 (Fischer 2011, Table 3.2); shown for illustrative purposes only. (***) Combined effect of climate change (A2 climate scenario) and biofuels (TAR-V3 and TAR-V1, the worst for undernourishment among various biofuel-use scenarios). CC=Climate change. Source: Fischer (2011, Tables 3.1, 3.2, 3.5, 3.18, 3.23, 3.26, 3.33 and 3.41, and p.115).

258

Hector Maletta

The prevalence of undernourishment (number of undernourished as percentage of world population) is based on the FAO reference population up to 2050 (Fischer 2011, Table 3.1) and the world population estimates for 2080 used by Fisher; the latter, not disclosed by Fischer (2011), is taken here to be 9752 million, based on FAO’s estimate for 2050 (9105 million) and the annual growth rate for 2050-2080 (0.23%) implied in the 2010 Revision of UN population projections. Once per capita food supply has been projected, the expected prevalence of undernourishment is estimated on the basis of a well-known curvilinear correlation between undernourishment and the degree of adequacy of per capita dietary energy supply (or apparent consumption) relative to minimum dietary energy requirement (i.e., ratio of kcal/person/day to MDER), as explained in the Appendix (Section 12.5.2). As shown in Figure 64, Table 55 and Table 56, these projections of future rates of undernourishment envisage a gradual tendency towards non-significant levels, starting with a worldwide prevalence of 13.8% in 2000 and culminating in a worldwide prevalence between 1.5% and 2.1% in 2080. Prospects may be only slightly altered (mostly for the better) by recent retrospective adjustments in undernourishment estimates (FAO SOFI 2012, 2013, 2014). Climate change and biofuels are thus projected to have a negative but very small impact on the prevalence of undernourishment and its decline during the 21st century, compared to a future without climate change and without expansion of liquid biofuels for transport. A somewhat larger difference between the various climate and biofuel scenarios is expected in the near future, causing undernourishment estimates for 2030 to vary across scenarios from 9.2% to 11.4%, a range of 2.2 percentage points; inter-scenario variability is narrower for 2050 (5.8%-7.0%, or 1.2 percentage points) and minimal for 2080 (1.5%-2.1%, or 0.6 points). Notably, the integrated assessment models used by Fisher, even if applied to a high GHG emission scenario such as A2, suggest that the inclusion or exclusion of climate change and biofuels makes little difference to the end result concerning undernourishment: all variants are close to each other and well within their own margins of uncertainty. This is due, on the one hand, to the consideration of agriculture as an adaptive system, whereby farming adapts to changes in the environment that are distributed over two or more generations of farmers, and on the other hand, data on the availability of suitable usable land (AB 2012) that is projected to be tapped whenever projected productivity growth is unable to cope with projected increasing demand.

259

Impact of climate change on agriculture

Fischer’s estimates of climate change impact are based only on the A2 emission scenario, which imply high carbon emissions and lower economic development in a fragmented world. In this model, the various biofuel hypotheses are (somewhat paradoxically) not allowed to influence the A2 emission pathway, based on the use of fossil fuels alone. In fact, under increased biofuel use, emissions and warming should be lower. However, the resulting estimates strongly suggest that climate change and biofuels will have only a limited impact on the generally declining trend in the prevalence of undernourishment, which will in all cases lead to a world with very low prevalence of hunger by 2050-2080. Indeed, the effect of climate change alone, projected by either Hadley or CSIRO models and not considering biofuels, is practically nil. Almost the entire projected impact, especially in 2030 and 2050 but not in 2080, comes from projected use of biofuels, because the only biofuels considered are those based on food crops (ethanol and biodiesel); the effect of climate change alone is not significant. If biofuels were not based on short-cycle food crops but on crop residues or permanent non-food crops, or on relatively marginal land not producing significant amounts of food crops or livestock food products, the impact will be much lower. The effect of biofuels, however, is almost irrelevant by 2080, even in projections considering only food-crop-based biofuels grown on existing cropland (i.e., not considering the addition of new cropland). Recent updates to undernourishment estimates (FAO SOFI 2014) indicate that the world rate had already fallen to 11.3% in 2012-2014, and by 2015 (or shortly thereafter) is likely to reach the MDG target of 9.2% (half the 1990-1992 rate); it is then unlikely to remain at 9.2% by 2030 as per Fischer’s forecast without climate change, which is based on older, higher and rather stagnant undernourishment estimates. In the light of the more recent FAO SOFI figures, Fischer’s 2011 projections seem slightly overstated. Even so, per Fischer’s 2011 estimates and including the effects of climate change and biofuels, world undernourishment will be just 6.6% in 2050 and clearly non-significant (maximum 2%) by 2080. Taking into account the more recent SOFI estimates, this situation will come about much earlier. Fischer (2011) also offers prospects for each major world region, but only to 2050. Unfortunately, regional figures for 2080 are not explicitly published in Fischer (2011). Some regional figures for 2080 have been added here (Table 56) based on Fischer’s worldwide results for that year and indications in his text on the impact in absolute numbers for some of the regions (Fischer 2011, Table 3.18). The regional prospects (Table 56) echo the worldwide decline shown in Table 55. 260

Hector Maletta

Table 56. Prevalence of undernourishment, 2000-2080, including impact of climate change (A2 emission scenario) and biofuels (TAR-V3 scenario).   Baseline projection (FAO-00 of IIASA) World total Sub-Saharan Africa Asia (developing) Rest of the world Climate change (A2), no biofuels effect  World total Sub-Saharan Africa Asia (developing) Rest of the world Climate change (A2) and biofuels (TAR-V3)  World total Sub-Saharan Africa Asia (developing) Rest of the world

2000

2030

2050

2080*

13.8% 29.9% 17.0% 4.5%

8.4% 21.2% 8.3% 2.7%

5.0% 13.9% 3.7% 1.9%

1.5% 3.6% 1.1% 0.6%

13.8% 29.9% 17.0% 4.5%

8.5% 21.2% 8.6% 2.7%

5.0% 13.9% 3.7% 1.9%

1.8% 4.3% 1.2% 0.8%

13.8% 29.9% 17.0% 4.5%

9.5% 24.0% 9.4% 3.0%

5.5% 15.2% 4.0% 2.1%

1.9% 4.5% 1.4% 0.8%

Source: Estimated from Fischer (2011, Figure 3.8(B), Tables 3.1 and 3.18, and p.115). Scenario A2 used only for emissions and economic development, but with the population of the FAO baseline. (*) Regional distribution of impact for 2080: estimated from world total, the percentage distribution of impacts by 2050, and indications about the impact in absolute numbers for some of the regions in Fischer (2011), Table 3.18). Population in 2080: projected from Fischer reference projection to 2050 (his Table 3.1) and growth rates 2050-2080 from the UN 2010 Revision of population estimates, which closely agrees with the projection implicitly used (but not published) by Fischer.

Undernourishment in Sub-Saharan Africa, according to these projections, will still be significant in 2050, but nonetheless will be less than half that of 2000, with Fischer’s estimates for those years falling from 29.8% to 13.9% (or 15.2% if liquid biofuel scenario projections are included). By 2080, the figures will fall lower than 5% and therefore will be non-significant by FAO standards due to the inherent variability of individual dietary energy needs and other uncertainties. Further decreases in 2050-2080 are expected in all regions under these scenarios, especially in Sub-Saharan Africa, since we know that the overall prevalence by 2080 is projected to be about 2% at the world level, implying about 3% for developing countries. Most regions were already expected to have low absolute undernourishment figures by 2050; the only regions with numbers large enough in 2050 to make a significant contribution to the anticipated reduction in 2050-2080 are Sub-Saharan Africa and South Asia. Fischer does not publish regional results for 2080, but available figures in his report suggest Sub-Saharan Africa in 2080 will have about 4.5% prevalence and Asia about 1.2-1.4% (with South Asia about 3%, and negligible levels in other developing regions). The developing world as a whole will fall from 6.6% in 2050 to about 3% in 2080, and the world total to about 2% as seen 261

Impact of climate change on agriculture

before; all these figures are below the 5% significance threshold estimated by FAO for its undernourishment indicator. According to FAO’s methodology, taking into account the inherent inter-individual variation in food needs and a possible lack of precision in the data, any prevalence below 5% is statistically indistinguishable from zero. In summary, Fischer (2011) projects a small overall effect of climate change and biofuels on the future prevalence of undernourishment relative to a baseline projection not considering such factors. These prospects are similar (and slightly better for Sub-Saharan Africa) than those obtained previously by Fischer et al. (2002a and 2002b). Most of the small projected effect will be caused by projected production of liquid biofuels based on food crops and not by climate change (besides using land suitable for food production, cropbased biofuels also push prices up, thus affecting economic access to food). After 2050, the bulk of the remaining undernourished will be in South Asia and Sub-Saharan Africa; however, the above projections for 2080 imply very low prevalence even in these regions (4.5% in Sub-Saharan Africa, 1.9% in Asia as a whole, and probably about 3% in South Asia).

10.3.4. Impact on cropland use Considering only climate change, without including any impact from cropbased biofuels, Fischer et al. (2011) estimate that cultivated agricultural land in 2080 is expected to expand by 10m Ha (Hadley) or 13m Ha (CSIRO) due to climate change in comparison with the referenced FAO no-climate-change scenario, more or less evenly divided between developed and developing countries (Table 57). Fischer (2011) uses ‘cultivated land’ as equivalent to FAO’s ‘arable land’, i.e., including land lying fallow. The harvested area (which equals arable land, minus land in fallow, plus land under multiple cropping) will increase by 14m Ha under both the Hadley and CSIRO models: see Fischer (2011:117), Tables 3.19 and 3.20. These figures exclude permanent crops. Table 57. Expected impact of climate change on total cultivated land and harvested area by 2080, in millions of hectares, compared to FAO-model baseline projection without climate change. Biofuels ignored. Climate model used Developed countries Developing countries World total

Cultivated land (million Ha) Hadley CSIRO +5 +6 +5 +7 +10 +13

Totals may not coincide due to rounding. Source: Fischer (2011:117).

262

Harvested area (million Ha) Hadley CSIRO +6 +6 +8 +7 +14 +14

Hector Maletta

These expected net increases in cropland due to climate change represent less than 1% of the otherwise expected amounts of land that will be cultivated or harvested in 2080, projected to be about 1.8 and 1.6 billion Ha respectively in the no-climate change scenario for 2050 (Fischer 2011: 104) and 2080 (though figures for the latter year are not explicitly reported by Fischer). The expansion of cultivated or harvested land also will be in the order of 1% or less in both developed and developing countries, much below the net expansion in suitable land (suitable for rain-fed crops) which will increase by 9% globally. This small net effect on cropland use worldwide is, however, the result of somewhat larger negative or positive shifts in specific areas. In particular, medium-high latitudes in the Northern Hemisphere (North America and much of Eurasia) will expand more in relative terms, though always moderately. In any case, expected growth in agricultural production, with or without climate change, is not expected to result from a significantly larger area of land devoted to crops, but mostly from increasing yields and crop mix changes, as in past decades. If biofuels are considered, the expected expansion in crop area will be larger, though still representing a small percentage increase. In the TAR-V1 scenario, total cultivated land (which will increase by 10-13 million Ha as a result of climate change alone) will expand by an additional 50 million Ha due to the land requirement of biofuels; thus, total additional agricultural land will be 59-63 million Ha. These projections are for the TAR-V1 biofuel scenario which has the greatest land requirements; the additional land required by biofuels in TAR-V3 will be some 30 million Ha. Harvested area, in turn, will increase by 76m Ha in TAR-V1 and by 46m Ha in TAR-V3 (Fischer 2011:147). These additional increases dictated by biofuel scenarios will expand cultivated or harvested land by a further (but small) percentage, about 2-4% relative to the reference projection without climate change (1.6 to 1.8 BHa). As noted above, unused and available crop-suitable land (about 1.4 billion Ha more than actually used, or projected to be used, for crops) will permit a much larger expansion if need be. Part of this anticipated increase in land use, due to climate change and biofuels, will not entail an expansion of arable land but only increased cropping intensity (multiple crops per year, or reduced periods of fallow), as has been occurring in the past. Thus, the actual increase in cropland will be lower. Fischer’s estimation includes only liquid biofuels for transport, but the use of extra cropland for biofuels may be reduced by substituting solid biofuels 263

Impact of climate change on agriculture

from lignocelullosic crops (mostly grasses and shrubs) grown on marginal land (Campbell et al. 2008) and crop residues, none of which require any more cropland. In summary, as per Fischer (2011), the total net increase in cropland due to climate and biofuels together (about 60-75 MHa depending on the scenario) will be about 4% of total cultivated land, compared with a scenario for 2080 without climate change and without more biofuels. Such an increase will not be unfeasible: as seen above, one half of all land suitable for rain-fed crops is currently not used for crops. In addition, Fischer estimates that land suitable for rain-fed cultivation will increase by 9% as a result of climate change, even when possible increases in irrigation and cropping intensity are discounted. The margin of unused suitable land will thus actually increase even after the additional land required by climate change and biofuels is accounted for. As in previous instances, it is clear that projected agricultural production, which involves a drastic reduction in undernourishment while feeding more than 9 billion people in the second half of this century - and without even reducing food supplies demanded due to overconsumption and obesity - will not imply much increase in cropland, even after considering the impact of climate change and biofuels. In all cases, it should be noted, the only land areas taken into consideration are Prime and Good land, located within adequate agro-ecological zones, and not encroaching into protected areas or marginal land. Part (and possibly most) of the increase in cropland may come from existing agricultural land, i.e., through conversion of grazing land into cropland, without requiring much of a net reduction of current non-farm land (the net result, however, may include some local expansion onto non-farm land, partly or totally offset by changes in the opposite direction elsewhere). Because the importance of crop-based fodder in livestock sustenance is bound to increase (as it did in recent decades), the slight anticipated expansion in cropland will be largely offset by declining use of grassland for livestock, even if livestock production is expected to increase in line with observed trends towards more intensive livestock feeding and a reduction in grazing. Thus, total farmland will expand to an even lesser degree than cropland; or not at all, as has already been the case since the late 1980s or early 1990s. Crop-suitable land will expand worldwide due to climate change, as large expanses of land become crop-suitable in temperate zones, making up for losses of crop-suitable land that may occur chiefly in parts of the tropics where precipitation is projected to decline. 264

Hector Maletta

All these projections, as explained above, are conditional on a number of assumptions built into the scenarios considered - assumptions which for the most part are conservative or pessimistic. For instance, biofuel scenarios are based on the extensive use of annual crops for liquid biofuels, and do not take into account the possibility of increasing use of permanent crops (e.g., cultivated grasses or shrubs) grown on marginal land, especially for solid biofuels. As another example, climate change impacts are based on unlikely scenarios of slow development with a large population in a fragmented world (e.g., IPCC’s A2 scenario), whereas in reality the world is on a path of rapid globalised development with slower population growth, especially in developing countries. The effect of CO2 on vegetation (fostering photosynthesis and economising of water) have been considered for the case of crops, but not for the case of pastures, although satellite evidence indicates a global ‘greening’ of arid or semi-arid zones since 1980, attributable to increased concentrations of atmospheric carbon dioxide (Donohue et al. 2013); the same effect of increased vegetation, albeit less detectable from satellites, must be also present in pastures outside arid or semi-arid zones.

10.3.5. Climate change, water, and irrigation The integrated assessments developed by IIASA and others that have been reviewed above have mostly concentrated on the effects of climate change on crops, and especially on rain-fed crops, factoring in climate change in temperature and precipitation but not making explicit the effects on water availability, nor incorporating prospects for irrigation. FAO projections anticipate a moderate expansion of irrigated land in line with previous trends and country-specific prospects. This should be complemented with analyses of the effect of climate change on water needs and water availability. As discussed earlier, climate change will entail a global increase in both temperature and precipitation. Warming effects will occur everywhere, but this will not be the case with rainfall: atmospheric convection, ocean circulation and related phenomena will cause some land areas to get drier and other areas to get wetter. Seasonal rainfall patterns, i.e., the distribution of annual rainfall over the year, may also change: some areas may experience increased rainfall during one season and reduced rainfall in another. Existing climate models generate projections on the amount and distribution of rainfall, although the uncertainty of the projected geographical distribution of rainfall is considerable. However, there is (for certain areas of the globe) widespread agreement between different scenarios, different models, and different versions of the models. 265

Impact of climate change on agriculture

Temperatures and water affect both rain-fed and irrigated agriculture through several causal pathways. Elevated temperature alone will increase evapotranspiration, and thus cause an increase in the water requirements of plants. On the other hand, local rainfall may increase or decrease. If it decreases, plants adapted to the previous rainfall regime will be under higher stress, possibly diminishing their yields. If it increases, the effect will be beneficial overall but in certain areas could cause problems of its own, in the form of excess flooding. Besides, increased CO2 concentrations reduce the water requirements of plants, especially those of the C4 type, like maize or sugar cane. Crops receive water directly from rainfall or from elsewhere, through natural runoff or various types of human-made irrigation systems that convey the water from sources such as lakes, rivers, springs, or underground aquifers. Part of the water withdrawn from these sources via an irrigation system may be lost along the way through evaporation or ground infiltration, with the result that only part of the water withdrawn actually reaches the crops. The ratio of water actually used by crops to water withdrawals at the source defines the irrigation system’s efficiency. This leads to the distinction between net and gross irrigation water needs: net needs are equivalent to evapotranspiration; and gross needs also include water losses occurring anywhere in the system. Cropping intensity is also important: each crop has its own needs, and thus double or triple cropping (two or three successive crops planted and harvested per year on a single area of land) will greatly increase the amount of water needed for that particular area. In addition, some areas equipped for irrigation are not actually used due to various factors (lack of water at the source; silting of reservoirs or canals; intentional or unintentional water diversion away from the fields; malfunctioning or disrepair of gates, sprinklers, or other irrigation infrastructures and equipment; and so on). To assess the future of irrigated agriculture, including impacts of climate change, it is first necessary to assess the multiple factors involved: ŒŒ Changes in the extent of land area equipped with irrigation. ŒŒ Changes in the mode of irrigation, e.g., gravity-driven vs. pressurised; earthen and open vs. lined and covered canals for water conveyance; flooding, sprinklers, drip, or other delivery modes. ŒŒ Changes in the land area that is actually irrigated, within the area equipped with irrigation systems.

266

Hector Maletta

ŒŒ Changes in the average efficiency of irrigation systems, which affect the withdrawal required in order to cover the net water needs of irrigated crops. ŒŒ Changes in cropping intensity of the area under irrigation (total area of irrigated crops harvested per year, divided by physical area actually irrigated, or equipped with irrigation systems). ŒŒ Changes in net water needs of crops caused by warming (higher evapotranspiration). ŒŒ Changes in net water needs of C4 crops caused by elevated CO2 concentrations. ŒŒ Changes in the amount of local rainfall directly falling on the crops. ŒŒ Changes in irrigation water availability at the source caused by changes in rainfall over the catchment basin (including, in some cases, changes in water supply from glaciers) The assessments of climate change impacts on agriculture that have been reviewed thus far say little or nothing about irrigation. The area under irrigation is usually assumed (explicitly or implicitly) to remain constant, as is its efficiency. Indeed, many Ricardian models do not consider irrigation at all. Some of them calculate different predictive equations for irrigated and rain-fed land, taking the presence or absence of irrigation as a given, i.e., assuming that each piece of land is (now and forever) either rain-fed or irrigated - essentially the same assumption underlying Fischer et al. (2002a). Other Ricardian models (e.g., some of those in Mendelsohn and Dinar [2009]) treat irrigation as a choice: it is assumed that farmers facing different climates will choose whether or not to employ irrigation, and this assumption underlies comparisons of land value or farm revenue under different climates and varying uses of irrigation. In both cases, there is no distinction between, on the one hand, facultative (and generally exogenous) public investments in basic infrastructure for irrigation or drainage systems, such as dams, reservoirs and primary canals, and, on the other hand, farm-level (possibly endogenous) decisions to invest in on-farm (or perhaps community-level) irrigation. In general, the cost of irrigation is not addressed. Besides the effect of water as such, turning some areas drier or wetter, it is also necessary to consider the effect of elevated carbon dioxide on the water requirements of C4 plants such as maize. Even in places where climate change may reduce the supply 267

Impact of climate change on agriculture

of water, elevated CO2 in the atmosphere may (at least partially) offset that reduction by diminishing the water needs of these crops. An adequate assessment of these problems should enquire as to: (1) whether crops in the future will demand more or less irrigation or drainage at different places, due to increased temperature, lower or higher precipitation, or higher atmospheric concentrations of carbon dioxide; (2) whether any additional irrigation water will be available at the various locations where it will be needed; and (3) the costs and benefits involved in any required increases in available water, irrigated area, irrigation efficiency, or other relevant factors. Some models have investigated the availability of water under climate change (Arnell 2004; Alcamo, Döll et al. 2003; Vorosmarty et al. 2000). A few authors have addressed the question of the impact of climate change on irrigation water demand, including Döll and Siebert (2001) and Shen et al. (2008). Demand may be expressed in terms of the net water requirements of crops, or in terms of gross withdrawals required at the source, which cover not only net crop requirements but also water losses due to irrigation inefficiencies. Fischer et al. (2007) is one of the few studies including an integrated assessment of climate change on irrigation demand, both gross and net, irrigation efficiency, and the cost/benefit implications. Climate change impact on irrigation water requirements and availability Döll and Siebert (2001) developed a global irrigation model by integrating simplified agroecological and hydrological approaches. Petra Döll (2002) uses this framework to investigate global impacts of climate change and variability on agricultural water irrigation demand by comparing the impacts of current and future climate on net irrigation needs, and her results are expressed in these terms. She explains: ‘Gross irrigation requirement’ is the total amount of water that must be applied by irrigation such that evapotranspiration may occur at the potential rate and optimal crop productivity may be achieved. Only part of the applied water is actually ‘used’ by the plant and evapotranspirates; this amount, the difference between the potential evapotranspiration and the evapotranspiration that would occur without irrigation, is the ‘net irrigation requirement’. The other part of the added water serves to leach salts from the soil, leaks or evaporates unproductively from irrigation canals, or runs off; this amount depends on irrigation technology and management. The ratio of the net irrigation water requirement and the total amount of water that needs to be withdrawn from the source, the gross irrigation requirement, is called ‘irrigation water use efficiency’. (Döll 2002:271)

268

Hector Maletta

Thus, if plants require 100 units of water, more than 100 units have to be taken from the source due to irrigation inefficiencies. Traditional irrigation methods such as flooding are less efficient than modern methods such as drip or sprinkler irrigation; water conveyance through pipes or cement-lined canals reduces water loss along the way (to evaporation or ground infiltration). These factors are absent from Döll’s analysis of net irrigation needs (measured at crop level). To associate projected climate change with local precipitation, Döll proceeds as follows: she uses an existing dataset with estimated precipitation from 1901 to 1995 on a 0.5° by 0.5° grid of the planet’s surface; two climate models (ECHAM4 and HadCM3) are retrospectively applied to 1950-1979 and 1960-1989 in order to calibrate expected precipitation levels associated with projected climate; by interpolation, those climate model results are downscaled from their own (coarser) spatial resolution to the 0.5° by 0.5° grid. The baseline climate for her study is the 1950-1979 mean (ECHAM4) or 1960-1989 (HadCM3). The two models were down-scaled to a grid of 0.5° by 0.5° cells (about 60 by 60 km each) to fit the water requirement map of Döll and Siebert (2001). Net irrigation needs were estimated for the baseline climate of either model and for two future periods: 2020-2029 and 2070-2079. Only projections to 2070-2079 are reproduced here for the sake of brevity; however, it is worth noting that projected changes in irrigation needs occur mostly in the first half of the century considered. Petra Döll found that changes in precipitation, combined with increases in evaporative demands, increased the demand for water by crops in irrigated areas worldwide in 1995, with small relative changes of about +5.5 to +7.8% by the 2070s at the world level (depending on the model used) and an average impact of +6.6%, as well as somewhat larger than average impacts, about +15%, in Southeast Asia and the Indian subcontinent - mostly due to the greater requirements of rice. This assessment of increases in water needs at crop level, derived from changes in local precipitation, reveals nothing about the availability of additional water at the source of irrigation systems or any expected change in the efficiency of its conveyance to the crops (estimated at a world average of 0.45 in 1995), or any change in the crop mix grown in irrigated areas. It also omits reference to any expansion of irrigated areas in the same or other regions and assumes that cropping intensity (crops per year) is unchanged. These factors, however, are important components of the final effect on crop production. However, the outcome of Döll’s work suggests that even if irrigation efficiency and crop mix are unchanged, the amount of water required by irrigated crops will increase by only a small percentage 269

Impact of climate change on agriculture

(6.6% on average for the two models used): at the baseline efficiency of 0.45, increasing water use by crops by 6.6% translates into a 3% increase in water withdrawals. Only a modest increase in average efficiency (to, say, 0.48) will make up for such an increase in net crop needs. The above refers to changes between the baseline climate and the 2070s, but an intermediate period (2020-2029) is also estimated. Net crop water needs are projected to increase from the baseline to the 2020s by 3.3% in the ECHAM4 model and 5.1% in HadCM3. The further percentage increase from the 2020s to the 2070s is 2.1% and 2.6%, respectively, which means that most of the impact occurs earlier: as climate change continues throughout the century, the effect on water needs tends to increase more slowly. Taking the average of the two models, two-thirds (4.2%) of the projected increase (6.6%) occurs from the baseline up to the 2020s, and one-third (2.3%) in the remaining half century up to the 2070s. The ECHAM4 model projects more global precipitation than the Hadley Centre model, thus generating a lower increase in irrigation water needs. Since the impact of climate change on the amount and regional distribution of future precipitation is affected by uncertainties, it is appropriate to use the average of the two models as presented here in Table 58. It should be noted that regional irrigation efficiency varies between 0.45 and 0.70 everywhere but in East and South Asia, where it is 0.35 in both cases due to the low efficiency prevailing in paddy fields. These two Asian sub-regions represent 45% of all irrigated areas in the world. Anticipated percentage changes in net water requirements will translate into the same percentage increase in gross irrigation water withdrawals, but only if irrigation efficiency is constant. However, irrigation efficiency cannot be assumed to remain constant if the growing prevalence of modern methods of irrigation is taken into account (especially those based on pressurised sprinkler or drip irrigation). Döll’s (2002) study is restricted to areas that were equipped for irrigation in 1995, and does not inquire into expansions of irrigation after that year. FAO projections of agriculture up to 2050 (as in FAO 2006, FAO 2011a, and AB 2012) do anticipate some expansion in irrigated areas, which is consistent with past trends. Moreover, cropping patterns are also extremely simplified in Döll’s analysis, which distinguishes only between rice and non-rice crops. In addition, the effects of CO2 fertilisation (which reduces plant demand for water) were totally ignored, which is regrettable in view of the known effects of CO2 on C4 plants’ water requirements. These limitations render Döll (2002) a rather rough approximation of the effects of climate change on irrigation requirements. Any effect of CO2 fertilisation (which cannot be ignored) or any increase in irrigation efficiency (which is 270

Hector Maletta

only to be expected) will lead to a reduction in projected gross irrigation requirements, and will thus reduce the estimated impact of climate change on irrigation. The increased water needs estimated by Döll might be covered by a relatively small increase in efficiency. With an average efficiency of 0.45, an increase of some 6-8% in net water needs could be accommodated by a small increase in efficiency to 0.48-0.49. Such an increase in world average efficiency to the small degree required in Table 58 is not difficult to achieve, nor unlikely to happen of its own accord, but is not equally feasible in all regions. Some increase in efficiency may already have been achieved since the baseline period of Döll and Siebert (1995), at least in some areas. Though there may be some old irrigation systems that have extremely low efficiency which may even be declining further due to silting at reservoirs and canals or land salinization, among other problems, in general, irrigation is gaining in efficiency as a result of the increasing prevalence of pressurised systems using sprinklers and drip irrigation. There are even advances in water management efficiency that are expected to lead to gradual improvements in paddy cultivation: for instance, replacing traditional transplanting with wet-seeding significantly improves water efficiency (Bhuiyan et al. 1994); the use of water tubing with multiple side inlets also saves a significant amount of water (Vories et al. 2005); and sprinklers are an important improvement for more efficient irrigation of rice (Vories et al. 2010). Tubing and sprinklers are more capitalintensive than wet-seeding, and are thus more suitable for commercial than for subsistence farmers. The gradual expansion of these systems in Asia is expected to improve the average water efficiency of rice irrigation on that continent.

271

Impact of climate change on agriculture

Table 58. Impact of climate change on net irrigation water requirements of land irrigated by 1995. Region Canada

Net irrigation water needs in past and future climates (km3/year) Irrigated Irrigated Irrigation 2070-79 Percentage change area 1995 cropping efficiency 1961-90 (000 Ha) intensity Baseline ECHAM4 HadCM3 Mean ECHAM4 HadCM3 Mean 710

1.0

0.70

2.4

3.3

2.9

3.1

37.5%

20.8%

29.2%

23,560

1.0

0.60

112.0

123.0

117.9

120.5

9.8%

5.3%

7.5%

Central America

8,020

1.0

0.45

17.5

18.1

19.7

18.9

3.4%

12.6%

8.0%

South America

9,830

1.0

0.45

26.6

28.2

29.1

28.7

6.0%

9.4%

7.7%

Northern Africa

5,940

1.5

0.70

66.4

56.0

57.7

56.9

-15.7%

Western Africa

830

1.0

0.45

2.5

2.4

2.6

2.5

-4.0%

4.0%

0.0%

Eastern Africa

3,580

1.0

0.55

12.3

14.5

14.3

14.4

17.9%

16.3%

17.1%

Southern Africa

1,860

1.0

0.55

7.1

6.4

7.2

6.8

-9.9%

1.4%

-4.2%

OECD Europe

11,800

1.0

0.55

52.4

56.5

57.8

57.2

7.8%

10.3%

9.1%

Eastern Europe

4,940

1.0

0.50

16.7

19.7

22.1

20.9

18.0%

32.3%

25.1%

Former USSR

21,870

0.8

0.60

104.6

104.4

108.7

106.6

-0.2%

3.9%

1.9%

Middle East

18,530

1.0

0.60

144.7

126.5

137.8

132.2

-12.6%

-4.8%

-8.7%

USA

-13.1% -14.4%

South Asia

73,460

1.3

0.35

366.4

410.7

422.0

416.4

12.1%

15.2%

13.6%

East Asia

49,250

1.5

0.35

123.8

131.3

127.1

129.2

6.1%

2.7%

4.4%

Southeast Asia

15,440

1.2

0.40

17.1

30.4

28.6

29.5

77.8%

67.3%

72.5%

Oceania

2,610

1.5

0.70

17.7

18.2

19.7

19.0

2.8%

11.3%

7.1%

Japan

2,700

1.5

0.35

1.3

1.4

1.5

1.5

7.7%

15.4%

11.5%

World

254,910

1.0

0.45

1091.5

1151.0

1176.8 1163.9

5.5%

7.8%

6.6%

Source: Döll (2002:282). Irrigation efficiency by region: Döll and Siebert (2001, Table 1). World averages of cropping intensity and irrigation efficiency added (calculated as weighted averages). Averages of the two climate models have also been added.

Increased water needs require extra water available for withdrawal. Projections of irrigation water availability have been produced by several studies, including Shen et al. (2008); Alcamo, Flörke and Märker (2007); Arnell (2004); Alcamo, Döll et al. (2003a, 2003b), and Vörösmarty et al. (2000). However, the study of water needs still needs to be improved. Fischer, Tubiello, Velthuizen and Wiberg remark:

272

Hector Maletta

Yet much remains to be done to improve the predictions of future irrigation water requirement in agriculture. First, biophysically and agronomically based hydrology computations, such as those used by Döll […], should be performed within a spatially detailed agro-ecological zone (AEZ) assessment model, so that water-demand estimates are consistent with predictions of crop biomass production and yield. Second, because many interactive processes determine the dynamics of crop production beyond agroclimatic conditions [Fischer et al. (2005)], studies that focus on irrigation water should also include, apart from climate change, the impacts of socio-economic scenarios (Fischer et al. 2007:1085).

Integrated assessment of climate change, agriculture, water needs, and irrigation Shen et al. (2008) estimate future water withdrawals for agricultural, industrial, and domestic uses, based on detailed projections for each of the main SRES scenarios. They include a projection of irrigated areas, based on past correlations with population and other variables, and also project improvements in water use efficiency (Table 59). Agricultural withdrawals will increase differently across scenarios, depending not so much on their climate as on their respective population projections. In particular, the A2 scenario, which assumes unaccountably high (and ongoing) population growth, would lead to agricultural water use increasing at 1.22% per year from 2000 to 2075, while in other scenarios the rate of increase would be a more modest 0.41-0.66%. Moreover, agricultural water withdrawals would decrease from 2055 to 2075 (and presumably also afterwards) in the A1 and B1 scenarios where population peaks at mid-century and then declines. The importance of agriculture in total water use decreases from 70.5% in 2000 to 33-55% in 2075, while other forms of water use increase at yearly rates of 1.56% to 2.57%, reflecting increased urbanisation, industrialisation, and domestic demand. The results produced by Shen and colleagues imply higher withdrawals in agriculture than estimated by Alcamo, Flörke and Märker (2007), especially because the latter do not include expansion of irrigated areas. Instead, Alcamo et al. predict somewhat higher domestic withdrawals.

273

Impact of climate change on agriculture

Table 59. Water withdrawals implied by SRES scenario, 2000-2075 (km3/year)   Agricultural use 2000 2025 2055 2075 All uses 2000 2025 2055 2075 % agriculture 2000 2025 2055 2075 % annual growth 2000-2075 Agricultural use Industrial and domestic uses All uses

A1B

A2

B1

B2

183.0 237.2 259.9 248.6

183.0 267.4 378.0 454.3

183.0 237.2 259.9 248.6

183.0 241.0 286.2 298.7

259.4 437.0 654.6 759.7

259.4 469.4 708.5 877.0

259.4 567.1 678.0 532.6

259.4 366.4 483.5 542.3

70.5% 54.3% 39.7% 32.7%

70.5% 57.0% 53.4% 51.8%

70.5% 41.8% 38.3% 46.7%

70.5% 65.8% 59.2% 55.1%

0.41% 2.57% 1.44%

1.22% 2.31% 1.64%

0.41% 1.77% 0.96%

0.66% 1.56% 0.99%

Source: Shen et al. (2008:24, Table 5).

An important study on water demand in relation to climate change and agricultural requirements is Fischer et al. (2007). Their work is an extension of the integrated models used to estimate the future of agriculture under climate change (Fischer et al. 2002a, 2002b, 2005), and is intended ‘to improve, within a coherent AEZ framework, estimates of irrigation water requirements under current and future decades brought about by changes in both climate and socio-economic conditions. As part of this methodology, regional renewable water resources were estimated as a function of precipitation and evapotranspiration’ (Fischer et al. 2007:1085). It is important to note that in this case, the focus of Fischer and his colleagues was not on the impact of climate change on irrigation, but on evaluating the savings in water and irrigation efforts that could be achieved through mitigation of climate change. Therefore, their main emphasis was on comparing two future scenarios: one characterised by more intense climate change, and another reflecting strong mitigation of emissions. They used a single socioeconomic scenario (A2r) prepared specifically for this purpose (Grübler et al. 2006; Riahi et al. 2007). The socioeconomic specifications of the original SRES A2 scenario were revised to reflect recent changes in the demographic outlook for world population growth. In this revised A2r scenario, world population reaches about 12 billion by 2100 (as opposed to 15 billion in the original SRES A2 274

Hector Maletta

scenario). There is still a ‘delayed fertility transition’ that is also mirrored in a delayed economic development catch-up, resulting in an initial stagnation in and a subsequent very slow reduction in income disparities. A2r population projections, as a revised version of A2, are certainly less outrageously off-themark than the original SRES A2 scenario, but are still clearly at odds with recent trends in fertility, recent considerable reductions in world poverty, increasing integration of the world economy, and faster growth on the periphery than the core of the world economy. Fischer and associates ran the socioeconomic A2r scenario under three climate hypotheses: (1) no climate change; (2) the strong warming envisaged in the original SRES A2 scenario; and (3) the mitigated climate change anticipated in the SRES B1 scenario. The difference between the latter two is used by the authors to estimate the benefits of mitigation as regards irrigation. However, comparing any of these two climate-change hypotheses with the baseline situation of no climate change is also useful for assessing the implications of (mitigated or unmitigated) climate change on water supply and irrigation. Estimates in this study by Fischer and colleagues were generated at decennial steps from 2000 to 2080, for each region included in the IIASA’s Basic Linked System (BLS) world food model. Fischer et al. based their analysis on the work at IIASA described in Grübler et al. (2006) and in Riahi et al. (2007), developing three revised scenarios based on the climate change pathways of the A2, B1, and B2 scenarios, downscaled to a grid of 0.5° by 0.5°. The main indicator used by Fischer et al. (2007) is the Water Scarcity Index (WSI), defined as the ratio of gross agricultural water withdrawals (AWW) to internal renewable water resources (WRI) by region. Available water resources are assumed to remain constant under a no-climate-change scenario (the A2r reference) but to change over time under climate change simulations. These simulations, under two hypotheses of global warming (A2 and B1), were prepared by means of two different climate models: Hadley and CSIRO, developed by the Hadley Centre at the University of East Anglia (UK) and the Industrial Research Organisation of Canada, respectively. The WSI represents the proportion or percentage of available water that is withdrawn for agricultural use. Other additional shares of renewable water resources are withdrawn for domestic or industrial use. There is no definite threshold of acceptable levels for the WSI, but the authors suggest that levels above 40% should be considered as worrying, as they would indicate heavy use of water for agriculture that may interfere with other uses. Internal renewable water resources (WRI) are estimated using data from AQUASTAT, the FAOSTAT section on water and irrigation; 275

Impact of climate change on agriculture

these resources totalled 43,006 Gm3/year at the world level for the year 2000 (Fischer et al. 2007: 1084). The corresponding amount was projected up to 2080, but the resulting projections vary in opposite directions depending on the climate model used. World water resources decrease if climate change is simulated using the Hadley model: the amount falls by -10.8% with unmitigated climate change as per A2, and -6.0% under a mitigated B1 scenario. On the other hand, WRI increases if climate change is simulated using the CSIRO model: it grows by +1.5% under A2, and +2.0% under B1 (Fischer et al. [2007:1100]). In either case, however, these global changes in water resources are relatively minor at the world level, albeit they may represent a higher percentage of water resources in specific regions and under specific models; for instance, in the case of Latin America, the Hadley model projects decreases of 28% (mitigated B1) and 41% (unmitigated A2), but for the same region, the CSIRO model projects only a minor effect: 0.0% under B1 and -2.8% under A2. Model structure and parameters explain these considerable differences. To estimate precipitation and its effects on irrigation, the authors used a dataset containing historical climate data throughout the 20th century (1901-1996), including monthly maximum and minimum temperatures, precipitation, cloudiness, and other relevant variables. This dataset provides the above data for each cell in a 0.5° x 0.5° world climatology grid; each side of these cells is about 55-60 km. This grid was superimposed onto a FAO global map of available irrigated areas with a higher resolution of 5’ × 5’ (about 10 km × 10 km). Crop water requirements were estimated for a combination of four crop types: wetland rice, a generic dryland crop, a generic perennial (fruit trees, citrus), and sugar cane. These estimates were also broken down by season for matching with monthly data on precipitation and other climatic conditions. In each grid cell, crop water requirements and average precipitation were used to determine the water deficits of crops under rain-fed conditions. Any deficit thus detected was used as an estimate of how much extra water should be supplied through irrigation. It should be noted that water requirements were estimated for a set of crops that are not representative of the actual crops grown in each region (no one would grow wetland rice in the Middle East, for instance); therefore, the results are not to be interpreted as expressing the actual proportion of water resources used for irrigation but as a reference index only. The calculations for each grid cell were then aggregated at the country level, and then, finally, at the regional level. Baseline water resources and crop water requirements were estimated for average climatic conditions in 19611990, with reference to the areas irrigated in 2000. On this baseline, climate model simulations were run at decadal steps until 2080 to project changes in 276

Hector Maletta

all relevant variables, under the two chosen paths of climate change (A2 and B1) and a single socioeconomic scenario (A2r) for population, income, trade, and other related variables. The authors, with more updated data than used by Döll, estimate a baseline world-average irrigation efficiency of 0.51 for 2000, with different values per region. The total irrigated area worldwide in 2000 was 271 million Ha (17.6% of cultivated land). Irrigated area is projected to increase, in line with FAO (2003a) for projections to 2015 and 2030. Irrigation efficiency is projected to increase by 10% from 2000 to 2030, varying by region, as per projections in FAO 2003a, and by a further 10% from 2030 to 2080 in all regions of the world (Fischer et al. 2007:1088). As a result of this exercise, the global (weighted) Water Scarcity Index was found to be 13.6% in 2000, which is considered as a low value that leaves ample room for increasing domestic and industrial uses of water. However, higher values were found for certain regions such as the Middle East. The global WSI remains quite low when projected to 2080, from an initial value of 13.6 and a projected value of 14.2 in the hypothesis of no climate change, to values of between 15.5 and 17.1, depending on the climate model used and the degree of mitigation of climate change. For the whole world (Table 60) and using the Hadley climate model, the water scarcity index was initially estimated at 13.6 for the year 2000, changing by 2080 to values within 14.2 (without climate change) to 15.5% (with mitigated climate change as in the B1 climate projections, applied to the A2r socioeconomic scenario) and to 17.1% (with unmitigated climate change as in the A2 climate projections also applied to A2r). With the CSIRO model, the global WSI increases in a similar manner, to 16.0 (mitigated) and 16.8% (unmitigated). At the regional level, for brevity, only baseline (2000) and final (2080) figures are presented, along with the unmitigated A2 warming scenario, setting aside the mitigated warming of the B1 scenario (Table 61).

277

Impact of climate change on agriculture

Table 60. Changes to irrigation water for world agriculture due to climate change, 2000-2080.   2000 2010 2020 2030 2040 Projected irrigated land (million Ha) 271 292 313 327 342 Without climate change: 1350 1453 1559 1630 1707 Net irrig. water requirements (Gm3) 2630 2750 2873 2924 3019 Gross agric. water withdrawals (Gm3) Irrig. efficiency (reqs./withdrawals) 0.51 0.53 0.54 0.56 0.57 9705 9418 9179 8942 8827 Withdrawals per hectare (m3) Requirements per hectare 4982 4976 4981 4985 4991 Water Scarcity Index (WSI) under A2r socioeconomic scenario (percentages)* Reference (no climate change) 13.6% 14.3% 14.9% 14.6% 14.5% A2r socioeconomic scenario, with A2 or B1 climate change projections With Hadley model: A2r, A2 climate (higher emissions) 13.6% 14.8% 16.0% 15.9% 16.0% A2r, B1 climate (lower emissions) 13.6% 14.4% 15.1% 14.9% 14.9% With CSIRO model:  A2r, A2 climate (higher emissions) 13.6% 14.9% 16.1% 16.1% 16.2% A2r, B1 climate (lower emissions) 13.6% 14.9% 16.2% 15.9% 15.8%

2050 356

2060 369

2070 382

2080 393

1773 3090 0.57 8680 4980

1840 3162 0.58 8569 4986

1903 3225 0.59 8442 4982

1961 3278 0.60 8341 4990

14.4% 14.3% 14.2% 14.2%

16.2% 16.5% 16.8% 17.1% 14.9% 15.1% 15.3% 15.5% 16.4% 16.5% 16.7% 16.8% 15.6% 15.7% 15.9% 16.0%

(*) Water Scarcity Index: percentage of gross agricultural withdrawals relative to total renewable water resources. Source: Fischer et al. (2007).

Table 61. Water scarcity index for irrigated land, 2000-2080, under A2 scenario and two climate models. Region World Developed Developing North America Europe + Turkey Developed East Asia/Pacific Eastern Europe + former USSR Sub-Saharan Africa LAC NENA East Asia South Asia Developing SE Asia

Irrigated land (million Ha) 2000 2000 2080 Increase Baseline 271 393 122 13.6 69 79 10 5.0 202 315 113 19.6 25 31 6 6.9 18 18 0 4.6 5 6 1 4.6 26 31 5 4.0 8 32 24 2.8 20 46 26 1.6 20 28 8 60.9 58 71 13 15.0 78 109 31 43.2 12 18 6 5.2

Water scarcity index (WSI) 2080, A2 climate WSI change Hadley CSIRO Hadley CSIRO 17.1% 16.8% 3.5% 3.2% 7.1% 6.7% 2.1% 1.7% 22.4% 21.8% 2.8% 2.2% 10.7% 10.1% 3.8% 3.2% 10.1% 7.3% 5.5% 2.7% 5.9% 5.3% 1.3% 0.7% 5.4% 5.3% 1.4% 1.3% 10.4% 10.3% 7.6% 7.5% 6.6% 3.5% 5.0% 1.9% 88.1% 96.1% 27.2% 35.2% 17.4% 21.6% 2.4% 6.6% 42.0% 42.0% -1.2% -1.2% 7.2% 5.9% 2.0% 0.7%

Source: Fischer et al. (2007, Tables 2, 4, 9[a] and 10[a]).

Only two regions have a high water scarcity index at the baseline year 2000: the Near East / North Africa (NENA) region (61%) and South Asia (43%). Values above 40% are regarded by the authors as evidence of significant water 278

Hector Maletta

stress, since water is also needed for other purposes, chiefly industrial and residential. Both regions with high WSI, however, have a high WSI now and this will be the case in the future even in the absence of climate change: their future predicament is not the result of climate change, which will accentuate water scarcity in NENA and maintain or somewhat alleviate the conditions prevailing in South Asia. In more precise terms: in the hypothetical reference condition of no climate change, the NENA region will keep its WSI at about 61% while South Asia will increase from 43% to 48%, due entirely to the explosive population growth assumed in the A2r socioeconomic scenario. Incorporating climate change into that projection will greatly accentuate the already high water scarcity in NENA: using the Hadley model, the region ends up in 2080 with a WSI of 76.5% (B1) to 88% (A2). With the CSIRO model the figures are 85% and 96%, and even the strong mitigation implied by B1 fails to produce a reduction in the regional WSI: it only achieves a somewhat less dramatic increase. In either case, what the projections show is an unsustainable combination of water resources, crop mix, and population pressure, imposed chiefly by the initial conditions of this semi-arid region, the mix of crops used for estimation of water requirements, and an assumed socioeconomic scenario of high population growth. On the other hand, and even under these demanding assumptions, no such predicament arises for South Asia, the other area with high WSI, which has a much larger population. In either A2 or B1, and under either Hadley or CSIRO, South Asia is projected to have an almost unchanged WSI of around 40-42%, as it had in 2000, because increased water requirements will be met by the combined effect of more availability and more efficiency. All other regions will end up with low WSI values and, moreover, with little change from the levels of 2000 in spite of anticipated growth in irrigated areas due to the rapid population growth of the A2r socioeconomic scenario. The only region with a WSI that will be both high and rising, i.e., NENA, is a semiarid region that represents just 7% of all irrigated areas, less than 10% of the areas irrigated in the developing world, and a very small percentage of the population in developing countries. Even if almost all regions (except NENA) have relatively low WSI values and are expected to record just small increases in this index, these are just net results for the average of entire regions or the whole world. Specific ‘hotspot’ locations may endure higher increases in water stress. Re-localisation of agricultural activity may be induced in these hotspots, and the gradual reduction in the population share living in rural areas may also be accelerated, if local water resources cannot bear the required increase in irrigation. Moreover, irrigation 279

Impact of climate change on agriculture

usually translates into an intensification of labour requirements in agriculture, since much irrigated land is used for labour-intensive activities such as growing fruits and vegetables, whereas much rain-fed land is used for extensive crops requiring smaller inputs of labour that are, moreover, more easily replaced by machinery than labour employed in the production of fruits and vegetables.

10.4. Other assessments 10.4.1. The SIMPLE model Ulris L. C. Baldos and Thomas W. Hertel (2013, 2014) have developed a partial equilibrium model that integrates biophysical and economic aspects in order to project agricultural production and undernourishment (which they call ‘caloric malnutrition’), taking into account technical progress, demography, economic development, climate change, and production of biofuels. These authors use the so-called SIMPLE model (Simplified International Model of agricultural Prices, Land use and the Environment). This is usually an aggregate model but here they disaggregate it into 15 regions, basing most of their analysis on the IPCC 2007 report and FAO estimates of undernourishment from before the recent methodological overhaul. In this respect, they share the information base of the IIASA-FAO studies. There are as yet no studies using the more recent updates from the IPCC (the 20132014 AR5 report) or from FAO (the new estimates of undernourishment, retrospective to 1990, that were introduced by FAO in FAO-SOFI (2012) and have been published annually since that year). Baldos and Hertel (2013) first use the SIMPLE model for the past (19612006) in order to assess whether it has the power to predict the actual facts of the past regarding agricultural production and other aspects (not including undernourishment). In their 2013 paper, they state that the SIMPLE model ‘is shown to do remarkably well at capturing observed global changes in crop production, area, yield and price’ (Baldos and Hertel 2014:556). Their success in this endeavour led them to include a food security (undernourishment) module and geographical disaggregation, as presented in their 2014 publication. They provide projections of population, incomes, technological progress (in the guise of Total Factor Productivity or TFP), and projections of temperature, precipitation, and CO2 fertilisation. They carried out several runs of the model, some considering only the impact of increased demand (population and income) at zero TFP, and others incorporating climate change with or without CO2 fertilisation. 280

Hector Maletta

One of the strongest aspects of Baldos and Hertel’s work is the use of the recent analysis by Keith Fuglie and others on factors of agricultural growth, mainly using (as does this book) the FAOSTAT database and multiple studies on agricultural investment and technological improvement via innovation and diffusion of new production techniques. According to studies on various countries included in the volume edited by Fuglie et al. (2012), and the overall assessment by Fuglie (2012), the global agricultural Total Factor Productivity (TFP) is accelerating, led by developing countries. This strong growth of TFP offsets a parallel worldwide decline in the growth of input intensification. These trends, along with acceptable projections of population, income and drivers of agricultural TFP, allow Baldos and Hertel to forecast a sustained increase in agricultural growth due to growing TFP over the coming decades. This includes not only further discoveries and innovations yet to be developed by agricultural scientists (Fuglie’s Innovation-Invention factor), but also growing adoption of existing technology by those farmers playing catch up around the world (Fuglie’s Technology Mastery factor). According to Baldos and Hertel’s projections, the worldwide prevalence of undernourishment will fall by 84% in 2050 compared with 2006, from 12.0% to a non-significant 1.9% in a baseline scenario not considering climate change. The baseline relies on projections of population, income, agricultural productivity (TFP), and biofuel use. The resulting 2050 figure for the entire world (1.9% prevalence) will increase by 0.3 points (to 2.2%) with climate change without CO2 fertilisation, and will decrease by -0.4 points (to 1.5%) if the effects of increased CO2 are included. At any rate, the increase or decrease caused by climate change is well within the uncertainty of these projections, it represents a small fraction of the projected change in prevalence, and the projected prevalence remains well below the 5% threshold of significance attributed by FAO to its undernourishment indicator (figures below 5% are not significantly different from zero). Based on the application of their model to the recent past (1961-2006) the authors are more confident in their global than their regional results. However, their regional projections also show similar reductions, all leading to statistically non-significant prevalence of undernourishment.

281

Impact of climate change on agriculture

Table 62. Baldos-Hertel projections of the prevalence of undernourishment under projected climate change, 2006-2050. Region

Baseline (2006)

Scenarios for 2050 Baseline a

CC, no CO2 fertl.b

CC with CO2 fertl.b

World

12.0%

1.9%

2.2%

1.5%

Sub-Saharan Africa

23.5%

2.4%

3.1%

1.6%

Central Asia

21.4%

1.0%

1.2%

0.8%

China/Mongolia

9.6%

0.9%

1.0%

0.8%

Southeast Asia

12.8%

2.8%

3.2%

2.3%

South Asia

20.2%

1.2%

1.5%

0.9%

Central America

10.1%

3.6%

3.8%

3.3%

South America

8.2%

0.9%

1.0%

0.8%

a. Baseline scenario based on projections of population, income, agricultural productivity (slower growth of TFP than historical record), projected use of biofuels, and no climate change. b. CC = Climate Change. Based on model projections in Muller et al. (2010), with and without the estimated effect on crops of increased CO2 concentrations. Source: Baldos and Hertel (2014:559, Table 1).

Baldos and Hertel rightly surmise that many studies on response elasticities to climate change are based on short term fluctuations in climate and are thus unable to capture long term elasticities of supply that involve changes in multiple aspects of the production system. Also, they remark on the necessity to incorporate projected growth of agricultural productivity (measured as Total Factor Productivity, as per Fuglie 2012). In fact, in the absence of productivity growth from 2006 to 2050, there would be a significant reduction in food production and a significant increase in undernourishment, with or without climate change. Of course, negating any growth of TFP in that period is just a theoretical hypothesis, since the very momentum of technological innovation coupled with the gradual diffusion of innovations across farming systems - even if both occur in the future at reduced rates - ensure that the hypothesis will not be confirmed. Moreover, the years after 2006 provide abundant evidence that productivity is not only increasing, but is doing so at an accelerating pace (Fuglie 2012). Baldos and Hertel do adopt a prudent hypothesis in assuming that future productivity growth will slow down in comparison with the years immediately before their baseline (1995-2006), and even under that rather pessimistic hypothesis, the projected reductions in undernourishment are very important.

10.4.2. The IMPACT model The International Food Policy Research Institute (IFPRI) has developed an integrated model for assessing the future of food called IMPACT. It is much less complete in scope than the IIASA models, and IFPRI publications are less 282

Hector Maletta

detailed. Methodological aspects of the model and its reports are discussed in the Appendix, Section 13.5.2. The model includes estimates of the biophysical impact of climate change on major crops, trends in technological change (which include basic spontaneous adaptations such as rotating crops and varieties or changing the planting date, as well as technological discoveries and innovations), as well as changes in the relative price of farm products and inputs. The model is described by Rosegrant et al. (2012) and the latest major report is Nelson et al. (2010). The IMPACT model is the main predecessor of the more recent AgMIP models, described later (§10.4.3). The IMPACT model and the reports based on it suffer from a number of shortcomings, discussed in more detail in the Technical and Methodological Appendix. Some of these include: ŒŒ The biophysical effects of climate change on crops are simulated on the basis of crop models, without actual empirical measurements of crop yields (with different varieties and cultivars) in different climates. ŒŒ These biophysical effects are simulated for a particular variety of each crop, without taking into consideration that different varieties and cultivars of each crop are grown at different latitudes and climates. ŒŒ These same biophysical effects are simulated at a constant atmospheric concentration of carbon dioxide, namely 369 ppm, below current levels (nearly 400 ppm) and well below the levels needed for the envisaged future global warming to occur. In other words, the effect of CO2 on crop yields and water needs is totally ignored (some modest CO2 effects are, however, included in some reports about the IMPACT model, such as Nelson et al. (2009). This contrasts with the conclusions in this regard found in the Fifth Assessment Report of the IPCC (2014b), where the effect of CO2 fertilisation is clearly recognised. ŒŒ The biophysical effects of climate change on the entire wide range of crops and the multiple varieties of each are inferred from the modelled effect on a single variety for each of a limited number of crops (wheat, rice, maize, soybeans, and groundnuts). The effect of growing those crops under different farming systems is also ignored. ŒŒ The tendency of yields is assumed to include the biophysical effects as well as an intrinsic trend in crop productivity. However, this intrinsic trend, even in the absence of climate change, and with slower population growth, is assumed to be only large enough to keep the current level of dietary 283

Impact of climate change on agriculture

energy practically constant from 2010 to 2050. This is not consistent with past trends in the growth of agricultural productivity and food demand, even when population was growing far faster than it will grow in the future, and is also below the productivity growth rates estimated in other projections. Because cultivated land area is expected to change very little, future per capita food availability (in the absence of climate change) is presumed to remain practically constant. The assumptions responsible for this result are those concerning impacts of climate change, as well as those concerning expected growth of crop production in the absence of climate change. ŒŒ The assumed growth of income, the main determinant of demand, is among the highest and fastest of several such projections, tending to overstate future demand for food. ŒŒ The model is essentially an econometric partial-equilibrium model, not seeking overall consistency but estimating the effect of marginal variation in one or a few particular variables while ignoring the rest or keeping them constant. This tends to produce results that are not necessarily consistent with one another at the scale of the whole system, especially when large changes are involved that would unfold along a relatively long span of time. Such large, long-term and slowly unfolding changes cannot be gauged through marginal impacts under ceteris paribus clauses. ŒŒ Much of the results depend on year-by-year projection of market-clearing prices for major food commodities. This kind of model has not previously shown a great ability to forecast changes in prices: econometricians were unable to predict either the price depression of the late 1990s or the surge in the 2000s on this basis. Practically all these peculiarities in the model tend to overstate the negative effects of climate change on agriculture. In general, the model has many shortcomings, but virtually none tend towards lessening a negative impact. These problems severely affect the validity of the results, as discussed in Section 13.5.2. Here, we present the main results, but with a strong caveat that such shortcomings need to be taken into account. The IFPRI model does not produce an estimate of the future growth of agricultural GDP. Its main emphasis is on a few staple crops, and many of its more detailed results have not been published. However, the model has produced some land use projections and some food security indicators such 284

Hector Maletta

as per capita dietary energy supply and child malnutrition in developing countries. We concentrate on these results here. Unfortunately, its reports do not explicitly disclose many of the most important figures, which have been estimated here based on the published results (see Section 13.5.2, for the various calculations involved). Harvested area. The IMPACT model publications include projections of total cultivated area. There is, at times, no clear distinction between total cultivated area (including temporary and permanent crops), arable land (area with temporary crops), and harvested area (where successive crops in the same area within a single year are counted as separate harvested areas). These projections envisage very little change in global cultivated land, just as other studies on this matter such as those by FAO and IIASA. Moreover, the envisaged change is greater when the model is run without climate change (+2.3%) than with it (+0.3%); in other words, at the world scale the IFPRI model expects climate change to moderate the expansion of cropland and to save on harvested land as compared to the alternative scenario without climate change (or ‘perfect mitigation’). Some of this reduction in harvested area is the result of lower cropping intensity (fewer harvests per hectare) and some comes from less expansion in arable land. The land use projections in the IMPACT model have both an exogenous (AGR) and an endogenous (price-responsive) component. AGR is the intrinsic growth rate of harvested area and is not dependent on prices (the general formula is Equation [1] in Rosegrant et al. (2012:5). The other component depends on the price of crops. Little detail is provided on the intrinsic area growth rates, though Nelson et al. (2010:31) briefly note that: The exogenous component reflects a combination of historical trends and assessments about future changes, including urbanization and other land use change. The AGR values typically decline throughout the period; they are greater than zero for crops in some countries and less than zero for others.

Nelson et al. (2010) generally envisage three major socioeconomic scenarios: the ‘baseline’ scenario is half-way between ‘pessimistic’ and ‘optimistic’ alternatives in regard to population growth and economic development. The baseline scenario is based on the medium variant of UN population projections and assumptions on per capita GDP growth adopted by the World Bank project on the Economics of Adaptation to Climate Change, with some minor revisions. In general, here we concentrate on the baseline socioeconomic scenario. The general worldwide outcome for the so-called ‘baseline’ or intermediate scenario of the IMPACT model is a very small 285

Impact of climate change on agriculture

increase in harvested area. This varies by country group: harvested area decreases by about 10% in developed countries, shows little change in middleincome developing countries, and increases by about 21-23% in low-income countries, resulting in little overall change at the world level. The difference induced by climate change is in all cases very small; most of the change is due to non-climate components, and occurs with or without climate change. This is qualitatively in keeping with IIASA models and FAO projections not considering climate change.30 The ‘pessimistic’ and ‘optimistic’ socioeconomic scenarios do not differ much from each other in terms of the projected amount of harvested land. In all cases, the expected variation in harvested land is small and within the uncertainty range of all these projections, i.e. within ±3 percentage points, and most are within ±2 points from the baseline projection. Note that the values for the baseline scenario are sometimes outside the range defined by the optimistic and pessimistic scenarios: the latter are optimistic and pessimistic in other regards (population and income), but not as regards land use. The main lesson from this projection is that scenarios of slow or rapid population increase and slow or rapid economic development do not require a significantly different expansion of harvested land, relative to the moderate baseline scenario, where harvested land varies very little. The world, as things have stood, uses less than half the Prime and Good cropland available, and even pessimistic projections on the future of agriculture do not require cultivation of much extra land.

The overall projection to 2010 and 2050 of the area harvested area in 2000, however, may be overestimated in Nelson et al. (2010). This is already perceptible in the first ten years of the projection. The FAOSTAT figures were 1142.3 MHa for 2000 (the IFPRI projection baseline) and 1254.5 MHa for 2010. The latter is 10% below the 1378 MHa projected by IFPRI for 2010. Thus, the projection algorithms in the IMPACT model seem to contain a built-in tendency to overstate the growth of harvested area, which is likely to produce an exaggerated projection of area to be harvested in 2050.

30

286

Hector Maletta

Table 63. IMPACT model: Projected harvested area, with and without climate change, 2010-2050, baseline socioeconomic scenario.   Country group World Without climate change With mean climate change Developed countries Without climate change With mean climate change Middle-income developing countries Without climate change With mean climate change Low-income developing countries Without climate change With mean climate change

Projected (MHa)

Projected 2010-50 change MHa Percentage

2010

2050

1386 1378

1418 1482

+32 +4

241 241

217 216

963 957 182 181

Other scenarios Optim.

Pessim.

+2.3% +0.3%

-1.2% -1.6%

1.1% 0.8%

-24 -25

-9.8% -10.3%

-12.1% -11.6%

-10.3% -9.9%

1005 948

+12 -9

+1.3% -1.0%

-0.5% -2.6%

1.6% -0.5%

225 219

+43 +38

+23.7% +20.9%

20.9% 18.1%

25.4% 22.7%

Source: Nelson et al. (2010:32). Mean climate change = average of projections with MIROC and CSIRO models under A1B and B1 SRES scenarios. Optimistic and pessimistic socioeconomic scenarios differ in population and income growth, not in climate. Harvested area based on FAOSTAT arable land for 2000 (excluding permanent crops), projected to 2010 and 2050 by the IMPACT model. Changes in harvested area result from changes in arable land, in the share of arable land lying fallow, and/or in cropping intensity. Extra cultivable land due to climate change (mainly in developed countries) was not considered.

Dietary energy supply. As regards dietary energy, the central or baseline projection of the IMPACT model is that daily per capita calories will remain practically constant from 2010 to 2050 (with a negligible increase of just 0.4%). This is the combined result of yield decreases (estimated to result from climate change) in addition to changes in area and yield (driven by climate, by farmer responses to changing product and input prices, and by an assumed exogenous improvement in agricultural productivity). All these assumptions were strongly biased against any positive results, mainly through the following features: (a) The projections take into account the projected loss of output from lands that will become unsuitable for cultivation, but exclude the increased projected output from new lands that will become cultivable or more fertile; (b) Wherever the model predicted a strong yield increase, the outcome was regarded as anomalous and suppressed by putting an exogenous cap on such increases; no similar cap was imposed on yield decreases; (c) The effect of higher CO2 on yields and water needs was ignored in Nelson et al. (2010); that is, not just estimated to be relatively low, as in previous reports such as Nelson et al. (2009), but directly excluded; 287

Impact of climate change on agriculture

(d) Intrinsic rates of crop productivity growth were assumed to become much lower than in the past. The authors’ actual estimates of dietary energy supply in developing countries in 2010 or 2050 are not disclosed; only the baseline scenario variation over forty years and the effects of various hypotheses about future productivity growth. However, based on the actual figures for dietary supply in 20002010 it can be inferred that the implicit level (both for 2010 and 2050, with just 0.4% change) is about 2700 kcpd. The projected level of 2050, a mere 0.4% above the 2010 projection, implies that increased productivity will be just enough to compensate for population growth and the estimated negative impacts of climate change. Projections of production, as seen above (and more extensively in the Appendix, Ch. 13), are heavily affected by several procedures and assumptions that are strongly biased against growth in per capita food supply, not only those relative to the impact of climate change but also those unrelated to climate (such as technical change, land use, economic growth and the like). It is obvious that modifying any of these several biased assumptions will result in greater improvement, e.g., by assuming even a small CO2 effect, refusing to impose an external cap on predicted yield increases (but not on yield decreases), or refusing to exclude the expansion of cultivable land at medium or medium-high latitudes resulting from a warmer climate at non-tropical latitudes (or altitudes). However, it should be noted that not even the adoption of such strongly biased assumptions and procedures (such as imposing caps on positive results) generates a reduction in future dietary energy. It should be noted also that actual energy supply in 2010 (above 2800 kcal/person/day) is already above the 2010 projection resulting from the IMPACT model, for a period (20002010) too short for the effects of climate change to have much bearing. This is a further indication that the projections tend to be too pessimistic, not just on climate but also on other aspects such as productivity growth. It should also be noted that dietary energy intake tends to flatten out after reaching a level above 3000-3100 kcpd, while most agricultural production growth and food consumption increases tend to show an increased supply and consumption of micronutrients and protein; even under the hypothesis that dietary energy stays constant, projected income growth will imply more growth in other components of food consumption. Unfortunately, the study concentrates on staple food, the least elastic part of food supply and demand, and thus this aspect goes unnoticed. In spite of being centred on staple food, however, these projections are predicated on total consumption of all foods; 288

Hector Maletta

supplies are, moreover, measured only in terms of calories, disregarding growth in other nutrients. As has been amply shown in historical trends, increases in dietary energy preferentially come from non-staple food, and there is increased consumption of foods with few calories but that supply essential micronutrients, as is the case of fruits, vegetables or dairy products. Child malnutrition. Underweight (low weight-for-age in children under five, the index of child malnutrition chosen in the IFPRI model) is also projected to decrease from an initial level of about 27.7% (projected for 2010 on the basis of data for 2000) to an expected level of 17% or 20% in 2050 (with or without an assumed 40% increase in intrinsic crop productivity growth rates in developing countries, respectively). Again, these figures are not directly reported, but laboriously inferred here from partial data disclosed in Nelson et al. (2010): see detailed calculation in Section 13.5.2 in the Appendix). These projections on child underweight in Nelson et al. (2010) thus indicate an expected reduction, though the projections seem rather conservative in view of recent sharper decreases in malnutrition (see historical trends of nutritional status indicators in Section 7.3). Many more results could be potentially extracted from the IMPACT model, but existing reports are quite economical on substantive and methodological details (see Section 13.5.2 in our Technical and Methodological Appendix). It is, however, remarkable that even with so many assumptions and methodological details heavily biased towards considerable negative effects of climate change and relatively poor progress in crop productivity, the conclusion is that dietary energy will not deteriorate (in fact, a modest improvement is projected) and the prevalence of child underweight will significantly diminish. Much more pessimistic assumptions would have been needed to obtain a decrease in dietary energy and a higher prevalence of malnutrition, which have not resulted as yet from the IFPRI model.

10.4.3. AgMIP models The AgMIP project is a collaborative effort between several international institutions including IFPRI, FAO and others, to integrate various models related to the biophysical effects of climate change on agriculture and the economic response of farmers facing these effects (Rosenzweig et al. 2013). The general approach is similar to the IMPACT model, but much improved in many respects. AgMIP studies are based on an ensemble of climatic and economic models, and thus are expected to be more robust in those conclusions that are consistent across models. On the other hand, they are still subject 289

Impact of climate change on agriculture

to some of the problems detected in the IMPACT model, such as ignoring CO2 fertilisation and centring on a limited array of (mainly staple food) crop varieties (a presentation of the main results can be found in Nelson et al. 2014). Another issue is that biophysical effects on agriculture are modelled as being independent of farmers’ economic decisions. The general outline of this approach is as follows: Climate General circulation models (GCMs)

Biophysical

Temp. Prec.

Global gridded crop models (GGCMs)

Economic

Yield

(Biophysical)

Global economic models

Area Yield Cons. Trade

The outcomes of the economic models are the effective resulting yield, area harvested, consumption, and trade. Farmers could change crop varieties, and a general improvement in technology is allowed for as an exogenous factor. In this approach, the biophysical effects of a changed climate on today’s crop varieties (without CO2 fertilisation) will be faced by future farmers who will then decide whether to grow these or other varieties (or species). The above chart shows the unidirectional nature of the framework: it is not foreseen, for instance, that the crop models themselves will be changed - as will the crop mix - as they ought to be because the crop mix, crop varieties and farming practices prevailing in 2100 (which may be used by future agricultural scientists as the basis of their own prospective crop models) will be different from models calibrated to conditions prevailing around 2000. The AgMIP project is an on-going endeavour and the main authors recognise its limitations. Nelson et al. (2014) use a more restricted set of crops (maize, rice, wheat, and soybeans) than those included in previous publications by these authors, but crop yields in this case are endogenous; the assumption of no fertilisation effect of CO2 is maintained; the climate scenario is the most extreme of the AR5 set of scenarios (RCP8.5), which stipulates that in 2100 world population will be 12 billion, CO2 concentration will be about 940 ppm, and the global temperature will rise by almost 5°C. The general conclusion of the exercise is as follows: The mean biophysical yield effect with no incremental CO2 fertilization is a 17% reduction globally by 2050 relative to a scenario with unchanging climate. 290

Hector Maletta

Endogenous economic responses reduce yield loss to 11%, increase area of major crops by 11% and reduce consumption by 3%. (Nelson et al 2014: 3274)

Given that the average per capita income in 2100 is estimated (in scenario RCP 8.5) to be about 20,000 USD (van Vuuren et al. 2011), staple food will represent a declining fraction of all food consumption, and, in turn, food consumption will represent a declining fraction of expenditure, the estimated 3% reduction in food consumption for 2100 will still represent a major increase in comparison with current levels, even if the projection is based on only four major staple commodities (three cereals and soybeans) – the demand for which is relatively inelastic. The ‘consumption’ referred to in the conclusions is most probably measured in metric tons, not in value, an approach that cannot be applied in a more general assessment that includes fruit, vegetables and animal products. The paper did not consider what would happen if the most extreme of the scenarios failed to materialise, or if some CO2 fertilisation effect actually occurred, as demonstrated in all experiments that have been carried out. The envisaged percentage effects on yields, crop area, and consumption should, of course, refer to the land use, production, and consumption to be attained by the target date (2050), and not to that prevailing on the baseline date (2000). It can be seen that even if the effects of elevated concentrations of carbon dioxide are ignored, along with the other factors already mentioned, this integrated climate-econometric model approach forecasts that the target level of food consumption (mainly of staple food) by 2050 will be reduced by 3%. In fact, consumption of staple food is already stagnant or declining, and this behaviour is accentuated in correlation with the level of income. Since per capita income in 2100 is forecast to be higher than today, it is only logical to expect that consumption of staple food will not increase by much, or could even decline. Such a development would be compatible with a general increase in the real amount of food consumed, in total or per capita terms. As in other similar analyses that take a rather pessimistic view, the future level of food consumption is not anticipated to be significantly reduced as a result of climate change. However, even a very modest effect of carbon dioxide fertilisation (and water saving) might easily turn that 3% reduction into an increase, as is evident in other studies such as Nelson et al. (2009), Fischer (2011), or Cline (2007). If non staple food were included, there would also be more positive results, since, as shown above, consumption of cereals and other staple foods reaches a saturation level (already reached or almost reached in most parts of the world) after which demand for such foodstuffs stalls or declines. 291

Impact of climate change on agriculture

10.4.4. Ricardian models Ricardian models (see technical aspects in Section 13.6) estimate the effect of climate conditions on land value or farm revenue, controlling for other variables, by means of regression methods applied on (usually cross-section) data from farms existing today and located in different climates. This approach estimates the effect of current climate differences over space (between different locations), and then uses the results to estimate impacts of climatic changes over time (at any particular location). It assumes that farmers in the future will achieve the same level of adaptation to local climate as today’s farmers achieve under their current climate. These models do include adaptation, which is quite laudable, but they are unfortunately unable to incorporate CO2 effects since all their data are taken from farms operating within a fixed atmospheric CO2 concentration. We discuss two sets of Ricardian estimates: some led by Robert Mendelsohn (Mendelsohn [2000], Mendelsohn and Dinar [2009], and other related papers) and those in Cline (2007). We regard the Mendelsohn studies as good examples of a Ricardian approach, and Cline’s as having severe shortcomings. However, we discuss the conclusions of both. Mendelsohn (2000). This study is best suited to our present purposes because it includes an estimate of agricultural GDP in 2100 for all major world regions, with and without climate change. A new group of similar studies undertaken more recently under the general direction of Mendelsohn has produced a number of regional results for 2006-2009 (such as Seo et al. 2007), and an overall report (Mendelsohn and Dinar 2009), but not (so far) global or regional estimates for future agricultural GDP or food supply. Mendelsohn (2000) uses cross-sectional data, analysing the net impact of differences in seasonal temperature and precipitation on normal farm revenue (or on farmland value as a proxy), controlling for other variables such as soil, latitude, and technology. Estimates are presented for three different scenarios of (local) warming, implying temperature increases of 1°, 2°, and 3.5° Celsius. A default projection of future agricultural output not considering climate change is based on exogenous OECD projections of economic growth during the 21st century. Modelled effects of climate change are predicated on that hypothetical future level of agricultural production. Farmers are expected to make spontaneous choices regarding land use, crop planted, use of irrigation, and other such issues. These choices are, however, left implicit in Ricardian models: predictors include climatic variables and (fixed) characteristics of the 292

Hector Maletta

farm and zone (e.g., farm size), as well as variables reflecting fixed macro context (e.g., national policies, taxes, exchange rates, etc.). The models estimate the net impact of changes in temperature and rainfall on farm revenue per hectare or the value of a unit of land. The study is based on detailed analysis at the district level for Brazil, the United States and India, and then generalised to the entire world. To evaluate the relative impact of climate change in terms of agricultural GDP, Mendelsohn (2000) uses OECD estimates and projections of total GDP growth, and, based on past data, assumes that agriculture will grow in each region at half the rate of total GDP growth; thus, in the absence of climate change, it is assumed that world agriculture will grow from 1990 to 2100 at 1.24% per year, about half the rate projected by OECD for total world GDP (2.46%). These OECD growth rate estimates for total GDP are lower than observed trends in the previous half century. As with FAO projections, Mendelsohn assumes an economic growth slowdown over the century. Mendelsohn’s GDP has been converted into USD at market exchange rates and consequently tends to underestimate the real output of developing nations where a dollar typically has more purchasing power. This unduly understates the future level of total and agricultural GDP in Asia, Africa, and Latin America. Therefore, hypotheses about the level and growth of both total and agricultural GDP are quite conservative. GDP tends to be underestimated for developing countries, GDP growth is assumed to decelerate, and therefore the agricultural growth rate may be underestimated. The results are shown in Table 64. In spite of the pessimistic assumptions, the overall impact found is positive (world agricultural output in 2100 would be 5% higher under a changed climate than otherwise expected, regardless of the degree of warming). The growth rates and the impact of climate change vary by region. Increasing amounts of local warming, from +1° to +3.5°, produce larger and increasing positive impacts in temperate regions (Europe, North America, and Oceania), and positive impacts of decreasing size (that in some cases become negative) in regions with large areas of subtropical and tropical climates (Asia, Africa, and Latin America). If local warming is around +2°C, effects are positive everywhere, except for a small negative effect in Africa (-0.2%) that is probably well within statistical margins of error (though this aspect is not reported).

293

Impact of climate change on agriculture

Table 64. Projected net impact of climate change on 2100 agricultural GDP in major regions of the world, under various hypothetical changes in mean local temperature. Value in billion USD (1990 prices)

Percent of 2100 agricultural GDP

Agric. GDP 1990

Agric. GDP 2100* +0°C

+1°C

+2°C

+3.5°C

+1°C

+2°C

World

1,230

4,759

223

237

239

4.7%

5.0%

5.0%

Africa

84

416

16

−1

−30

3.8%

−0.2%

−7.2%

492

2,259

74

66

47

3.3%

2.9%

2.1%

89

441

19

11

−1

4.3%

2.5%

−0.2%

 

Region

Asia Latin America

Residual impact of climate change in 2100 (not including CO2 fertilisation)

Residual impact of climate change in 2100 (not including CO2 fertilisation), relative to absence of climate change +3.5°C

West Europe

191

542

12

15

18

2.2%

2.8%

3.3%

East Europe

231

694

63

95

137

9.1%

13.7%

19.7%

North America

127

360

36

49

66

10.0%

13.6%

18.3%

16

47

4

3

1

8.5%

6.4%

2.1%

Oceania

(*) Reference projection without considering climate change. Projected at half the rate of GDP growth as per OECD projections. Effect of CO2 fertilisation is not included. Regional or worldwide agricultural output obtained by aggregation of national outputs according to market exchange rates prevailing in 1990, and thus not correcting for differences in purchasing power. Source: Adapted from Mendelsohn (2000: 12, 26).

Table 65. Projected agricultural GDP in 2100 and 1990-2100 growth rates, under various hypothetical changes in mean regional temperatures. Agricultural GDP (billion USD at 1990 prices) 1990

Agricultural GDP (annual rate of growth)

2100

1990-2100

Region

Baseline

+0°C

+1°C

+2°C

+3.5°C

+0°C

+1°C

+2°C

World

1230

4759

4982

4996

4998

1.24%

1.28%

1.28%

1.28%

Africa

84

416

432

415

386

1.47%

1.50%

1.46%

1.40%

492

2,259

2,333

2,325

2,306

1.40%

1.42%

1.42%

1.41%

89

441

460

452

440

1.47%

1.50%

1.49%

1.46%

Asia Latin America

+3.5°C

West Europe

191

542

554

557

560

0.95%

0.97%

0.98%

0.98%

East Europe

231

694

757

789

831

1.01%

1.08%

1.12%

1.17%

North America

127

360

396

409

426

0.95%

1.04%

1.07%

1.11%

16

47

51

50

48

0.98%

1.06%

1.04%

1.00%

Oceania Source: Table 64 above.

Thus, according to Mendelsohn’s impact estimates (Table 64) and assuming farmers in 2100 will have the same degree of adaptation to the prevailing climate and the same degree of efficiency in the use of available technology as existed in 1990, climate change would cause a net increase in world agricultural GDP of about 5%, compared to the agricultural GDP that is projected in the absence of climate change. Regions in temperate climates (including those in North America, Southern LAC, Europe, Asia, and Oceania) will benefit most, 294

Hector Maletta

while regions in tropical and subtropical climates will suffer some damage (especially in scenarios that imply higher increases in local temperature). In all regions, the percentages (especially for losses) are relatively small (from a maximum loss of 7.2% to a maximum gain of 19.7% for the various world regions). These losses or gains are relative to the 2100 output in the absence of climate change, which by all accounts will be substantially higher than the level of the base year (1990), even in per capita terms. In all cases, even in Africa, production in 2100 will be several times greater than in 1990 once the impact of climate change is taken into account (Table 65). As regards the expected growth rate of agricultural production over the century, Mendelsohn estimates imply that impacts of climate change will be practically irrelevant at the world level: world agricultural GDP is expected to grow at an average annual rate of 1.28% under the three levels of warming considered, slightly above the growth rate (1.24%) that was assumed in the absence of climate change (Table 65, right-hand panel). As shown in Table 65, far from implying increased food scarcity, food output worldwide and in all regions is projected to be a good deal higher than the levels of 1990 (about four and five times higher, respectively) even in the worst of these scenarios (Table 65, left-hand panel). In 2100, the worst-hit region (Africa), in the worst case, will have an agricultural GDP 4.6 times as large as that of 1990. Even in per capita terms, these projections would represent significant progress. According to Mendelsohn’s projections of agriculture and UN population projections, per capita agricultural GDP worldwide will at least double from 1990 to 2100. According to the 2010 revision of UN Population Projections (Medium Variant), world population in 2100 will be 10.9 billion, whereas the population in 1990 was estimated at 5.3 billion. If these figures for population are combined with Mendelsohn’s (2000) projections of agricultural GDP, per capita agricultural GDP in constant 1990 prices will be 232 USD in 1990 and will have grown by 2100 to values of between 457 USD (with 1°C warming) and 458 USD (with 3.5°C warming), i.e., doubling over 110 years at a rate of 0.6% per year, regardless of the level of warming. The above implies that global per capita farm output in 2100, with or without global warming, will be about twice as large as it was in 1990. In a world increasingly interconnected by trade, local growth in per capita production is less relevant, as increased trade may make up for any local imbalance between food output and food demand. The current definition of food security, in fact, is centred on access and recognises trade as an essential component of food 295

Impact of climate change on agriculture

security. It should be noted, therefore, that if agricultural growth happens to be lower than average in a given region, regional food insecurity (i.e., inadequate access to food) may not increase. If income levels grow as expected, even if the projected rates of economic growth are modest, food trade may cover any remaining deficit between local output and local demand. As indicated before, regions where output grows more slowly than population will probably become more reliant on food imports, while other regions will provide the necessary exports. Trade, in fact, is expected to increase its share of world food supply, which would constitute a continuation of past trends that show a faster increase in food trade than food production (see Ch. 5). Per capita output at the regional level, however, would also increase to a similar degree, doubling or even trebling its 1990 level, except in Africa where it (as per these projections) would be stagnant and slightly declining. It should be noted, however, that these projections for Africa were based on the dismal performance of African agriculture up to the early 1990s, and thus do not account for its much faster growth in the 1990s and 2000s; and they refer to output, not to access to food. The poor outcome in per capita food production obtained for Africa is in fact not due to climate change: it is almost equally disappointing for all levels of warming, from +0° to +3.5°C. The determining factor for Africa’s outcome is not so much the impact of climate change, as the poor output of Africa in 1990 and its expected level of output in 2100, even without climate change. Since the 1990 level of GDP was estimated in US dollars at market exchange rates, it was strongly influenced by differences in purchasing power between developed and developing countries, and especially in Africa, where exchange controls and misaligned exchange rates were the norm up to the 1990s (such policies have been largely modified by reforms undertaken in the 1990s and 2000s). Since Mendelsohn assumes agriculture will grow at half the rate of GDP and African agriculture is expected to grow at 1.47% per year in the 0°C scenario, it is clear that the economic growth of Africa from 1990 to 2100 will progress at an annual rate of 2.94%: twice the rate of agriculture. With a projected population growth rate of 1.017% (as per the latest UN revision), Africa’s per capita GDP will grow at a yearly 1.92% from 1990 to 2100, thus increasing by a factor of eight over that 110 year period. That kind of increase in per capita GDP (a close proxy of per capita income) and the observed correlation between per capita income and levels of food consumption and undernourishment will ensure access to adequate food for practically the entire population of Africa, irrespective of how much food is produced on that continent. This is remarked on to highlight the difference between 296

Hector Maletta

projections of agricultural output and projections of access to food. Besides, for countries where agriculture represents a relatively high share of total GDP, as in Africa, Mendelsohn’s assumption that agriculture in all regions grows at half the rate of GDP may be inadequate: in such countries agriculture is likely to grow at a rate closer to (albeit somewhat lower than) that of total GDP. This implies faster growth for African agriculture, which is in fact already growing at rates significantly higher than the world average (see Section 3.4). Mendelsohn’s results find, under the strong condition (typical of Ricardian models) that no increase in the degree of adaptation occurs from 1990 to 2100, and without considering gains from increased atmospheric carbon (which will typically imply a higher agricultural output as a result of both increasing yields and a significant reduction in crop water requirements). This is crucial; even a very modest (say, 10%) beneficial impact of CO2 on output will more than offset the most negative impact envisaged in the worst case scenario (-7.2% in Africa under +3.5° local warming). Not even wider diffusion or increased efficiency in the use of available technology is permitted by this projection, because Ricardian estimates are based on the average level of adoption and adaptation prevailing today, whereby some farms adapt better than others, as is usually the case. The average percentage difference existing in 1990 between actual average farm performance and the theoretical performance achievable with existing technology is implicitly expected to persist. In reality, however, this is not likely to happen. In general, production by more advanced (i.e., more skilled and educated) farmers tends to fare better, and hence to represent an increasing share of total output; this factor alone will imply an increasing average prevalence of more advanced farming practices and technology (the average rate of advance may be slower or faster in different zones, depending on other conditions). Any technical progress involved in agricultural growth is, in Ricardian models, assumed to be climate-neutral, in the sense of not improving farmers’ average level of adaptation to their local climate, and this is also unlikely to happen. Many new techniques or innovations will address the challenges posed (at each location) by ongoing or projected climate change; thus technical progress is likely to be adaptive. Besides, an increasing share of total output will come about with the more efficient alternatives. Even under these adverse hypotheses, Mendelsohn’s study leads him to draw a positive conclusion: The literature to date suggests that global warming will not damage aggregate global food supplies over the next century. Existing models predict that there will be sufficient food to feed future populations, even with global warming. 297

Impact of climate change on agriculture

Global warming is not expected to affect aggregate production in most developing countries. On the other hand, it is likely that global warming will increase production in most temperate and polar countries leading to small increases in overall supply. (Mendelsohn (2000:27)

Mendelsohn and Dinar (2009). Mendelsohn (2000) was based on aggregate district-level data from the US, India, and Brazil. More recent studies by Mendelsohn (reported in Mendelsohn and Dinar [2009] and related papers) were based on samples of individual farms in various countries. Their 2009 book summarises the theory and statistics of Ricardian models, reviews older studies based on aggregate data such as those used for Mendelsohn (2000), and examines in detail more recent studies based on farm samples. These newer sample-based studies by Mendelsohn and associates have not produced an estimated impact of climate change on future agricultural GDP, either regional or global, to which these effects should be applied. More details are given in §13.6. Mendelsohn and Dinar (2009) analyse sample data from various countries in Africa and South America as well as China, estimating the impact of climate change in monetary terms.31 In the case of Africa, the marginal effect of a 1°C increase in temperature (p.104) is expected to be negative for rain-fed farms (-26.7 USD per Ha) and positive for irrigated farms (+35.04 USD per Ha), both not quite large. For a unit increase in monthly precipitation (+1 mm/ month) the effect is positive but also small in both types of farm (+3.82 USD and +2.70 USD). In the case of livestock, the marginal effects of climate change on the value of owned livestock are different for small and large farms, not because of different impacts of climate, but because small farms ‘shift to crops in wetter locations’ (p.107). The marginal effects are small in relative terms: per dollar of livestock revenue, an increase of 1°C will cause a reduction of 0.024 USD (-2.4%) in small farms and 0.033 USD (-3.3%) in large farms; an extra mm/month of rainfall will cause even smaller percentage reductions of -0.4% and -0.6%, respectively (Mendelsohn and Dinar 2009:108). The authors do not disclose the relative frequency and mean size of small and large livestock farms, nor of rain-fed and irrigated farms, hence it is not possible to All the amounts are in USD but the authors (as well as the background paper by Kurukulasuriya and Mendelsohn (2008) do not specify the exact unit of measurement or base year. In one example of a model dealing with the U.S. (p. 96), they state that the amounts are in 2008 USD, updated by the US GDP deflator. In results for other continents there is no such specification; amounts in local currencies are presumably converted into USD at market exchange rates (as in Mendelsohn [2000]) though no explicit statement is made about this point.

31

298

Hector Maletta

estimate an overall impact on African agriculture as a whole, but it is evident that the total impact will probably be a relatively small negative number, which will affect the future level of output of the target year (which in their case is 2100). Even at very low rates of agricultural growth, far below the recent performance of over 3% per year, African agriculture will greatly expand its agricultural GDP by 2100 compared to the beginning of the century, even accounting for the negative impact of climate change. For South America, the results are based on farmland value rather than farm revenue. This is in itself a limitation since land markets are quite imperfect in much of Latin America, especially in Andean countries (the study included Colombia, Venezuela, and Ecuador). Land values are influenced by imperfect titling and by non-agricultural land use. Models found positive marginal impacts for precipitation and negative marginal effects for temperature in the region as a whole. Land-value elasticities relative to temperature were estimated (p.112) at -1.21 for small farms and -1.89 for large ones, and rainfall elasticities of land value were positive but much lower in absolute value (+0.07 and +0.35, respectively).32 In the case of China, the annual elasticity of all farms with respect to temperature is 0.09, and 0.80 for rainfall. A climate change scenario introduced by the authors, with +2.5°C and an 8% increase in precipitation, generates a loss of about 6% in agricultural revenue (though this rough scenario applies the same climatic change to the entire territory, not downscaling for local differences). The extent to which this will be a percentage of future agricultural production in 2050 or 2100 is impossible to tell, because no projection of total production is offered. The kind of impacts resulting from these studies by Mendelsohn and his colleagues, if applied to the degree of climate change envisaged in IPCC The authors calculate temperature elasticities of revenue (or land value), i.e., the percentage change in revenue or land value per a one percent change in temperature; this is technically incorrect because temperature scales (Celsius or Fahrenheit) have an arbitrary zero point, and thus a percentage increase makes no sense. Elasticities (and percentage increases in general) are only valid for ratio scales, i.e., interval-level measures with a non-arbitrary zero, such as age or distance, and not for scales with arbitrary zeroes like temperature or longitude. Thus, 20°C is not one half of 40°C, nor is it twice 10°C. One degree Celsius does not equate to a certain percentage of a given temperature (say, 10% of 10°C) simply because the arbitrary zero in the Celsius scale is the freezing point of water; it would be a different percentage if calculated in Fahrenheit degrees. For the same example, an extra 1°C on a baseline of 10°C, interpreted as +10%, when expressed in Fahrenheit terms would be 1.8F over a baseline of 50F, representing not 10% but 3.6%. Percentage change in temperature could possibly be defensible (though still debatable) for the Kelvin scale, where the zero level is the absolute zero located at about -273°C, but Mendelsohn and Dinar (2009) do not use the Kelvin scale for their elasticity calculations.

32

299

Impact of climate change on agriculture

projections, may at times be relatively large, but the agricultural output of 2100 will be, by all accounts, large enough to withstand such effects and still remain much larger than today in per capita terms. For instance, if agricultural growth proceeds from 2000 to 2100 at rates like those contemplated by Mendelsohn (2000) or other similar projections, and population grows according to the UN (Medium) projections, production per capita in 2100 will be much larger than today. Mendelsohn and Dinar (2009) do not estimate a worldwide impact for climate change, but for the climate change projected by IPCC in business-as-usual scenarios such as A1 or A2, a rough average of these Ricardian estimates may indicate a loss of between 10% and 20% of the output that would otherwise be attained, not including the effect of CO2 (its inclusion would mean a lower detraction or even an increase). Growing at 1.25% per year (as in Mendelsohn [2000]), much slower than in the past half century), an index number of agricultural GDP (with a base of 100 in 2000) would be 346 in 2100 before accounting for the effects of climate change. This GDP, reduced by, say, a very pessimistic 20% due to climate change, will be equivalent to 277, an increase of 177% over its 2000 level. Meanwhile the population would grow by 78% (i.e., at 0.58% per year) from 2000 to 2100, as per the Medium Variant of UN projections, and thus agricultural GDP per capita in 2100 would be 55% higher than it was in 2000, even in this extremely pessimistic hypothesis on the overall damage caused by climate change. In summary, Mendelsohn’s set of studies based on farm samples, summarised in his 2009 book with Dinar, concentrates on the sensitivity of farm revenue (or land value) to marginal changes in climate, not allowing for any impact of CO2 concentrations, nor accounting for the impacts of economic growth, technical change, changes in agro-ecological zoning, or changes in the area of farmland. The size of the farm samples on which the new studies are based (as explained in our Technical and Methodological Appendix) is relatively small, and the samples are non-random, not properly weighted, and do not cover all agro-ecological zones. Large portions of the globe (such as all of Asia outside China, Oceania, Europe, North America, several major South American countries, the whole of Central America, Mexico, and the Caribbean) are excluded from the studies based on farm samples reported in Mendelsohn and Dinar (2009). The data are taken from non-random clustered samples, but they are treated as if they came from simple random samples, though the authors recognise that they do not (see Section 13.6). Ricardian studies, on the other hand, do not consider the effect of increased CO2 on crop production via increased photosynthesis and (for C4 crops) a reduction in water needs. For all these reasons, estimates in Mendelsohn and Dinar (2009) 300

Hector Maletta

are, in our opinion, not as far-reaching as those in Mendelsohn (2000), which include a projection of agricultural and total GDP growth and were based not on the self-declared revenue of farmers but on aggregate statistics per district, which usually reflect effective production more precisely. However, some of the problems are intrinsic to Ricardian models and thus they also affect Mendelsohn (2000). Cline (2007) uses the Ricardian approach combined with agronomic crop models and considers the beneficial impact of atmospheric carbon dioxide on plants as a separate factor. The Ricardian model used by Cline has several defects. For instance, it does not fare well when applied to actual data: it retroactively predicts negative output values for many countries and regions when applied to the recent past, as illustrated by Cline himself in his methodological appendix (Cline 2007:123-124). When a model has such a limited capacity for predicting the past, it should hardly be relied upon for predicting the future. However, for what it is worth, it is discussed here. Cline arrives at less than dramatic conclusions: world agricultural production by the 2080s will be about 3.2% less than it would otherwise be in the absence of climate change (Cline 2007:71).33 The purpose of this estimate is to measure the biophysical impact of climate change but it does not take into account future technical change, and ignores the expansion of suitable farmland in middle and mid-high latitudes due to warming (only currently existing farmland is considered). The estimated percentage impact is to be interpreted as an impact not on today’s production but on future output, relative to future population, produced by future technology, none of which are investigated in Cline (2007). In summary, estimates based on Ricardian models show small or moderate effects of climate change on future agriculture output, even if they frequently ignore the beneficial effect of increased CO2. Some studies show a small positive effect, some a small negative one, always to be added to or detracted from future hypothetical production levels without climate change, which, by all accounts, will be much higher than today’s. However, these methods are rather crude, and do not fully consider the various factors at play; allowing for some (even limited) effects of increasing CO2 concentration on crops, or a modest improvement in technology, or changes in crop mix and agro If carbon fertilization due to climate change were not considered, the reduction projected by Cline would be 18.9% (Ricardian models) or 15.9% (crop models), but of course carbon fertilization must be considered.

33

301

Impact of climate change on agriculture

ecological zoning, would make up for much of the envisaged negative effects. To achieve this, an integrated assessment is needed. So far the most important and complete integrated assessment programme is provided in the set of models developed by IIASA (Section 10.3). In these more complex models, the net effects of climate change are very small, either negative or positive, and prospective agricultural development (even accounting for climate change, use of land for biofuels, and allowing for a general slowdown of technical progress) would wipe out undernourishment in all regions during the coming decades. In fact, the most challenging period is the coming two decades or so, not because of climate change (which will not yet have had a significant impact) but because the prevalence of hunger is still high in some parts of the world, and many factors (war, state failure, inefficient economic systems, and so on) impede faster progress, e.g., in many African countries.

302

11. Concluding remarks

11.1. Summary of prospects We have only considered projection exercises that actually provide estimates of total agricultural production (not only cereals or other staple foods) and estimates of hunger or malnutrition. These exercises, in spite of their insufficiencies and varied assumptions and methods, generally share (or are consistent with) a number of expectations with regard to food and hunger in the 21st century: ŒŒ Agricultural output, which is already large enough to feed the world’s entire population and more, is projected to be more than enough to feed the world population up to 2050 or beyond (2080). Very little expansion of farmland would be needed for this purpose, even if technological progress slows and allowing for excess food consumption by the increasing number of overweight and obese people. ŒŒ The net effects of climate change on agriculture at the world scale would be limited, even under scenarios of severe climate change and intense use of biofuels. The net effect may be positive in some regions and negative in others, and the overall net effect may be slightly positive or slightly negative (relative to otherwise expected future output) depending on the assumptions adopted and projection methods. ŒŒ World per capita dietary energy supply by 2050-2080 is expected to be about 3000-3200 daily kilocalories, up from about 2800-2900 at present. Even in the most pessimistic projections it is projected to increase, however slightly. Trends towards increased dietary diversity and

303

Concluding remarks

expanded consumption of non-staple food (meat, milk, eggs, fruits, and vegetables) will continue. Production and consumption projections are based on projections of income and demand, and therefore production is also expected to be large enough to sustain the current tendency towards increased overconsumption and obesity, although of course it is desirable for such trends to be discouraged or reversed by appropriate policies. ŒŒ Trade in food and other agricultural products is expected to increase, both in absolute terms and as a percentage of world production. Economic development in the peripheries of the world economy would generate not only increasing local production, but also (and in some regions mainly) larger trade flows in farm-related products. ŒŒ Undernourishment is consistently projected to diminish, reaching non-significant levels by mid-century or shortly thereafter, both worldwide and in all major regions, even under the worst scenarios for biofuel development, and for emissions and climate change. Overall, the conclusion of this review of historical and prospective estimates is twofold: (a) world food security has been greatly improving in recent decades, and (b) it is expected to keep improving in the future, even under quite conservative hypotheses. Hunger, however, although globally on the wane, is still a scourge in our time, afflicting hundreds of millions of people in the developing world, and actions to diminish it must continue. Increasing levels of overweight and obesity should also be addressed.

11.2. Challenges ahead Even if our intent here is chiefly to examine trends and prospects rather than to offer policy proposals, the above review clearly suggests some key policyrelevant insights or hypotheses: ŒŒ Those most affected by food insecurity are concentrated in South Asia and Sub-Saharan Africa. This is already recognised by international donors that concentrate their efforts in those areas. As the situation improves in the coming decades, and nutritional deficiency in Latin America further diminishes, South Asia and especially Sub-Saharan Africa would even more clearly be recognised as the last bastions of hunger and food insecurity in the world.

304

Hector Maletta

ŒŒ Producing enough food is not the main challenge. The key to reducing undernourishment and malnutrition is improving the earning capacity of the poor by enhancing their capabilities, diversifying their livelihoods, facilitating their access to education and better jobs, and improving their access to health care and sanitation: all factors leading to higher incomes, better standards of living, easier access to food, and better biological utilisation of food. ŒŒ Low income and extreme poverty are the main factors behind lack of access to food. Income inequality among people is more a function of differences between countries with respect to per capita income, than a result of domestic inequality, which explains only a fraction of total inequality. Reducing domestic inequality will also be helpful, but may have less of an overall impact than convergence between countries towards higher levels of income. Such convergence already has been taking place through faster growth in developing countries as compared to richer ones, and its continuation requires ongoing rapid growth in developing countries especially an acceleration of economic growth and social development in South Asia and Sub-Saharan Africa. ŒŒ Violence and corruption hinder the reduction of malnutrition and undernourishment, as does the lack of preparedness for recurring natural disasters. Such factors divert valuable resources and preclude valuable opportunities for economic growth and livelihood improvement, as well as perpetuating food problems. Good governance is thus another key factor: most countries where undernourishment has increased are subject to corrupt or tyrannical governments, violent strife, or state failure. The eradication of hunger should happen sooner than projected. Fighting hunger is an on-going battle. The key challenge in the years ahead is not the production of sufficient food but rather ensuring access to adequate food by everyone and at all times. The sooner this goal is met, the better. Its achievement primarily depends on economic growth and social development, but according to forecasts, it may take several decades (to 2050 or beyond) for social and economic development to reduce undernourishment to negligible proportions in the worst affected parts of the world. More effective policies may (and should) help achieve this goal earlier since this is a very pressing need, especially – as already noted – in South Asia and Sub-Saharan Africa. The most important determinant of food security is economic development. Higher income means less hunger. There is practically no

305

Concluding remarks

undernourishment in nations with per capita incomes above 12,000-15,000 USD per capita (in 2011 PPP USD). At the world level, developing nations are growing faster than developed ones, leading to higher growth of average income in these countries. Even if domestic agriculture is not enough to cover domestic demand, higher incomes enhance the capacity to import food from other countries, allowing trade to play its key role in achieving food security. Food trade has grown much faster than food production, and will probably keep increasing as a proportion of food supply. However, development itself tends to determine an increase in intra-country inequality because in many cases the average income of the poor tends to grow more slowly than their country’s per capita income. Specific social protection policies should be enacted for the most vulnerable sectors of the population, to ensure universal access to food and other essential goods and services. Even if, on the whole, faster growth of income in the periphery lifts many people above the poverty line, thus enhancing food access and reducing undernourishment, other less visible problems remain, such as a deficit of micronutrients and the growing scourge of obesity. The goal of eradicating hunger and preventing the growth of obesity will be achieved sooner if more is done to incorporate additional distributional equality in the development process, to enhance nutritional awareness, and to improve governance, allowing for better policies to be enacted and implemented. Even if the developing world as a whole is growing fast, certain countries are lagging behind, mostly due to internecine or international war, and corrupt and inept systems of government. Enhancing the chances of peace, tolerance, and better governance are important ingredients for faster eradication of hunger. In the meantime, investments in human capital and early nutrition, which may include conditional or unconditional cash transfers to the poor, are one way of easing the path to getting more people above the hunger line more quickly. Besides, these measures will ensure that children’s cognitive development is not irretrievably damaged by inadequate early nutrition. Public transfers are by no means a sustainable way of getting people out of poverty, but they may be an acceptable step along the often difficult road to prosperity. Improved public services in education, sanitation, and health care are at least as important as (and probably more than) cash transfers in producing new generations of healthier and more educated people who could put the growth process on a faster track, and reduce poverty and hunger more quickly. 306

Hector Maletta

Subsistence (or infra-subsistence) smallholders are probably among the groups most severely affected by hunger. Paradoxically, the very fact of producing food for on-farm consumption often results in a severe inability to ensure access to sufficient (and adequately varied) food for a healthy life, as per the definition of food security. Smallholders may constitute a declining share of the total population, as the developing world rapidly urbanises and rural people diversify their livelihoods, but subsistence smallholders are still among the hungriest of all. For poor peasants to escape from poverty it is necessary, on the one hand, to achieve improvements in their agricultural activity, and on the other, and possibly more importantly, diversification of their livelihoods. To a certain extent, livelihood diversification may entail not only off-farm activities but also diversification of the subsistence farm itself, to include cash crops that secure easier access to food on the market. Purchasing food is, in any case, an inevitable part of the smallholder subsistence system, because each farm is only able to produce a limited array of food products, while other foods are accessed through the market, and the small size of such farms often precludes producing in them more than a fraction of the food a family requires. Food purchases may be afforded through the sale of farm products or through revenue accruing from other income sources (wage labour, non-farm selfemployment, seasonal labour migration, remittances, or public transfers). Hence, livelihood diversification is a key element for the poor smallholder family to achieve a more decent level of income and thus be able to overcome chronic or seasonal hunger, and to enrich food consumption beyond the mere supply of dietary energy. Of all public interventions aimed at improving the agricultural productivity of subsistence farmers, one of the most widely undertaken is probably adaptive water management infrastructures, especially small irrigation and drainage systems. In many areas of the developing world, the critical factor for agriculture is not land but water, and not just the annual amount of rainfall but also its seasonal distribution. Thus, construction of reservoirs and irrigation systems may often be the key element along the path towards enhanced agricultural production, introduction of cash crops, and adoption of sustainable farming systems. Adoption of better water management practices is also very important for water conservation and improved efficiency. Microwatershed management and soil conservation practices, along with small community reservoirs, are amongst the avenues to improve the productivity of subsistence farming across the developing world, especially in hot and relatively dry areas in South Asia, Africa and Latin America. 307

Concluding remarks

The local effects of climate change on water availability are expected to vary widely, and their projection is fraught with uncertainty. On the whole, global warming will cause more precipitation. However, precipitation is not evenly distributed; many projection models forecast a drier environment in certain parts of the world, notably Mexico and Southern Africa, and timely public interventions to tackle this issue will be crucial for the communities involved. Agricultural R&D aimed at basic foodstuffs as well as cash crops, and especially tailored for the use of smallholders, is another key ingredient for enhancing subsistence production. The Green Revolution of the 1960s, though often criticised in some respects, has undoubtedly done much to improve crop yields around the globe. In the 21st century, a second Green Revolution is in the offing, probably driven by genetics (including genetic engineering), and will likely be coupled not with on-farm consumption but with greater integration into world markets. Enhancing smallholder production of staple food, however, does not mean promoting household self-sufficiency in food supply; it is often the case that many subsistence farmers prove to be better off when they produce for the market and use the revenue to procure their food also through the market. However, this depends on local conditions and resources, and will vary widely from one place to another.

11.3. Pending research A number of technical issues and data shortcomings have been pointed out at different points of this book and they merit further investigation. Here, we will briefly set forth what we believe to be the main points of a future research agenda regarding the matters discussed in this book, with a view to filling data gaps and addressing shortcomings in current analyses. Land, capital and labour. We have examined agricultural (and especially food) production and the amount of land (and water) used or required to that end. But we have said little about the corresponding amount of reproducible capital goods used as means of production, such as farm infrastructure and equipment. We have also said next to nothing about farm employment. This is primarily due to a lack of adequate data with worldwide coverage. In the case of land, FAOSTAT land statistics include the total amount of agricultural land, with major uses such as arable land, land with permanent crops, and permanent pastures. There is also information on the area equipped with irrigation systems, though a breakdown of how much of each crop is grown on rain-fed or irrigated fields is not provided. There are also other 308

Hector Maletta

areas for which data are absent or insufficient, such as the distinction between natural and cultivated permanent meadows and pastures, which is missing in FAOSTAT for most countries. FAO is responsible for establishing the international standards for agricultural censuses, and one of the main topics of these censuses is land tenure and the size distribution of farm holdings. FAO has potential access to the database of every agricultural census in the world, and may also ask national governments for detailed information on the agrarian structure of their countries. However, this information has not been systematically consolidated and made available in FAOSTAT. Agricultural censuses, the main source for land tenure and farm size data, are often scarce or non-existent in many developing countries. As a result, the proportion of small, medium or large farms in total agricultural land area and their share of farm production remain largely unknown on the world level, as does the area and output of each crop and each kind of livestock corresponding to farms of various sizes. Where farm size distribution is known, it is often difficult to ascertain land quality and suitability according to farm size. Other related matters are likewise difficult to gauge on a world scale, including: the relative importance of corporate versus family farming; the structure of farm ownership (including the extent of corporate or family ownership of multiple farms in different locations, or the same farm being co-owned by several partners); the extent and forms of tenancy; the extent of the process of ‘land grabbing’ whereby corporations and governments take possession of land (mainly in Africa) formerly held by smallholders (who often lack formal ownership titles); and other related issues. As regards agricultural employment, FAOSTAT contains a number of series starting in 1980 with projections up to 2020, provided by the International Labour Organisation (ILO) and purporting to reflect the number of workers employed in agriculture, as well as the number of people (workers and their families) dependent on agriculture for their livelihood. We chose not to use these data, however, primarily because they do not cover the time span of our analysis (where most variables go back to 1961 and projections extend to 2050 or beyond). In addition, these estimates do not tell us much about the characteristics of agricultural labour; the estimated number of farm workers is reported, but there is no information about their employment status (how many are smallholder producers, or provide unpaid family help; how many are wage earners, or employers), their level of education, or the number of days or hours per year they devote to farming. Most of these missing data are ordinarily collected in rural household surveys or population censuses, and also in agricultural surveys or censuses, but they are collected at irregular 309

Concluding remarks

intervals and not in all countries. To date, international bodies have not been able to construct a coherent and complete data series on the size of agricultural employment that covers the same time span as other FAOSTAT series. In addition, most projections of the future performance of agriculture give only limited consideration to agricultural labour. With respect to agricultural capital, FAO recently has devoted considerable efforts to developing a statistical database on reproducible means of production for agriculture, including land improvements, irrigation works, tools, and machinery, all valued at 2005 prices. This series, however, is incomplete and defective. It starts in 1975 and only extends to 2007 - also a shorter period of coverage than other areas of FAO statistical databases. More importantly, valuation principles are quite simplistic, since they do not take into account the quality or productive capacity of the various means of production included. For instance, all tractors around the world are given the same unit value, regardless of horsepower, date of manufacture, or other characteristics. Likewise, all irrigation is given the same value per hectare, regardless of irrigation method, efficiency, or other particularities. Thus, a hectare with computerised and pressurised drip irrigation where water is pumped out from underground aquifers is given the same value as a hectare with traditional flood irrigation or gravity-driven canals, and growth in irrigation implicitly assumes that the composition of irrigated areas according to the efficiency of irrigation methods is constant. As to hand tools, a fixed sum (20 USD) is assigned per agricultural worker (as per ILO employment figures) in all countries around the world, from Gambia to Germany, from the United States to Uzbekistan. Some authors have used these data to estimate the contribution of capital investment to agricultural productivity growth, e.g., Fuglie (2012), Baldos and Hertel (2014), and von Cramon-Taubadel et al. (2011). In the case of Fuglie and also Baldos and Hertel, other complementary sources are used besides FAOSTAT data, which generally improves the quality of the calculations. We use Fuglie’s results (as applied by Baldos and Hertel) and the von CramonTaubadel study in our analysis of agricultural productivity and future prospects for undernourishment. All these authors divide output growth into increases in the use of inputs (labour, land, and capital), and increases in Total Factor Productivity (estimated as a residual). A common characteristic of these works is that improvements in the quality of capital goods or materials, not properly measured in FAOSTAT’s investment figures, is implicitly conflated with Total Factor Productivity. This in turn implies that technical progress in industries designing or producing capital goods or materials (such as tractors and seeds) 310

Hector Maletta

is confounded with technical progress in the agricultural sector itself, where these inputs are used in agricultural production. This approach (resulting in a sort of ‘vertical integration’ of total factor productivity) is a reasonable compromise, though it introduces some ambiguity into the interpretation. The estimates are to be understood as indicating net growth in productivity in the ensemble of activities that comprise agriculture itself and its backward linkages with other industries, such as those devoted to the production of seeds, pesticides, or farm machinery. Further research is needed to refine these data and analyses. Value chains in agriculture and food. FAOSTAT contains information on primary crop and livestock production, and data on major processed products such as vegetable oils, meat or sugar. However, data on industrial products based on agricultural raw materials, on the value chain of agricultural or food products, and on the agribusiness sector structure are generally insufficient. Some information can be found regarding the food and beverage industry in the statistical databases of the UN Industrial Development Organization (UNIDO), including output and employment, but these data mostly concern the corporate sector and tend to ignore artisanal or informal processing. Existing databases with world coverage do not afford an analysis of the agricultural value chain, which should include information about the structure of the chain and the proportion of each stage in the final price of the product. Taxation on agriculture and food is also not clearly accounted for. Therefore, our analysis concentrates on the value of primary production, valued at producer (farmer) price, and the unit value of farm-based exports and imports (including primary and processed products). Micronutrient consumption and deficiency. There are data on apparent consumption of dietary energy and protein, but not on apparent consumption of vitamins and minerals, nor on the prevalence of micronutrient deficiency on a world scale. The consumption of micronutrients can only be indirectly inferred from the consumption of certain foods that are rich in specific substances, such as fruit, vegetables, or milk. Micronutrient deficiency can be measured biochemically, and the extraction of blood samples for this kind of assessment is included in some nutrition surveys, but such measurements are expensive and seldom carried out. Hidden hunger thus remains exactly that: hidden, and will probably survive the eradication of dietary energy deficiency. Even after undernourishment is reduced to non-significant levels, micronutrient deficiency may persist, and detailed data will probably not be sufficiently available to assess its prevalence and depth.

311

Concluding remarks

Waste and losses. Data about waste and losses are included in FAOSTAT food balance sheets and commodity balances, but these data cover only postharvest losses occurring on-farm or at the wholesale stage and only provide partial information about losses and waste occurring at the retail stage or in households. The methodological overhaul of undernourishment presented in FAO-SOFI (2012) includes new estimates of retail losses that are not yet incorporated into the mainstream FAOSTAT series. Household waste is still generally ignored, except for a few studies on specific countries. Available food waste statistics in the United States and other developed countries are usually expressed in crude weight terms, without detailed consideration of economic or nutritional value, or non-food re-use, of wasted products. A detailed study of food waste is likely to reveal insufficiencies in data and concepts. Endogenous population. Most projections (of climate, agriculture and other matters) use the UN population projections, combining them with whatever projections of income are chosen for each particular study. This approach disregards the universally acknowledged interdependence between per capita income and population growth. The UN population projections are simply based on past demographic trends and some simple assumptions about the future, applied uniformly to all countries regardless of their particular characteristics, and ignoring projections of income, education, health care, and other variables affecting demographic variables. Together with income, education is another key variable for population projections; they concurrently affect demographic variables, especially fertility and migration, and (through income) also have an important influence on mortality. Population and income are causally inter-linked. Population should be considered as an endogenous outcome of demographic, educational, and economic factors, but this is seldom, if ever, the case. Biophysical and human factors in agriculture. Crops are not natural vegetation and livestock is not animal wildlife: they are the result of adaptive human activity involving interaction with the environment. This simple and self-evident idea is, however, too often forgotten or reduced to an afterthought or a footnote in analyses of natural constraints acting on agriculture, and especially in analyses of the impact of climate change on agriculture. As abundantly discussed in this book, this tends to be misleading. Expectations about the future of agriculture must include expectations about natural processes (such as climate change and its biophysical effects on plants and animals) but also expectations about the human side of agriculture. Facultative interventions such as building a new irrigation system with a large dam and primary irrigation canals may be considered separately, in order to 312

Hector Maletta

evaluate their costs and benefits, because they are actually facultative and the decision is (to some degree) exogenous; but spontaneous adaptive behaviour on the part of farmers is an essential component of agriculture, and should be considered on a par with biophysical processes when projecting the future of agriculture or the impact of projected climate change. Agriculture and adaptation. The theoretical underpinnings of adaptation to climate change are often formulated in a sketchy manner, and require more theoretical elaboration and methodological guidance for research. This includes adaptation as a normal component of agriculture as well as a possible response to climate change. The evolutionary notion of adaptation, here, as in other fields, is best understood as a population-level process whereby the composition of the relevant population changes over time under some selective pressures. This notion fits uneasily with some deeply ingrained analytical traditions. In agricultural science, the emphasis is mostly on the biophysical side of farming, taking for granted the corresponding agricultural human behaviour; in many cases the human side is approached from a normative viewpoint, positing what agronomists recommend doing instead of looking at what farmers actually do or may be expected to do. In the social sciences, Agricultural Economics (like other areas of economic inquiry) is largely dominated by neo-classical marginal analysis, which examines a static equilibrium situation to ascertain how that equilibrium compares to the new equilibrium that will result from a marginal (and usually exogenous) change in a particular variable, keeping other things constant and assuming that agents make choices that maximise their subjective utility so that intertemporal equilibrium is regained. In many cases, the estimation rests on partial equilibrium models. Besides other shortcomings, this approach is not well suited to analysing the actual behaviour of people in out-of-equilibrium situations, nor the development of systemic changes spanning a long period of time in which all variables will probably change and no ceteris paribus clause is appropriate. Some recent developments in Economics are more suitable for thinking about adaptation and the evolution of production systems, including Evolutionary Economics (e.g., Hodgson and Knudsen 2010), Experimental Economics (e.g., Smith 2008), or Behavioural Economics (e.g., Camerer et al. 2004). In these theoretical perspectives, economic agents are assumed to be purposive but the outcomes are not necessarily optimal (except in some trivial sense that fits any outcome), and analytical style evolves from comparative statics to more dynamic (and even historical) perspectives where the economic system is in perpetual flux and general equilibrium is never likely to be reached. Full application of these perspectives to the field of agricultural evolution and adaptation is another pending task. 313

TECHNICAL AND METHODOLOGICAL APPENDIX

12. Measuring historical trends

This part of the book explains the various sources, measures and indicators used throughout the text, especially those concerning measuring and projecting food availability and access. It also discusses methods for projecting the future of agriculture and food security, including the impact of projected climate change, among other factors. This chapter discusses technical issues related to past trends, while Ch. 13 refers to projections of the future.

12.1. Regions The main source of data on agriculture and food with worldwide coverage is the compilation and homogenisation of national statistics by the UN Food and Agricultural Organization (FAO). Our analysis of past trends is mostly based on the FAOSTAT database kept by FAO in consistent series going back to 1961. FAOSTAT includes detailed national data, the world total, and subtotals for various continental and sub-continental regions. The continental regions in FAOSTAT are Africa, the Americas, Asia, Europe, and Oceania. Examples of FAOSTAT sub-continental regions are South America, North Africa, or Polynesia. FAOSTAT data are regrouped in this book into five regions: Africa; Asia/Pacific; Latin America and the Caribbean (LAC); North America; and Europe. This required some rearrangement of FAOSTAT data: ŒŒ We divided FAOSTAT’s ‘Americas’ region into two sub-regions: North America (US and Canada); and Latin America/Caribbean (LAC), comprising FAOSTAT sub-regions ‘Central America’ (which includes Mexico), ‘Caribbean’, and ‘South America’.

317

Measuring historical trends

ŒŒ We merged FAOSTAT regions ‘Oceania’ and ‘Asia’ because of Oceania’s close relationship with Asia and its relatively small size in terms of population and food output. For the sake of brevity, we refer to ‘Asia/ Oceania’ as ‘Asia’ throughout this book. ŒŒ The whole area of the former USSR, even those areas technically located in Asia, until 1991 were counted as part of Europe. The Russian Federation (including its Asian parts) is still counted in Europe as well as the Western post-Soviet states (Ukraine, Belarus, and the three Baltic states). However, the Asian post-Soviet states, counted as being in Europe until 1991, have been classified as being in Asia since 1992; these include Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, and Uzbekistan. To ensure comparability over time, these countries are included in Europe (and subtracted from Asia) throughout this book. Thus, ‘Europe’ in our case includes not only the Asian parts of Russia (as is customary), but also the USSR’s Asian successor states. ŒŒ Sub-regions are only occasionally discussed in this book (e.g., Eastern Europe, Sub-Saharan Africa, or South Asia). In particular, some tables refer to NENA (Near East and North Africa), which includes countries in FAOSTAT’s ‘Western Asia’ and ‘Northern Africa’ regions. Individual countries are sometimes mentioned, but the analysis is centred on major regions only. ŒŒ ‘Developed’ and ‘developing’ countries, as separate groups, are not currently available in FAOSTAT, although such groups can be constructed by aggregation of national or regional data. Some other sources provide data for developed and developing countries as separate global regions (e.g., WHO data on nutrition or World Bank Development Indicators where ‘developed’ is equivalent to ‘high income’ and ‘developing’ comprises middle and low income countries). Developing countries comprise all countries in Africa and LAC, and most of Asia except Japan, South Korea, and some minor high income Asian countries such as Singapore. It is important to recall that the current status of a country in the development scale may change when it is projected into the relatively distant future, e.g., 2100.

318

Hector Maletta

How to get detailed information by country and by product Given its wide scope, and for reasons of brevity, this book presents and analyses information only at the world level and for large regions, and refers mostly to aggregate production and consumption or to some specific products (or groups of products) such as cereals or tubers. However, the main sources of information (e.g. FAOSTAT) also supply most of the data at the level of individual products and countries. FAOSTAT data is freely available to access, browse, and download at http://faostat.fao.org.

12.2. Real agricultural and food output Food comes chiefly from the main components of agricultural production: crops and livestock. Food also includes fish and seafood, but (as we shall see) statistical information about fisheries production is somewhat different from the data about crops and livestock. In addition to data on physical production, output from crops and livestock (but not fish) is also estimated in terms of monetary value at current and constant prices. Monetary valuation is important for purposes of aggregation, for comparing the output of different countries, and for monitoring growth in aggregate food output over time.

12.2.1. Product classifications FAOSTAT databases for crops and livestock includes both primary and processed products. For instance, wheat trade include the primary product (wheat grain) as well as wheat-based processed items such as flour, bread, breakfast cereals, or pastries. To incorporate trade flows into food commodity balances and food supply, some processed products are converted by FAO into their primary components. This is not easily done since a specific processed product (e.g., pastries) may involve more than one primary product (e.g., wheat, sugar, and milk). Products, whether primary or processed, are also classified as food or non-food. Food products are those known to be consumed as food by humans, even if part of their output is used otherwise; for instance, cereals are regarded as food products, even if part of the cereal output is fed to livestock or used for making biofuels. For the purpose of classifying products, FAO defines ‘food’ in a very strict sense, involving some decisions that are not immediately obvious. Non-food

319

Measuring historical trends

products include, of course, a number of products clearly not ingested by humans, such as fibres (cotton lint, wool, or jute), hides, rubber, oilseed cakes, or beeswax. More controversially, some primary products actually ingested by humans, such as stimulants like tea or coffee, are not classified as ‘food products’ in the strict sense used by FAO (because they do not provide nutrients) although they might be included in a broader definition of ‘food’ that includes food and stimulants. Salt, a mineral substance, is also not included in FAOSTAT statistics, which only cover products of vegetable and animal origin (which may, however, contain sodium).

12.2.2. Aggregation across products and countries Agricultural output comprises hundreds of different products and therefore any discussion of its size or growth must involve a method for aggregating those products into a coherently defined total. The output of single agricultural products such as wheat or lemons may be measured in physical units such as tonnes. A group of similar products (e.g., all cereals) may also be meaningfully measured in physical units. However, when heterogeneous products are considered (wheat, oranges, beef, milk, eggs, wool, and so on) a common denominator (other than physical weight) is required. Tonnage might be sufficient for calculating transportation costs (except if special care or refrigeration is needed for some particular products), but volume in cubic meters may be a more relevant measure for the purpose of storage. For the specific purpose of measuring the supply of dietary energy, the various food items may be weighted by their energy content measured in calories or joules. But dietary energy is not the only relevant aspect of food. The same food items may be aggregated according to their protein content, their Vitamin A content, or their iron content, among many other possibilities. Each of these and other metrics may be adequate for one purpose or another, but hardly for a meaningful measure of aggregate food supply or demand. For more general purposes, many characteristics are considered at the same time. Each food product may contain various nutritional components (energy, protein, and various vitamins and minerals), and their value is also affected by consumer preferences and cost of production. The real aggregate output of agriculture, as in the case of other sectors, is best measured in economic terms, i.e., by multiplying each physical quantity by a corresponding price. Prices reflect the net result of all valuations made by people in regard to a particular product of any kind, including valuations made by producers (based on costs of production using the resources and 320

Hector Maletta

technology available for its production) and those made by consumers (based on preferences), and therefore prices are the natural candidate for a meaningful aggregation of heterogeneous items such as the various crop and livestock products. Market prices, however, are affected by inflation (over time) and by price and currency differences (across countries). One unit of any currency, used in its home country, may be used to buy different amounts of wheat or milk in different years, due to general domestic inflation and also to changes over time in the relative price of those products (in relation to each other or to other products). If that same unit of domestic currency is exchanged for foreign currencies, it may also buy differing amounts of a given food item in different countries, due to inter-country differences in relative prices and to variations in the purchasing power of the original currency if exchanged into different foreign currencies. Measuring changes in supply or demand for food while ignoring these problems is surely misleading, since it may confuse a real change in output or consumption with a mere monetary difference over time (due to inflation) or across countries (due, for instance, to different levels of taxation, or misalignment of exchange rates). Aggregating agricultural output to make meaningful comparisons over time and space thus requires correcting for these problems (domestic inflation, differences in relative prices over time and space, and variable purchasing power of a currency when used in different places). This leads to the concept of real output comparisons over time and space, involving not only the use of constant prices over time but also converting currencies at rates ensuring Purchasing Power Parity (PPP). Under this approach, which is standard in economic comparisons, products are aggregated at world-average prices that are constant over time and uniform across countries. This approach is particularly suitable for a discussion of food security, a concept centred on economic access to food: the real economic value of food, compared to the level and distribution of real incomes, is an adequate measurement of access to food. By the same token, since farms may grow a variety of crops or produce a variety of livestock products, subject to technical possibilities of production as well as market prices and other economic factors, meaningful aggregation of agricultural or food output should be done in real economic terms, as is also the case for other sectors of economic activity, for the purpose of measuring GDP, or National Income. Consistent with this approach, FAOSTAT provides estimates of the value of agricultural production, covering crops and livestock valued at producer prices 321

Measuring historical trends

of a chosen base period, by product and country. These values are available in FAOSTAT for every year since 1961, expressed in constant worldaverage producer prices of the reference period (converted into USD at PPP conversion rates). For years after 1990, value of production is also available at current prices in local currencies and in current USD calculated at market exchange rates. The concepts and methods involved in these valuations are discussed in the next section.

12.2.3. Agricultural PPP conversion rates and world-average prices A good measure of real agricultural output, allowing for meaningful aggregation or comparison across borders and over time, should correct for variation in the purchasing power of money, both over time and across countries. Even if different measures of real output may be expressed in different units of measurement (e.g., in dollars or pounds, normalised at constant prices of any chosen base year), a good measure of real output should be a reliable measure of real growth that is as independent as possible from the above factors (inflation, differences in purchasing power, and changes in relative prices). This is achieved by using a single set of constant world-average prices converted into a common currency by means of a set of Purchasing Power Parity (PPP) conversion rates that adjust for differences across countries in the purchasing power of money, and also for inflation within each country. To comply with this methodological requirement, in this book we use FAO estimates of the real food (or agricultural) output, valued at constant and uniform producer prices. These prices are the output-weighted world average of domestic producer prices in 2004-2006, calculated originally in national currencies and converted into international USD at market exchanged rates, adjusted with agricultural purchasing power parity (PPP) conversion rates.34 Making real comparisons over space is conceptually similar to making real comparisons over time. Both purposes entail the construction of index numbers. An index number is the ratio of a target value to a base value, such as VA/VB for the ratio of the values of output in countries A and B, where A is the target country, and B the base country. This ratio measures the output of country A in units of B output. If the two values pertain to the same country, The PPP conversion rates calculated by FAO for agricultural products do not necessarily coincide with the PPP conversion rates calculated for all goods and services, or for all consumption, as those produced by the International Comparison Programme of the United Nations (for its 2005 version see WB [2008]; for the 2011 update, WB [2013]).

34

322

Hector Maletta

or two countries using the same currency, all prices are denominated in the same measurement unit (the single currency). For international comparisons between countries with different currencies, however, price comparisons involve dealing with two elements: the various prices (expressed in each domestic currency), and conversion rates of all domestic currencies into a common reference currency. A common instance of an index number is the Consumer Price Index (CPI), which compares the price level of period t relative to a base period (t0). The price level, in this case, is defined as the price of a certain basket of goods and services. Evaluating the price level involves computing the value of this basket at the prices prevailing at time t, and the value of the same basket at the baseperiod prices. Taking the quantities consumed at time 0 as constituting the reference basket, the price index is:

This ratio compares the price of a basket of N goods, taken in fixed quantities qi (i=1, 2, ..., N) at prices of period t, to the value of the same basket at prices of period 0. Since the quantities are the same, the ratio can only vary due to changes in prices. Likewise, a quantity index compares two different baskets, corresponding to times 0 and t, valued at the same prices of period 0. If the base prices are the ones prevailing at time 0, the quantity index is:

This measures the ratio of quantities of time t, valued at prices of time 0, to quantities of time 0 valued at the same prices. In summary, a price index measures change in prices for fixed quantities; a quantity index measures change in quantity for fixed prices. Measuring agricultural or food output requires a quantity index. The above formulas for price and quantity indexes are Laspeyres indices, because they use the prices or quantities of the base period (time 0) as the reference. Indexes using the quantities or prices of the target period (time t) as the reference are called Paasche indexes. Each is to some degree affected by the choice of the reference quantities (or prices). An average of the two (the so-called ‘ideal’ index) is often preferred, though not always calculable on a 323

Measuring historical trends

regular basis. Besides price and quantity indexes, there is also the value index, which is simply the ratio of the two values, such as the value of the basket of time t, valued at prices of time t, to the value of the base basket of time 0 valued at prices of time 0; let us call this index Wt:

All these indexes equal 1 at the base year: 1. One desirable property of quantity and price indexes is factor reversal. It requires that the value index should equal the product of a quantity index multiplied by a price index, in the form of Wt=QtPt. Thus, the quantity index should be Qt=Wt/ Pt and the price index should be Pt=Wt/Qt. However, for this property to hold exactly in this type of index, the price and quantity indexes cannot both be of the same type (i.e., both Laspeyres or both Paasche). Given a value index, i.e., a ratio of final to initial value, Wt, any Laspeyres price index has an implicit quantity index, . This implicit quantity index readily emerges from its definition:

This shows that, if the factor reversal property must hold, the quantity index implied by a Laspeyres price index is a Paasche quantity index, i.e., a quantity index based on target-period prices. If the price index is Laspeyres, the implicit quantity index should be Paasche. The converse is also true. Thus, the Laspeyres and Paasche indexes are closely related in this particular sense: if you have a Laspeyres price index you automatically have a Paasche quantity index, and vice versa; from a Laspeyres price index you cannot get a Laspeyres quantity index so that between them they satisfy the factor reversal condition, and therefore the Laspeyres index does not have the factor reversal property.35 The factor reversal property holds for bilateral indexes of the same type, such as the Fisher index, which is the geometric mean of a Laspeyres and a Paasche index; if L is a Laspeyres index and P a Paasche index, the ‘ideal’ Fisher index is The factor reversal property holds for Fisher’s index: a value index W=Wt /W0 equals the product of two Fisher indices for prices and quantities: W=FqFp. Fischer’s index and other similar indices are bilateral since they use weights from both the base and the target years. 35

324

Hector Maletta

Considerations exemplified above for indexes over time also apply to indexes measuring differences over space (across borders). One only needs to replace periods 0 and t with countries A and B or, more generally, replace the t indicating time by a j denoting one of k countries (j=1, 2, ..., k). The Purchasing Power Parity (PPP) approach for international comparisons rests on principles similar to the factor reversal property. The ratio of values for two countries should equal the product of a quantity ratio and a price level ratio. Thus, the quotient between the value of output in two countries (WA/ WB or WAB) should equal the product QABPAB. The PAB factor measures the price level in country A as a proportion of the price level in B. QAB measures the real level of output of A as a proportion of the output produced in B. This real output ratio is QAB=WAB/PAB. However, in this case, the construction of such indexes involves not only quantities and (domestic) prices, but also the use of adequate conversion rates to translate the various currencies into a common currency in such a way that the aforementioned property is obtained. As with the other quantity index, QAB should be calculated on the basis of a common set of prices for evaluating the output of both countries. At the same time, since each country uses a different currency, and market exchange rates are volatile and affected by many factors, it also involves the estimation of a set of PPP conversion factors to modify market exchange rates, and thus to express all monetary figures in a common currency in such a way that real comparisons can be made between values of different countries. To achieve this we have to determine a set of prices and a set of conversion rates. There are several ways of achieving this goal, with different virtues and shortcomings. FAO aggregates agricultural output values using the so-called Geary-Khamis method (Geary 1958; Khamis 1970, 1972, 1984), which simultaneously calculates international prices πi for all commodities and PPPj conversion factors for all countries. In this approach, the uniform prices πi are the world average of domestic prices, weighted by output quantities and translated into a common currency at PPPj rates, ensuring uniform purchasing power. These prices are calculated for a reference period (currently, FAOSTAT uses the mean prices of 2004-06). These international prices (π) of constant and uniform purchasing power, and expressed in US dollars through PPP conversion rates, are interchangeably referred to as expressed in PPP USD or ‘international dollars’. The iterative procedure used by FAO for this purpose is explained in detail by Rao (1993). We provide a summary here. We deal with many commodities Xi, where i varies from 1 to N, and various countries Zj where j varies from 1 to K. Each country has a national currency, and the price of each commodity at 325

Measuring historical trends

time t, expressed in units of local currency, is pij. The corresponding quantity is qij. Prices and quantities always refer to a period of time (t), but, for the moment, the time reference can be omitted because we ignore changes over time in quantities or prices; instead, we concentrate on spatial comparisons and aggregations at a single time. The domestic price of commodity i in country j, originally expressed in the local currency, can be converted to a reference currency (e.g., the US dollar) where r is the at the market exchange rate. This dollar price is denoted by reference currency. However, since market exchange rates may be misaligned, and dollar-equivalent prices differ across countries, this would imply that the same physical amount of product may be expressed in different amounts of dollars depending on the country involved. To achieve equality of purchasing power, market exchange rates should be adjusted by so-called PPP factors, leading to the world-average price of a commodity i (over the K countries considered), expressed in international dollars of uniform purchasing power. Such international prices for a given base period are defined for each commodity by equation [1]: [1] To estimate these prices we need an estimate of the PPPj conversion factors to turn prices in US dollars at market exchange rates into prices in international dollars of uniform purchasing power. These PPP factors indicate the relative price level of a country. Purchasing power parities in the Geary-Khamis approach are defined for each country by equation [2]: [2] This definition states that the price level of country j relative to the price level of the reference country (e.g., the United States) equals the ratio of the local output of all commodities valued at domestic prices at market exchange rates, , to the value of the same output valued at the international prices πj. A complete calculation of the N prices and K conversion factors involves solving an unwieldy system of N+K simultaneous equations with N+K unknowns. Khamis (1970, 1972) and Rao (1971) have shown that this system has a unique positive solution if one of the unknowns is fixed at any

326

Hector Maletta

arbitrary level (for instance, by assuming that one of the PPPj or one of the πi equals 1). Once this or a similar decision is taken, all the PPPj factors and all the international prices πi are uniquely determined. Typically, one currency is chosen as the reference currency, and therefore its own PPP will equal 1 (e.g., one US dollar will equal one international dollar), and this determines the rest of the unknowns. Solving these equations, however, can be cumbersome when many countries and commodities are involved (FAOSTAT value-of-production data refer to about 130 countries and about 190 commodities). A simpler iterative algorithm is usually applied to approximate the solution. For this iterative algorithm, one starts with arbitrary values for the PPPj conversion factors. For instance, the initial values (where the superscript indicates the iteration cycle) may be set initially at unity for all countries (j=1, 2, ..., K):

This initial assumption is equivalent to assuming that market exchange rates are aligned at purchasing power parity, but that is just a starting point. Using these or any other initial values for the conversion factors in equation [1] leads to an initial estimate of the international prices, . These initial price estimates are then used in equation [2] to get a new estimate for each PPPj, possibly different from the uniform unity value initially adopted. If this second set of estimates for prices and PPPs is different from the first, a new iteration is performed, obtaining a second set of prices and conversion factors . Iteration continues until every PPPj conversion factor and every international price πi converge to stable values. Convergence is assured, as demonstrated by Rao (1971) and Khamis (1972), and is normally attained with enough precision in relatively few rounds of iteration. Uniform purchasing power refers to the set of commodities considered, i.e., in this case the set of agricultural commodities included in the calculation. The reference ‘basket’ for the price level index is composed of all agricultural commodities, taken in the quantities produced in the world at the reference period and valued at the world-average prices prevailing during the same period.

327

Measuring historical trends

12.2.4. Alternative base periods for PPP rates and producer prices As with any index number based on a given reference period, the FAO series of agricultural value of production at 2004-2006 prices (converted into USD at PPP rates) are, to a degree, dependent on the choice of reference period. If the base year is changed, the consequent changes in relative prices (between products) and in relative purchasing power of money (between countries) may affect key results, e.g., the rate of growth of agricultural production, or the share of each country or region within world agricultural output. The possible effect of the choice of base period can be analysed by computing the same results (e.g., the growth of agricultural output) using some other reference period. FAO data in international dollars at world-average producer prices are derived from FAOSTAT data on domestic prices in both local currency and current US dollars (at market exchange rates), which are only available for the more recent period starting in 1991. There are no values at current prices before that date. Moreover, the economic and monetary upheaval caused by the collapse of the Soviet block and the subsequent creation of new countries with new currencies, many with imperfect statistical reporting and poor price statistics, makes it unadvisable to use data from the early 1990s. For our purposes, we developed a new series for total real agricultural output with two alternative base periods: 1995 and 2010, which can be compared with the official FAO series based on the 2004-2006 period (centred on 2005). The new sets of prices and PPP conversion rates were estimated by means of the iterative procedure described in Section 12.2.3. The procedure converged quite rapidly. In the case of 1995, by the sixth iteration the maximum percentage change between prices of consecutive iterations was down to 0.22%; by the eighth iteration, the maximum percentage change was 0.05%, and by the tenth iteration it was 0.01%. We used the prices and PPP conversion rates resulting from the tenth iteration. Results for 2010 were calculated in the same way and converged in a similar fashion. The list of countries and products included for 1995 and 2010 was practically the same as for 2004-2006, with only some minor changes due to data availability. The results of this exercise (Table 66) show that changing the base from 20042006 to 1995 or 2010 induces only negligible changes in the real growth rates of world agricultural production. Even if relatively large swings occurred between these periods in agricultural commodity prices and in the relative worth of currencies, the change of base period would cause only minor and generally non-significant changes in the growth rate of real agricultural output. In view of this, we use FAOSTAT estimates based on 2004-2006 328

Hector Maletta

throughout this book; the conclusions would not be significantly altered by the use of alternative reference periods. Table 66. Annual growth rates of gross real agricultural output at the world level, 1961-2011, at world-average producer prices converted into international dollars at PPP conversion rates, with three different reference periods for producer prices and PPP conversion rates. Reference period 1995 2004-2006 2010

1961-2011 2.37% 2.37% 2.34%

1961-1971 2.74% 2.77% 2.79%

1971-1981 2.16% 2.19% 2.17%

1981-1991 2.13% 2.11% 1.98%

1991-2001 2.21% 2.15% 2.22%

2001-2011 2.61% 2.61% 2.55%

Source: Results based on 2004-2006: FAOSTAT data on value of gross agricultural production in international 20042006 dollars. Results based on 1995 and 2010 reference periods calculated from FAOSTAT data by product and country (physical output and value of gross production in current dollars at market exchange rates).

12.2.5. FAO world-average 2004-2006 prices in PPP dollars FAO has not published the 2004-2006 world-average prices of agricultural products on which the value of production is based. They may be estimated, however, as unit values (ratio of output value to output tonnage), using FAO figures for tonnage and value of production (Table 67). Tonnage is available at the world level for most products included in value of production, and products without tonnage are of very limited importance in terms of value of production. Products with value but not tonnage data are included, however, in the total value of production. Table 67. World average farm producer prices, 2004-2006, in international dollars per metric ton, at agricultural PPP conversion rates. Product Wheat Rice, paddy Barley Maize Rye Oats Millet Sorghum Buckwheat Quinoa Fonio Triticale Canary seed Mixed grain Cereals, n.e.s. Nuts, n.e.s.

USD/MT 157.78 278.66 118.98 141.67 112.06 114.21 181.45 153.80 217.07 645.23 391.05 137.09 272.68 154.44 261.68 1833.28

Product Potatoes Sweet potatoes Cassava Yautia (cocoyam) Taro (cocoyam) Yams Roots/tubers, n.e.s. Sugar cane Sugar beet Sugar crops, n.e.s. Beans, dry Broad/horse beans, dry Peas, dry Chick peas Cowpeas, dry Melonseed

329

USD/MT 168.78 75.53 104.46 487.00 212.09 255.04 171.01 32.84 43.01 29.49 601.39 355.45 190.88 484.07 335.78 430.05

Product USD/MT Pigeon peas 534.23 Lentils 408.36 Bambara beans 196.81 Vetches 176.12 Lupins 173.57 Pulses, n.e.s. 555.83 Brazil nuts, in shell 896.84 Cashew nuts, in shell 875.31 Chestnuts 777.68 Almonds, in shell 2950.97 Walnuts, in shell 1552.62 Pistachios 3284.14 Kolanuts 599.40 Hazelnuts, in shell 1602.91 Arecanuts 1747.31 Onions and shallots, green 204.06

Measuring historical trends Product USD/MT Product Soybeans 274.29 Kapokseed in shell Groundnuts, in shell 451.14 Cottonseed Coconuts 110.57 Linseed Palm kernels 258.12 Oilseeds, n.e.s. Palm oil 435.06 Cabbage/brassicas Olives 800.70 Artichokes Karite nuts (sheanuts) 136.38 Asparagus Castor oil seed 390.25 Lettuce and chicory Sunflower seed 275.22 Spinach Rapeseed 278.94 Tomatoes Tung nuts 137.72 Cauliflowers, broccoli Safflower seed 315.18 Pumpkin/squash/gourd Sesame seed 676.93 Cucumbers/gherkins Mustard seed 410.22 Eggplants (aubergines) Poppy seed 811.89 Chillies/peppers, green Plantains 206.46 Strawberries Oranges 193.26 Raspberries Tangerine/mandarine/clem. 247.02 Gooseberries Lemons and limes 396.48 Currants Grapefruit (incl. Pomelos) 224.84 Blueberries Citrus fruit, n.e.s. 452.05 Cranberries Apples 422.91 Berries, n.e.s. Pears 408.83 Grapes Quinces 394.82 Watermelons Apricots 552.11 Melons, cantaloupes Sour cherries 610.62 Figs Cherries 1271.27 Mangoes, guavas Peaches and nectarines 544.42 Avocados Plums and sloes 596.77 Pineapples Stone fruits, n.e.s. 799.56 Dates Pome fruits, n.e.s. 217.24 Persimmons Anise/badian/fennel/coriander 5527.22 Natural rubber Ginger 677.23 Gums natural Spices, n.e.s. 695.16 Cow milk Pyrethrum, dried 951.87 Cattle meat Cotton lint 1429.20 Buffalo milk Flax fibre and tow 471.61 Buffalo meat Hemp tow waste 227.52 Sheep milk Kapok fibre 217.70 Wool, greasy Jute 283.21 Sheep meat Other bast fibres 479.96 Goat milk Ramie 594.45 Goat meat Sisal 592.68 Pig meat Agave fibre, n.e.s. 489.62 Hen eggs, in shell Manila fibre (abaca) 738.81 Other eggs, in shell Fibre crops, n.e.s. 498.73 Duck meat Tobacco, unmanufactured 1592.76 Geese meat

USD/MT 82.23 330.04 303.21 304.69 149.64 720.39 910.16 467.51 234.19 369.56 239.57 175.34 198.55 213.80 470.76 1357.28 1935.00 1407.69 894.97 2531.49 829.38 1742.28 571.62 113.92 184.09 597.00 599.17 692.96 285.05 510.70 322.57 1143.83 336.19 312.06 2701.38 398.87 2691.74 389.41 1913.12 2722.80 335.58 2396.09 1537.24 829.39 2884.22 1647.57 1642.32

Product USD/MT Onions, dry 210.03 Garlic 526.34 Leeks, other alliaceous veg 894.73 Beans, green 355.56 Peas, green 330.95 Leguminous veg, n.e.s. 343.79 String beans 953.54 Carrots and turnips 249.50 Okra 639.49 Maize, green 413.82 Mushrooms and truffles 1804.26 Chicory roots 776.81 Carobs 210.33 Vegetables, fresh, n.e.s. 188.44 Bananas 281.63 Cashewapple 943.66 Kiwi fruit 815.69 Papayas 283.80 Fruit, tropical fresh, n.e.s. 408.67 Fruit fresh, n.e.s. 349.03 Coffee, green 1074.36 Cocoa beans 1038.49 Tea 1063.48 Maté 130.89 Hops 3401.65 Pepper (piper spp.) 2084.30 Chillies and peppers, dry 1095.43 Vanilla 16601.35 Cinnamon (canella) 1392.09 Cloves 2216.58 Nutmeg/mace/cardamom 2082.24 Turkey meat 1306.94 Chicken meat 1424.41 Bird meat, n.e.s. 2930.54 Horse meat 1557.65 Ass meat 550.44 Mule meat 527.91 Camel milk 340.97 Camel meat 2095.75 Rabbit meat 1857.96 Other rodent meat 818.69 Other camelid meat 2137.84 Game meat 2175.87 Meat, n.e.s. 1321.56 Honey, natural 2509.43 Beeswax 9363.76 Silk-worm cocoons 3148.98

‘n.e.s.’ = not elsewhere specified. Prices estimated as the ratio of value of production to tonnage produced, as per FAOSTAT (Oct. 2013). Prices are worldwide output-weighted means of national producer prices in domestic currency (average of 2004-2006), converted into USD at Purchasing Power Parity conversion rates. See Rao (1993) for methodology. All milk prices refer to fresh whole milk. Meat prices refer to carcass weight of indigenous meat (excluding imported meat). Mangoes include mangosteens (a Malayan fruit). Non-food products are in italics. For unclear reasons ‘kiwi fruit’ and ‘leeks and other alliaceous vegetables’ are mentioned in FAOSTAT in both the food and non-food categories; they are considered as food products here.

330

Hector Maletta

By using the national value of agricultural production in 2004-06 dollars at market exchange rates, compared to the corresponding PPP values, one may also derive the PPP conversion factors used by FAO, associated with the prices in Table 67; for brevity’s sake these factors are not reproduced here. The fact that the prices used for aggregation are producer prices implies that they reflect the value of primary farm products and the revenue of farmers, not including in principle the added value derived from marketing and processing. The fact that prices are constant over time adjusts for inflation, i.e., for price changes over time. The fact that these prices are also uniform across countries, i.e., that one single set of prices is used for all countries, corrects for differences in prices across countries, thus valuing each product at the same price in all countries at all times. The fact that these prices are expressed in a single currency with uniform purchasing power over agricultural products, ensures the results are not only expressed in a common currency but also adjusted for differences in the purchasing power of international dollars over agricultural products in different countries (such differences may emerge if local prices were converted into dollars at official or market exchange rates). The result is a measure of real agricultural output, comparable over time and across nations. Most figures regarding past trends in agricultural output presented in this book are FAO data on value of production, expressed in PPP terms, and cover only crops and livestock, excluding fish and seafood. FAO does not produce PPP estimates of the value of fish and seafood, which can only be aggregated in physical terms, or at some other set of prices, e.g., international export prices (for instance, the FAO Fish Price Index); the same occurs with the value of agricultural trade flows expressed in current or constant dollars at market exchange rates without PPP adjustment. By the same token, some projections of future trends, especially those concerning the impact of climate change, have used market exchange rates to aggregate national GDPs (total or agricultural) into regional or world totals. Estimating world or regional production growth in US dollars aggregated by means of market exchange rates tends to understate the relative weight of poorer countries, where items are cheaper in dollars, and to overstate the weight of more developed countries with higher price levels. Since developing countries usually grow faster, their reduced weight also translates into an understatement of aggregate growth rates. Analysis of past trends can be carried out with PPP valuation of real output, but it should be noted that many available estimates of future output and income rely on the market exchange rates of the respective base period, which are also used for future dates; they are therefore vulnerable to distortion due to this 331

Measuring historical trends

source. For instance, the same physical growth in, say, maize production will have a different impact on total output depending on whether it occurred in the United States or in China. To avoid these misleading results, PPP exchange rates and world-average prices should be used.

12.2.6. Other related technical issues Countries included in agricultural PPP rates estimation. FAO estimates value of production for practically all countries where physical production is reported; only very few are excluded, due to inadequate statistics. However, the estimation of PPP conversion rates and the corresponding prices in PPP dollars are based on a large set of countries in which some countries are also not included. These exceptions include mainly countries with poor or non-existent producer price statistics, such as Afghanistan, Mauritania, Sierra Leone or the Democratic Republic of Congo, as well as a number of small countries and territories (chiefly island countries such as Saint Lucia, the Solomon Islands, or the Seychelles), and some Gulf states with little agricultural production such as Oman, Qatar or Kuwait. The estimation of 2004-2006 prices was based on 130 countries and 189 products. The resulting world-average prices were also used to estimate the value of agricultural production in all years and countries, even for the above-mentioned countries that lack domestic price data but have information on physical production. Product coverage for agricultural PPP value of production. The agricultural sector, broadly defined, includes crops, livestock, hunting, gathering, forestry, and fishery. However, FAO data on the value of agricultural production include only products from crops and livestock. In these sectors, the coverage of physical production is practically complete, with only few exclusions due to lack of consistent data across countries; for instance, there are no data on the production of flowers; cultivated meadows and pastures are not considered as a separate category of agricultural production but are merged with naturally grown meadows and pastures. Some minor products or by-products included in FAO estimates of physical production are not included in FAO estimates of value of production, especially in cases where several countries fail to report on the quantity, the price, or both; products not included in the series on value of production include animal hides or skins, offals, snails, cassava leaves, hempseed, and a few more; their exclusion does not have a significant impact on total value of production, and even less so on growth rates. Moreover, the main purpose of the FAO series on PPP value of production is to provide an index of real growth rates of production and comparable estimates of agricultural output across countries, even if some minor products are not included. 332

Hector Maletta

Some hunting and gathering products are in fact included. ‘Game meat’ is a category in production and value-of-production series (see average PPP price in Table 67) but the coverage is probably incomplete as much hunting goes unreported; likewise some products included in physical and value terms are in part gathered (e.g., mushrooms and truffles). Fish and seafood is not included in FAOSTAT value of agricultural production, but the corresponding physical production statistics are available. The value of production of fish and seafood is available for recent years (e.g., FAO (2012a) and, since the 1990s, FAO regularly issues a Fish Price Index; however, these price and value data are based on export prices, not on producer prices, and do not use PPP conversion rates. Food and non-food products. Farms produce food and non-food products (examples of non-food products include: wool, hides, tobacco, jute, and cotton lint). FAO’s data on value of production show total agricultural production and food and non-food products separately (as well as product groups such as cereals or vegetables, and specific products such as milk or wheat). A product is classed as a food product if it is used for food, even if it is also used as animal feed or for other purposes. Thus, for instance, maize is classified as a food product, although much maize is used as animal feed. Stimulants (coffee and tea) are counted as non-food products, though, of course, they might conceivably be included in a broader category of farm food and stimulant products. Other infusions such as maté (ilex paraguariensis) and herbal teas are counted as food. As remarked in the note below Table 67, ‘kiwi fruit’ and ‘leeks and other alliaceous vegetables’ appear listed in FAOSTAT as both food and non-food items, but the latter is probably a misprint; they are actually treated as food (and not as non-food) in FAOSTAT and in this book. Primary and processed products. FAO’s definition of ‘agricultural output’ for its series on value of production refers mainly to primary farm products such as cereal grains or oilseeds, but also includes a number of products that have undergone some degree of industrial processing, e.g., meat and offals, mostly produced not on farms but at slaughterhouses. Some animals are in fact slaughtered on farms, but most livestock producers usually sell live animals and are paid per head or per kilogram of live weight, whereas meat is valued per kilogram of carcass weight. Prices of meat as sold by slaughterhouses cover not only the producer price of the live animals but also a margin for transportation to the slaughterhouse and the cost of killing the animals and of producing meat and other by-products such as offals, hides, or skins. Likewise, the cotton value of production is expressed by the separate value of cotton seeds and cotton lint, i.e., ginned products, not by ‘seed cotton’ which is the actual product harvested in cotton fields (including seed and lint). 333

Measuring historical trends

Cotton seed is food and cotton lint is not; hence the distinction is necessary to calculate the total value of production of food and non-food products. Producer prices are often thought of as ‘farm-gate prices’, but not all products are actually sold at the farm’s gate; the more sensible operational definition is ‘prices received by producers’. Even for primary farm products such as grain or vegetables, data may in fact refer to prices received by producers at wholesale markets or other points of first sale, and are thus likely to include some margin for transportation to the point of sale; in some cases the price may include some other post-harvest costs borne by the producer, such as storage, refrigeration, or classification, so the price is thus to be interpreted as ‘as sold’ and measured at the place of first sale. Gross and net output. FAO provides figures at constant prices in agricultural PPP dollars for the value of gross and net agricultural (and food) output. Gross output value is simply the value of the total amount produced (e.g., the tonnage of maize that was harvested, multiplied by its price). The net output value of a given product in a given country equals gross output minus the intermediate use of the product for seed or feed within the agricultural sector of the same country. Exported products that end up used as feed or seed in other countries are included in the net output of their country of origin, but deducted in the country of destination. The amounts used as seed or feed in a given country may have come from imports; the FAOSTAT metadata provides this explanation: ‘Deductions for seed (in the case of eggs, for hatching) and for livestock and poultry feed apply to both domestically produced and imported commodities’. Agricultural inputs that do not originate in the farming sector (such as pesticides) are not deducted. Hence FAOSTAT’s ‘net output value’ is not equivalent to agriculture’s value added; the only purpose of computing net output is to measure the value of farm output as it leaves the farm sector. Depending on the goals of an investigation, analysis may focus either on gross or net value of output. Agricultural land. Farm production involves the use of land for growing crops or raising livestock. It includes three mutually exclusive major categories of land use at any particular time: - Arable land - Land with permanent crops - Permanent meadows and pastures 334

Hector Maletta

Arable land is land where temporary crops are grown, i.e., crops cultivated and harvested within one agricultural year; for any given year, arable land includes land cultivated during the year as well as land that is lying fallow that particular year but is regularly cultivated. Land with permanent crops includes land covered with any crop lasting more than a year (e.g., fruit trees, coffee, and sugar cane), except cultivated pastures such as alfalfa. Permanent meadows and pastures, in turn, include naturally grown grasses as well as cultivated pastures. FAO statistics on permanent meadows and pastures include separate categories for natural and cultivated meadows and pastures, but few countries report separate figures for these two kinds of meadows; thus, for world or regional totals, the two categories cannot be separated. On the other hand, livestock is fed in various ways. It may graze on permanent meadows and pastures (natural or cultivated), graze on temporary crops intended for forage or silage (e.g., barley), be fed fodder harvested from permanent or temporary pastures (e.g., hay), or fed products such as coarse cereal grain (e.g., maize) or soybean cakes (usually delivered at feedlots). Cropland is defined as the sum of arable land plus land with permanent crops, and is unequivocally related to crop output. Livestock production, on the other hand, is not easily associated with a particular area of land, since livestock feed may come from permanent pastures but also from (permanent or temporary) crops, forested land, or other sources. Cropping intensity is defined as the number of annual harvests per unit of arable land. Fallow land is not harvested at all in a given year; most land under crops is harvested once per year, but some is harvested twice or more, from two or more successive crops grown on the same piece of land; this is most often done under irrigation and for short-cycle crops. Harvested land is the sum of all areas harvested in a year; for double or triple cropping, each hectare is counted as harvested every time it is harvested during the year. Thus, the total harvested land may be larger than arable land (though it may also be smaller due to fallow land, crop failure, or permanent crops still in the growing stage). To count the number of harvests per year on a single piece of land, the general principle is to count one harvest for each successive crop planted and grown during the year (i.e., successive temporary crops). When there are several gatherings from the same standing plants, as is often the case with permanent crops (e.g., when tea leaves are collected from the same tea plants at various times during the same year), the whole process is counted as a single harvest. The same principle applies when a temporary crop, e.g., manioc, is gradually harvested as the products mature over the year from the same standing plants. 335

Measuring historical trends

Value of agricultural capital. FAOSTAT has recently added a database on the stock of agricultural fixed capital, though it is still preliminary. At the time of writing, it covers the period 1975-2007. The agricultural capital stock series in FAOSTAT includes land improvements, livestock, farm equipment and infrastructure, and permanent crops, valued at 2005 prices. Among other insufficiencies of the data is that some of the estimates are based on simple proxies such as land area or number of machines, without adjusting for differences in quality (e.g., between different soils across various models and sizes of machines, between different permanent crops or between irrigation works of different type and efficiency). FAO’s methodological documentation on its fixed agricultural capital stock database (http://faostat.fao.org/site/660/ default.aspx#ancor) explains that ‘it was not possible to make any adjustment for quality change/variation on physical assets, particularly regarding the varieties of machinery’. This means, for instance, that a tractor or harvester existing in a given country in 1975 is assumed to be equivalent in value to a tractor or harvester existing in any other year or country, and is assigned the same price (in constant 2005 dollars), regardless of any qualitative differences between them. For instance, the tractor stock of 1975 might have a smaller average horsepower than the units existing in 2007; likewise, tractors in Vietnam or China may have a different mean horsepower and capabilities than those in the United States or Brazil. Because of these inherent problems, the usefulness of the FAOSTAT capital stock series is, for the time being, quite limited. Population. The UN Population Division estimates are used by FAO (and other organisations and researchers) to estimate the population of the world or its regions. These estimates are updated every two years. These biennial revisions only slightly modify past estimates, as new censuses or surveys are incorporated into the database; the adjustments usually affect only the most recent years, where information for some countries had been only preliminary. Most of the analyses in this book use the 2008 or 2010 revisions of the UN estimates.

12.3. Agricultural and food trade Trade as a general economic process includes both domestic and international trade. However in the present context most data are at the scale of nations, and thus trade data refer to international trade. From the point of view of food security (interpreted in terms of food access by individuals) domestic trade is equally important, but in the present context, we mostly refer to foreign trade. 336

Hector Maletta

Total world exports and total world im­­ports must (in theory) be equal. However, they do not coincide in practice. Trade is ordinarily valued at bor­der pri­ces: exports are thus valued by countries in FOB (free-on-board) terms, and imports in CIF (cost, insuran­ce and freight) terms. Correct comparison of real agricultural ex­ports and im­ports involves using the same valuation principle for both; for instance, using FOB prices for both flows. CIF prices inclu­de the price of the traded agricultural good at the point of origin, plus the cost of transportation and insurance, and thus reflect something more than the value of tra­ded agri­cultural products; notice that transportation and insurance costs for an imported good may be themsel­ves imported (supplied by foreign transportation and insurance companies) or pro­vided by domestic companies. Besides the differences arising from CIF and FOB va­ luation of trade flows, several other factors may pro­ duce small sta­ tis­ tical errors and inconsistencies in trade data provided by the various countries, as described in FAOSTAT metadata (http://faostat.fao.org/site/362/DesktopDefault. aspx?PageID=362). These factors include ti­me lag (goods exported in December may reach destination in January, cau­sing the shipments to be classified in different years for exports and imports); reporting period (most coun­tries report by calendar year, January to December, but a few use other periods); mis­class­ification (a commodity may be classified differently at origin and destination countries); free zones (in some countries, goods may or may not be regarded as imported or exported when they enter or exit a free zone); in route losses (some shipments are lost or des­troy­ed be­fore reach­ing destina­tion); reporting errors (some reports may contain involuntary typos or errors); con­fi­dentiality (tran­sact­ions not included in statistics due to various reasons); smuggling and inform­al trade (not passing through cus­toms); and misinvoicing (deliberately over-pricing or un­der-pricing ship­ments in order to avoid taxes or currency controls, or other similar reasons). The net effect of these various factors may be to amplify or reduce the FOB-CIF gap between ex­ports and im­ports. Thus the world’s total exports do not exactly coincide with the world’s total imports, creating the false impression that the world’s trade is slightly off balance. As a net result of these various statistical problems, linked to the way foreign trade is accounted for in exporting or importing countries, the reported value of world imports is slightly higher than the reported value of world exports (Figure 65 shows the discrepancy).

337

Measuring historical trends

Figure 65. World agricultural exports and imports at constant 2004-2006 prices. 1000 Billion USD, at 2004-06 prices

900 800 700 600 500 400 300 200 100 0

1961

1971

1981 Exports

1991

2001

2011

Imports

Source: FAOSTAT

Intra-regional trade. FAOSTAT trade data come from national statistics. This implies that, for a given continent or re­gion, aggregate exports or im­ports in FAOSTAT are the sum of the ex­ports of the respective coun­tries, including flows within the sa­me region; thus total agricultural exports for a region (e.g. Afri­ca) in­clu­de also intra-region trade, and should not be construed as the amount of exports ‘lea­ving the region’. Likewise, total imports in­to a region’s countries should not be seen as imports ‘en­tering the region’ but ‘entering the countries’. Regional imports or exports are just the value of products entering or leaving the coun­tries of that region. Units of measurement. For most commodities, FAOSTAT statistics provide the physical amount traded and the corresponding value. Physical flows are expressed generally in terms of weight (in tonnes), though a few products are expressed in number of physical units (e.g. head count of live ani­mals traded). The monetary value of trade flows of agricultural (food and non-food) pro­­­­­ ducts is provided by FAO in cur­rent US doll­ars. FAOSTAT does not provide an account of trade flows at constant and uniform prices, as it does with production. In this book, an estimate of real flows is presented, in the form of imports and exports at constant 2004-06 prices (i.e. the average unit value of each commodity in the reference period 2004 to 2006).

338

Hector Maletta

As we have seen in the case of output, a measure of aggregate real quan­­ tities would imply adjusting for price changes over time and price differences across countries. Sin­ce ex­port/im­port prices are ordinarily de­no­mi­na­ted in USD for most traded commodities, and goods have the same value for the exporter and the importer country, mis­­­al­ign­ments in the pur­chasing power of currencies at market or official ex­chan­­ge rates are of limited im­port­an­ce. Thus, for evalua­ting real trade flows, data in current USD should only be corrected for price chan­ges over ti­me. For this purpose we have computed the mean unit values of the period 2004-2006, the same pe­riod on which real output series are based, and applied those unit values to physical export and import flows avail­­­able since 1961, to get a series of real exports and imports for agricultural or food pro­ducts, at con­stant 2004-06 prices. This approach requires some adjustments for products lacking information about quantities (or va­lues) in 2004-06. These special cases include at least the following categories: ŒŒ Products not traded in 2004-06 (but traded at other pe­riods since 1961). ŒŒ Products traded in one or two of the reference years (2004-06) but not in all three. ŒŒ Products traded in 2004-06 but in small quantities, which get rounded to zero in the FAOSTAT da­tabase. Quantity data are in tonnes, or head of livestock, or thousand head in the case of chicken and other small animals, all given in integer amounts; thus quan­tities below 0.5 tonne or below 500 units are rounded to zero. Value, in turn, is given in thousand USD, again in integer format; values below 500 USD are rounded to zero. As a result, some small flows may be rounded to zero in both quantity and value, whilst trade in some other items shows a po­si­tive value and a (re­ported) zero quan­tity, or the converse. ŒŒ Some heterogeneous product categories given only in value but not in quantity. This includes, for instance, ‘Crude materials’ (item code 1293 in FAOSTAT’s classification) designating a hete­ro­­­ge­neous coll­ection of commo­dities (bulbs, live plants, cut flo­wers, ivory, feathers, pig bristles, and many more) for which only the total mo­ne­ tary amount is reported with­out any disaggregation into individual com­modities or any reference to their quantity. Another such example is ‘Live animals, unspecified’ (FAO­­STAT item code 1171): it may include small and large animals and it does not speci­fy quantities.

339

Measuring historical trends

The general procedure we followed for both exports and imports was as follows: ŒŒ Tables for exports and imports by product were obtained from FAOSTAT, at world le­vel, with value in current USD and the respective quantities involved, for each product and year. This co­vered over 450 distinct items, with yearly data from 1961 to 2011. There were a few categories of a miscellaneous or residual nature for which only values were reported but not quantities. ŒŒ Average reference prices (unit values) for 2004-2006 were computed for items for which quan­ti­ty and value were available in the three years from 2004 to 2006. This gene­rated a set of unit values for exports and another set for imports. Both sets included very si­milar but not exactly the sa­me list of goods. Goods with 2004-06 prices covered typically 95% of total trade (the remain­der co­r­responds to goods lacking quantity or value data for any of the three years 2004 to 2006, inclu­ding also those few categories for which no quantity is reported at any year). ŒŒ Trade flows at constant 2004-06 prices (unit values) from 1961 to 2011 were com­pu­ted for each region of the world, for those goods for which 2004-06 unit values were availa­ble. This entailed downloading the series of quantities for the various products traded at each region, and multiplying by the res­pective 2004-06 world unit values. ŒŒ The annual values of imports and exports at current prices was computed at each re­gion for that sub­set of goods for which 2004-06 unit values were available. This value was com­pared with the FAOSTAT va­lue of total agricultural trade flows (also at current prices) in the respective region, which includes items without unit va­ lues. The per­ centage difference found in the series at current prices was then ap­plied as an adjustment to the regional series at constant prices, in order to get an estimate of real trade represent­ing all products (i.e. including also those pro­ducts for which no quantity data exists). This adjustment represented just a small percentage of total trade. Trade as a share of production. The traded share of total agricultural and food production is not easy to determine, because the product coverage and the valuation principle is not exactly the same for both. First, exports are valued in FOB terms and imports in CIF terms. Traded items are often previously processed, whereas agricultural production is accounted for in terms of the primary products coming from farms. 340

Hector Maletta

However, an approximate assessment can be prepared. We have compared the value of world agricultural exports, which include manufactured products based on farm products (such as flour, pastry, wines, beer, and cigars), to the output of the agribusiness chain, i.e., the sum of (a) the gross value of primary agricultural products (at producer prices) plus (b) the value added by manufacturing food, beverages, and tobacco. The former is taken from FAOSTAT, the latter from the industrial statistics included in the World Bank’s World Development Indicators.36 This tends to overstate the value of production since some industrial processing is already included in FAOSTAT’s value of agricultural production (e.g., value added by slaughterhouses to produce meat). Since value added in agro-industry as well as the value of exports is only available in dollar figures based on market exchange rates, we have taken agricultural production valued on the same basis (FAOSTAT data on value of production in current USD, with local currencies converted at market exchange rates). Table 68. Estimated traded share of agricultural production 1994-1996 to 2009-2011 (million USD and percentages) 1994-1996 5,823,991 13.7% 796,693 1,637,816 2,434,509 432,302 17.8%

Total world manufacturing value added (current USD) a Food bev. tobacco: % of manufacturing value added b Food bev. tobacco manuf. value added (estimated) c Gross value of agricultural production d Total output of primary and processed farm products e Total agricultural exports f Exports as percentage of production g

1999-2001 5,574,622 13.9% 772,470 1,454,337 2,226,807 414,168 18.6%

2004-2006 7,449,576 18.3% 1,361,431 2,035,364 3,396,795 660,630 19.4%

2009-2011 9,402,636 22.2% 2,086,860 3,386,062 5,472,922 1,114,734 20.4%

a. World Bank WDI, indicator code NV.IND.MANF.CD. Value in million current USD. b. ‘Food bev. tobacco’ refers to the food, beverages, and tobacco manufacturing industries. Based on World Bank WDI, indicator code NV.MNF.FBTO.ZS.UN. World percentage based on countries with values available for both indicators in at least one year of each period. c. Product of the two preceding rows. d. FAOSTAT gross value of agricultural production (primary products of crops and livestock) in current USD. e. Sum of the two precedent rows. f. FAOSTAT total value of agricultural exports in millions of current USD. Fish excluded. Includes processed products. g. Ratio of the two preceding rows.

The value added in the ‘food, beverages and tobacco’ industries has been calculated especially for this purpose. The WDI provides, on the one hand, the total value added in manufacturing in current USD (indicator code NV.IND.MANF.CD); and on the other hand, the percentage represented by ‘food, beverages and tobacco’ within total manufacturing value added (indicator code NV.MNF.FBTO.ZS.UN). Both series start in the 1990s. These two indicators are not available for some (typically small) countries, and not for every year. We have calculated the total of each indicator for countries where both are available in at least one year within four periods: 1994-1996, 1999-2001, 2004-2006 and 2009-2011 (taking the average of those years for which data was available within each 3-year period). The resulting percentage was then applied to the estimated world total manufacturing value added for the same years.

36

341

Measuring historical trends

The main conclusion from this approximate estimation is that agricultural products are traded in increasing proportion relative to production. At the world level, the share of exports increased from 17.8% to 20.4% in about 15 years. These figures can be readily calculated just since the 1990s, but it is likely that the relative importance of agricultural trade has been increasing since the 1960s. Based on the growth of agricultural production and agricultural trade from 1961-1963 to 2009-2011 (as per Figure 23), it may be estima­ted that in the early 1960s, agricultural trade represented less than 8% of agricultural (and agro-industrial) production, as compared with 20% in 2009-2011.

12.4. Food supply-demand accounting The aggregate amount of food available and consumed, either worldwide or in a particular country or region, is usually expressed using certain accounting methods such as supply-utilisation accounts and food balance sheets. These are accounting devices; as such, they consist of definitional identities calculated ex post. For instance, in any given country, the total supply of a given food must necessarily equal all uses of that food (including waste), just as supply must always equal effective or realised demand. Conventions usually applied in food supply-demand accounting are reviewed in this section: supplyutilisation accounts (SUA) and the food balance sheet (FBS). It is important to note that these accounting devices are invariably ex post identities, and are also non-normative but descriptive. In particular, they do not involve norms about how much or what kind of food is required to supply necessary nutrients or to maintain good health: the figures simply summarise what is actually produced, traded or consumed.

12.4.1. Supply-utilisation accounts Within the boundaries of a given country, sub-national region, community, or household, and in a given period, food becomes available in three different ways: it can be (a) produced internally during this period, (b) acquired from outside sources, or (c) drawn from pre-existing stocks. Food thus obtained is subsequently used (or wasted) in various forms: it may be (a) consumed as food, (b) fed to animals, (c) lost or wasted, (d) stored away for future use, or (e) moved out of the boundaries considered (e.g., exported). This gives rise to the notion of a food budget, comparing availability and utilisation of food products, also known as a supply-utilisation account (SUA) or commodity balance. These balances or SUAs may also be (and actually are) estimated for non-food products such as wool or tobacco (see for instance FAO [2003c]). 342

Hector Maletta

The SUA is a flow measure, representing the amount of a product made available and utilised during a given period (usually a year or a season). The annual SUA of a given commodity i (e.g., wheat), expressed in physical terms (e.g., in tonnes) and commonly calculated at the national level and for a given year, is expressed by the following ex post identity: Supply-Utilisation Account (SUA) for a given product, period and country Domestic supply of product i = Domestic utilization of product i DSijt = DUijt Soijt + Pijt + Mijt – Xijt = Fijt + Uijt + Scijt

The domestic supply DSijt of a specific product i (e.g., wheat), during a certain period t (normally a year) and in country j, is the sum of the opening stocks of that commodity (Soijt), plus domestic production made available during the period (Pijt), plus imports (Mijt), minus exports (Xijt). Domestic utilisation (or effective domestic demand) DUijt in the same country and for the same product and year is the sum of human food consumption (Fijt) plus amounts allocated to other uses (Uijt), plus any final or closing stocks (Scijt). Other uses (Uijt) include all non-food uses of the product, i.e., its use as seed, animal feed, and input for non-food industries (e.g., bio-fuels), plus waste or losses. For some purposes these various non-food uses may be explicitly separated, e.g., to analyse food waste. A SUA may potentially be calculated for each specific product, within any specific geographical unit (country, region, or world), and for any given year. The joint balance for various food products (which may be also added up, for instance in terms of dietary energy supply) makes up a food balance sheet. For the moment we refer to SUAs, separately compiled for specific commodities; food balance sheets (FBS), which assemble information on all foods, are discussed later in Section 12.4.2. Commodities in SUAs and FBSs are normally primary commodities, such as wheat grain. Export or import data may refer to processed products such as bread or pasta, but within SUAs these products are usually expressed in equivalent amounts of the primary commodity or commodities involved (e.g., exported or imported bread and pasta are expressed in wheat-equivalent terms). The availability of food items for human consumption in FAOSTAT SUAs and food balance sheets reflects apparent food consumption, i.e., amounts of 343

Measuring historical trends

food products delivered to consuming units, chiefly households but also other places where people eat, e.g., restaurants, hospitals, schools, barracks, jails, and the like, which are for this purpose equated to households. Food available for consumption is measured ‘at the household gate’, once other uses (as well as losses during the marketing and processing chain) have been deducted. The formula takes into account post-harvest waste and losses, but also only ‘up to the household gate’, as explained, for example, in FAOSTAT’s Frequently Asked Questions section (http://faostat.fao.org/site/565/default.aspx, question 7): ‘Consumption in the Food Balance Sheets refers to consumption at the household gate’. In the same vein, FAO’s methodology for estimating undernourishment is based on the daily dietary-energy supply (DES) at the household level: ‘the daily per person DES refers to food acquired by (or available to) the households rather than the actual food intake of the individual household members’ (FAO 2008:9). Thus, apparent consumption includes food waste occurring within households or within other consumption units such as restaurants or hospitals. Such waste may include food decaying while held in household storage, food lost to vermin, kitchen and plate leftovers, food used to feed pets, etc. There are some estimates of household food waste, but as yet none have been incorporated into FAO food balance sheets. For a given estimate of apparent consumption, actual food intake by individuals is thus likely to be somewhat lower on account of household waste. F in the formula above is thus a measure of food available to households, i.e., food the household has access to, as measured ‘at the household gate’, albeit not necessarily consumed in its entirety by household members. It is variously called ‘food supply’ or ‘apparent consumption’, and is often calculated as a residual once the other quantities are estimated empirically. Thus, the apparent consumption of a given food product i for country j at year t is: Fijt = Soijt + Pijt + Mijt – Xijt – Uijt – Scijt Instead of being estimated as a residual, direct estimates of Fi based on household surveys are sometimes used, but such surveys (often based on verbal recall or on household record-keeping on a particular day or week) are known to underestimate actual consumption (FAO 2003b). All figures in the SUAs are annual flows during period t, except for opening and closing stocks. Obviously, the closing stocks of year t equal the opening stocks of year t+1. In practice, stock change is the relevant flow variable, and 344

Hector Maletta

the formula, as it is usually provided, replaces both initial and final stocks with a single flow variable (S) representing stock change: Sijt = Scijt – Soijt Likewise, foreign trade may be also simplified in a similar manner, by replacing Mijt and Xijt with net exports or trade balance for each product, denoted here by Tijt: Tijt = X ijt – M ijt Using these synthetic expressions for stock change and net exports, the formula for Fi becomes: Fijt = Pijt – Tijt – Uijt – Sijt The above equation is in fact an accounting identity representing the ex post demand-supply balance of a given commodity. It may refer to any level of aggregation: the world, world regions, countries, sub-national regions, and local communities. An analogous identity also holds at the level of individual households: P would stand for household food production, if any, e.g., output from a family farm or kitchen garden, while M and X would stand for food entering or leaving the household, respectively (including purchases and sales, and also transfers and donations). In the same vein, stock change S at the household level would refer to changes in the contents of the household’s pantry. At the world level, the trade balance of each commodity (Ti) is in principle (or theoretically) zero, since world exports must equal world imports (except for statistical discrepancy). Instead, the trade balance of a product for a single nation (or household) should not in principle be (and is usually not) zero; the two trade flows (M and X) are important when the definition is applied to specific countries. As an ex post identity, the supply/demand balance of a product illustrates things as they actually are, not as they ought to be. Thus, for instance, it accounts for the existing level of waste and losses (included in the U term), regardless of whether such waste is avoidable or unavoidable. By the same token, the level 345

Measuring historical trends

of apparent food consumption (F) that emerges from the balance is not meant to signify whether such consumption is adequate, excessive or insufficient, nor is it sensitive to possibly unequal levels of consumption across households or people. Total supply of a food product in a given period (usually a year) for a given geographical area (country, region, world) may be divided by the corresponding population to estimate per capita supply, usually presented in terms of the physical yearly quantity of each product (kg/person/year) or in terms of its contents of protein and fat (in grams/person/day) and dietary energy (kilocalories/person/day). FAO SUAs refer to calendar years, but the original information refers to the agricultural year, variously defined by different countries (e.g., July 1 to June 30, or April 1 to March 31); FAO ascertains the distribution of production by month or season to estimate calendar-year figures. SUAs estimate the total and per capita amount of food available for human consumption, but not its (possibly unequal) distribution across the population. Distribution among households can be gauged in household consumption or expenditure surveys, or estimated by other means. As SUAs and household surveys estimate the amounts of food available for consumption at the household gate, they do not reveal the intra-household distribution of food among household members. Household consumption or expenditure surveys seldom investigate actual intake of food by individual members.

346

Hector Maletta

Food aid and ex-ante food balances. Humanitarian food aid may or may not be counted as part of aggregate food availability, depending on the purpose of the analysis. Normal ex post SUAs count food aid as part of trade flows, but food aid is often excluded in ex-ante or prospective food balances. These prospective balances aim at estimating whether expected food supply will be sufficient to meet expected demand in the short term (e.g., over the current or next agricultural year). If the goal is to measure the autonomous capacity of the reference unit (nation, region, household) to have enough food, or to determine how much food aid will be needed, then food aid should not be included in food supply. In ex-ante SUAs, usually restricted to cereals or other staple foods, the terms are arranged differently, isolating the requirement of aid as an expected uncovered gap of commodity i for country j and year t, denoted as Geijt: Geijt = Soijt + Peijt + Meijt – Xeijt – Ueijt – Fhijt – Scijt Geijt refers to the expected gap of commodity i in the coming season, to be covered by food aid; Soijt stands for the opening stocks of commodity i; Phijt is the expected level of production; Fhijt is the expected habitual level of food supply of i (defined as habitual per capita supply in normal years, multiplied by expected population in the target period); Ueijt is the expected amount to be devoted to other uses (including losses); Meijt and Xeijt are the expected commercial trade flows, perhaps including ‘structural’ or ‘program’ food aid that is regularly received but not related to a current emergency; and Scijt is the desired or planned level of closing stocks at the end of the future period envisaged. See methodological details in the FAO-WFP guidelines for crop and food supply assessments (FAO/WFP 2009). This book is not concerned with ex-ante food balances, intended to determine food aid needs. We restrict our analysis to ex post SUAs and food balances, showing actual (realised) availability and apparent consumption. FAO ex-post SUAs (commodity balances) and food balance sheets, included in the FAOSTAT statistical system, do include food aid as an integral part of trade flows (separate FAOSTAT data are available on food aid flows). Unlike ex-ante or prospective balances, ex-post SUAs are identities, which are always balanced by definition: ex-post total supply equals ex-post total utilisation, without any ‘uncovered gap’.

12.4.2. Food balance sheets FAOSTAT contains separate SUAs for each product (by country and year) as well as annual comprehensive food balance sheets covering all food products in a given year and for a given country or region. Food balance sheets (FBS) are calculated for each particular food in physical terms (usually in metric tons), and are not valued or aggregated in economic terms. Physical quantities cannot be aggregated, except for similar products (e.g., cereals). However, 347

Measuring historical trends

food balance sheets include aggregate estimates of daily per capita supply of fat, protein, and dietary energy. Food balance sheets, like SUAs, are often expressed in terms of primary commodities (e.g., wheat). Imports of processed products (e.g., flour or pasta) are converted into the amount of the primary commodity or commodities that went into their production, by application of certain conversion coefficients. Thus, imported wheat flour is converted into wheat grain equivalent, and imported powder milk into fluid milk. Sometimes a processed product contains more than one primary product; for example, imported cookies may be converted into the equivalent amounts of wheat, butter, and sugar. A typical food balance sheet includes one line per product, and columns for domestic production, trade flows, domestic availability (P-X+M), losses or waste, use as feed or seed, other non-food uses, and food supply. All flows are expressed in tonnes per year. The food supply of each food item (e.g., wheat, green peas, or beef ) is also calculated in per capita terms (kg/person/year) and broken down into the corresponding daily amounts of energy, protein or fat according to food composition tables. Estimates of per capita availability of major nutrients, such as protein and fat, are provided in grams/person/day, and dietary energy in kcal/person/day. FAO food balance sheets do not provide estimates for vitamins and minerals, and also fail to provide an explicit account of the supply of carbohydrates, but the latter can be estimated by difference, deducting from total dietary energy the energy contained in protein (4 kcal per gram), fat (9 kcal per gram), and alcoholic beverages (as indicated in the respective line of the FBS). The energy provided by carbohydrates, once estimated, can be converted into carbohydrate quantity at a rate of 4 kcal per gram. Daily per capita amounts of dietary energy, protein, and fat are also aggregated in the first few lines of the FBS, for all food items and also for major subsets such as vegetable and animal products.

348

Hector Maletta

Calories and joules: How dietary energy is measured Dietary energy alternatively may be measured in calories or joules. A kilocalorie is defined as the amount of energy required at sea level to heat one litre of pure water by one degree Celsius (more precisely from 14.5°C to 15.5°C). One kilocalorie equals 1000 ‘chemical’ or ‘small’ calories, which have the same definition but refer to a millilitre of water (a volume of one cubic centimetre). However, in common parlance in the context of food, kilocalories are often called ‘calories’ for short. This book discusses dietary energy in terms of kilocalories (abbreviated as kcal) because they are the most commonly used unit in this context, but the official international measure of energy is the joule and it is recommended that joules should be used also for food (FAO 2004). A joule is defined in terms of the international unit of force, the newton, which is the force required to accelerate a mass of one kg by one meter in one second. A joule is the work performed by a force of one newton to displace its point of application by one meter in the direction of that force. One kcal = 4.184 kilojoules (kJ), and one kJ = 0.239 kcal. The estimated dietary energy content of a particular food item is based on the energy the body can extract from its contents of carbohydrates, fat, protein and ethylic alcohol – the only substances the human body can use as sources of energy. These substances provide energy according to Atwater coefficients (approximately 4 kcal/gram for carbohydrates and protein, 7 for alcohol, and 9 for fat). For details see FAO (2003b, 2004), and Shetty (2005).

12.5. Access to food The centrality of access in the current definition of food security necessitates the use of suitable definitions and measures of food access. Of course, access is different from (actual or apparent) food consumption, i.e., from individual intake. For instance, a household may have an income that is deemed more than sufficient to acquire adequate food and cover other necessities as well, but internal household decisions may allocate that income in such a way that food is not acquired by the household in the required amounts (e.g., if the household head uses the money for other purposes), or food is not consumed by individual household members according to their needs and possibilities (e.g., if intra-household food distribution is not equitable). In fact, the definition of food security in use since the 1990s does not emphasize actual food intake but only food access, i.e., the capability or entitlement to acquire sufficient and adequate food, irrespective of the actual application of this entitlement or capability. Thus, people with income above the cost of a basket of basic goods and services (including adequate food) are deemed to 349

Measuring historical trends

have good economic access to food, even if the money is ultimately used for something else. This is the same approach used to measure poverty in terms of a poverty line: the line itself may be derived from the cost of a basket of goods and services, but earning an income at or above the poverty line does not ensure that it will be used to purchase that same basket of goods and services. Household access is not equivalent to individual access (or actual consumption), due to household waste and also due to unequal intra-household distribution of food access or intake (relative to individual needs). Data on the distribution of food consumption comes in most countries from household-level surveys, which typically ignore individual consumption. Most measures of food access (and also measures of poverty) are thus defined and evaluated at household level, and applied equally to all household members. Thus, if a household has access to 20% less (or more) dietary energy than the sum of the needs of individual members, it is implicitly assumed that every individual member also consumes 20% less (or more) than their needs. Actual individual consumption of food is indeed rarely measured on a massive scale. Very little information is available on individual consumption or access to food, or on intra-household distribution (of food, income, welfare or other consumption).37 Several approaches have been used to measure food access in order to operationalize the concept of food security, in tandem with the evolution of the concept itself. In the 1970s and 1980s the indicators most often used concerned national self-sufficiency, such as the share of consumed food calories that are produced domestically, or the gap (if any) between domestic production and domestic consumption of staple food. After the 1996 World Food Summit re-defined food security in terms of access to food, the most important internationally-used indicator became the prevalence of undernourishment. This indicator is supplied by FAO in its reports on The State of Food Insecurity in the World (SOFI), published annually since 1999, and was also estimated for prior periods in FAO’s fifth and sixth World Food Surveys (FAO 1987; FAO 1996). To understand the FAO indicator of the prevalence of undernourishment, a brief review is necessary of how per capita dietary energy needs are determined.

12.5.1. Dietary energy needs Body cells generate and expend energy for various purposes: See Haddad et al. (1997) for discussions of theoretical, practical, and methodological issues associated with intra-household distribution of resources.

37

350

Hector Maletta

ŒŒ Basal metabolism (energy expended to stay alive at rest). ŒŒ Metabolic response to food (energy expended to ingest, digest and metabolize food). ŒŒ Physical activity. ŒŒ Pregnancy and lactation. ŒŒ Child/adolescent growth.38 Energy is generated by burning (oxidizing) a simple sugar (glucose) that is transported in blood cells along with the necessary oxygen. Glucose is manufactured in the body by metabolising carbohydrates, protein, fat, or alcohol, and transformed into energy according to bodily needs. Sugar not used for energy is transformed into fat and stored away. Body fat stores are necessary to a certain extent, but to avoid harmful effects on health, stored fats should be kept within a safe range as a percentage of total body mass. Deficient or excess body fat increases the risks of disease and mortality. On this basis, a range of acceptable body weights are determined for given heights. The recommended levels of dietary energy intake for a population group have been established by the UN since the 1950s, and are periodically updated by a joint group of experts under the aegis of WHO, FAO, and the United Nations University: see the latest version in FAO (2004), and background papers in FAO (2003b) and Shetty (2005). Dietary energy requirements for a population group are based on the average total energy expenditure (TEE) of healthy people with acceptable body weight for their sex and age, engaging in various levels of physical activity (and growing normally in the case of children). TEE is usually measured in kilocalories (kcal/person/ day, abbreviated as kcpd). Even among healthy individuals, good health is compatible with a range of body weights and a range of physical activity levels. The range of acceptable body weights and total energy expenditure by sex and age is usually calculated separately for children under the age of five and for older individuals. In the case of healthy children under five, direct measurements of energy expenditure per kg of body weight have been carried Energy needs for pregnancy, lactation, and growth include: (a) potential energy that could have been generated by food nutrients (e.g., protein) that are instead converted into new tissue; (b) energy content of mother’s milk; and (c) energy expended in the metabolic activity required for new tissue deposition and milk secretion.

38

351

Measuring historical trends

out, and tables have been prepared to show the distribution of height and weight for age and sex. The weight and height attained by well-nourished healthy children under five vary little across ethnic groups all over the world, and thus a single height-for-age growth curve and acceptable margins have been determined (WHO [2006]: Tables 19-20 for boys and Tables 27-28 for girls). Likewise, there is a single weight-for-age growth curve and acceptable margins (WHO [2006]: Tables 38-39 for boys and Tables 49-50 for girls). WHO (2007) also prepared standard curves for weight by height (Tables 57-58 for boys and 66-67 for girls).39 These standard growth curves give the percentile distribution of height and weight by age and weight by height for healthy children growing normally and performing normal physical activity for their age. Normal TEE for these children is given for the median weight and height, i.e., at the 50th percentile of the distribution; this median level of TEE is used for normative purposes for a population group (FAO [2004], chs. 3 and 4). Thus, for children under five, needs depend on a normative height and corresponding weight, to be attained at each age. This is not the case for older age groups, where acceptable weights are determined according to attained (rather than normative) height for each age and sex. The main indicator of acceptable weight for attained height is the Body Mass Index (BMI), defined as the ratio of weight (in kg) to the square of height (in meters). The reference BMI distribution reflects the range of weights of healthy people, which has a normal distribution. The acceptable range of BMI values corresponds to the amount of fat deemed not to endanger health that people store in their bodies. WHO (2007) provides the reference distribution of BMI values for each sex at ages 5 to 19; the values for age 19 are also used for older individuals. During recent decades (from the 1980s to the early 2000s) norms regarding acceptable weights relied on a range of BMI values, typically 18.5 to 24.9. People below 18.5 were regarded as too thin, while those with BMI≥25 were classed as overweight (and as obese if BMI≥30); these ranges, however, were valid only for adults, and more importantly, they seemed to differ across ethnic groups (some groups seemed perfectly healthy though they were outside the acceptable range, typically below 18.5 in several Asian 39 The stature of children up to 24 months of age is measured as length (lying down); above 24 months it is measured as height (standing). Because of gravity, height at 24 months is about 0.7 cm shorter than length; growth curves thus have a ‘step’ of 0.7 cm at 24 months, reflecting the change in measuring technique. Graphical growth curves for ages from 0 to 59 months are sometimes standardised in terms of length, i.e., actual length up to 24 months, and height+0.7 cm at ages over 24 months, to avoid the appearance of a ‘step’ or ‘jump’ in length at that age (e.g., WHO [2006, Figures 64 and 79]).

352

Hector Maletta

populations). Though this issue has yet to be completely resolved, more recent norms are more tolerant of different criteria for establishing limits. Besides, there is a tendency to use the reference BMI distribution as a guide rather than as fixed values. Thus, current practice regards as acceptable all BMI values between the 5th and the 95th percentile of the BMI reference distribution, as well as the corresponding range of weights for each sex, age group, and attained height, for both children and adults. It is understood that good health can be maintained at any weight within this range. The use of the BMI table requires measurements or estimates of the average height attained by sex and age, from childhood to adulthood. Current estimates of energy expenditures (and thus energy needs) for each sex and age are provided in FAO (2004). The estimation method of these energy needs varies with age: ŒŒ For infants under one year of age, direct measurements of total energy expenditure (TEE) are obtained from healthy infants, yielding energy needs by age and sex (FAO 2004:14). ŒŒ For older children up to 17 years of age, total energy expenditure per kg of body weight is estimated for each sex and age for children performing moderate physical activity (FAO 2004:26-27) and for children performing light or heavy physical activity (FAO 2004:29-30). To estimate TEE for each age and sex, based on TEE per kg, it is also necessary: (a) to ascertain or estimate the average attained height by age and sex; (2) to choose a reference weight for that height; and (3) to establish a reference level of physical activity. Average heights by age and sex may be measured or estimated on the basis of health surveys or other sources. The range of acceptable weight-for-height by sex is derived from the reference distribution of BMI, as explained before; the normal reference weight corresponds to the median of the reference BMI distribution, while the minimum reference weight corresponds to the 5th percentile. ŒŒ For adults, a factorial method is used: TEE is defined as the product of the Basal Metabolic Rate (BMR) multiplied by a factor representing a Physical Activity Level (PAL): TEE = BMR × PAL The BMR is the amount of energy expended by the body merely to stay alive at rest: keeping all internal organs working (heart, kidneys, liver, brain, and so 353

Measuring historical trends

on), maintaining a stable body temperature, and other basic bodily functions. It is measured, for instance, in kcal/day per person, but for some applications it is expressed as kcal/day per kilogram of body weight. Standard predictive equations estimate the BMR for each sex and age group as a linear function of body weight (FAO 2004:37). Therefore, to estimate BMR, reference weights by sex and age should be used. The most common reference weights are those corresponding to a given percentile of the BMI distribution; the 5th percentile is usually regarded as determining the minimum acceptable weight; the midpoint or normal weight is at the median of the BMI distribution, and the maximum is commonly considered to be at the 95th percentile. The PAL factor indicating level of physical activity (which also includes metabolic response to food) may range from 1.55 for light activity, to about 1.80 for normal or moderately active lifestyles, and to 2.20 for vigorously active lifestyles. To estimate the total dietary energy needs of a population, mean individual needs by sex and age should be calculated in kcal/person/day and multiplied by the size of the respective age-sex group. The mean daily energy requirements for a certain age-sex group (other than young children) requires identification of the mean height of that group to be identified, then estimation of the reference weight for that height, the BMR for that weight, and a normal or average level of physical activity (usually estimated by a moderately active PAL of around 1.80). The total daily needs of all age-sex groups (in kcal/ day) are then added up to obtain the total daily energy needs of the whole population. A small additional allowance should be included to cover the energy required for pregnancy and lactation, but this does not add much to the total energy needs of the population as a whole. The allowance per birth is usually estimated at 210 kilocalories per day (FAO 2004), which is multiplied by the number of births estimated on the basis of population size and the crude birth rate. Finally, the resulting total energy needs are divided by total population to express them in per capita terms as kcal/person/day.40 Average energy needs for a typical population, calculated for median reference weight and moderate physical activity, is usually about 2100 kcal/person/ day, but varies across countries (typically ranging between 2000 and 2300) depending on mean height by sex and age, and the age-sex distribution: rich countries usually have (on average) taller people and a higher proportion of adults, thus requiring more energy per capita than poor countries where The printed edition of FAO (2004) comes with software for calculating the per capita dietary energy needs of a population based on its age-sex structure, average adult height, and birth rate, for various levels of physical activity.

40

354

Hector Maletta

people are typically shorter and children represent a higher share of the total population. Average per capita needs tend to grow over time as countries attain higher levels of economic development.

12.5.2. The FAO undernourishment indicator Undernourishment is based on the Minimum Dietary Energy Requirement (MDER), which corresponds to a habitual level of food energy intake (by sex and age) that allows for maintenance of minimum acceptable body weight and performance of light physical activity, plus special needs related to pregnancy, lactation and child growth. The indicator is defined as the probability that a population member has habitual access to less than the minimum amount of dietary energy compatible with good health.41 The calculation of this indicator involves estimating: (a) the probability distribution of food consumption (in terms of dietary energy) across the population, based on data about household food acquisition or food expenditure, or, when these are not available, on household income or expenditure plus Engel coefficients showing the share of income or expenditure on food at each income level; and (b) a Minimum Dietary Energy Requirement (MDER) for each country and year. Unequal food access. The distribution of consumption across the population is usually approximated by the lognormal curve, meaning that the logarithm of dietary energy supply is normally distributed (FAO 2008). In the latest revision of the indicator (FAO-SOFI 2012) this has been modified for the use of either the skew lognormal or the skew normal distributions. The probability distribution requires two or three main parameters (depending on the chosen model): the mean (based on the average per capita supply of dietary energy, as provided by food balance sheets) and the standard deviation and skewness of access to dietary energy (estimated from available household income and expenditure surveys). As per FAO methodology, the MDER of a given country at a certain date is estimated as follows (FAO 2008:7):

Following epidemiological terminology, prevalence equals total cases existing at a given date divided by total population (the date, in practice, may be the average over one week, month or year). Incidence refers to new cases occurring during a period (typically a year) as a percentage of total population at risk in the same period. Prevalence is a stock (at a given time) and incidence is a flow (per year), both relative to a reference population. 41

355

Measuring historical trends

ŒŒ For children below the age of 10, the reference body weight is the median of the range of acceptable weights (WHO 2006, 2007). In this case, energy needs reflect normal (moderate) activity level. ŒŒ For adults and children aged 10 and above, reference body weight is the minimum, estimated on the basis of the 5th percentile of the BMI distribution (WHO, 2007), for the average attained height by age and sex. MDER refers to moderate physical activity for ages 10-17, and light physical activity for ages over 18. The use of ‘light’ physical activity for adults as part of the definition of undernourishment may be debatable. People at risk of hunger may need to perform (on average) more than light physical activity in order to procure food and other necessaries. In fact, it is precisely the poor who are often required to do the heavier work; for them, this work is an obligatory expenditure of energy, and therefore they may not be able to make a living with a habitually light level of physical activity. The 2012 methodological overhaul of FAO’s undernourishment measurement offers an alternative indicator based on normal (rather than light) physical activity; however, it is just a rough approximation, yet to be perfected, because there continues to be insufficient information to determine the average amount of obligatory extra energy that people in each country must expend in order to do the work involved in their livelihoods. Throughout its history, this FAO indicator has undergone several revisions, including a major one in 2012 (retrospective to 1990). These revisions included: ŒŒ Updating (often retrospectively) the estimates of total population; this was important for some countries (mostly in Africa) where previous estimates were no more than rough approximations due to the lack of a regular series of reliable censuses. ŒŒ Improving the food-balance-sheet estimation of food supply by revising estimates of food production, food trade, losses, and non-food uses. In particular, the 2012 revision included a major revision of estimates of retail losses (previous estimates reflected mainly losses occurring at the production and wholesale levels, omitting much of the loss occurring at retail level).

356

Hector Maletta

ŒŒ Updating the estimated shape (variance and skewness) of the probability distribution of food access within the population, using new household expenditure surveys in various countries. ŒŒ Updating energy requirements by sex and age, as per the latest meeting of experts meeting convened by FAO and WHO (published as FAO [2004]) as well as the new WHO growth standards for children (WHO 2006) and the new BMI reference distributions (WHO 2007), all of which have modified older standards adopted in the 1980s and used until the mid or late 2000s. ŒŒ Adjusting the mathematical model used to estimate the probability distribution of food access. Up to 2011, the model was the lognormal distribution; in the 2012 revision, the models adopted included the skewlognormal and the skew-normal distribution, which require not only the mean and standard deviation (as was the case with the lognormal) but also a skewness parameter (also derived from household surveys where they were available, or otherwise estimated).42 As a consequence of these changes, the series has been homogenised since 1990. Estimates for previous periods, such as those supplied by the World Food Surveys (for 1969-1971 and 1979-1981), are not strictly comparable to the more recent ones. In this book, only the revised figures (SOFI 2012) are used for 1990-2012. The undernourishment indicator concentrates on dietary energy, and does not reflect in detail the availability or consumption of protein, vitamins or minerals. However, there are some additional indicators available for many countries worldwide that point in that direction, including a number of rough indicators of dietary diversity (e.g., percentage of cereals and tubers in total energy supply). Consumption of micronutrient-rich foods is also a useful indicator, and it is used to this end in this book. The indirect method used by FAO, based on aggregate data and a theoretical distribution model, is vulnerable to uncertainties and error arising out of the various sources mentioned above. Besides, it refers to the probability of undernourishment for the average person, but people differ in their energy needs due to normal variation in height, body shape, sex, age, and physical A summary is in FAO-SOFI (2012). See also FAO (2012b); Cafiero and Gennari (2011); and Cafiero (2012, 2013).

42

357

Measuring historical trends

activity. Thus, individuals who are shorter than average may be consuming below the average MDER for their age and sex (calculated for the average attained height), and yet may be covering their minimum food needs. By the same token, individuals who are taller than average may not be satisfying their minimum dietary energy needs even if they access food energy above the corresponding average MDER. However, these and other inter-individual differences in needs (for example, genetically-based variation in individual metabolic efficiency) are expected to be randomly distributed with a near zero net effect. Estimates of per capita food availability are regularly updated as new data becomes available on yearly output, trade, and stocks. However, the standard deviation and skewness of the distribution of food supply across households, as well as coefficients used to estimate non-food uses (seed, feed, losses, etc.) are not usually so well established and are taken as constant in the short term, to be updated only when new evidence becomes available, e.g., when a new health survey indicates a change in average attained heights. Even in a (real or simulated) population where everyone covers their minimum individual needs, the estimation procedure will always show a positive level of undernourishment; these ‘false positives’ may include as much as nearly 5% of such population, due to the combined effect of the different sources of variation involved. In view of all these uncertainties and the underlying variation, FAO regards undernourishment prevalence estimates of below 5% as highly imprecise and statistically undistinguishable from zero. These low estimates are reported simply as ‘less than 5%’ in FAO tables; the actual (and unreported) value could be calculated by replicating the underlying estimation.43 Since dietary energy deficiency is the key determinant of malnutrition in the most affected parts of the world, the undernourishment indicator concentrates on dietary energy and ignores insufficient consumption of protein, vitamins, and minerals to the extent that these are uncorrelated to the consumption of dietary energy; countries (or persons) who cover their energy needs might still have an insufficient supply of other nutrients, sometimes called ‘hidden hunger’. Survey-based scales of dietary diversity, Five percent is also the relevant minimum level for evaluating progress toward Target 1-C (related to hunger) of the Millennium Development Goals. The target for undernourishment is considered to be achieved if a country has either reduced the percentage prevalence of undernourishment by one half or more, relative to the 1990-92 baseline, or has brought the estimated prevalence to a value below 5%.

43

358

Hector Maletta

which count the number of food groups (cereals, pulses, fruits, etc.) and the number of specific food items within each group (i.e., the various cereals, pulses, fruits, vegetables, etc.) that are actually consumed by a given household, as well as the frequency with which they are consumed, tend to reflect adequacy or deficiency in protein and micro-nutrients and have been increasingly used in field surveys to complement measures of dietary energy supply. They have been shown to have a significant correlation with actual consumption of dietary energy, protein and micro-nutrients (Hoddinott and Yohannes 2002). Their correlation with energy intake supports the use of energy deficiency as an approximation, especially in poor countries where undernourishment is more widespread. However, direct and systematic data on the distribution of dietary diversity (or micronutrient consumption) among people and households are not available on a regular country-bycountry basis. Since staple food consumption is relatively inelastic, increases in dietary energy usually indicate an increase in dietary diversity and a reduction in micronutrient deficiency, but that is not always the case (some people, especially in rich countries, add calories chiefly in the form of extra fat and oil instead of fruits, vegetables, meat or milk, thus creating the paradox of obese people with hidden malnutrition). Unfortunately, no indicator of micro-nutrient deficiency covering the world and its regions is regularly produced. The situation in this respect can only be imperfectly gauged from the scattered cases for which data are available, and from consumption of micro-nutrient-rich foods such as fruit and vegetables. The undernourishment indicator, based on the estimated distribution of habitual consumption, and reported as an average for three-year periods, ignores seasonal or transient situations of food shortage at the household or national levels, and may thus fail to reflect food insecurity arising from short-term instability of food supply. Besides, based as it is on ex post data, it also ignores insecurity of future access, i.e., the risk component of food security. Moreover, given that it is also based on objective facts, it ignores the subjective dimension of food insecurity (e.g., fears that not enough food will be available in the near future), which may prompt people to adopt pre-emptive coping mechanisms of various sorts. The availability and distribution of such subjective feelings and coping mechanisms vary across households and as such are not well monitored on a worldwide basis. Despite their shortcomings and because of their wide availability and direct relationship with estimated agricultural production via food balance sheets, FAO undernourishment measures are widely used, this book included, along

359

Measuring historical trends

with a number of other indicators, especially anthropometric measures of malnutrition and data about food production and trade.

12.6. Stability of food availability and access Food availability and food access should be stable, and in addition, reliably stable. Nations or households should be reasonably confident that both food and the ability to acquire it should continue to exist, in both the present and the future. Stability problems include both inter-seasonal and interannual variability. Some extremely poor communities, relying on subsistence agriculture to cover food needs and without much monetary income, may face seasonal shortages (especially near the end of the agricultural year after a poor harvest) and inter-annual shortages due to poor harvests, droughts, floods, hurricanes, unusual outbreaks of pests and diseases, and other climaterelated events, and sometimes also due to violence that makes it difficult for farmers to care for their crops or livestock, or for food to be imported and transported to the areas where it is needed. Inter-annual instability may also be caused by economic downturns affecting employment and livelihoods; macroeconomic problems may impair a nation’s ability to import food, just as microeconomic trouble may reduce a household’s ability to purchase food. For instance, the wave of soaring food prices in 2007-2008 caused sharp increases in food prices and left some people without access to food in many poor countries; the 2008-2009 international financial crisis also caused increased unemployment and reduced access to food in various parts of the developing world (FAO 2008d, 2009a, and 2009b). Well-nourished people can withstand a spell of limited access to food by using up their bodily reserves of fat, storable vitamins, and minerals, but in many instances this does not apply: some people are already chronically underfed, or have an unbalanced diet. In these cases, a spell of reduced access to food may seriously affect their health and even their survival. Stability of access to food depends crucially on stability of income and access to markets. Excessive reliance on domestic self-sufficiency, curtailing links to the market, or macroeconomic conditions in which import capacity is reduced (e.g., an excessively devalued currency or a huge current account deficit, without ready access to international credit), may create unnecessary instability in that the whole food supply would be exposed to the vagaries of climate (especially in semi-arid countries reliant on rain-fed agriculture or on precarious irrigation systems that require upstream rainfall). Combining market and domestic production adds stability: shortages in world markets are not likely to coincide with domestic crop failures. 360

Hector Maletta

At the level of households, short-term instability of income does not immediately translate into instability of consumption in general, or food access in particular: consumption may be smoothed through the use of credit, household stocks or savings, or asset liquidation. Many poor populations lack access to credit and have only the most elementary assets to liquidate, and, to compound the problem, asset liquidation may endanger their future ability to procure food (and other items). In the absence of income, households may still use their food stocks, if they have any, until, say, income is restored or more food is harvested; for instance, subsistence farmers with no other income may consume their own food stocks from the previous harvest, to be replenished after the next crop. Thus, running out of income and stocks, without credit or savings, may cause seasonal hunger. In years past, many governments kept huge emergency stocks of staple food, at a high cost (including physical and financial storage costs, and the cost of constant renewal of perishable foodstuffs). The stocks were often created when the government purchased output at guaranteed support prices, at times of low farm prices, and were used at times of food shortages to stabilise prices and ensure additional supply. This practice has largely been reduced or discontinued due to its high fiscal cost and the ever more viable alternative of relying on increasingly liberalised and integrated world markets. As another alternative, some countries keep anti-cyclical funds, drawn from extra tax income accruing to fiscal coffers during good times, to be used at times of emergency or recession (to provide economic stimulus so as to expand welfare programmes for the poor, and/or pay for emergency food imports intended for domestic food assistance to poor households or individuals). Currently, very few governments keep huge emergency food stocks.

12.7. Measurement of nutritional status Besides undernourishment, which refers only to food access, there are other measures that point to the effects of inadequate food utilisation by individual organisms, chiefly through anthropometric indicators of nutritional status. As mentioned above, utilisation depends not only on food access and consumption, but also on health factors such as infections, which hinder biological utilisation. Food access is an input into people’s nutrition. Nutritional status is an outcome, usually measured by anthropometric indicators, especially height and weight. Attained height depends on both genetic and environmental factors; the chief environmental factors are (1) food intake and (2) health conditions permitting biological utilisation of food.

361

Measuring historical trends

Genetic factors and other random influences determine normal variation in individual heights around their average value. It has been established, however, that genetic factors do not influence average attained height in the earlier years of life; healthy and well-fed children under five from various ethnic groups attain similar average heights, and a similar distribution of heights, regardless of their ethnic group (WHO 1995, 2006). This finding prompted WHO to establish a universal reference growth curve for preschool children (WHO 2006), now based on an international study of healthy and well-fed children from several continents and various ethnic groups (the Multicentre Growth Reference Study, MGRS). The impact of ethnic (i.e., genetic) factors on average attained height, however, gradually increases with age, thus determining different average heights for adolescents and adults in various ethnic groups, even among healthy and well-fed individuals.44 Thus, height at ages over five is not used as an indicator of malnutrition since differences may also be caused by genetic factors unrelated to nutrition or health. For each gender and age, the height of healthy and well-fed children is normally distributed around the reference mean. In the reference (MGRS) population of healthy children, only a small percentage of children (2.3%) differ from the mean by more than two standard deviations. In an actual population, the percentage of such children is used as an indicator of stunting (i.e., shortness due to malnutrition). This indicator reflects the cumulative effects of long term exposure of children to insufficient food intake and/or frequent infections that hinder biological utilisation of food. For each height (and for each age), in turn, healthy and well-fed children who do not have excessive fat in their bodies have an average normal weight; in the reference population, these weights are normally distributed around the reference mean, with only 2.3% at a distance of over two SD above or below the mean. In an actual population, the proportion of children with weights below -2 SD of the reference mean weight is used as an indicator of malnutrition. Two such indicators are useful: wasting (insufficient weight for attained height) and underweight (insufficient weight for age). Wasting refers to thinness relative to the child’s current height, and reflects short term Thus, for instance, in spite of similar normal growth in childhood, well-nourished individuals descended from Nilotic tribes (originally living in Kenya, Ethiopia, and neighbouring countries) attain on average a taller adult height than other African ethnic groups. A similar difference is observed between people from Hokkaido vs. other islands in Japan. Besides these inter-ethnic differences, which are regarded as having a genetic origin, average stature for a given ethnic group may increase over time with economic development, as has in fact occurred in many countries during recent decades and centuries (Fogel 2004; Komlos 1994; Steckel and Rose 2002).

44

362

Hector Maletta

food insufficiency (due to lower intake or poor biological utilisation of food); underweight reflects the combined effect of long-term linear growth (height for age) and short-term weight for height. Nutritional status in the case of adults or adolescents measures only weight adequacy, relative to attained height. The main indicator is the Body Mass Index (BMI). In recent years, WHO has published reference BMI distribution tables by sex and age, for normal individuals aged 5 to 19 (WHO 2007); the figures for 19-year-olds are considered suitable for adults of all ages. Anthropometric insufficiency (in height, weight, or both) are indicators of the probability of malnutrition at population level; they are not indicators of individual malnutrition per se, especially in the case of stature but also for weight. To determine whether a particular child is actually malnourished other indicators are needed about individual health status (e.g. evidence from blood samples, presence of nutrition-related diseases such as scurvy or night blindness, etc.). Even among healthy and well-fed individuals there is a normal range of variation in height and weight; there are always some individuals who are naturally short, slender, stocky or very tall, irrespective of nutrition. Trends in anthropometric indicators of malnutrition (chiefly stunting and wasting in children under 5) are regularly collected by countries through health and nutrition surveys, and assembled by the World Health Organization in a world database (www.who.int/nutgrowthdb/estimates/en/). The WHO also maintains a world database on overweight and obesity, though it has not yet achieved sufficient world coverage. Worldwide trends in nutritional status are produced by means of multilevel regression models using information from successive surveys carried out in multiple countries, translated into worldwide or regional trends (e.g., de Onís et al. 2000, 2010, 2011).

12.8. Numbers and percentages In recent years it has become fashionable to refer to the number of affected people, e.g., the number of undernourished persons, or persons at risk of hunger, instead of focusing on their percentage relative to the total population. For instance, FAO often refers to the number of undernourished persons, and several Millennium Development Goals are stated in terms of the number of persons affected. In fact, both approaches are valid, but absolute numbers give rise to an ambiguity resulting from the fact that total population is not constant; a change in the absolute number of the undernourished may be caused by changes in the prevalence of hunger, changes in population size, or a combination of both. 363

Measuring historical trends

If an increase in the absolute number of hungry or poor people were an indicator of a worsening social situation, then poverty in the United States of America would have to be regarded as greater today than at the time of the country’s independence in 1776: at that time, total population was less than four million people, of which the vast majority were undoubtedly poor by today’s standards; recent poverty statistics in the US show a poverty level of 15%, with 46 million people below the official poverty line - over ten times as many as the total population of 1776. U.S. infant mortality, if measured by the number of infant deaths, will be greater now than in 1800 or 1900 due to population increase alone, indicating larger numbers of infant deaths even though infant mortality rates (as well as birth rates) have been greatly reduced in the intervening years. If absolute numbers are in any way a relevant measure, this will also be true for people free from hunger. The numbers of those in this category have greatly increased, even in countries or periods in which the numbers of the malnourished were increasing. Recent (FAO-SOFI [2014]) estimates find that the number of undernourished people in the world fell by 209.2 million, from 1014.5 million in 1990-1992 to 805.3 million in 2012-2014. In the same period, the number of non-undernourished people increased in far greater proportions, having grown by 1963 million, rising from 4393 million in 1990-1992 to 6356 million in 2010-2012. If one chooses to use absolute numbers, the increase in the number of non-hungry people should be the most relevant: they are far more numerous than the hungry and increasing at a faster rate than the numbers of the undernourished are decreasing. Thus, in the period from 1990-1992 to 2012-2014, for every reduction of one in the number of undernourished people, there were nine new non-undernourished people in the world. This whole approach based on absolute numbers is, however, profoundly misleading. Neither the number of hungry people nor the numbers of the well-fed are relevant indicators. Absolute numbers may have headline value and may be required for some purposes, e.g., for calculating how much food is necessary in an emergency in order to provide targeted food assistance to people in need. But for most purposes, the most appropriate indicator is the percentage or prevalence, in this case, the number of undernourished people as a percentage of total population at a given time. This is also valid for poverty, infant mortality, illiteracy, or unemployment, especially for comparisons across countries or over long periods of time where the relevant population may also vary significantly.

364

13. How to see the future: projection methods

13.1. General methodological background Prophecies, extrapolations, predictions, and projections. Forecasting future events may take many forms. The first two (prophecies and extrapolations) are easy to dismiss, but a key conceptual distinction should be made between the other two (projections and predictions), both of which could constitute valid modes of forecasting but which differ from each other. A prophecy is an unconditional statement about the future, without much scientific evidence to support it; it may purportedly be based on extrascientific grounds such as the voice of an oracle, words in a holy book, or a pattern of tea leaves in a cup. An extrapolation is also an unconditional statement about the future, but is usually quantitative, and commonly based on the belief that the future will be an unaltered continuation of past trends: if one magnitude grew at a certain annual rate in the past, it is believed that its future value can be forecast by extrapolating that rate of growth. But just as prophecies are highly fallible, so too are extrapolations; the future need not be a passive replication of the past. Unlike prophecies and many extrapolations, predictions and projections are statements about the future that are based on more solid scientific theory. Predictions are statements forecasting what will happen under specific conditions according to accepted theory. For instance, known facts and accepted astronomical theories about planetary motions inform predictions of the date of future solar eclipses. A good prediction must take into account all relevant factors in order to anticipate their future status and implications.

365

How to see the future: Projection methods

They may be precise (as in astronomy) or imprecise and probabilistic (such as the life expectancy of a patient with terminal cancer). Projections, on the other hand, are not unconditional; they are conditional on a hypothetical scenario containing contingent events that may or may not materialise. A projection simply makes explicit the logical implications of a series of scenario assumptions; the assumptions themselves may fail to materialise, and their suspected consequences might be of an uncertain nature or magnitude; when many aspects are involved, it is easy to omit or misunderstand the effects of a certain key variable that is essential to the matter. Projections are appropriate when dealing with a complex evolving system with many interacting variables and multiple uncertainties concerning the value of its parameters and the precise mathematical relationship between them. Projections describe the expected consequences (according to a given theory) of a specific combination of hypothetical developments that may or may not actually occur, on the basis of a mathematical model that is assumed to be appropriate for these purposes. How the future will unfold is uncertain, and thus projections are usually made for multiple hypothetical scenarios, quantified using multiple models, in an attempt to cover a variety of possible future situations. The fact that a particular scenario ultimately fails to materialise does not imply that the basic theory must be rejected or adjusted, since the scenario in question was only one of many possible ones, all compatible with the relevant background theory, and it may have been simulated with the aid of a specific mathematical model (or suite of models) selected from among many possible models that differ in various details. One important consequence of this distinction, therefore, is that predictions can be falsified, but projections cannot. If Newton’s theory of gravitation implies that a planet must be orbiting the Sun at a certain distance beyond the orbit of Uranus, and the planet is eventually observed (as actually occurred when Neptune was discovered on the basis of such a prediction), the theory is thus corroborated. Conversely, a failed prediction (e.g., no such planet is found at its predicted location) means that the background theory, the data, or even both were wrong.45 Corroboration does not entail verification of the background theory on which the prediction was based: finding Neptune corroborated Newton’s gravitation theory, but that theory was ultimately disproved when other Newtonian predictions failed, e.g., when (in 1919, during a solar eclipse) the light from a distant star was seen to bend towards the Sun, as predicted by Einstein’s General Relativity, thus falsifying Newton’s Universal Gravitation theory, according to which light cannot be affected by gravity due to its absence of mass; the 1919 finding thus reduced the Newtonian theory to the status of a sound empirical approximation subsumed under Einstein’s more accurate and more general theory.

45

366

Hector Maletta

A scenario’s failure to materialise, instead, does not disconfirm the basic theory. However, when a scenario anticipates events that ultimately fail to materialise, it will be legitimate to infer that some of the scenario’s assumptions or hypotheses were wrong, even if the projection based on such assumptions was correctly inferred. A dismissal of a scenario on this basis, however, is usually only possible ex post, after the course of events has diverged from the scenario’s assumptions and implications. Another way of rejecting a scenario that may be possible even in advance is to prove that some of the assumptions or parameters were wrong from the beginning. For instance, climate change models assume a value for certain unobservable or uncertain parameters such as the sensitivity of global temperature to a doubling of atmospheric carbon dioxide concentrations. If further research shows that the assumed sensitivity value was wrong, a change should be made in the model’s parameters, probably modifying projections of future climate change based on that model, though not necessarily casting doubt on the basic theory about greenhouse gases and climate. Projections about the future of agriculture and food security usually rest on a large number of assumptions, and on theories linking assumptions and consequences. For instance, these projections assume certain biophysical models whereby crops grow according to the species cultivated, the specific variety of that species, the soil in which it is grown, the prevailing climate at the reference site, and a set of farming techniques such as time of planting, use of fertiliser and pesticides, etc. Projections about future demand for farm products must assume some future growth of population and income, and also include assumptions on how demand for each agricultural product will respond to the growth of population and income, based usually on the idea that quantitative responses of this kind will be similar to the ones observed in the past. For instance, if per capita demand for milk in the past has been observed to increase as a function of per capita income, and the respective income elasticity of demand (the percentage increase in milk demand for every 1% increase in per capita income) has also been estimated from past data, then a scenario containing an assumed path of per capita income growth, coupled with the assumption that the observed elasticity will still be valid at a chosen future date (such as 2050), implies a certain per capita demand for milk at the chosen future period. Such assumptions are hypothetical: the effective level of income attained by 2050 may turn out to differ from the projected one, and people in the future may also modify their tastes and preferences, leading to a change in the income elasticity of demand for milk. Once future demand is projected in one way 367

How to see the future: Projection methods

or another, a projection of future supply to meet that demand should also be produced, using assumptions about future land use, land productivity, and agricultural technology. Because the future is uncertain and scientists deal with an imprecise body of theories, it is often the case that several scenarios are developed and several projections prepared, any (or none) of which might possibly materialise. For example, the Intergovernmental Panel on Climate Change uses several scenarios, with different levels of greenhouse gases emissions, and runs them with the aid of various computational models in order to make projections about the world’s climate. The purpose of having several scenarios (and various models for each) is usually to provide some guidance for policy-making under uncertain conditions. How to build scenarios. A scenario that is worth considering is not just any combination of the factors at play but a plausible one. Plausibility implies consistency and feasibility. Consistency means that all the assumptions defining the scenario are coherent with one another. For instance, the rate of demographic growth is related to the level of economic development: typically, high income countries have lower fertility rates than developing countries, and thus a scenario about the future of a country or region should not (normally) assume it will be simultaneously rich and have high fertility (possible exceptions are rich conservative societies such as Saudi Arabia). If a scenario assumes a certain path of population growth and separately assumes a certain path of economic development, the exercise may well result in an implausible combination of income level and fertility rates. Feasibility requires that scenarios avoid assumptions regarded as materially impossible or extremely unlikely (for instance, assumptions requiring implausibly extreme values of a crucial variable, or a sudden and unwarranted reversal of past trends). A plausible scenario of future development in a complex system involves not just positing the final state of the system, but a plausible path whereby each intermediate stage evolves into the next through the interaction of all its moving parts. Agriculture and food systems involve a combination of many human activities and natural processes in a complex adaptive system, and any scenario regarding the future development of that system should take these essential features into account. Once the ‘storyline’ of a scenario is established, it is implemented in a mathematical model purporting to reflect or simulate the main features of the underlying processes. There is usually a wide range of possibilities regarding the mathematical shape of the model that quantifies a 368

Hector Maletta

storyline. Given a formal model, in turn, there is usually a range of possible numerical values for each of the multiple parameters that must be fed into the model in order to obtain the desired projections. Modellers frequently use several scenarios based on various sets of assumptions and parameters that are regarded as both plausible and feasible. Projections about agriculture and food security. Future prospects of food and hunger depend primarily on projections of food demand, availability and access. These in turn depend on other factors. Food demand depends mainly on population and income. World food availability depends on the amount of food production that is actually available for human consumption (i.e., not expected to be used for other purposes such as animal feed or biofuels). Local availability in particular regions or countries also depends on food trade. Access to food by households and individuals chiefly depends on the level and distribution of income and expenditure. Future food production to meet projected demand may be constrained by the expected availability of resources (like land and water), climate projections, expected technological progress, and other factors.

13.2. Population, income, and food demand Future food demand depends chiefly on the expected number of people, their expected income, and their expected food-demand response to income. Projection of future food demand is thus understood as a function of population and income projections. Each additional human being implies additional food consumption, but the amount and composition of the extra food demand depends on income (and local cultural preferences). Most projections assume a certain exogenous rate of population growth, and a given path of economic development. Estimating the prospects of hunger thus involves not only an assessment of future agricultural growth, but also assumptions about future growth of population and income, and about how per capita demand for food (and for each specific foodstuff) is affected by income and associated factors such as urbanisation and education. All this should be projected not on the world level but by country or region. Data suggest that the higher the income, the lower the growth rate of per capita consumption of dietary energy, rising from around 2000 kilocalories per day in the poorest countries to stabilise somewhere around 31003300 kilocalories per day once income reaches relatively high levels. Diet composition also changes with income. At very low levels of income, simple and rather monotonous diets prevail: usually well over one half of the dietary 369

How to see the future: Projection methods

energy consumed by people in this category, comes from cereals and tubers, with low consumption of meat, dairy, fruits, and vegetables. At the chemical level, this means that low income diets are mostly based on carbohydrates (sugars and starch). At higher income levels, consumption of staple foods stabilises or declines while consumption of fat, protein or micronutrients (from meat, milk, vegetable oil, fruit or vegetables) tends to grow with income even beyond the saturation level of staple food. These broad trends, in turn, vary greatly across world regions; for instance, the enormous increase in sugar consumption observed in the United States during recent decades is much more moderate in Western Europe and practically absent in Japan. In practice, however, existing country-level data on the correlation of income, education, and food consumption are frequently used to arrive at projections of future demand for food (considering either total food or specific food items) once a scenario has been adopted regarding future population and income growth. FAO and other international institutions, as well as academic researchers, have accumulated considerable data and experience in the analysis of the relationships between income and food demand. FAO projections of future food consumption (e.g., FAO 2006, 2011, and AB 2012) are based on this store of knowledge.

13.2.1. Population projections The standard source for population information is the UN Population Division series of estimates, which start in 1950, and projections for the future that extend to 2050, 2100 or beyond. The UN usually prepares three variants of its projections: high, medium and low. Available projections of agriculture are based on the medium variant of UN population projections and they are cited or used in this book without modifications. UN estimates and projections are updated every two years, based on country censuses, surveys, and demographic records (births, deaths, and migration). In even years such as 2010 or 2012, projections of total population (by sex and age) are updated, while rural-urban estimates and projections are updated in odd years such as 2011 and 2013, based on the population revision of the previous year. The most recent projections of future agricultural production or food consumption available at the time of writing are based on either the 2008 or 2010 revisions of the UN population projections. UN population projections are based on a purely demographic approach. The future course of fertility rates, for instance, is not explicitly related to expected growth in fertility determinants such as income or education, but only to past trends and exogenous assumptions about the future, e.g., 370

Hector Maletta

the hypothesis of convergence of all countries towards a common fertility rate. These assumptions on fertility (as with those concerning mortality or migration) are not clearly linked to projections of income, education or other relevant factors. This is important, because projections of agricultural growth do assume a certain future course of economic development. It might be the case that the levels of future fertility assumed in population projections are not compatible with some of the hypotheses on economic growth employed in agricultural (or other) projections. The replacement level of fertility is 2.1 children per woman. In recent UN projections of population, the fertility rate of all countries is assumed to converge towards the replacement level (except in the 2008 revision, where it was assumed to converge to 1.85, with the process of convergence starting almost immediately after the baseline period). Fertility has been observed to decline well below 2.1 (to levels around 1.2-1.5 children per woman) and to remain at that low level for many years in various developed countries, e.g., in Western Europe and Japan, or in less rich but deeply secularized countries such as many East European countries. In many of these low-fertility countries it is still declining. Hence it seems debatable whether women in all these countries will suddenly change their behaviour and start having more children in the coming decades, as assumed in UN projections. It is also not obvious that countries with increasing incomes and education and declining fertility, which are now somewhat above two children per woman, will stabilise at replacement level instead of continuing their decline in fertility down to levels previously observed in richer or more secularized countries. Evidence in this regard (Myrskylä et al. 2009) suggests that sub-replacement fertility tends to persist, but eventually to rise back towards replacement level (without necessarily reaching it) only at very high levels of income and living standards, as measured by extremely high values on the UN Human Development Index; such high levels of human development have been reached by just a handful of very rich countries, and the observed increases in fertility rates are still well below the replacement level. Some countries, e.g., Japan, continue to have low fertility despite reaching the highest echelons of the Human Development Index years ago. Rich countries may stay below replacement level for many years, and there is little evidence that all countries would ultimately converge to the same level. Migration assumptions in UN projections are also ‘blind’ to economic, social or cultural considerations. The net balance of international migration, for instance, is expected to decline rapidly over time for all countries, including those with either a negative or a positive migratory balance at present. In 371

How to see the future: Projection methods

most countries, net migration would become zero in two or three decades. This assumption is not related to any assessment of known factors driving international migration, which in reality has been increasing in recent decades. Voluntary international migration motivated by economic reasons is widely expected to keep increasing due to the rapid development of globalisation and inter-dependence, causing increased mobility of goods, services, capital, and people across national borders. Violent conflict or other severe problems also cause flows of forced migration (refugee flows), and this kind of migration is also showing a tendency to increase over time. Immigration, in turn, is a possible factor affecting a country’s fertility rate. If millions of fertile-age women from high-fertility countries move to Western Europe, average fertility possibly will increase in Europe. These immigrant women, with their culture of higher fertility, may bear more children than their native counterparts during their remaining fertile years in their new country. Moreover, their daughters, when they reach childbearing age, may also have higher fertility than the surrounding native women of their generation. In the case of the US, immigrant women are notably more fertile than their native counterparts (Preston and Wang 2008). But in UN projections, migration has been assumed to remain low or diminish, while fertility is expected to be rising, thus implicitly negating any influence of immigration on the recovery of fertility. UN projections, besides, do not distinguish between native and immigrant women as regards fertility, nor include immigration as a factor in fertility projections in any explicit way. Population growth is in large part conditioned by economic and social development. It should not be projected independently. Ideally, agricultural projections should emerge from an integrated model, where demographic, economic, social, cultural and environmental projections are coherently combined in plausible scenarios. Some such integrated models do exist, including the ‘integrated assessment’ models developed by IIASA and FAO, which jointly produce climatic, agricultural and economic projections. But even these integrated models use the UN population projections as an exogenous and separate input, instead of creating a model in which population is endogenously determined along with income, greenhouse gas emissions, and other intertwined variables that are assumed to be related to one another. In the absence of such integrated schemes, the UN demographic projections are usually taken as a given in most existing projections related to agriculture, climate, and food demand. It should be noted, however, that past experience with successive rounds of UN projections suggests that they tend to overstate 372

Hector Maletta

population growth. Actual demographic growth in the coming decades may then be expected to be somewhat below the projected figures. As a result, projections of future food demand are likely to be overstated. The consequent differences may not be significant up to 2050, but could be more significant when population projections are extended to the second half of the century.

13.2.2. Projections of income Per capita income growth rates are closely (and inversely) related to per capita income level (see empirical evidence on correlates and determinants of economic growth in Barro [1998]): the speed of growth typically decreases as income grows higher. Per capita income in rich countries rarely grows faster than 1-2% per year for any extended period of time, while rates over 5% are quite common in fast-growing emerging countries. Income growth may, however, be very low or negative in ‘failed states’ or very poor countries, such as Haiti or Somalia, where minimal institutional conditions for sustained economic growth are not met, and demographic growth rates tend to be high. The most usual projections of economic growth, however, are rather ad hoc, and are not based on any theory about the determinants thereof. After adopting some hypothesis about population growth, another hypothesis is separately adopted about future economic growth. In this book, we cite and use several existing projections of climate, agriculture, and hunger that are based on exogenous and unrelated assumptions about income and population growth, all of which regrettably follow this sort of methodological misconception, though we are aware of their intrinsic limitations and point them out when necessary.

13.2.3. Projections of poverty The prevalence of extreme poverty up to 2010 and projections for 2015, as estimated by Ravallion (2013), suggest that a little more than one billion people endure such appalling living conditions. Ravallion then asks ‘how long it will take to lift one billion people out of poverty’ (actually meaning a reduction in the numbers of the poor by one billion), under different assumptions regarding income growth and income inequality. He uses two hypotheses: in the more pessimistic one, the developing world outside China will return to the (lower) rate of reduction in the prevalence of poverty that was observed in the 1980s and 1990s while China stays on track; in this case, it will take about 50 years to lift one billion people out of extreme poverty. In a more optimistic scenario, poverty rates in developing countries outside 373

How to see the future: Projection methods

China would keep diminishing at the (faster) rate prevailing since 1990, and China would also retain its historical performance in this regard. Under this second scenario, one billion people will be lifted out of poverty by 20252030. Under current population projections, this absolute reduction of one billion in the ranks of the poor would mean reaching a poverty prevalence rate of 3%, sharply down from the 16% projected for 2015 for developing countries as a whole. In practice, a reduction in the overall rate to 3% allows for some countries (which are now developing) reaching levels approaching zero poverty, while others remain at higher levels, as is the case today. Even if on the world level extreme poverty becomes insignificant in the future, it will still be a pressing reality in some parts of the world. In this vein, some authors (including some who, like Ravallion, work with the World Bank) have criticised Ravallion’s projection to 2030. Kaushik Basu (2013) noted that by 2030 (even with global extreme poverty decreasing to just 3%) there will still be a number of countries with high poverty rates. This would be so even assuming that the poorest countries grow at unprecedented high rates in the years up to 2030. His central thesis is that growth alone will not suffice, and ‘shared prosperity’ (a recent World Bank catchphrase implying a strong policy-propelled reduction of inequality within and between countries) is a necessary element for the eradication of poverty. Along similar lines, Yoshida et al. (2014) remark that Ravallion’s projections assume that the developing world will remain at the current pace of poverty reduction without showing that the developing world can actually continue the current pace of poverty reduction. Yoshida et al. propose replacing the use of a global population growth rate, as in Ravallion (2013), with specific projections of demographic rates for each particular country (whereby the population of poorer countries tends to grow faster). In a second correction of Ravallion’s work, they assume that income (or consumption) inequality will not remain constant but will keep changing in the way it has been doing in recent years (i.e., towards more domestic inequality albeit with less inequality between countries). They conclude that as poverty prevalence reaches lower values, its growth elasticity diminishes, so that reductions in poverty due to growth become increasingly more difficult to achieve, with the result that ever higher rates of economic growth are required to achieve a given reduction in poverty. In particular, per capita income (or expenditure) in developing countries, which is estimated to grow at about 2.7% in 2010-2012, will have to accelerate to 374

Hector Maletta

an annual 5% by 2022-2024, and then climb to impossible rates of over 40% per year to achieve the goal of 3% poverty by 2028-2030. This non-linearity expresses the diminishing returns of income growth on poverty reduction, and the increasing difficulty of eradicating the last of the poor (usually located in the most difficult-to-reach places): in their scenario projections, global poverty by 2030 will be about 8.5% instead of the 3% expected using Ravallion’s more simplified methodology. To achieve this goal, the authors conclude, growth alone is not enough: overcoming the deleterious effect of increasing domestic inequality will also be necessary. These conclusions are probably correct. Of course, reducing world poverty to 8.5% in 2030 would be a tremendous success, even if this still represented a significant level of poverty. But this analysis suggests that to accelerate the eradication of poverty, something more than mere economic growth is required.

13.2.4. Projections of food demand Given projections of income and population, demand for specific products may be projected on the basis of past trends and using demand functions that predict per capita demand on the basis of per capita income at country or regional levels. A large number of these functions have been estimated for particular products, geographical areas, and periods. Typically, the demand for each product (especially food) is more elastic at lower levels of income, but the effect of successive income increases leads to diminishing returns in terms of demand; for some products, demand may actually fall above certain levels of income, while in other cases demand simply stabilises or grows very slowly once certain income levels are attained or surpassed. The various FAO reports on the future of agriculture such as FAO (1995) or FAO (2003) make reference to a variety of studies about demand elasticity for each particular food product in various countries or internationally. Other than income, demand is also influenced by cultural factors (such as the strong vegetarian preferences of the population of India, or preference for one particular cereal such as rice in East Asia, maize in Mexico and wheat in Europe) and by relative prices, including the price of some foods relative to other foods, or the price of all food relative to other goods and services. Some projection models (but not all) include the capacity to project relative prices, though up to now these projections have not been particularly good (e.g., practically no model in the 1980s or 1990s, or indeed up to 2002, predicted the food price rises observed in the late 2000s and early 2010s).

375

How to see the future: Projection methods

13.3. Projections of agricultural output 13.3.1. General methodological background From a very simple point of view, agricultural production may be expressed ex post as the product of agricultural land area (i.e., land used for agricultural production) multiplied by agricultural output per unit of agricultural land: Agricultural production = Agricultural land (in hectares) × Agricultural output per hectare

If this formula is used for a single product such as wheat or apples, the land involved is the area used for that single product, and then the output might be measured in physical terms, such as harvested area of this product, multiplied by tonnes obtained per harvested hectare. This is, however, not adequate when the formula is applied to an ensemble of products. In such a case ‘agricultural output’ should result from the aggregation of multiple products on the basis of some aggregation principle (or ‘metrics’), a common unit for all products involved. For assessments linked to food security, and also in general, as discussed in Section 12.2, the economic value of food output is preferable to other aggregation principles such as calories or tonnage. Real economic value of output requires controlling for inflation over time and for differences across countries in the purchasing power of money. Potential and effective agricultural production. One question frequently posed concerns the maximal output obtainable in a given country or territory with given land resources (and assuming a certain level of production technology). Potential agricultural production is composed of two factors: total land suitable for agricultural production, possibly of various qualities and types, and potential production per hectare for each type of land. Production per hectare, however, may refer to a variety of crops or livestock kinds; the same piece of land may have low potential for one crop but high potential for another, and likewise for various kinds of livestock. Potential production can only be defined under very restrictive conditions, for example the potential output of traditional varieties of rain-fed cereals under traditional techniques of production. Using higher-yield cereal varieties or applying better technologies may alter the results, probably resulting in a higher potential output from the same land. Similarly, allowing farmers to choose among a wider variety of crops may also cause potential output to vary, probably increasing the economic value of the maximum output obtainable from the same land.

376

Hector Maletta

In practice, production is not projected on the basis of the maximal or potential land use or land productivity, but determined by expected demand. Any notion of maximum production would act only as a constraint: demand could not be met if it exceeds potential production, but given the possibility of trade, this restriction is hardly binding at the local level. It could only become binding at the planetary level, but this is not something that is expected to happen in a world where population is about to stabilise, land use for agriculture is already nearly stable, food production exceeds requirements, and half of the usable land remains untapped. The idea of estimating maximum potential production historically emerged out of Malthusian concerns about how many people the Earth could feed or sustain; see, for instance, a long series of estimates of the maximum population carrying capacity of the Earth, from 1679 to 1994, in Cohen (1995:402-418). However, available studies for agriculture in the 20th and 21st centuries show that: (a) potential production has been increasing over time due to improved technologies; and (b) production potentials (at least at the world level) are in fact much higher than required to meet demand. For instance, AB (2012) and other previous FAO studies suggest that only one half of all Prime and Good land is actually used for crop production, and that little more land will need to be cropped to meet future growth in demand, even if productivity grows more slowly than at present or in the recent past. Given these considerations, we shall restrict our attention to projections of effective, rather than potential, farm production. Here, effective farm production is defined as production actually obtained from land that is actually being used for farm production. Actual production within a certain geographical area is not usually at the maximum potential level, no matter how the latter is defined. The reasons for not producing at maximum potential are broadly twofold. On the one hand, not all producers know how to obtain the maximum yield, can afford the latest technology, or are able successfully to apply it to its full extent. On the other hand, effective demand for food products may be satisfied with less than the maximum level of production achievable on a world scale. Actual production will tend to match effective demand (mediated by market forces such as incomes and prices), and demand is likely to be lower than the theoretical potential, which is itself increasing over time due (mainly) to technical progress. If demand for a particular product reaches levels close to the potential, the corresponding price is likely to rise, reducing demand and fostering supply of that product or its substitutes.

377

How to see the future: Projection methods

Projections of effective demand and investigation of the probable ways in which production (either domestic or foreign) will meet such demand are thus more important than potential production. A given amount of food could be produced with different geographical distributions, with different technologies, and using different amounts of each type of land. The final outcome will be the result of existing resources and constraints at each geographical location, and decisions will be (to a large extent) made by farmers motivated by market conditions, and also affected by public interventions (e.g., investment in irrigation schemes, agricultural subsidies or taxes, etc.) and other factors external to the farm (agricultural R&D, macroeconomic conditions, supply of inputs, etc.). In all probability, in view of existing data and projections, production of the food output required to meet demand will be feasible but will not require all available land to be used at the maximum possible level of productivity. Any projection exercise may have some leeway to assume various possible combinations of increased farmland and/or improved technology (i.e., increased output per hectare), always remaining below the potential output allowed by each combination of land and technology. In most cases, such exercises foresee little future increase (during the present century) in the area devoted to crops or livestock. Local imbalances of supply and demand are adjusted by markets through foreign trade: if agricultural growth in a particular country does not match growth of domestic demand for agricultural products, some surplus domestic products may be exported and/or some foreign products imported, as needed. Trade enters the picture along with local production at a level that depends on the macroeconomic environment, which may be more or less conducive to the country’s integration into global markets. Imports of food products may be afforded by exports of other agricultural products, by non-agricultural exports, or by other sources of foreign currency revenue (such as remittances). At the world level, global production may be expected (on average) to match effective global demand. Effective demand, of course, means solvent demand: even if supply equals effective demand, some people may remain hungry due to inequalities in economic access to food. However, it has been repeatedly observed that undernourishment is inversely correlated with per capita food consumption (or its ratio to per capita food needs). Moreover, undernourishment in most cases becomes not significant above a certain ratio of food consumption to food needs. These correlations are often used to project future levels of undernourishment. 378

Hector Maletta

Expansion of farmland. Farmland comprises land used for annual or permanent crops, and land allocated for livestock grazing. At the world level, farmland has expanded relatively slowly in recent decades and at decreasing rates; in fact, it is estimated to have peaked in the early 1990s and to have declined slightly in the 2000s. The main factor driving growth in agricultural production in the recent historical record is not the addition of extra land, but a sustained increase in output per hectare. In the future, a slower pace of productivity growth may motivate an expansion in the land area used by agriculture, if the growth of demand justifies it. Besides the possible expansion or contraction of the land area actually used for agriculture, a parallel issue is the amount of land that is suitable for agriculture. As shown in Ch. 4, only one half of all land usable for crops is actually used for that purpose. Changes in the land area suitable for crops can be expected and might be caused by several factors. One of these factors is climate change: global warming may cause some currently unsuitable land, located in relatively cold climates, to become apt for agriculture, e.g., by gaining a longer no-frost period and better overall conditions for agriculture; this would expand the land suitable for crops, but not necessarily land actually used for crops. At the same time, climate change may cause some agricultural land to become unsuitable for crops (e.g., desertification of some tropical semi-arid agricultural land, if climate change results in even drier conditions in the corresponding zone). At the world level, the net change in land potentially usable for agriculture will depend on the balance between increases and reductions in various regions. Most available projections in this regard conclude that the net effect will be small, mostly agreeing that it would imply an expansion of potential farmland (i.e., land suitable for agriculture) worldwide due to climate change. Another factor for potential farmland expansion is deforestation, as has been occurring (albeit at decreasing rates) in the Amazon, Indonesia, and elsewhere. Whether such increases in potential farmland (land suitable for agriculture) do result in an actual expansion of farmland will depend on effective demand for agricultural products, which are mostly driven by population and income; it will also depend on the rate of growth of land productivity, since rapidly increasing yields in existing farmland, or increasing cropping intensity, may discourage expansion into new (and possibly inferior) land. In recent decades, for example, strong growth in demand was matched by strong growth in production without much expansion of farmland. Most existing projections of agriculture tend to project a very small increase in the use of land for agriculture: even if productivity grows more slowly, population growth (and 379

How to see the future: Projection methods

the growth of food demand) would also be slower; the net result would be, if anything, just a small increase in agricultural area. Technological change. Productivity per unit of land may increase due to application of new technological knowledge, or to more extended (or more efficient) use of existing technological knowledge. Besides, since output comprises many products, the value of production per hectare (at constant prices) may also change due to shifts in the product mix, i.e., in the allocation of land to a different mix of various crops and livestock types. Such changes in product mix will mostly depend on the structure of demand. Efficiency in the use of existing knowledge depends on effective adoption and adaptation of such knowledge by farmers. New technological knowledge may come from various fronts such as more efficient machinery, better plant protection, improved farming practices, or the creation of better seeds or breeds through adaptive research or genetic engineering. Besides proper technological innovation, much improvement in output is brought about by more efficient use or wider adoption of existing technology, which in turn is affected by farmer education and agricultural extension. As in other industries, not all farmers operate at the technological frontier due to various factors such as poor education, lack of access to markets and many others, all making for incomplete and slow adoption of existing knowledge. In most developing countries, there is a huge ‘catch-up gap’ due to insufficient or inefficient use of existing technology, which permits productivity in these countries to advance faster than is the case in developed countries where most farmers use up-to-date technology. Even if there are no new breakthroughs in agricultural technology, wider diffusion of existing technology at the rates of adoption observed in the past will lead to a sustained increase in the productivity attained by agriculture. Breakthroughs, for their part, are not likely to stop being made, and thus the technological frontier is likely to keep expanding. Changes in land use and the product mix. Real agricultural output per hectare consists of many different products and is usually measured in terms of value of output, at constant prices. A change in the value of production may involve changes in physical yield per hectare for any particular crop or animal, or changes in land use or the product mix. Changes in land use may involve changes in the area of agricultural land devoted to different crops, in the share devoted to crops vs. livestock, or in the amount of land allocated for agricultural and non-agricultural purposes. It may also involve more intensive 380

Hector Maletta

use of land (e.g., passing from one to two crops per year, or changing the proportion of fallow land) and changes in water management (from rain-fed to irrigated crops, or from traditional to modern methods of irrigation). Changes in the product mix, both at the farm and aggregate levels, normally follow the dictates of demand, as well as the changing possibilities afforded by available technology. Demand may increase for certain products and be stagnant or shrinking for others; for instance, per capita demand for cereals has been close to stagnant since the 1970s, while demand for oil crops has grown enormously, as has demand for dairy products and vegetables. These changes are chiefly produced by changes in income levels and world markets, and may secondarily be influenced by changes in tastes or preferences (e.g., a reduction in the prevalence of smoking may cause a reduction in tobacco crops). On the supply side, the development of new seeds with higher yields, or new techniques that save on resources (such as no-tillage cultivation) may encourage production and lower the price of some crops more than others, thus affecting production and demand. To sum up, agricultural growth may be affected by changes in demand (e.g., income growth triggering shifts in demand towards more expensive foods like meat, fruit, and vegetables, as compared with staple food such as cereals or tubers), by changes in environmental conditions (temperature, water availability, atmospheric CO2 concentration), and by invention or adoption of improved technologies (pressurised irrigation, no-tillage or low-tillage practices, better seeds, confined animal husbandry, soil conservation techniques, better plant protection, more efficient equipment, and so on). Changes may or may not involve an increased level of on-farm capital investment (relative to land area or to farm revenue), usually require improved (or different) skills on the part of farmers, and various off-farm (public or private) developments in roads, processing plants, power supply, storage and trading facilities, or other aspects necessary for agricultural production and marketing. Projecting future agricultural growth requires the projection of all these factors. On the demand side, the main factors are population and income. On the supply side, the key factors are land expansion, technological progress, and climate change. The transformation of these factors into actual growth in production may involve additional capital investment and improved farmer education and skills.

381

How to see the future: Projection methods

13.3.2. FAO projections of farm output The most serious and systematic exercises for projecting agricultural output have been undertaken by FAO, based on its huge store of detailed knowledge about the agricultural sector of most countries in the world, and on every crop and livestock. FAO has produced a series of projections of world agricultural production starting with World agriculture towards 2000 (FAO 1981) and Alexandratos (1988) and proceeding to intermediate updates and then to projections for more recent target dates (2010, 2015 and 2030), as in FAO (1995) and FAO (2003a). The most recent projections, to 2050, are those in FAO (2006), an interim report revised and updated in AB (2012). FAO projections are based on agro-ecological zoning, crop models, economic models, and exogenous assumptions about economic and demographic growth, but do not include the presumed effects of climate change (which are mentioned and briefly discussed, e.g., in FAO (2003, Ch.13), but not integrated into the projections). Fischer (2011) uses a variant of FAO projections as a ‘no-climatechange’ baseline in order to assess the presumed effects of climate change and biofuels under various scenarios. FAO projections constitute a valid attempt to reflect expected developments in agricultural demand, trade, and production under (existing or expected) natural and technological constraints. However, it has some specificities and limitations that are worth noting. This review of its methodology reflects the details provided in FAO (1995, Ch. 42), and FAO (2006, Appendix 2). The most basic tool in FAO’s projection is the commodity balance, or Supply-Utilisation Account (SUA), which we have introduced in Section 12.4.1. These balances, in FAO prospective studies about the future, cover 32 commodities (or commodity groups); some products were included individually (e.g., wheat) while other products (e.g., peas or beans) were aggregated into a group (pulses). These 32 basic commodities or commodity groups are: Crops: wheat; rice (paddy); maize; barley; millet; sorghum; other cereals; potatoes; sweet potatoes and yams; cassava; other roots; plantains; sugar (raw); pulses; vegetables; bananas; citrus fruit; other fruit; vegetable oil and oilseeds (in vegetable oil equivalent); cocoa beans; coffee beans; tea; tobacco; cotton lint; jute and hard fibres; rubber.

382

Hector Maletta

Livestock: bovine meat; sheep and goat meat; pig meat; poultry meat; eggs; milk and dairy products (in whole milk equivalent). However, for estimates about the developing world, the analysis of some of these commodities was further disaggregated. Thus, vegetable oil (one of the 32 commodities) was disaggregated into eight categories for oils from various crops (soybean, coconut, sesame seed, groundnut, sunflower, rapeseed, palm, and ‘all other oilseeds’). Sugar was disaggregated into sugar cane and sugar beet. Cow milk and goat/sheep milk were separately analysed. This raised the total to 41 commodities (34 crop and 7 livestock products), all of which are farm products: fish and other aquatic products are excluded, though consumption of fish and seafood is included in FAO food balance sheets. Some products not included in the list of basic commodities (e.g., camel meat) are nonetheless added for the purpose of calculating certain total quantities such as total dietary energy or total agricultural GDP; the addition of these other products is based on the observed (or in some cases, projected) proportion of the omitted products within the relevant total. The basic products were analysed on a country-by-country basis (though some small countries were grouped). The study considered 120 geographical units (countries or country groups) including 26 developed nations and 94 units in the developing world: 41 in Sub-Saharan Africa (40 individual countries plus a group of six small countries called ‘Sub-Saharan Africa, Other’); 15 units in North Africa and the Near East (including 14 individual countries as well as ‘Near East, Other’, a group of 3 small countries); 25 units in Latin America and the Caribbean (LAC) including 24 countries and one group of 11 small Caribbean countries grouped under ‘LAC, other’; 6 countries in South Asia; 13 units in East Asia, which also covers developing countries in Oceania/ Pacific (including 12 individual countries as well as one group of 9 small countries, ‘East Asia, Other’). FAO projections are not a normative blueprint for what ‘ought to occur’ to reach a predetermined goal, nor are they a mere ‘extrapolation of trends’. The projections are based on detailed knowledge of specific products and specific countries, and on interdisciplinary analysis with the participation of many specialists and drawing on various sources. They are not the automatic outcome of a mathematical model; though models are used as a projection tool, the results are checked for inconsistencies with existing country-level and product-level knowledge. FAO documentation provides further details:

383

How to see the future: Projection methods

The variables projected in the study are (i) the demand (different final and intermediate uses); production and net trade balances for each commodity and country; and (ii) key agroeconomic variables, i.e. for crops: area, yield and production by country and, for the developing countries only, by agroecological zone (irrigated and rain-fed with the latter subdivided into dry semi-arid, moist semi-arid, subhumid, humid land, and fluvisols/gleysols); and for livestock products: animal numbers (total stock and offtake rates) and yields per animal (FAO 2003a:378). The bulk of the projection work [to 2015 and 2030] concerns the drawing up of SUAs (by commodity and country) for the years 2015 and 2030, and the unfolding of the projected SUA item ‘production’ into area and yield combinations for rain-fed and irrigated land and, likewise, for livestock commodities into the underlying parameters (number of animals, offtake rates and yields).[...] The overall approach is to start with projections of demand, using Engel demand functions and exogenous assumptions on population and GDP growth [...] In addition, but only for the cereal, livestock and oilseeds sectors, a formal flexprice model was used (FAO World Food Model; FAO 1993c) to provide starting levels for the iterations and to keep track of the implications for all variables of the changes in any one variable introduced in the successive rounds of inspection and adjustment. The model is a partial equilibrium model, composed of single commodity modules and world market feedbacks leading to national and world market clearing through price adjustments (FAO 2003a:379-380). The World Food Model is interactive (i.e., allowing for the simultaneous determination of commodity supply, demand, trade, stock levels and prices) and dynamic (i.e., allowing for the outcome of one year or a sequence of years to influence the outcome of future years). Fundamentally it is a price equilibrium model, which means that commodity price is determined at the level where world supply is equal to world demand and all variables are simultaneously determined. [...] The model consists of a set of demand, supply and stock equations for each commodity and each country with the levels of production and demand determined by factors including population and income growth rates, income elasticities, own and cross-price demand and supply elasticities, demand and supply shift variables and various assumptions about economic trends and policies. However, price assumes a central role in the model, because it enters in the determination of all supply and demand equations for all countries and all commodities. Domestic prices are linked to world prices, which in turn are determined by world demand and supply (FAO 2003c:166).

384

Hector Maletta

The World Food Model covered only a limited number of basic foods (cereals, livestock, and oil crops). Commodities covered outside the World Food Model included sugar; tropical beverages like tea and coffee; fruits; and agricultural raw materials. Basic foodstuff production scenarios largely assumed the continuation of trends in area, livestock numbers and crop and animal yields, as modified by the interaction of prices generated through a market clearing mechanism and checked for technical feasibility. It assumes, with no exceptions, a continuation of current national and international policies affecting production, consumption and trade. It also assumes ‘normal weather’, that is, the absence of any particular climatic condition, either favourable or unfavourable, which could affect yield or harvested area (FAO 2003c:166; our italics). The projections for the commodities not included in the World Food Model, namely sugar, tropical beverages, fruit and agricultural raw materials were made using methodologies, ranging from detailed econometric commodity models to supply and demand projections based on past trends supplemented by expert judgements of commodity specialists. In a number of cases, the projections were prepared jointly or in cooperation with international commodity bodies and international commodity experts. (FAO 2003c:166)

It is important to note that, on the one hand, FAO expertise has been employed for many years in developing countries, largely in efforts to ensure food selfsufficiency (considered at the time as the hallmark of food security). As a result, projection methods tended to produce as much as possible within the country in the first instance, when it was deemed technically feasible to do so. Net trade was obtained as a residual. Domestic production, in turn, was modelled as sensitive to additional inputs (e.g., more fertiliser or pesticides, more tractors, more investment in irrigation) and certain exogenous assumptions on technological progress for each kind of crop or livestock. The model did not systematically include relative cost functions to help decide whether additional supply would come from imports when domestic production was more costly, or whether more land should be devoted to products for export at the expense of products for domestic consumption. As a result of these assumptions and methods, FAO projections are generally oriented towards self-sufficiency, and therefore tend to understate the growing importance of food trade. Nonetheless, they end up anticipating an increase in food trade, which is probably understated but still significant.

385

How to see the future: Projection methods

The FAO system includes detailed assessments at the country and product level, and is thus not based on universal or uniform equations applying to all countries. It also involves expert judgment and is thus not a standard statistical model generating automatic results: it is checked to ascertain that projections for every country and product do not produce clearly inconsistent or unrealistic results. Relying mostly on past trends (which for the most part were those from the previous thirty or forty years, from 1961), projections prepared in the 1980s or 1990s were not able to predict the rapid increase in productivity that occurred in many countries during the 1990s and 2000s, including Africa and Asia but also (most importantly) exporting countries such as Brazil or Argentina, nor were they able to anticipate the consequent increase in world food trade. In particular, since technical progress was understood as a function of increased inputs per hectare, FAO projections were not able to forecast productivity changes that reduce the amount of inputs per hectare or per tonne (e.g., no-tillage techniques that reduce the use of machinery and fuel per hectare and per tonne). Most of these exercises did not anticipate the rise in agricultural prices in the 2000s, prompted by rapid demand growth in emerging countries, especially China, and made possible by new technologies like bio-engineering and no-tillage farming; nor did they predict the subsequent reduction of such prices after the spikes observed in 2007-2008 and 2010. One result of these shortcomings is that FAO projections have systematically underestimated future growth in agricultural production, as illustrated in Section 9.2.4. It may logically be surmised that FAO projections are also prone to underestimate future growth, since they are based on similar assumptions and procedures.

13.4. Impacts of climate change on agriculture 13.4.1. Climate change projections The Intergovernmental Panel on Climate Change (IPCC), established in 1988, has produced a series of reports on possible evidence of anthropogenic (man-made) climate change, caused chiefly by human emission of greenhouse gases (GHG), mainly carbon dioxide (CO2) from burning fossil fuels (like oil or coal) and from other sources (e.g., deforestation). Changes in world climate have different local manifestations and may affect agricultural activity. As these changes are expected to continue in the future, the IPCC has published a series of projections on the future climate, based on a range of possible scenarios of socioeconomic development and what is known about the natural and human processes involved. Based on these climate projections, estimates 386

Hector Maletta

also have been produced of their expected impact on agriculture. This section examines basic concepts about climate, climate change, and climate change projections. Defining climate. Climate is usually defined as the long-term average weather condition in a particular location. The conventional definition of climate (adopted by the World Meteorological Organization) defines it as the average level, seasonal variation, and inter-annual variability in temperature, precipitation, and other such conditions, measured over a 30-year period. Three decades may seem sufficient to define normal weather; however, it may not be enough to identify normal natural variability; some natural oscillations last longer than 30 years. It is now recognised that multi-decadal natural variation occurs in certain important processes such as the El Niño Southern Oscillation, or the Pacific Decadal Oscillation. These multi-decadal cycles apparently last for about 60 years, with wide repercussions for world temperature and precipitation. There are also secular variations (at the scale of centuries) such as the so-called Little Ice Age from the late Middle Ages to the mid-1800s, as well as the gradual warming after 1850 (long before human GHG emissions were noticeable, but compounded by GHG in more recent times) and the warmer climate experienced during the previous Medieval Warm Period. Any further climate changes induced by human activity (such as those caused by emission of greenhouse gases) are superimposed over such natural processes. In addition, there is a wide range of short-term natural variability around these long-term trends and oscillations. The total effect (i.e., an observable change in the Earth’s climate) is the outcome of all these sources of variability, averaged over a period of suitable length (in the order of several decades) compared to similar periods in the past. To isolate human factors from natural variability, natural cycles of various lengths should be taken into account. The greenhouse effect. The main factors affecting the Earth’s climate are the sun and the atmosphere. The Earth’s surface temperature is kept relatively warm (with a current mean temperature of about 15°C) due to the so-called greenhouse effect of gases in the planet’s atmosphere. As solar radiation hits the planet, greenhouse gases (GHG) capture part of the infrared back radiation, keeping it from escaping to outer space. In the absence of an atmosphere and its greenhouse effect, the Earth would be much colder, with a probable mean temperature of about -18°C, as on the Moon. The main greenhouse gas is water vapour, representing about 98% of the mass of greenhouse gases; atmospheric water vapour is part of the general water cycle involving evaporation from oceans and other water bodies, plant evapotranspiration, and precipitation. 387

How to see the future: Projection methods

Other greenhouse gases besides water vapour include chiefly carbon dioxide (CO2) and minor amounts of other gases such as methane. Some other atmospheric components such as aerosols (solid particles in the air) as well as clouds may reflect incoming sunlight back to space, thus keeping it from heating the surface of the planet. The same reflecting effect, or albedo, is also caused by bright surfaces such as areas covered by ice or snow. In the absence of interfering factors, the planet’s mean energy balance should be zero: as much energy is received as is radiated back to space. This balance may be altered by factors such as fluctuations in solar activity, occasional volcanic eruptions (filling the atmosphere with sun-blocking aerosols), or human GHG emissions. The latter cause an increase in the atmospheric concentration of CO2 and other greenhouse gases, tending to prevent some heat from escaping back to space. Atmospheric CO2 concentration was about 270-280 parts per million (ppm) in preindustrial times, but has grown to nearly 400 ppm and continues to rise – most notably since 1950. Reference to GHG often concerns CO2 only, or includes all GHG in terms of their CO2 equivalent. The direct effect of increasing CO2 concentration is relatively mild: doubling the atmospheric concentration of this gas is estimated to cause a direct increase in mean temperature of about 1°C. But this directly-induced effect of CO2 triggers a number of feedback effects. Some of them are positive feedbacks that exacerbate the effect of emissions, while other feedbacks are negative, i.e., they attenuate the direct effect of GHG. For instance, the increase in temperature directly caused by elevated CO2 and other GHG will cause an increase in water evaporation, determining increased atmospheric water vapour, another greenhouse gas, and thus exacerbating the greenhouse effect of CO2 (a positive feedback); but some of the increased evaporation will also increase cloud cover, which will reflect more solar energy back to space, a negative feedback. The net effect of doubling CO2 on mean world temperature, once all feedbacks are allowed for, i.e., the equilibrium climate sensitivity to changes in GHG (mainly CO2) concentration, is uncertain; the IPCC (2007a) estimated it to be somewhere between 2°C and 4.5°C, with the likely value around 3°C, but a number of more recent estimates point to somewhat lower figures. The recent IPCC report (2013) gives a range of 1.5°C to 4.5°C and abstains from providing a ‘likely’ value within that range. Besides, this estimation refers to the equilibrium long term effect on global temperature, which may take many years to occur after a given increase in CO2, and may never materialise as an equilibrium condition if CO2 concentrations keep increasing or are reduced in the meantime. The transient sensitivity (to 388

Hector Maletta

materialise within decades) is smaller than the equilibrium one. It should also be noted that there are diminishing temperature returns to any given increase in CO2 concentrations. For example, assuming for the sake of argument a sensitivity of 3°C, and starting at the preindustrial concentration (275 ppm), the first doubling of that CO2 (i.e., adding 275 ppm to attain 550 ppm) may cause an increase of 3°C; the next 3°C will require another doubling, to 1100 ppm, i.e., a much larger increase, 550 ppm this time, instead of 275 ppm. Thus, each additional degree of temperature will require a larger absolute increase of CO2 concentration and, of course, a larger amount of GHG emissions. IPCC climate change projections. The Intergovernmental Panel on Climate Change (IPCC) has produced a series of reports on the prospects of climate change over the 21st century. The latest is the fifth report, known as AR5 (IPCC 2013; IPCC 2014a-c). The fourth report, or AR4, was published in 2007 (IPCC 2007a-d). Its climate projections were based on scenarios described in the Special Report on Emission Scenarios or SRES (IPCC 2000) and they are the latest to have been widely used as a basis for agricultural projections. The number of SRES scenarios is, essentially, four. Their development included two methodological phases: first a ‘storyline’ describing key features of the future world imagined by each scenario in qualitative terms; second, a quantification of the scenario storylines by inserting adequate parameters into specific climate models. The storyline includes a socioeconomic scenario (population and income growth, degree of urbanisation, sources of energy, etc.) and its repercussions in terms of land use, energy use, and GHG emissions, which are entered as inputs into a climate simulation model. A given scenario may be quantified by different climate models. The two key dimensions distinguishing the four SRES storylines are the different degrees of societal concern for the environment (in comparison to concern for economic growth and other considerations) and the degree of world socioeconomic integration and convergence. In the dimension of concern for the environment, two possibilities were distinguished, known as A and B. Variant A scenarios reflect a future centred on economic development with little concern for environmental protection; variant B scenarios represent a future with more concern for the environment, even at the expense of growth. In the second dimension, world integration, choices are termed 1 and 2. Variant 1 represents a more integrated world economy, with stronger growth everywhere and convergence of countries towards more equal levels of income; Variant 2 is a more fragmented world with more closed economies, less growth, and persistently unequal levels of 389

How to see the future: Projection methods

development. The combination of these two dimensions gives rise to four ‘possible futures’ that are reflected in the four main SRES scenarios (Table 69). Table 69. Main dimensions of SRES scenario families. Priority A. Growth B. Environment

Degree of convergence and integration 1. Integrated world 2. Fragmented world A1 A2 B1 B2

The A1 scenario assumes a world of rapid economic growth, with the world population peaking in mid-century and declining from that point on; rapid introduction of new, more efficient technologies; and the income of poorer and richer countries converging. A2 describes a very heterogeneous world with high population growth, slower economic development and less technological change. Economic development is mostly regionally oriented; international trade is less important; and economic growth and technological change are more fragmented and slower than in other storylines. B1 describes a convergent world, with the same population growth as A1 but more rapid changes in economic structures toward a service and information economy, with reductions in material intensity (i.e., less use of materials and energy per unit of GDP), and wider diffusion of clean and resourceefficient technologies. The emphasis is on global solutions to economic, social, and environmental sustainability, including improved equity, but without additional climate initiatives. B2 describes a world with intermediate population and economic growth, emphasising local solutions to economic, social, and environmental sustainability, and less rapid and more diverse technological change than in the B1 and A1 scenarios. B2, like B1, is oriented toward environmental protection and social equity, but focuses on varied local and regional (rather than global) trends (IPCC 2000, §4.2-4.3). These socioeconomic scenarios generate different levels of GHG emissions. The climate models used to quantify the scenarios take these emissions as input, and include estimates of the direct effect of GHG emissions on climate, plus the net effect of multiple (positive or negative) feedbacks. Some model quantifications have been designated by the IPCC as ‘marker’ or standard variants, and are widely used. In the case of A1, several sub-scenarios were created that differ in their mix of energy sources; one of them (A1B) is the standard one; another variant (A1F1) envisages an intensification of GHG emissions, and therefore more warming. Climate models divide the Earth’s surface into a large number of cells (the size of which is typically 0.5×0.5 390

Hector Maletta

degrees of latitude and longitude), and project the climate of each cell year by year, up to a target period. Global temperature change is projected from the average of 1980-1999 to the average of 2090-2099, with the results summarised in Table 70 (ordered by magnitude of expected warming). Table 70. Global mean temperature increase across SRES scenarios, from 19801999 to 2090-2099, according to the Fourth Assessment of the IPCC (2007). Temperature increase (degrees Celsius) Scenario

Best estimate

Likely range

B1

1.8°

1.1° to 2.9°

B2

2.4°

1.4° to 3.8°

A1

2.8°

1.7° to 4.4°

A2

3.4°

2.0° to 5.4°

Source: IPCC (2007a:45). A1 quantified in the A1B version. Each scenario has been quantified with a number of climate models; figures reflect the median and range of the ensemble of models, as reported in IPCC (2007a).

Projected warming is not expected to be uniform across the planet. Arctic regions will warm the most, middle latitudes and the Tropics by intermediate amounts, while Antarctic temperatures will remain practically stable or increase by a small amount. There will be more warming in winter than in summer, more at night than during the day, and more at the daily minimum than the maximum. Local conditions may also vary according to a number of other factors, especially sea currents and convection of warm and cold air. World rainfall will increase: a warmer planet will produce more water evaporation, which in turn will cause an increase in cloud coverage and rainfall. However, changes in rainfall will be uneven across the globe: some areas are expected to experience reduced rainfall while others become rainier, in some or all seasons of the year. RCP climate change scenarios. In the fifth IPCC report (AR5) published in 2013-2014, the SRES scenarios are abandoned and replaced by a set of Representative Concentration Pathways (RCP), which are described in IPCC (2008) and van Vuuren et al. (2011). RCPs mainly consist of a path of change in atmospheric concentrations of GHG, still based on certain storylines as a starting point, involving certain levels of population or income, but accepting that the same concentration trajectory might be produced by other possible combinations of changes in population, income, energy sources, and other factors. The preparation of these RCPs involves examination of many existing scenarios with various assumptions about technology, population, income, and other factors but in the end, the RCPs are simply what their name indicates: a path for future change in GHG concentration: 391

How to see the future: Projection methods

RCPs are referred to as concentration pathways in order to emphasize that while they are based on existing scenarios in the literature that have underlying socioeconomic assumptions and emissions outcomes, they are being selected on the basis of their emissions pathways and associated concentrations of radiatively active gases and aerosols, and their primary purpose is to provide these concentration pathways to the CM community to produce new climate change projections (IPCC [2008]:12).

This approach defuses (or dilutes) some of the criticisms levelled against some of the SRES scenarios based on extremely unlikely storylines and assumptions (e.g., the rapid population growth of the SRES A2 scenario seems incompatible with its assumed income growth or with observed population trends). However, projections of agricultural production based on the AR5 RCPs are still scarce, and we therefore do not dwell much on them here.46

13.4.2. Climate change and agriculture Climate and vegetation. Plants depend on the environment: soil, sunlight, temperature, and rainfall. They grow by capturing carbon dioxide from the atmosphere through photosynthesis, absorbing water, nitrogen, and various minerals from the soil, and turning all these materials into vegetal tissues (carbohydrates, fats, and protein). As a special case, leguminous plants can also capture atmospheric nitrogen with the help of some symbiotic bacteria, while other plants need to find nitrogen (an essential component of proteins) in the soil. Any plant variety or cultivar that successfully grows in a certain location is adapted to a certain range of variation in the environmental conditions prevailing there. There is a range of temperatures and rainfall that is compatible with near-optimum growth of the plant in question, but as temperatures and rainfall move beyond that range the plant may be stressed in one way or another, and its growth will probably diminish. If the plant was subject to sub-optimal conditions to begin with, certain climate changes might actually improve its growth, while for those plants within their optimum range, any significant alteration of the temperature or precipitation regimes will negatively affect their growth and yield; this is because the fit of the plant to the prevailing climate indicates that it has been naturally or artificially selected One early study based on an AR5 RCP is Nelson et al. (2014), using the IMPACT model under the most extreme of the AR5 pathways (RCP8.5). We briefly review (and find wanting) this and other related models in Section 13.5.2 below.

46

392

Hector Maletta

for the normal range of environmental conditions prevailing in that location. In some extreme cases (e.g., total desertification) vegetation may cease to be viable in certain locations. Of course, there will normally be other plants, or other varieties or cultivars of the same plant species, better suited to the new climate in the given location. Gradually unfolding climate change may induce gradual but tentative replacement of some plants, varieties or cultivars by others, or more generally a change in the plant mix at each location. When some conditions deviate from the optimal range, a given plant variety or cultivar will normally reduce its normal growth or yield. At even greater deviations, that particular variety or cultivar will simply not grow, failing to germinate, flower, and produce fruits or grains, due to temperatures that are too hot or too cold at a certain stage of the plant cycle, or due to too much or too little humidity for that particular cultivar of a certain variety of a given crop species. The new conditions, however, may be adequate for some other known cultivar of the same variety, for some other variety of the same crop, or for some other crop. Since changes in climate are not instantaneous, new cultivars or varieties gradually may be tried and selectively adapted (by farmers or experimental agronomists) to the evolving climate conditions of each particular place. Most changes in the statistical distribution of crops, varieties and cultivars in a given area will tend to be adaptive, in line with agricultural practices. Non-adaptive changes, as is normal in a trial-anderror process, will be short-lived as those responsible will promptly abandon them, others will choose not to imitate them, or, in some cases, the resulting varieties might survive as minority variations with a diminishing share of total output. The above is valid for all climate changes. In the case of climate change associated with GHG emissions, the biophysical impact on plants operates through two channels: (a) by changing the mean and distribution of temperature and rainfall, and (b) by changing the atmospheric concentration of CO2 (which plants capture through photosynthesis). Both sets of factors will affect the plants. This is valid for all plants (i.e., wild vegetation as well as crops). In the case of agriculture, i.e., plants grown (and animals raised) by people, there is another factor to take into account: (c) human farming activity, without which agriculture would not exist, and which is essentially adaptive, especially for slow changes like those affecting the prevailing climate, atmospheric CO2, and available farming technology. Human activity in this regard includes farming practices and technology, as well as the possible use of crops to make biofuels with the purpose of reducing the demand for fossil fuels (the chief culprits of GHG emissions). 393

How to see the future: Projection methods

Vegetation models and crop models. A model, in this case a mathematical construction representing the relationships of climate, soil, and plants, can be used in order to simulate the impact of climate change on plants. The two main types are vegetation models (for wild vegetation) and crop models (for agriculture). A model of wild vegetation incorporates factors such as soil, climate, plant species and varieties, and their interactions in a given ecosystem (where they may also interact with local animals). For most applications, wild vegetation models do not consider human activity as a key factor, although some such models may include certain human effects such as deforestation or foraging. Curiously enough, ecosystem models usually include the interaction of plants and animals but fail to consider the interaction (of plants and animals) with humans, in spite of the fact that Homo sapiens is a numerous and widely distributed species that is present in most habitats around the globe and interacts with most ecosystems in a very powerful way. But if overlooking humans when considering wildlife may be forgivable, disregarding human involvement in agriculture is not. A model of agricultural production, e.g., a crop model, obviously must incorporate the same basic factors (soil, climate, and plant species and varieties) as well as human farming activities. Unlike the case of a natural ecosystem, agricultural production requires people (farmers), who must choose particular crops to grow and perform a series of farming practices (ploughing, planting seeds at a certain depth, with certain spacing, and on a certain date, applying or not applying various amounts of fertiliser, managing water, weeding, protecting plants from pests and diseases, and harvesting). All these activities may take various forms depending on multiple factors: the objective conditions of the crop; available technology; the extent of the farmer’s technological awareness; the availability of the necessary equipment and infrastructure; the variable ability and willingness of farmers to learn how to perform the activities involved, and to perform them with greater or lesser efficiency and effectiveness; the process of change in farming systems, in farm sizes and organisation, and in prevailing production techniques; and the process of gradual (demographic and economic) replacement of the people involved in managing farm production as the climate undergoes long term change extending over several generations. This distinction between wild vegetation and agriculture when assessing the impacts of climate change is extremely important for our present purposes. For crops, the biophysical effects of climate on plants are not sufficient to assess the ultimate impact of climate change on agriculture.

394

Hector Maletta

Nevertheless, evaluations of the impact of climate change that centre only on biophysical effects on plants, treating the possibility of adaptation as a separate issue instead of looking at it as an integral part of the agricultural production system are still quite frequent, especially because the biophysical effect is presented as a necessary outcome while adaptation is regarded as a mere option or possibility. For instance, using studies based on existing crop models calibrated to specific locations, the IPCC AR5 report (IPCC 2014a :498) presents the results of many simulated yields of major cereals when several theoretical increases in temperature (up to 5°C) are introduced into these crop models. The increase in temperature is normally introduced with a ceteris paribus clause whereby it is the only factor that changes, leaving location, soil, and crop constant, with or without adaptations in choice of variety and other farming practices. These simulations suggest that ‘without any adaptation’ (which is a flawed approach to modelling agriculture, a human adaptive activity) the yields of, say, wheat in temperate regions (each study assuming a certain variety of wheat grown at specific locations under certain technology) will vary very little, with temperatures simulated to increase by up to 2°C, and will decrease by about 10% if temperature is increased by more than 2°C and up to +5°C. It is worth noting that when incremental adaptations by farmers are considered, the projected decrease in wheat yields turns into an increase of roughly the same magnitude, i.e., 10%, even at temperatures 5°C higher. Since these are simulations conducted using mathematical crop models, where only temperature is allowed to vary and (almost) everything else is kept constant, they tell us next to nothing about global production of wheat at a future point in time (50 or 100 years ahead) as the world gradually reaches the higher long-term average temperature projected. There will be local adaptations by those future wheat farmers to grow the same crop at each location (e.g., using a different variety, adopting new measures to combat pests and diseases, or providing the crop with more water or fertiliser); however, anticipating future output involves many other considerations. In the course of the time that will elapse until such increases in temperature come about, there will doubtless be many other developments: new seeds and varieties will be developed and marketed for the same and other crops; new farming techniques will be invented and/or become more widespread; the geographic distribution of the crop will change, possibly towards higher latitudes; and the relative importance of different wheat varieties, and of wheat itself relative to other crops, will also change. Most importantly, three or more generations will elapse, with a total replacement of today’s farmers by their grandchildren 395

How to see the future: Projection methods

or great-grandchildren, and, in all likelihood, by complete strangers, whether individual persons or agribusiness corporations, that happen to acquire the land. Even if wheat yields in all locations decrease under conditions of climate change as compared with no climate change (which is unlikely; they are expected to rise at higher latitudes, and to rise everywhere through modest incremental adaptations), future global production of wheat, or agricultural output in general, may still be much higher in the late 21st century than is the case at present, based on the dictates of demand and the possibilities offered by technology and human ingenuity. The question of whether or not this will actually materialise cannot be answered by simulations of yields based on crop models alone, without simultaneous consideration of other parallel changes in population, income, technology, crop mix, crop geographical distribution, changes in land tenure and farm management, and other factors. Demand for wheat, for instance, will indeed change: as seen in Section 3.6, per capita cereal demand has been stagnant for decades at the world level, and decreases with the growth of income; it is still on the rise in poorer countries, stable or falling in developed ones. It is to be expected that as more countries attain higher income levels, per capita demand for cereals (wheat included) will tend to stabilise and eventually decrease, as has already occurred in Europe and North America. This decreasing per capita demand will combine with slower population growth to reduce the pace of growth in total demand for wheat or other cereals, relative to the overall growth of demand for agricultural products. In turn, decelerating demand will cause a slowdown in production, while agricultural science will be making efforts to expand the yield potential of each crop (such as wheat) by developing new varieties with higher yields that are adapted to future climatic conditions (e.g., more heat-resistant or more resilient to pests and disease). The new possibilities offered by agricultural technology, e.g., new seeds or new farming practices, will gradually spread, as has occurred in the past, as available knowledge and market conditions allow. The mere expectation of climate change causing lower yield of current varieties in current locations is not enough to foresee the future of that crop, or other crops generally.

13.4.3. Effects of atmospheric carbon dioxide effect on plants Atmospheric CO2 concentration affects crop physiology and is thus an important driver of the impact of climate change on agriculture, alongside temperature and precipitation. It has two major manifestations: increased photosynthesis, and reduced water needs,47 and affects both crops and wild Allen et al. (1996) provide a general description of these factors and how they work.

47

396

Hector Maletta

vegetation. Regrettably, however, these effects of CO2 are often ignored in assessments of the impact of climate change on agriculture. Some commonly applied methods to assess the probable impacts of climate change on agriculture (e.g., agronomic crop models, as well as Ricardian models that compare farm revenues between locations with different climates) are synchronic or cross-sectional, with data under current CO2 concentration, and thus fail to take this factor into account, except in specific experiments in which increased carbon concentrations are exogenously introduced. Some more complex models that could include the CO2 factor often fail to do so, or afford it only cursory attention. This approach is lacking in rigour, since the effect is recognised to exist and to be positive for plant production. Even if different methods yield different quantitative results, all find a positive effect for increased photosynthesis and/or a reduced water requirement of plants under elevated atmospheric CO2 concentrations. Impact of carbon concentration on crops. Experimental results show that increasing atmospheric CO2 concentrations from current levels (375-390 ppm at the time of most experiments on record) to about 550 ppm (i.e., twice the pre-industrial concentration of about 275 ppm) could increase crop yields by 10% to 30% and could also reduce the water requirements of C4 plants to varying degrees. Since most assessments of potential negative impacts of climate change on crops are of the same order of magnitude, including or excluding this factor will greatly affect impact assessments. The inclusion is not optional, since the effect of increased atmospheric carbon is a necessary aspect of the impact of anthropogenic climate change on crops, is well grounded in plant science, and has been proven by a large and wide-ranging number of studies (observational, experimental, and based on model simulations). Before examining these studies, it is important to note a common cause of ambiguity or imprecision related to the baseline CO2 concentration. To evaluate impacts on crop yields of increases in CO2 concentrations, the common measures are yield response ratio, RR=Y(C)/Y(C0) and enhancement factor EF=RR-1, where C is the expected (increased) concentration and C0 is the baseline or reference concentration. The choice of baseline concentration C0 is not irrelevant: different studies use different baselines, making the results non-comparable. The authors of a paper summarising various such studies note that: Response ratios or enhancement factors discussed herein were re-scaled to a reference CO2 concentration C0=350 ppm, unless noted otherwise. This is necessary for proper data comparisons across many studies. To this end, we note that because atmospheric CO2 concentration has risen in the last several decades at about 0.5% per annum […] from 330 ppm in the mid-1970s,

397

How to see the future: Projection methods

[…] to over 380 ppm today […] comparing yield response ratios of similar experiments made in different periods requires re-scaling of the apparent, smaller responses obtained using higher reference CO2 concentrations (Tubiello et al. 2007:216).

However, such a normalisation to a single baseline concentration does not feature widely in reports of relevant experiments. Considering a concentration of 550 ppm, i.e., about twice the preindustrial level (2×CO2), the 2007 IPCC report describes the results as follows: On average across several species and under unstressed conditions, recent data analyses find that, compared to current atmospheric CO2 concentrations, crop yields increase at 550 ppm CO2 in the range of 10-20% for C3 crops and 0-10% for C4 crops (Ainsworth et al. 2004; Gifford, 2004; Long et al., 2004). Increases in above-ground biomass at 550 ppm CO2 for trees are in the range 0-30%, with the higher values observed in young trees and little to no response observed in mature natural forests (Nowak et al. 2004; Korner et al. 2005; Norby et al. 2005). Observed increase of above-ground production in C3 pastures is about +10% (Nowak et al. 2004; Ainsworth and Long, 2005). For commercial forestry, slow growing trees may respond little to elevated CO2 (e.g., Vanhatalo et al. 2003), and fast-growing trees more strongly, with harvestable wood increases of +15-25% at 550 ppm and high N (Calfapietra et al. 2003; Liberloo et al. 2005; Wittig et al. 2005). Norby et al. (2005) found a mean tree net primary production (NPP) response of 23% in young tree stands; however in mature tree stands Korner et al. (2005) reported no stimulation.48 (IPCC 2007b:282; our italics)

It should be noted that nowhere in this paragraph is the baseline clearly indicated. Fieldwork for the cited studies was carried out during different periods in the 1990s and early 2000s. Since concentrations have increased from about 350 ppm in 1990 to about 380 ppm in 2005, the actual baseline of different studies may vary. As a result, the (absolute or percentage) increase in CO2 corresponding to each of the stated increases in yields is not selfevident. However, even allowing for the ambiguity related to the choice of baseline, the above citation from the IPCC Fourth Assessment Report of 2007 summarises the scientific evidence then available regarding the expected effects of carbon dioxide concentrations on the yields and water needs of plants. Since that time new evidence has come to light, including not only Complete references for the works cited in this paragraph can be found in the IPCC AR4 report, available online. 48

398

Hector Maletta

the effect on crops but also on wild vegetation. Moreover, it is well known that farmers using greenhouses habitually provide a carbon-rich atmosphere to increase the yields of their crops. In addition, there have been developments in relation to the methodology for measuring the effect on crops of CO2 fertilisation, evolving from earlier Closed Chamber (CC) methods to more recent Free Air CO2 Enrichment (FACE) experiments (Long et al. [2006] and Tubiello et al. [2007]. However, neither has proven to be perfect. CC methods involve growing plants within plastic cubicles with an artificial atmosphere at higher CO2 concentrations, but these artificial chambers modify other factors too (exposure to pests and diseases, wind, sunlight, and others). On the other hand, FACE experiments create an elevated CO2 concentration in an open environment by means of an array of CO2-emitting rods installed in the crop field; this strategy avoids the problems related to closed chambers, but has its own difficulties: the achieved concentrations at plant level may vary according to plant-to-rod distance, time of day, wind, and other factors. In both cases, extraneous influences must be factored out before reaching any conclusions. According to analyses of differences between the two methods in Tubiello et al. (2007), such influences can be controlled for; the unexplained differences between the results of the two methods are not significant if appropriate adjustments are made. However, not all studies are careful about reporting methods and standards, or about making adjustments when comparing the results of different studies. One important contribution to the debate on this matter is the estimation by Long et al. (2006) that FACE experiments show only half the effect of CO2 as compared with previous closed-chamber experiments. This half-effect is still considerable, but has frequently been used as an excuse to minimise or dismiss the CO2 effect (for instance in Nelson et al. [2010]). Tubiello et al. (2007), in a response to Long et al. (2006), dispute such allegations by means of an extensive meta-analysis of existing studies: Long et al. (2006) stated that results from free-air CO2 enrichment (FACE) were 50% lower than found in other studies. By contrast, previous metaanalyses of both FACE and enclosure studies by several authors, including these, had shown that FACE data may be consistent with observations in non-FACE experiments, such as glasshouses, closed and open-top field chambers, and laboratory studies: at 550 ppm enrichment, mean yields increased 17–20% in FACE, compared to 19–23% in non-FACE experiments. (Tubiello et al. 2007:217)

399

How to see the future: Projection methods

These authors emphasize, first, the importance of the baseline CO2 concentration in measurements of percentage impact, and specifically identify instances of different baselines in figures reported by Long et al. (2006), as well as other methodological differences (including some that artificially increased yield gains under older non-FACE experiments). Tubiello et al. state that once specific experimental differences are taken into account, FACE and non-FACE experimental results are consistent with each other. At any rate, FACE experiments are superior to older methods, and research in this field is on-going. Tubiello and Ewert (2002) provide a useful review of the literature and of the problems to solve. Ainsworth et al. (2008) have proposed a ‘new generation’ of FACE studies to test the impact of higher CO2 concentrations in a more systematic and reliable way. However, these new-generation studies have not yet materialised or have yet to be comprehensively reviewed. One possible approach is to use FACE experiments with the same crop grown in different climatic zones; another would be to grow crops in greenhouses at different temperatures, humidity levels, and carbon dioxide concentrations (though doing so in greenhouses may affect the results by protecting plants from winds and other factors present in open-air cultivation). Combining greenhouses and open field FACE experiments in different locations may enable a more profound understanding of the various factors at play and of their mutual interactions. All analyses to date show a significant impact of carbon fertilisation on crops. In the case of C4 crops (such as maize) the main effect is not so much a higher yield (except in drought conditions) but a reduced demand for water, which also occurs in C3 crops, albeit to a lesser extent. This is because larger amounts of available atmospheric CO2 allow for faster absorption of carbon and lower water loss at the plant’s stomata. This will have important implications for areas where climate change will create drier conditions, as is the case for maizeproducing areas in tropical or subtropical latitudes. The capacity to economise on water will translate into increased projections of maize production (if CO2 fertilisation is factored in) in those areas estimated to become more arid (Leakey 2009). However, the yield-equivalent of this lower demand for water has not been widely studied or reported; on the one hand, it is not always clear which areas with C4 crops will become drier or wetter; on the other, studies regarding the impacts of climate change on agricultural demand for water (such as Fischer et al. 2007) have not considered this factor in detail; although integrated assessment models used by Fischer and colleagues do take CO2 fertilisation into account, they do not fully account for the indirect effect of water-saving on area planted and yields. 400

Hector Maletta

All told, the effects of carbon fertilisation on food production and food security may be important. Ainsworth et al. (2008), who strongly advocate for more research on crop response to increased atmospheric carbon concentrations, explain: Estimates of the potential benefit of elevated [CO2] to global food supply suggest it will reduce the number of malnourished people in 2080 by between 12 and 580 million individuals, depending on the socio-economic scenario and the crop models considered (Parry et al. 2004; Schmidhuber and Tubiello 2007).

A large share of the envisaged adverse effects of climate change on agriculture may be offset by CO2 fertilisation, because estimates of the potential (negative) impact of climate change on crops are comparable in magnitude to the estimated beneficial impact of carbon fertilisation. The state of knowledge in this regard may be gauged from recent meta-analyses of the literature such as Vanuytrecht et al. (2012) who examined 53 papers with 529 independent observations, and found that ‘Overall, crops benefit from elevated [CO2] by improving water productivity (+23% for biomass production and +27% for yield production), which is achieved through production increases in biomass (+15% for aboveground biomass) and yield (+16%), in combination with a decrease in seasonal evapotranspiration (–5%)’. These results are for both C3 and C4 crops, and little difference was found between the two types of crop, but C4 crops were poorly represented in the set of studies included in the meta-analysis, which may have affected the results. The authors say: ‘Future FACE experiments on C4 species are desirable to draw robust conclusions about responses of staple crops like maize and sorghum to elevated [CO2].’ The IPCC AR4 concludes: Recent re-analyses of FACE studies indicate that, at 550 ppm atmospheric CO2 concentrations, yields increase under unstressed conditions by 10-25% for C3 crops, and by 0-10% for C4 crops (medium confidence), consistent with previous TAR estimates (medium confidence). Crop model simulations under elevated CO2 are consistent with these ranges (high confidence).49 (IPCC 2007b:276)

These increases in percentage yield, as discussed before, do not correspond to CO2 doubling from preindustrial levels, but to an increase of about 45% from levels current at the time of experiments (~380 ppm) to 550 ppm. The precise baseline is not reported.

49

401

How to see the future: Projection methods

This indicates a rather high positive impact of CO2 fertilisation (midpoint (+17.5%) for very important crops like wheat and soybeans, and more moderate impacts (midpoint +5%) for other major crops such as maize (not counting water savings), to mention but three of the most important food crops in the world. The size of these effects is large enough to offset (or substantially attenuate) any foreseen reduction in yields that ignores CO2 effects. On top of this, it should be stressed that the main issue as regards C4 crops is not their more modest increase in yields under climate change, but the continued viability of C4 crops under increased temperatures and drier conditions, provided that CO2 effects on water needs are considered. Even if yields may ultimately remain stable, grow very little, or decrease, any such outcome will be much more favourable than the severe decrease in yields resulting from warmer and drier conditions that will occur if the reduction in water needs due to elevated CO2 concentrations is not accounted for. Carbon concentration and general vegetation. The effect of CO2 on plant growth has also been studied using other approaches. Besides close-chamber and FACE experiments linking CO2 with crop yields and water needs, other studies have focused on the more general impact of atmospheric carbon dioxide on vegetation (including both crops and wildlife). This general approach has used modelling, observational, and experimental methods. As an example of the modelling approach, large scale simulations with multiple climate models, which entail a doubling of the concentration of CO2, appear to have a large positive impact on vegetation (in terms of Net Primary Production or NPP), as found by Hemming et al. (2013). These authors thus describe their work and conclusions: Net primary productivity (NPP) is often modelled explicitly in general circulation models (GCMs) utilising process models that may include plant photosynthesis, respiration, allocation of photosynthates, phenology, mortality and competition between plant functional types. It is an important measure for understanding the role of terrestrial vegetation in the global carbon cycle, and useful for gaining insights into the large-scale, integrated effects of climate and atmospheric changes on potential plant productivity and associated impacts, i.e. food security and carbon cycle feedbacks. However, there are simplifications and uncertainties in GCM projections of future climate change, as well as further uncertainties involved in modelling the associated terrestrial vegetation responses. In particular, it is important to highlight that many GCM simulations, including the ones used in this study, 402

Hector Maletta

do not model nutrient limitation, even though primary plant nutrients, e.g. nitrogen and phosphorus, are key limiting factors on plant productivity. Here, we examine sensitivities and uncertainties in large(global)-scale modelled NPP to climate and atmospheric carbon dioxide concentration [CO2], utilising a relatively large perturbed physics ensemble (PPE) of simulations generated from the HadSM3 GCM under equilibrium doubling of pre-industrial atmospheric [CO2]. We also exploit the ensemble design to highlight the relative importance of two, often opposing, forcings on NPP: (i) plant physiological responses to CO2, termed ‘Phys’; and (ii) plant responses to physical drivers of climate, termed ‘Rad’. It is important to note that this is a sensitivity study that provides useful guidance on the relative importance of the Rad and Phys drivers and their uncertainties. The results cannot be considered quantitatively realistic, particularly because the equilibrium experimental design and lack of nutrient limitation in the model are important limitations that prevent such interpretation. We find that doubled [CO2] and associated climate changes ultimately increase potential global average NPP by 57%, from 0.293 kg cm-2 yr-1 (∼36 PgC yr-1) to 0.460 kg cm-2 yr-1 (∼57 PgC yr-1). Spatially, the largest decreases (∼−0.45 kg cm-2 yr-1) occur across the north-east of South America in association with the largest decreases in precipitation. The largest increases (up to ∼0.75 kg cm-2 yr-1) occur across tropical Africa and Indonesia, where NPP is already high, and both temperature and precipitation increase under doubled [CO2]. In most regions where NPP shows an increase the changes are significantly larger than the ensemble standard deviation, indicating that increases in global NPP under doubled [CO2] are reasonably robust. However, in some regions, particularly north-eastern South America and Central America, where NPP decreases are projected, the standard deviation across the ensemble is larger than the average NPP change, indicating that even the sign of the NPP sensitivity to doubled [CO2] and climate is uncertain. These uncertainties are shown to be highly dependent on the relative sensitivities of NPP to the Phys and Rad forcings. (Hemming et al. 2013).

Some authors, like Hemming et al. in the first paragraph of the above passage, have warned that the favourable effect of CO2 on yields may be constrained by low availability of nutrients (like nitrogen) in soils (except in the case of legumes, which take their nitrogen from the air). However, the possible limiting influence of nitrogen availability in the soil may be overrated, since the same process of climate change causes an increased availability of mineral nitrogen in soils. In an independent study, Melillo et al. (2011) artificially 403

How to see the future: Projection methods

warmed a section of forest in Massachusetts for seven years, and found that warming resulted in increased availability of soil nitrogen, thus stimulating plant growth in comparison with control areas. This effect was independent of the effect of higher atmospheric CO2 concentrations, which during the experiment increased slightly but equally in both the experimental and control sections of the forest. Soil warming accelerated the decay of organic matter in soils, and also accelerated the mineralization of the nitrogen released by this decay. Even if the short-term result of soil warming was initially (in Melillo’s experiment) a net carbon loss from the forest, this net loss diminished over time due to the parallel and increasing effect of warming on nitrogen availability in interaction with the carbon fluxes; the net loss had practically vanished by the seventh year of the experiment: Soil warming has resulted in carbon losses from the soil and stimulated carbon gains in the woody tissue of trees. The warming-enhanced decay of soil organic matter also released enough additional inorganic nitrogen into the soil solution to support the observed increases in plant carbon storage. Although soil warming has resulted in a cumulative net loss of carbon from a New England forest relative to a control area over the 7-y study, the annual net losses generally decreased over time as plant carbon storage increased. In the seventh year, warming-induced soil carbon losses were almost totally compensated for by plant carbon gains in response to warming. We attribute the plant gains primarily to warming induced increases in nitrogen availability. This study underscores the importance of incorporating carbon–nitrogen interactions in atmosphere–ocean–land earth system models to accurately simulate land feedbacks to the climate system. (Melillo et al. 2011)

The Melillo study only covers seven years, but the process of carbon gains and nitrogen releasing is presumed to continue beyond the end point under a persistently warmer environment, thus converting the initial net losses into net gains. On top of the process described by Melillo and colleagues in relation to artificially induced warming, the increased availability of atmospheric carbon will generate extra photosynthesis and more efficient use of water, thus improving on the outcomes. Moreover, the Melillo study tends to reduce the importance of the nitrogen constraint for the materialisation of the Hemming model predictions.50

Although the Melillo study did not address the issue, it is possible that warming also favors mineralization of other chemical elements required by plants, such as phosphorus or potassium. More research is needed in this area.

50

404

Hector Maletta

The beneficial effect of increased carbon dioxide on the water needs of vegetation has also been found empirically by Donohue et al. (2013) in an observational study based on satellite data for warm arid zones around the globe. It examines recent changes in the so-called maximum cover edge. This concept is based on a planetary land-surface grid with 0.08° cells from which major irrigation areas, lakes, and wetlands were excluded. For each cell, fractional foliage cover and annual precipitation were measured. The analysis was restricted to the warm, arid grid cells where water was the main constraint for vegetation, including all cells with a continuous rainfall record during the period (1982-2010) for which satellite data on vegetation were available; the selected areas were in all continents, including cells in the Western United States, Mexico, Southern Africa, South and Central Australia, Western Argentina, Northeastern Brazil, Central Asia, and others. All these areas had rainfall regimes below 400 mm/year with fractional foliage cover mostly below 0.5, and they were considered as the ‘edge’ in the joint distribution of foliage and precipitation.51 For theoretical reasons, the authors’ expectation was that past increases in carbon dioxide concentrations in 1982-2010 should cause an increase of between 5% and 10% in fractional foliage cover at the edge. From 1982 to 2010, atmospheric CO2 concentrations increased by 14% and this caused a fractional foliage cover increase of 11.3% at the edge, just above the upper bound of the range derived from theory (10%). There was also a 10% increase in precipitation in the warm, arid zones covered by the study, which in general produced a ‘greening’ effect on vegetation, but this could not have caused the increase in foliage cover at the edge, the authors argue, because the statistical estimation of the increase in foliage kept precipitation constant so the estimated increase in foliage is therefore independent of any possible effect of precipitation. The authors discuss and discard some other possible explanations for the increase in foliage cover at the edge, concluding that it must have been caused by increased CO2 acting on the water use efficiency of photosynthesis, as suggested by prior literature:

The formal definition of the ‘edge’ is as follows. The joint distribution of cells by fractional foliage cover and annual precipitation shows that the median cell has increasing foliage cover as precipitation increases, tending to stabilise at around 85% cover for rainfall above 2000 mm/yr. A band was defined around the median covering the 5th to the 95th percentile of cells. The ‘edge’ is defined as the line corresponding to the 95th percentile, at low precipitation levels (