World Market Price of Oil: Impacting Factors and Forecasting 3030114937, 9783030114930

This book develops new econometric models to analyze and forecast the world market price of oil. The authors construct A

293 69 3MB

English Pages 194 Year 2019

Report DMCA / Copyright


Polecaj historie

World Market Price of Oil: Impacting Factors and Forecasting
 3030114937,  9783030114930

Citation preview


Adalat Muradov Yadulla Hasanli Nazim Hajiyev

World Market Price of Oil Impacting Factors and Forecasting


SpringerBriefs in Economics

SpringerBriefs present concise summaries of cutting-edge research and practical applications across a wide spectrum of fields. Featuring compact volumes of 50 to 125 pages, the series covers a range of content from professional to academic. Typical topics might include: • A timely report of state-of-the art analytical techniques • A bridge between new research results, as published in journal articles, and a contextual literature review • A snapshot of a hot or emerging topic • An in-depth case study or clinical example • A presentation of core concepts that students must understand in order to make independent contributions SpringerBriefs in Economics showcase emerging theory, empirical research, and practical application in microeconomics, macroeconomics, economic policy, public finance, econometrics, regional science, and related fields, from a global author community. Briefs are characterized by fast, global electronic dissemination, standard publishing contracts, standardized manuscript preparation and formatting guidelines, and expedited production schedules. More information about this series at

Adalat Muradov • Yadulla Hasanli Nazim Hajiyev

World Market Price of Oil Impacting Factors and Forecasting

Adalat Muradov Azerbaijan State University of Economics Baku, Azerbaijan

Yadulla Hasanli Azerbaijan State University of Economics Baku, Azerbaijan

Nazim Hajiyev Azerbaijan State University of Economics Baku, Azerbaijan Harvard University Cambridge, MA, USA

ISSN 2191-5504     ISSN 2191-5512 (electronic) SpringerBriefs in Economics ISBN 978-3-030-11493-0    ISBN 978-3-030-11494-7 (eBook) Library of Congress Control Number: 2019932991 © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland


In the book, the dynamics of the world market price of oil has been anaylized in the historical aspect, influencing factors classified and new trends in development of the world economy in the modern world analyzed. Medium and long-term forecasting was carried out by means of econometric models under the theoretical and methodological bases of forecasting oil price. The book can be used for adoption of proper decisions of governmental authorities, including in forecasting of the state budget. Also, it may be useful for economist scholars, doctoral students and higher school students.



In the light of new trends and challenges in the global economy, development of scientifically and practically substantiated forecast indicators of the factors influencing the price of crude oil of great strategic importance in the world markets for the medium and long-term periods, as well as understanding of the existing and expected processes and assessment of their effectiveness, are necessary. In this respect, it is necessary to perform a complex review of reliable forecast outcomes, considering the main economic and noneconomic factors affecting the world market prices of oil, to obtain such forecasts and to develop an adequate forecasting methodology. To develop adequate forecasting methods, firstly, the dynamics of the scope of demand and supply and the prices of oil in the economic–world market; oil reserves and dynamics of exploitation of major oil-offering subjects (Organization of the Petroleum Exporting Countries (OPEC), Russia, the USA, etc.); economic development dynamics of major oil-requesting countries (the USA, China, Japan, Germany, India, etc.); creation of alternative energy sources and their applicability to production; assessment of the impacts of the transmission mechanisms of natural–environmental–solar activity indicators on world oil prices; and study of perspectives on political conflicts, terror, polarization, globalization, and democratization processes are the main requirements. In the first chapter of this volume, the factors influencing oil prices, the impact of solar activity on oil prices, and forecasting by econometric models—as well as theoretical aspects of trend models and forecasting methods for oil prices—are analyzed in a broad spectrum from scientific and practical viewpoints, and analytical summarizations are made. In the second chapter, the details are collected and processed to build econometric models under the theoretical analyses carried out in the first chapter in a logical sequence. It should be noted that the influence of solar activity on oil prices and the trend models of the daily oil price are established and tested. In the third chapter, autoregressive integrated moving average (ARIMA), Holt, and trend models are set up to forecast oil prices in the world markets for the




medium and long-term periods. Moreover, oil prices in the world markets and the price of Azeri Light oil in the world market are forecast for the long-term period. It should be noted that the information collected for forecasting the factors influencing crude oil prices in the world markets for the medium and long-term periods are processed in Microsoft Excel and the econometric models are executed in the EViews 8 applied software package. The scientific and practical forecast indicators obtained from the models set up to forecast the world oil market price and the price of Azeri Light oil in the world market, and the proposals made, may be used in speeding up socioeconomic development, in the processes of preparation of state budgets and production and implementation of state programs, in the actions to be taken in this regard and the respective assessments to be carried out, in preparation of the action plans of the Azerbaijan Oil Fund and the State Oil Company of the Azerbaijan Republic for the medium- and long-term periods, in formation of the country’s macroeconomic policy and related scientific research works.


1 Theoretical–Methodological Principles of the Problem ����������������������    1 1.1 Necessity of the Problem, Study Level, and Methodological Aspects����������������������������������������������������������������������������������������������    1 1.2 Classification and Analysis of Factors Influencing Oil Prices����������    2 1.2.1 Theoretical Aspects of the Influence of Solar Activity on Oil Prices��������������������������������������������������������������������������    5 1.3 Forecasting Through Econometric Models��������������������������������������    8 1.3.1 Short-Term and Long-Term Econometric Models for Oil Price Forecasting������������������������������������������������������    9 1.3.2 Trend Models and Forecasting����������������������������������������������   11 1.3.3 Time Series and Nonstationary Problems����������������������������   13 1.3.4 Checking the Stationarity of Time Series: The Dickey–Fuller Test��������������������������������������������������������   16 1.3.5 ARIMA Models��������������������������������������������������������������������   20 2 Creating Econometric Models: Evaluation and Analysis��������������������   21 2.1 Data Collection and Processing��������������������������������������������������������   21 2.2 Econometric Modeling of the Impact of Solar Activity on Oil Prices��������������������������������������������������������������������������������������   27 2.3 Building and Testing of the Trend Model of Daily Oil Prices����������   36 2.4 Conclusions��������������������������������������������������������������������������������������   39 3 Forecasting of the World Market Price of Oil ��������������������������������������   41 3.1 Building of an Autoregressive Integrated Moving Average Model to Forecast Oil Prices����������������������������������������������   41 3.2 Building of a Trend Model to Forecast Oil Prices����������������������������   47 3.3 Long-Term Forecasting of World Oil Market Prices������������������������   52 3.3.1 Forecasting of World Market Prices of Azeri Light Oil��������������������������������������������������������������������������������   58 3.4 Conclusions��������������������������������������������������������������������������������������   59




Appendix A: Dynamics of a Number of Indicators Affecting Oil Prices������������������������������������������������������������������������������������   61 Appendix B: Prices of Crude WTI Oil����������������������������������������������������������   65 Appendix C: Maximum, Minimum, and Simple Average Oil Prices, 1986–2017����������������������������������������������������������������  127 Appendix D: Daily Oil Price—Statistical Characteristics of the Trend Autoregression Model and Proper Tests������������  129 Appendix E: S  tatistical Characteristics of the Oil Price Trend Model and Tests ������������������������������������������������������������������������  135 Appendix F: Oil Prices in the World Market (Months) ������������������������������  139 Appendix G: S  tatistical Characteristics of ARIMA (1.1.26) Model (3.1) and Tests����������������������������������������������������������������  157 Appendix H: S  tatistical Characteristics of Logarithmic Linear Trend Model������������������������������������������������������������������������������  161 Appendix I: Time Series of Oil Prices (Years)����������������������������������������������  165 Appendix J: S  tatistical Characteristics of the ARIMA Model and Forecast��������������������������������������������������������������������������������  167 Appendix K: S  tatistical Characteristics of the Forecast Model of Average Prices of Brent and WTI Oils—Tests and Forecast������������������������������������������������������������������������������  173 Appendix L: S  tatistical Characteristics of the Forecast Model of Prices of Azeri Light Oil in the World Market—Tests and Forecast����������������������������������������������������  179 References ��������������������������������������������������������������������������������������������������������  183

Scientific Editors

Rawi  Abdelal  Herbert F.  Johnson Professor of International Management, Harvard Business School/Director of the Davis Center for Russian and Eurasian Studies. Ziyad  Samadzadeh  Azerbaijani academician, Chairman of Economic Policy, Industry and Entrepreurship Committee of the Parliament (Milli Mejlis) in the Republic of Azerbaijan, Doctor of Science in Economics, Professor.

Reviewers Khoshbakht Yusifzadeh  Azerbaijani academician and the First Vice-President of the State Oil Company of Azerbaijan Republic (SOCAR), Doctor of Science in Geology and Mineralogy, Professor. Gorkhmaz Imanov  Corresponding Member of the ANAS, Azerbaijan, Institute of Control Systems of ANAS, a member of Royal Academy of Finance, Spain, Doctor of Science in Economics, Professor.


About the Authors

Adalat Muradov  studied at the School of Economy at Kiev State University (which is named after Taras Shevchenko) in Kiev, Ukraine, between 1985 and 1990 and graduated from that school with honors. He was awarded second place at the USSR Olympiad on Organization of the Economy and Production, held among students in 1989. He received a doctorate of philosophy in economics in 1993 and a doctorate of sciences in economics in 1998 at Kiev State University. He is an author of three textbooks, three monographs, three education programs, and more than 70 scientific articles. He has participated as a guest speaker at more than 100 international and local conferences. He worked as a professor at the Academy of Public Administration under the President of the Republic of Azerbaijan during the 1999–2004 period and at the Ministry of Economic Development (now called the Ministry of Economy) of the Republic of Azerbaijan between 2002 and 2014. He headed the Department of Macroeconomic Analysis and Forecasting during 2002–2007, the Department of Foreign Trade Policy and World Trade Organization between 2007 and 2009, and the Department of Economic Policy, Analysis and Forecasting between 2009 and 2014. He has been the rector of the Azerbaijan State University of Economics (UNEC) since 2014 by a decree from the President of the Republic of Azerbaijan and is the head of the Department of Economics and Business Administration at the UNEC Business School. He is the editor-in-chief of the Journal of Economic Sciences: Theory and Practice (in English) and scientific reviews at xiii


About the Authors

UNEC. He is an editorial board member for the Journal of Global Economy Review (Kozani, Greece). The major fields of his current research interests include macroeconomic policy, macroeconomic forecasting, the World Trade Organization (WTO), and oil and gas price forecasting. Yadulla  Hasanli  graduated from the School of Applied Mathematics at Azerbaijan State University (now called Baku State University) in 1981. He received a doctorate of philosophy in economics in 1989 and a doctorate of sciences in economics in 2007, and he has been a full professor since 2012. He was a postgraduate student at the Institute of Cybernetics at the Azerbaijan National Academy of Sciences between 1983 and 1987. He has been a participant and a winner at the national Olympiad in Mathematics in different years. Since 1981 he has worked at the Institute of Control Systems at the Azerbaijan National Academy of Sciences. He is the head of the Laboratory of Modeling of Social and Economic Processes at the Institute of Control Systems (part time). Between 1994 and 2005 he served as an assistant professor and head of department at the Higher Diplomacy College (now called the University of Eurasia). Between 1998 and 1999 he worked as an associate professor at the Institute of Political Science and Administration (now called the State Academy of Administration) under the President of the Republic of Azerbaijan. Between 2002 and 2005 he was a scientific advisor to the National Bank of Azerbaijan. In 1994 he has served as an assistant professor and a professor at the Department of Optimization and Administration at Baku State University. He became a local expert for the World Bank in Azerbaijan in 2000 and the Asian Development Bank in 2001. Between 2002 and 2005 he served as a scientific advisor to the Monetary Policy Department at the Azerbaijan Republic Central Bank. Since 2016 he has been a director of the Scientific Research Institute of Economic Studies, and since 2017 he has been a Professor at the Economics and Management department at the Azerbaijan State University of Economics (UNEC). He is an author of nine books (including four monographs) and about 167 scientific

About the Authors


works, 32 of which have been published abroad. He has attended about 20 international symposiums and conferences. He is an editorial board member for the Journal of Economic Sciences: Theory and Practice (in English) and the editor of scientific reviews at UNEC. The major fields of his current research interest are socioeconomic modeling, macroeconomic policy, macroeconomic forecasting, and oil and gas price forecasting. Nazim  Hajiyev  studied at the Baku branch of the Leningrad Financial Economics Institute (now called the Azerbaijan State University of Economics (UNEC)) between 1986 and 1990. He received a doctorate of philosophy in economics in 2000 and became an associate professor in 2014 with a decision from the Higher Attestation Commission under the President of the Republic of Azerbaijan. He was a winner at the national Olympiad in Mathematics in 1980. He has published one monograph and is an author of two textbooks for higher education, more than 90 scientific articles (including 25  in foreign countries), and the draft “Competition Code” of the Republic of Azerbaijan. He is responsible for executing 12 research works and more than 10 international and local grant projects (as a coordinator and expert). He has participated as a guest speaker at more than 20 international and local conferences. His research has been supported by the German Academic Exchange Service (DAAD), the US Agency for International Development (USAID), the United Nations Development Program (UNDP), the East–West Management Institute (EWMI) in New York (NY, USA), and the Science Fund of Azerbaijan Republic. He was a postgraduate student at the Center of Economic Reforms (now called the Institute for Scientific Research on Economic Reforms) between 1993 and 1996 and worked at the Ministry of Education between 1990 and 1992. He has been a researcher, a leading researcher, and a head of department at the Center of Economic Reforms under the Ministry of Economic Development (now called the Ministry of Economy). He worked at the Department of Antimonopoly Policy (State Service for Antimonopoly Policy and Consumer Rights Protection) under the


About the Authors

Ministry of Economy between 1996 and 2003. Since 2015 he has delivered lectures on “The Economics of Oil and Gas Resources” and “Project Management of Oil and Gas” in Greece and lectures on economics, managerial economics, and international management in the framework of UNEC and the University of Business and International Studies (UBIS; Geneva, Switzerland) Dual Degree Graduate Program. He is the managing editor of the Journal of Economic Sciences: Theory and Practice and scientific reviews at UNEC, and he also serves as an editorial board member for the Journal of Global Economy Review (Kozani, Greece). Since 2014 he has been the director of the UNEC Business School, and he has been an associate professor at the Department of Economics and Business Administration since 2017. He was a visiting scholar at the Davis Center for Russian and Eurasian Studies at Harvard University (Cambridge, MA, USA) during the 15th of  October 2017–31th August  2018 period. His last project was “Hydrocarbon Resources (Oil and Gas) in the Global Market and their Perspectives.” The major fields of his current research interests are applied economic issues, competition policy, socioeconomic modeling, macroeconomic policy, macroeconomic forecasting, and oil and gas price forecasting.

Chapter 1

Theoretical–Methodological Principles of the Problem

1.1  N  ecessity of the Problem, Study Level, and Methodological Aspects The roles that hydrocarbon resources, including oil products as energy carriers, play in the economy and in the life of people are undeniable. Thus, a demand arises for oil and oil products as energy carriers. Oil has been extracted for many years, and it is already produced in more than 100 countries that contribute to the supply of oil and oil products. The changes in supply and demand for oil continuously maintain the volatility of the oil price, increasing the interest of producers and consumers in oil prices. Changes in the price of oil, as an energy carrier, affect production indicators and prices in all fields of the economy at different levels. Political factors also affect the oil price because oil and oil products have a very prominent role in important areas of the economies of countries. Therefore, oil prices are always the focus of attention of both oil-producing and oil-consuming countries, and of politicians. In this respect, forecasting of oil prices is always on the agenda as an urgent issue. Changes in the oil price depending on the level of substitution elasticity also affect the volume of demand for, and prices of, the products of alternative energy sources. The role that the price of oil, as an energy carrier, plays in the development of countries, and of the world, is not in doubt. However, a method for providing a fairly accurate forecast of world oil prices is not yet known to science. In practice, oil price forecast information issued by influential international organizations deviates significantly from the actual prices. Nevertheless, research in this direction continues. The complexity of forecasting the oil price is associated not only with economic factors affecting it but also with the presence of political factors. As noted, forecasting of oil prices is carried out by a number of international organizations. However, for certain reasons, the forecasts provided by these organizations deviate from the actual prices. These deviations, including the political ambitions are being exacerbated in the times of crisis.

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2019 A. Muradov et al., World Market Price of Oil, SpringerBriefs in Economics,



1  Theoretical–Methodological Principles of the Problem

The main objective of this research is to provide reliable medium- and long-term forecasting by defining the independent variables (explanatory variables) and laws (patterns) of the distribution to forecast the oil price and then applying modern methods and methodologies (including an econometric modeling method). To provide forecasting for oil indicators in the world market, it is primarily necessary to define and analyze the factors affecting oil prices, to study the appropriate forecasting methods (trend, autoregressive integrated moving average (ARIMA), etc.) and to gather and process the data. Furthermore, forecasts will be made for pessimistic, medium, and optimistic scenarios under the best models determined after establishment of the short-, medium-, and long-term econometric models and testing of their adequacies. During the building of the econometric modeling, the following stages will be realized. Theoretical bases of econometric model Specification of econometric model

Statistical data

Assessment of the parameters of the model Testing the quality of the model No

Is the model adequate? Yes Provision of forecast

As mentioned above, forecasting of oil prices is provided by some international institutions such as the World Bank (2018) and the US Energy Information Administration (2018a, b). However, the changes occurring in the world economy and the global crisis in recent years have led to considerable deviations of these forecasts from the actual prices in some cases. For this reason, the need to carry out research in this direction is even more evident.

1.2  C  lassification and Analysis of Factors Influencing Oil Prices There are a number of factors that affect oil prices, and we can divide them into the following groups: –– Economics: Growth rate of the world gross domestic product (GDP); growth of the economies in the USA, China, and India; volume of oil production; levels of

1.2  Classification and Analysis of Factors Influencing Oil Prices


reserves in warehouses and stocks; volatility of the currency rate; scientific–technical progress, innovations, and new technologies; alternative energy resources; etc. (Braginsky 2008) –– Natural climatic conditions and solar activity: Relative air temperature, natural disasters, the Wolf number, etc. –– Politics: Interstate conflicts, etc. Analysis of the data shows that the growth rate of the demand for energy in China and India in recent years has exceeded the GDP growth rate. Currently, the process of industrialization and the formation of the middle class are creating high demand for energy resources, including oil, in these countries. In the industrialized countries, multinational companies and large corporations create the conditions of competition, innovation and new technologies, and scientific–technical progress, and, on the other hand, increase the effectiveness of the use of alternative energy sources and the process of development of efficiency in the use of funds at a higher level of venture business development. For example, the automotive industry— which has created considerable demand for petroleum products in the USA, Germany, and Japan—has also brought about much less fuel consumption through the development of new vehicles. Experts have come to the conclusion that the growth of the energy demand in China and India is a decisive factor in the formation of oil prices (Stiglitz 1975). The demand for oil in the world economy affects the world oil market price. The high oil price stimulates use of alternative energy sources. This leads to replacement of oil with other types of fuels (such as natural gases for generation of electricity and fuel energy). Eventually, this ultimately leads to a relative decrease in the demand for oil. Scientific and technical achievements in oil production and transportation costs reduce maintenance costs, thus enabling lowering of oil prices. The use of new technologies and new discoveries has a positive effect on improvement of the production of oil in the world market and, as a result, increases the proposal, and that stimulates the decrease in oil prices. The ongoing developments in oil production (horizontal and gas platforms and other offshore oil extraction and oil platforms for new construction, with no prone), reducing the level of costs, have caused a decline in oil prices. In some cases a factor of the problem comes from the fluctuation of the oil prices. In times of rising oil prices, oil prices move in the opposite direction with the discovery of new oil deposits. However, the depletion of a major factor is less important. The market is reacting calmly to the depletion of oil reserves. The possibility to increase oil production in a number of countries and the development of new technology for production of unconventional oil (bituminous sand in Canada and bitumen schists in the USA) have led to provision of optimistic forecasts by the International Energy Agency. Though oil is an exhausted natural resource, it will retain its priority importance for decades because of the volume of the discovered oil reserves and the economic efficiency of its use (Braginsky 2008).


1  Theoretical–Methodological Principles of the Problem

Climate conditions and solar activity significantly impact the demand for oil. First of all, the relative air temperature in the northern hemisphere is affected. It has been observed that the more the temperature goes down, the greater the demand for the oil during the winter time. This happens because the demand for oil products increases for generation of heat. It can also be expected that in the summer months in the southern hemisphere, especially in Eastern countries, the more the air temperature goes up, the greater the demand for oil products for cooling; conversely, the demand for oil products decreases in a mild winter. Thus, demand for oil products increases in the northern hemisphere during the winter months and in the southern hemisphere in the summer months, and vice versa. This balance has sometimes been disrupted with the background of global climate change over the past few years, which is reflected in oil prices. A number of studies have been conducted on the impact of indicators characterizing solar activity, such as the Wolf number, on the oil price (Hasanli and Ismailov 2014). A detailed investigation on this topic is described in the next section. The list of key factors affecting oil prices may vary depending on the time. First of all, it should be noted that the impact of military conflicts on the price of oil has weakened. Although the intervention of the US army in Iraq could previously create panic in the oil market, the recent military intervention of developed western countries (the USA, France, Britain, and others) and actions against Islamic State (ISIS) in Iraq and Syria have not affected the oil markets. Some experts believe that “the oil market is tired of politics” (Kravchenko 2007). However, others believe that reduced oil reserves in the USA will bring about an increase in oil prices and possible risks will be incurred in this way. Oil prices are affected by the expectations of market participants about political developments. For example, the release of sanctions against Iran, which were imposed by the USA and other developed countries, affected the decrease in oil prices. Financial structures are also among the factors that affect the dynamics of oil prices. The impact of financial structures such as hedge funds, banks, investment companies, and securities (fund) markets on oil prices has strengthened. These financial structures purchase shares in giant oil companies; therefore, they are interested in the preservation of high oil prices (Sharpe 1994). It is therefore no accident that the financial crisis in mid-2008 started in financial institutions, banks, and funds in the USA and resulted in a decline in oil prices. At the third Organization of the Petroleum Exporting Countries (OPEC) summit in November 2007—organized in Riyadh, the capital of Saudi Arabia—it was stressed that at present the formation oil market is located outside OPEC. The participants at the summit came to the conclusion that reductions in oil extraction, the reduction of world oil reserves, and political events have significant financial impacts on oil producers. Although oil is just a commodity, it has become the subject of speculation in financial markets (Dziuba 2008).

1.2  Classification and Analysis of Factors Influencing Oil Prices


1.2.1  T  heoretical Aspects of the Influence of Solar Activity on Oil Prices For a long time, researchers have paid great attention to studying the activity of the Sun. The reason is that the Sun has a strong influence on the Earth and on human life. Increases in solar activity excite the magnetic surface of the Earth and thus affect the physical and mental health of people. Considering that human beings play important roles in economic, social, and demographic development, the effects of changes in indicators characterizing the Sun’s impact on humans’ can also be considered natural. The great encyclopedist scholar of the East, Nasreddin Tusi, who lived in the thirteenth century and created the Maragha Observatory, showed Hulaku Khan with his experiences that the cause of processes occurring on the Earth needs to be sought in the sky (Muradov et al. 2018). Solar activity is characterized by various indicators. The most widespread of them is the Wolf number. Solar activity has periodicity, and there is information that the duration of this period is 10 or 12 years (Chizhevsky 1924; Obridko and Oraevsky 1993; Chertkov 1985; Sytinsky 1998; Lupachev and Chizhevsky 1996). These figures have mainly been evaluated by empirical methods. Changes in a number of indicators characterizing demographic development (natural increase, migration, marriage, divorce, life expectancy at birth, etc.) occur on a regular basis. There have been a number of scientific studies on the effects of solar activity on demographic, economic, and other processes (Hasanli and Ismailov 2014; Pakhalov 2011; Chizhevsky 1995; Charikov n.d.; Interfax 2010; Maddison 1964; Yandiev 2010; Hasanli and Ismayilov 2012; Hasanli et al. 2015; Belkin 2017; Muradov et al. 2016). The French physician Clement Juglar, who lived in the nineteenth century, studied statistics on birth, death, and marriage over a period of many years and concluded that regularities—more precisely, periodicities—exist in these statistics. Periodicity (increases and decreases) in demographic processes causes tendencies toward increases and decreases in other socioeconomic processes. Angus Maddison showed a correlation between world economic processes and solar activity on the basis of statistical data (Maddison 1964). The availability of statistics on long-term dynamics in the Wolf number characterizing solar activity and on the world population (birth, death, and natural increase), GDP, and world oil prices has allowed evaluation of these issues in research using econometric models (Dougerty 1992). As mentioned above, considering the increases and decreases in solar activity over the years—more precisely, its periodicity—its time series is approximated by the following regression equation:

 2π   2π   2π  SA = a0 + a1 f (t ) + a2 sin  t  + a3 sin  t  + a4 sin  t  + u,  T   2T   3T 



1  Theoretical–Methodological Principles of the Problem

Solar activity↑

Human activity↑

GDP per capita↑


Fig. 1.1  Indirect mechanism of the impact of solar activity on the life-span of people

where SA is solar activity; f(t) is the function of the time trend; a0, a1, a2, a3, and a4 are parameters; and T is the period of solar activity. The main result of the computer realization of the regression model (1.1) is the determination of the period length (t-parameter) of the Wolf number characterizing the solar activity. It may be more reliable to think that solar activity influences the life-span of people indirectly but not directly. Thus, solar activity has a positive effect on human activity (e.g., natural increase, work), increasing human activity raises the volume of production per capita, and the increase in productivity per capita increases the welfare and living standards of people and, consequently, the life-span of people (Fig. 1.1). The role that the price of oil, as an energy carrier, plays in the development of countries, and of the world, is not in doubt (Maddison 1964). The study of statistical data on oil prices under the demand–supply law provides grounds to say that changes in the world GDP volume affect world oil prices. The impact of another factor, solar activity,1 on world oil prices is studied and evaluated here. First of all, the following question can be posed: What might be the theoretical basis of the impact of solar activity on oil prices? Oil extraction does not have a long history at all, but solar activity has existed since the world was formed. Thus, an increase in the volume of production increases the consumption of energy carriers, including the demand for oil; however, according to the demand law, this, in turn, leads to an increase in oil prices. It is true that there are a lot of other reasons for changes in oil prices. However, changes in production volume remain one of the key factors affecting oil prices, and the change in solar activity is reflected in oil prices on the background of changes in the world GDP. In other words, there is a transmission mechanism for the impact of solar activity on human activity, GDP, and world oil prices, as shown in Fig. 1.2. Solar activity has a positive impact on human activity (e.g., natural population growth, work, health), increasing human activity increases the production of goods per capita, and the increase in per capita production increases oil demand and stimulates oil prices. The transmission mechanism for the impact of changes in solar activity on world oil prices is shown in Fig. 1.3. To achieve solution of the set issue, calculation of the impact of the indicator of solar activity on world oil prices is carried out using the two-step smallest squares

 A widely used indicator characterizing solar activity is the Wolf number, which is the number of spots and groups of spots on the Sun. It is named after Rudolf Wolf, who lived in Zurich, Switzerland, and was the first person to quantify the activity of spots and group of spots on the Sun. 1


1.2  Classification and Analysis of Factors Influencing Oil Prices Solar activity↑

Human activity↑

World GDP per capita↑

World oil prices↑

Fig. 1.2  Transmission mechanism for the impact of solar activity on human activity, the gross domestic product (GDP), and world oil prices

Solar activity↑

World GDP↑

World oil price ↑

Fig. 1.3  Transmission mechanism for the impact of solar activity on world oil prices

method (SSM) in the EViews applied software package. First of all, the parameters of the regression equation are evaluated:

WOP = C (0) + C (1) ∗ LOG ( W_GDP ) ,


where WOP denotes world oil prices, W_GDP is the volume of the world GDP, and C(0) and C(1) are parameters that characterize the impacts of the stable factor and the world GDP on world oil prices, respectively. Note that as W_GDP is the indicator of the amount of the world GDP, its impact on world oil prices is studied by LOG(W_GDP), the world GDP natural logarithm (under e—the Eiler number, e ≈ 2.72)—that is, by indexing. Then, using:

LOG ( W_GDP ) = a(0) + a(1) ∗ SA + a(2) ∗@ TREND,


the parameters of the regression equation are evaluated, where SA is the indicator characterizing solar activity; t is the time indicator; and a(0), a(1), and a(2) are parameters characterizing the impact of the stable factors, solar activity (SA), and the time factor (@TREND), respectively, on the world GDP volume (W_GDP). If (1.3) is written in place of (1.2), we obtain an econometric model of the impact of solar activity on world oil prices:

WOP = c(0) + c(1) ∗ a(0) + c(1) ∗ a(1) ∗ SA + c(1) ∗ a(2) ∗@ TREND.


Considering the transmission mechanism of the Sun as shown in Fig. 1.3, the impact of solar activity on world oil prices can be evaluated directly by the regression equation:

WOP = b(0) + b(1) ∗ SA + b(2) ∗@ TREND


WOP = b(0) + b(1) ∗ SA.




1  Theoretical–Methodological Principles of the Problem

In this case, the results obtained from evaluation of the parameters b(0), b(1), and b(2) of regression equation (1.5) should be a value equal to or approximating the parameters of (1.4):

b ( 0 ) ≈ c ( 0 ) + c (1) ∗ a ( 0 ) ,


b (1) ≈ c (1) ∗ a (1) ,


b ( 2 ) ≈ c (1) ∗ a ( 2 ) .


If the evaluation of regression equation (1.6) is adequate, then if (1.5) is correct, then there must be a correlation between (1.7) and (1.8) because of the different specifications of the conventional models and the presence of indicators with different units (inhomogeneity), and the economic interpretation of the b(0) and b(2) coefficients loses its importance.

1.3  Forecasting Through Econometric Models There are a number of forecasting methods. Econometric forecasting is the most widely applied one among them in practice. Building of econometric models serves two purposes. The first is to carry out the analysis, and the second is to provide forecasts. Although some statistical characteristics of econometric models are considered useful in the analysis, it is not expedient to provide forecasts with these models. The models built for forecasting require a number of additional statistical characteristics to be checked. For example, if the model to be used for analysis requires more attention to the meeting of t-statistics and homoscedasticity, primarily the determination coefficient, Durbin–Watson statistics, and other criteria are focused in the models for forecasting. Therefore, forecasting models differ from analysis models in some cases. The more common case is that all of the statistical characteristics and criteria of the built model are desirable, and such models are useful for both analysis and forecasting. However, it is impossible to continuously use the built model for both analysis and forecasting. One reason for these differences is that theoretical knowledge does not coincide with experimental knowledge in some cases. Thus, while theoretical knowledge is preferred in analysis models, forecasting models include more practical knowledge and specific features. If the dependence of the oil price on the factors affecting it is studied, then it is necessary to take into account theoretical knowledge, as well as experience, including specific situations. In practice we sometimes see several models that meet these criteria and are faced with the problem of how to choose the best of them. In such cases, the following indicators in the existing models that are valid for forecasting are compared: • Root mean square error • Mean absolute error

1.3  Forecasting Through Econometric Models


The model in which these indicators are smaller is more appropriate in forecasting. On the other hand, models in which the Theil inequality coefficient is small, bias proportion is small, variance proportion is small, and covariance proportion is greater provide more adequate forecasts of the given indicator series. Note that the bias proportion indicates the degree of deviation of the center of the forecast from the actual center, the covariance proportion indicates the concentration of the bias proportion, and the variance proportion shows how far the forecast change is from the actual one.

1.3.1  S  hort-Term and Long-Term Econometric Models for Oil Price Forecasting As mentioned above, medium- and long-term reliable forecasting of oil prices has always been observed with problems. This is connected not only to the broad spectrum of factors affecting it—economic, natural, and political factors—but also to the lack of quantitative characteristics of a number of indicators characterizing these factors. Changes in the price of oil, as an energy carrier, affect the economies of almost all countries—in other words, both oil-importing countries and oil-exporting countries—because the economies of all countries are interconnected as a result of import–export operations and international capital flow. These mutual relationships are more obvious from the world “input–output” model (Hasanli and Abasov 2014). When the oil price is modelled by means of a time series, changes and levels of the data on variables may have a short-term and long-term nature. For example, if the data on variables have new εt information at t time (that is not known before t time)—i.e., Δyt = εt—then εt represents the new data on the change in the yt variable; that is, the event or any change takes place in t  time within a short time. Therefore, models built with inclusion of the increase in variable (Δyt) are commonly called short-term models. On the other hand, in the case of yt = εt + εt − 1 +  εt  −  2  +  ⋯  +  εt−T, it can be seen that the level of yt consists of collection of the information on the previous time levels, events, or policy changes. Therefore, the value of the correct statistical analysis using (Δyt) changes instead of (yt) levels is the loss of old (long-term) data accumulated at the level of trends in the added value of the oil sector. Factors affecting the outcome in long-term econometric models are involved with their actual units of the measurements, and the indicator of the results, as a rule, depends on time. These factors include nonstationary time series that are able to reflect long-term fluctuations. The essence of the relationship between dependent variables and independent variables is of great importance in long-term models. These relationships are evaluated by referring to the standard deviations (errors), and strongly linked time series are chosen. In econometrics, simultaneously changing time series that with a certain function have a joint standard deviation of zero or about zero are called joint series


1  Theoretical–Methodological Principles of the Problem

or cointegration series. If there is a strong and lagged link between the series included in the model (the series may also be nonstationary), then such variables are considered to be joint or cointegrated with the dependent variable. In addition to the existence of close and lagged relationship between the variables in analysis of the joint relationships (analysis of cointegration), the essence of the mutual dependence between the variables (straight or inverse proportion) is also of significant importance. For example, the econometric model would be characterized by a longterm cycle if the oil price in the world market was a dependent (explained) variable, and there would also be factors and independent (explanatory) variables influencing the oil prices in the analysis (for example, the volume of the world GDP, tendency toward change due to time, etc.). A reliable forecast of oil prices is interesting for various business circles and various professionals. As the budgets of the oil-exporting countries are heavily dependent on oil prices in the world market, oil prices in the world market are accepted as the main factors in forecasting of revenues and expenditures of the state budgets. In general, the forecast of the oil price in the world market is considered a major variable with regard to its impact on the expectations and decisions of producers, consumers, financial institutions, and governments. For this reason, forecasting of oil prices in the world market is an important issue for a number of researchers. Several methods are used to forecast oil prices in the world market: trend, ARIMA, ARIMA–generalized autoregressive conditional heteroscedasticity (ARIMA–GARCH), Holt, Box–Jenkins, vector autoregression (VAR), vector error correction model (VECM), autoregressive fractionally integrated moving average–GARCH (ARFIMA–GARCH), ARFIMA–fractionally integrated generalized autoregressive conditional heteroscedasticity (ARFIMA– FIGARCH), etc. The results of the studies conducted with VAR and VECM models show that there is cointegration between the real (spot) oil prices and monthly futures contracts (Koehler et al. 2012). Considering the existence of a long-term balance here, the correction of the error vector of the VECM model was evaluated and the obtained forecast was compared with the actual price of oil, the result of which determined that the information provided in the oil futures market could explain a significant portion of the fluctuations in the oil price. The main idea of ​​the researchers implementing forecasting of the oil price in the world market by mean absolute percentage error (MAPE)  model is that the residuals in the time series may affect the evaluated parameters of the model and the forecasts in exponential smoothing. Therefore, more attention should be paid to the last residuals of the time series. The model considering this factor increases the accuracy of the forecast values (Koehler et al. 2012). It was concluded by the studies of Kang and Yoon (2013) that ARFIMA– FIGARCH models better reflect the previous behavioral peculiarities of oil prices than other models. However, the researchers assumed that none of the aforementioned forecast models can be useful for all three types of futures contracts. Thus, for example, an ordinary ARIMA–GARCH model is selected for the West Texas Intermediate (WTI) oil price, and ARFIMA–FIGARCH models are preferred for mazut and petrol futures (Kang and Yoon 2013).

1.3  Forecasting Through Econometric Models


Use of ARIMA models in forecasting is Box–Jenkins methodology (Box and Jenkins 1976). The Box–Jenkins methodology consists of four stages: model identification, parameter evaluation, diagnostic tests, and model forecasting ability. In his article, Chaido Dritsaki (2018) mentioned that an exact forecast is provided by this model in the short term but it cannot smooth the volatility of the time series; therefore, hybrid ARIMA–GARCH models have been considered to predict oil prices (Dritsaki 2018). As is known, one of the factors that cause sharp fluctuations in oil prices in the world market is crises in the world economy. On the other hand, there are some variables in these factors for which quantification of the characteristics is lacking, and so forecasting is much more difficult. Examples are the global financial crisis in 2008, the change in US–Iran relations in 2015, the coming to power of President Trump in the USA (who has pursued different politics from those of his predecessor, Obama), and so on. Therefore, in the present study, we conduct forecasting of the oil price by considering these factors as a fictive variable (DUMMY) in the trend and ARIMA models, where these factors have been predicted for oil prices.

1.3.2  Trend Models and Forecasting A trend model in the economy means a regression equation (1.10) characterizing the change in any economic variable, such as oil prices, depending on time:

y = f ( t ) + u,


where y is the result indicator for which forecasting is conducted, t is the time factor, u is the accidental error, and f(t) is the function providing the dependence of the result indicator on time (this function may be linear and nonlinear). We can provide the simplest linear trend function as follows:

y = α 0 + α1t + u,


where y is the result variable (oil price) and t is the time factor (month, term, year). The α0, α1 parameters in (1.11) can be found from the following linear equation system with the Method Least Squares (MLS):

 nα 0 + α1 ∑ti = ∑yi  i i .  2 + = α t α t ti yi ∑ 1∑ i  0∑ i i i i 


We can write the other descriptive trend functions as follows:

y = α 0 + α1t + α 2 t 2 + u ( parabolic ) ,



1  Theoretical–Methodological Principles of the Problem

y = α 0 + α1t + α 2 t 2 + α 3 t 3 + u ( cubic ) ,


y = α 0 + α1t + α 2 t 2 + α 3 t 3 +  + α n t n + u ( spline ) ,


y = α 0 + α1 / t + u ( hyperbolic ) ,


y = α 0 ∗ eα1t ∗u ( exponential ) ,


etc. If these trend functions are being accepted as denotations then they are brought into a linear shape; for example, if we accept t2 = x in (1.13), the parabolic function will become a linear function like y = α0 + α1t + α2x. The α0, α1, α2, α3, …, αn parameters are found by the MLS. Usually the (1.17) trend models with an exponential image and exponential function are used in forecasting the relationship between time and human factor. This feature can easily be linearized; if the logarithm of both sides of (1.17) is taken,

ln ( y ) = ln (α 0 ) + α1t + ln ( u )


is obtained. Equation (1.18) is called a logarithmic linear function, where y is the explainable variable and t is the explaining variable. If ln(y) = z and ln(α0) = α0 symbolizing is done,

z = a0 + α1t + e,


the linear function is obtained and the α0 and α1 parameters of the (1.19) trend function are found by MLS. To provide a forecast by using trend models, first of all, the parameters are found with MLS. If there is adequacy of the trend model and its found parameters, then we find the forecast for the result indicator by placing the value of the t time factor for the forecast period in the respective trend model for the said period. Interpretation of the α1 regression coefficient: Note that in (1.18) the α1 coefficient logarithmic linear dependence is called semielasticity. It is obvious that the elasticity coefficient indicates a change in the percentage of the explainable variable as a result of a 1% change in the explaining variable. If the elasticity coefficient of the y variable relative to the t  time variable is marked Ety, then:

Ety =

dy t ∗ . dt y


For interpretation of the α1 regression coefficient:

dy = α1 ∗ dt , y


1.3  Forecasting Through Econometric Models

α1 =


dy . y ∗ dt


where dt = 1 is adopted; that is, if both sides of the (1.22) equality are multiplied by 100 to show the percentage change in y—the result indicator in the unit change of time (year, term, month, day)—then:

α1 ∗ 100% =

dy ∗ 100% y


is obtained. The (1.23) equality indicates a α1∗100% change in the explaining y variable in the unit change of the t time factor. The α1 regression coefficient shows the growth coefficient of y in each unit increase in t. The α1 coefficient needs to be multiplied by 100 to express the change in y in a percentage in the interpretation.

1.3.3  Time Series and Nonstationary Problems Note that the econometric models that we will use in forecasting the oil price are realized in the time series. The mutual impact of the economic indicators as a matter of fact and the dependence of the current prices on those in previous times (years) lead to disorder in the stationarity of the series, including residuals. Therefore, let us provide a brief description of the nonstationary problem of the time series, its removal, and the testing methods. Stationarity means the stableness of a number of the properties of the variable due to time. If the density function of the distribution of a variable remains unchanged in spite of a change in the price due to time, then the process is called seriously stationary. In other words, the density function of the distribution remains unchanged when a seriously stationary series moves because of time. Sometimes this type of stationarity is also called serious stationarity. Serious stationarity may also be characterized by more frequently used statistical properties (mathematical expectation, dispersion). In a narrow sense, a stationary (seriously stationary) process has the following properties: • Its mathematical expectation does not depend on time. • Its dispersion does not depend on time. • Autocovariation and autocorrelation functions (a) depend only on time differences; and (b) are dual functions. The collection of covariation values at all possible intervals between the time moments is called the autocovariation function of the coincidental process. It is known that the correlation coefficient is obtained from the division of the c­ ovaluation into the square root of the product of the two dispersions. Since the dispersion is stable, we obtain the dependency of the correlation on the covaluation only. This defines the autocorrelation function of the time series. The autocorrelation function


1  Theoretical–Methodological Principles of the Problem

shows the degree of the statistical importance of the value of the time series in different moves (for example, 1 year or 2 years for annual data, etc.). If the mathematical expectation of the coincidental process and dispersion exist and are not dependent on time, the autocorrelation (autocovariation) function depends on the time differences only; then such a process is called a poorly stationary process, not a seriously stationary process, or, in a greater sense, stationarity. Any seriously stationary process is a poorly stationary process. In contrast, it does not mean that the poor stationary process is not a serious stationary process. Both of these definitions of stationary stability for normal processes are equivocal. Therefore, strong stationarity is obtained from poor stationarity due to normal or Gauss processes. It should be noted that the autocorrelation function in stationary processes cannot have any value. When setting up the values of the autocorrelation function, we have to ensure this property has been met. In econometric modeling of time series, white noise and random walk (exiting) processes are frequently encountered (Kantorovic 2012). If the conditions of Gauss–Markov (if E(ei) = 0, var(ei) = σ2; i ≠ j, cov((ei, ej) = 0) are met by the residuals (ei), it shows the process is poorly stationary or, in a greater sense, stationary. This process is called white noise. If we add the fact that the ei coincidental quantity is distributed by law normally to the list of the conditions, then the process is seriously stationary in a narrow sense. In this case, the statistical features of strong stationarity (stationary in a narrow sense), poor stationarity (stationary in a larger sense), as well as white noise processes, coincide. This case is called a Gauss white noise process in the literature and is written as ei − WN(0, σ2). White noise is very important in analysis of time series, and it causes more complex processes. The random walk process: Sometimes this process is called Braun movement. This process is provided as follows:

xt = xt −1 + et ,

where et is white noise. The random walk process can be viewed as a process with a regression coefficient equal to 1. The term “random walk” was derived with regard to an issue in a joke at the beginning of the nineteenth century: If a completely drunk person is left in a clean area, then where should we look for him a while later? Answer: If the drunk has accidentally exited, then you should look for him in the place where he was left in the area. More precisely, the drunk will eventually return to that previous place as a result of random walking (in the middle). The random walk process has the following properties: • E(xt) = x0 + 0 = const, x0—the value of the process at the start moment. • var ( xt ) = t ⋅ σ e2 . This shows that the dispersion of the random walk process changes depending on time. More precisely, it will increase proportionally to time.

1.3  Forecasting Through Econometric Models


We may conclude that the random walk process is not even poorly stationary (stationary in a larger sense), because the dispersion is not stable and changes depending on time. However, it is possible to turn a coincidentally exiting process into a stationary one by means of simple conversions. We can write the equation xt = xt−1 + et as xt − xt−1 = et or Δxt = et. We can sign the Δxt growth or the first difference (in Russian: первую разность) as zt and see it as another time series. Then, we obtain zt = et. Thus, if we see the first difference in the nonstationary series, we can obtain a new stationary series as a result. This is the most widespread method to turn a nonstationary times series into a stationary series. Evaluation of the coefficients of the regression equation when stationarity is violated by the method of the usual smallest squares causes certain distortions. Violation of the stationarity of the process, first of all, shows itself in the presence of autocovariation and autocorrelation.  Autoregression Assume that Y(t), as a variable, has autoregression (AR) dynamics; in other words, the dynamics of Y(t) in t time depend on the dynamics in t − 1 time. Mathematically this dependence may be provided for as follows:

Y ( t ) = a ∗ Y ( t − 1) + e ( t ) ,

where Y(t) is the value of the variable in t time, Y(t − 1) is the value in t − 1 time, and e(t) shows coincidental errors. If the condition |α|