The Effects of Globalization on National Labor Markets: Diagnosis and Therapy [1 ed.] 9783428522996, 9783428122998

Citation preview

Applied Economics Quarterly Supplement

2006

Beihefte der Konjunkturpolitik Issue 57 The Effects of Globalization on National Labor Markets: Diagnosis and Therapy

Edited by Dennis J. Snower Rainer Winkelmann and Klaus F. Zimmermann

asdfghjk Duncker & Humblot · Berlin

The Effects of Globalization on National Labor Markets: Diagnosis and Therapy

Applied Economics Quarterly Supplement Beihefte der Konjunkturpolitik Issue 57

The Effects of Globalization on National Labor Markets: Diagnosis and Therapy

Edited by

Dennis J. Snower Rainer Winkelmann and Klaus F. Zimmermann

asdfghjk Duncker & Humblot · Berlin

Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at .

Technical editor: Deborah Anne Bowen All rights reserved. No part of this book may be reproduced, translated, or utilized in any form or by any means, electronic or mechanical, without the expressed written consent of the publisher. # 2006 Duncker & Humblot GmbH, Berlin Typesetting and printing: Berliner Buchdruckerei Union GmbH, Berlin Printed in Germany ISSN 1612-2127 ISBN 3-428-12299-2 978-3-428-12299-8 Printed on no aging resistant (non-acid) paper ∞ according to ISO 9706 *

Internet: http://www.duncker-humblot.de

Editorial This supplement to Applied Economics Quarterly presents contributions to the 69th Annual Meeting of the Association of German Economic Research Institutes (ARGE-Institute), which took place in Berlin on April 27, 2006. The topic was “The Effects of Globalization on National Labor Markets: Diagnosis and Therapy” By taking on a theme of major current importance – particularly in the context of high unemployment – the conference contributed highly relevant economic policy advice to the federal government on a number of key issues. The lectures and discussions dealt in large part with the diverse implications of globalized labor markets: the consequences for monetary and fiscal policy, the importance of (de-)regulation in a globalized labor market, and the labor market effects of outsourcing. Several presentations focused on issues of income equalization and economic efficiency, federal support to the low-wage sector, and human capital formation as a component of economic policy. We are very grateful to the German Federal Government and in particular to the Federal Ministry of Economics and Technology (BMWi) for their support to this conference, to the numerous attendees for their enthusiastic participation, and to all those who provided interesting and thought-provoking contributions to the discussion. We would like give special thanks to Frank Oskamp (IfW Kiel) for the conceptual preparation of the conference and to the organizer, Ralf Messer (Secretary General, ARGE-Institute). Next year’s annual meeting is scheduled to take place on April 19, 2007, in Berlin and will deal with “Energy Markets”. June 2006

Dennis J. Snower Rainer Winkelmann Klaus F. Zimmermann

Contents Richard Layard Full Employment for Europe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

Justin van de Ven and Martin Weale Equality and Efficiency: Policy for Globalisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

23

Christian Dreger and Stefan Kooths The Effect of Globalization on Aggregate Labour Demand in EU countries . . . . . . . . . .

35

Comment: Herbert S. Buscher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

47

Robert Jäckle The Impact of FDI on the Skill Structure in German Manufacturing . . . . . . . . . . . . . . . . .

55

Comment: Alexander Lipponer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

79

Peter Nunnenkamp Relocation, Offshoring and Labour Market Repercussions: The Case of the German Automobile Industry in Central Europe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

83

Comment: Andre Jungmittag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 Alexander Spermann Basic Income Reform in Germany: Better Gradualism than Cold Turkey . . . . . . . . . . . . . 113 Comment: Steffen J. Roth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Full Employment for Europe By Richard Layard*

Abstract Unemployment, not productivity, is the problem in Europe. So unemployment does not reflect some general weakness of the European economy. If it did, we would also see poor secular productivity growth, which is not the case. Unemployment is a specific problem, having specific causes, and for which there are specific remedies. Many European countries have reduced their unemployment rates to United States levels or below, including some, such as Denmark, which have very high tax rates. It is precisely the variation of experience among the different countries, which helps us to understand what must be done by those large continental countries where unemployment remains so shockingly high. By the early 1990s the evidence already showed that the keys to reducing unemployment were welfare-to-work policies for the unemployed and more flexible wages. Those countries like Denmark, the Netherlands and Britain which acted on this evidence have halved their unemployment since then. Those which have not taken action have continued to have high unemployment, even at the peak of the European boom in 2000. In that year both France and Germany had record levels of vacancies despite massive unemployment – showing that the main reason for unemployment was a failure to mobilise the unemployed. Almost any job is better than being unemployed. Research on happiness shows that being unemployed is as bad for a person’s happiness as being divorced, and three times worse than losing one third of your income. So it is good for unemployed people that, after a while, they should be expected to fill most types of vacancy. It is also important that after some period they automatically receive offers of activity, which they are required to accept rather than staying at home on benefit. This “activation principle” has been a major factor in lowering unemployment in many countries. But this whole approach requires an active and energetic service which combines job search assistance and benefit monitoring. The Hartz reforms in Germany are a massive step in the right direction but they need to be energetically monitored and followed in other countries.

* Richard Layard, London School of Economics, is Professor of Economics (Emeritus), and founder and former Director of the Centre for Economic Performance. This paper draws heavily on Layard et al. 2005, which evaluates the experience of the 1990s, and on de Koning et al. 2004, which examines in detail the experiences of Britain, Denmark and the Netherlands. This paper originally appeared in (eds) A. Lopez-Claros, M. Porter and K. Schwab The Global Competitiveness Report 2005 – 2006: Policies Underpinning Rising Prosperity, World Economic Forum, 2005.

10

Richard Layard

Wage flexibility is also vital in regions where productivity is lower than elsewhere and needs to be adequately reflected in lower wages. This applies to east Germany, southern Italy and southern Spain. But other elements in the “flexibility” package often advocated, such as lower job security and lower taxes, would make little difference and there is no need to abandon the whole European model in order to deal with the specific problem of unemployment.

Introduction Unemployment, not productivity, is the problem in Europe. Europe’s four largest countries have unemployment rates far higher than the United States (see Table 1). As I shall show, this is a specific problem, having specific causes, and for which there are specific remedies. Unemployment does not reflect some general weakness of the European economy. If it did, we would also see poor secular productivity growth, which is not the case. (See Table 2).1 Moreover, many European countries have reduced their unemployment rates to United States levels or below, including some, such as Denmark, which have very high tax rates, and others, such as the Netherlands, which have tax rates typical of the EU as a whole. So there is no general “European unemployment problem”. It is precisely the variation of experience among the different countries, which helps us to understand what must be done by those countries where unemployment remains so shockingly high. Table 1 Unemployment Rate (percent) 1993

2004

7

5.5

Britain

10

4.8

Denmark

10

5.4

6

4.6

USA Reformed

Netherlands Did not reform France

11

9.7

8

9.5

Italy

10

8.0

Spain

17

10.8

Germany

Source: HM Treasury Pocket Databank, 24 May 2005.

1 See also Blanchard, 2004. As regards levels, productivity per hour worked in northern Europe is broadly similar to the United States.

Full Employment for Europe

11

Table 2 Productivity per Hour Worked in United States and Europe (growth percent per annum)

United States Europe (15)

1980 – 2004

1990 – 2004

1.8 2.0

2.0 1.8

Source: Groningen Growth and Development Centre; Conference Board.

The required remedies were advocated by many writers as early as 1990 (for example, Layard et al, 1991). Some countries pursued these remedies and others did not – a perfect natural experiment. Those countries which implemented the remedies have halved their unemployment, while in those countries which did not: France, Germany and Italy, unemployment remains almost as high as ever. In what follows, I begin by assessing the importance of reducing unemployment. I then discuss the main factors affecting unemployment rates, and how these explain the experience of different countries. I end by discussing what specific policy changes are needed, if there is to be real progress.

1. The Cost of Unemployment The present level of unemployment in some European countries is totally unacceptable, for the simple reason that unemployment is one of the clearest sources of human misery, one which we already know how to reduce. The high cost of unemployment is less a matter of the loss of output and income than of the pain of rejection. We can see this from the numerous studies of human happiness carried out in the last ten years, using Eurobarometer, the United States General Social Survey, or the World Values Survey. Table 2 reports the results of the World Values Survey, as analysed by John Helliwell (2003a).2 In this survey of 46 countries, individuals report their happiness on a scale from 1 to 10, as well as various other features of their life. These features are then used to explain happiness. To avoid incorrect inferences, the effect of each feature is examined, while holding the other features constant. Some of the effects are given in Table 3, which shows that a one third drop in family income reduces happiness by 0.2 points. But if, in addition, a person is unemployed, happiness falls much more steeply: by a further 0.6 points – equal to the effect of divorce. In both cases, the person ceases to be needed. 2

See also Layard, 2005, p. 63 – 68.

12

Richard Layard Table 3 Effects on Happiness Fall in happiness (points) Family income falls by one third Unemployed Divorced Separated

0.2 0.6 0.5 0.8

Source: Helliwell, 2003a. For the other variables which are held constant, see his paper, or Layard, 2005, p. 64.

In further work, Helliwell (2003b) has investigated the factors affecting the difference between nations in average happiness and in suicide rates. Again, the unemployment rate shows up as an important factor. Clearly, unemployment really does matter. And, a reasonable corollary to Table 3 would be that a job at almost any wage is better than no job at all, even if unemployed people do not always foresee this. 2. Expanding the Number of Jobs But how is the number of jobs determined? In the medium term, it depends on the number of people actively seeking work. Many people find this difficult to believe, and it should be stressed that many policy mistakes arise from a failure to understand it. Before the Pilgrim Fathers landed in New England, there were no jobs; but after they landed, jobs sprang up. Or, to take a less obvious example, the labor force in Britain has grown by 212 percent since 1851;3 over the same period, the number of jobs has grown by 212 percent. So, a market economy – ignoring the business cycle – always provides more jobs, if there are more people actively seeking work. The same point emerges if we compare countries. Since 1960, the labor force has grown at very different rates in different countries, mainly due to demographic factors. If jobs were determined independently of the supply of labor, this would have caused massive problems in countries with the fastest growing labor forces: the United States, Ireland, and Japan. Yet, as Figure 1 shows, wherever the number of job-seekers has grown faster, the number of jobs has grown at a roughly corresponding rate. Similarly, countries which have artificially reduced their labor forces by requiring early retirement have simultaneously reduced the number of jobs, with no change in the unemployment rate.4 3 Mitchell, 1992, p. 81, and National Statistics, Labour Market Trends, Table A.11, March 2005. 4 de Koning et al., 2004, Figure 3, and related discussion.

Full Employment for Europe

13

Thus, in the medium term, the number of jobs is not determined independently of the supply of labor. It is determined by a) the size of the labor force, and b) the equilibrium unemployment rate, which is itself influenced by how hard unemployed people look for work. In the short run, things are of course completely different: the number of jobs is determined by the level of aggregate demand. Since it is easy to expand the level of aggregate demand by fiscal and monetary policy, one has to ask why unemployment cannot always be diminished. The answer is that inflation would eventually set in; the labor market would become too tight, due to too many unfilled vacancies. This is, in fact, exactly what happened in Europe in 2000, forcing the ECB to raise interest rates, and dampen the boom. So the key to reducing unemployment is to mobilize the unemployed to fill emerging vacancies as quickly as possible.

Employment Source: OECD Labour Force Statistics; European Economy.

Figure 1: Percentage Growth in the Labor Force and Employment 1960 – 2000

3. Unemployment and Vacancies According to the simplest theory of unemployment, inflation will rise when a specific level of vacancies has been reached. This is the non-inflationary vacancy rate (see Figure 2).5 In addition, there is an inverse relation between vacancies and unemployment (see Figure 3): the higher the vacancy rate, the lower the un5

See, for example, Layard et al., 1991, p. 275.

14

Richard Layard

employment rate.6 Thus, equilibrium unemployment is whatever level of unemployment is consistent with the non-inflationary level of vacancies. This rate depends on human institutions. It can be reduced by mobilizing the unemployed, so that they are better at filling vacancies, thus moving the Vacancy / Unemployment (VU) curve to the left. Vacancy rate Vacancy rate

‘Non-inflationary’ ‘Non-inflationary’ vacancy rate vacancy rate

Change in inflation inflation Change in 0 Source: Author.

Figure 2: Vacancies and Inflation Vacancy rate Vacancy rate

V

‘Non-inflationary’ ‘Non-inflationary’ vacancy rate vacancy rate

U

Unemployment rate Unemployment rate Equilibrium Equilibrium unemployment unemployment Source: Author.

Figure 3: Vacancies and Unemployment 6

This assumes a fixed degree of turnover in the market.

Full Employment for Europe

15

At least six factors have been suggested as possible influences on equilibrium unemployment rates: 1. how unemployed people are treated, 2. the flexibility of wages, 3. the flexibility of employment (i.e. firing), 4. the size of the labor force, 5. hours of work, 6. taxes on labor. But by 1990, the evidence supported the overriding importance of the first two, and this still remains the case. Layard et al. (2005) show that institutional changes in these first two variables explained nearly half the variance in unemployment change across OECD countries between the 1960s and early 1990s, and similarly between the 1980s and 2000 – 2001. And there is still no consistent evidence that any of the other four variables have any major influence on the unemployment rate (see Layard et al. 1991, and Nickell and Layard 1999). So reform should be focused mainly on the first two factors and they are the focus of my argument.

4. How Unemployed People are Treated If unemployed people are more actively looking for work, and being helped in their efforts, there will be fewer unfilled vacancies at any level of unemployment. So the equilibrium level of unemployment will be lower. Therefore, one key issue has to do with how unemployed people are treated. Countries differ to an astonishing degree in how they treat unemployed people (OECD, 2001). Although the level of actual benefits is significant, the following are more important: 1. how long people may receive benefits without working, 2. what job search is required, with what intensity, and what kind of jobs must be accepted, 3. what help is provided with job search and employability. The importance of the duration of benefits can be seen from Figure 4. The longer you pay people for doing nothing, the longer they will do just that. But conditionality is also vital. In some countries, an unemployed person must visit the employment office every two weeks, report on the job search done; at this time, he or she will receive a list of possible vacancies to approach in the next two weeks. In some other countries, the unemployed have only to call the benefit office every three months to confirm that they are still out of work. In Britain, once a person has been unemployed for 13 weeks, he or she must accept any type of job

16

Richard Layard

Maximum duration of benefits (years), 1992

within one hour’s travelling time to work. In Germany, even after the Hartz reforms, such a requirement comes into effect only after a year.

0

20

4+

40 NZ

AL

UK

60 GE

IR BE

4+

SP FR

3

3

DE FN

2

AU

NE

2

NO SW CA

1

US

SZ

1

JA

0

0 0

20

40

60

% of unemployed people out of work for over 12 months, 1989

Source: Nickell and Layard, 1999.

Figure 4: Long-term Unemployment and the Duration of Benefits

Needless to say, conditionality is impossible unless the employment service is actively providing job-search assistance. But even in that case, it is not easy to enforce: if you don’t actually want a job, you are not likely to be offered it. For this reason – and on humanitarian grounds – many countries have concluded that the employment service, within the first year, should offer each unemployed person some sort of activity. This activation principle was included in the 1997 Employment Guidelines adopted by the EU heads of government. But it has so far been implemented by only a few countries, especially Denmark and the Netherlands. The case of Denmark is the most striking. As Figure 5 shows, in 1993, despite its 10 percent unemployment, the level of vacancies was at a record high. After concluding that the problem was a failure to mobilize the unemployed to fill available job vacancies, they adopted the principle of activation after a specified period of time, which is now one year. The results were remarkable: unemployment dropped by 6 percentage points, and the employment rate increased by 6 percentage points, with the number of people actually working in “activation” positions no higher than before activation became compulsory. The policy has worked, because people have become more willing to fill the existing vacancies. Thus, despite the huge increase in labor demand, there has been no increase at all in the number of vacancies.

Full Employment for Europe

17

Denmark 0.14 1990

Vacancy Rate (%)

0.12

1993

1995

0.1 0.08 1985

0.06 1975

0.04

1980

2001

0.02 0 2

4

6

8

10

12

Unemployment Rate (%)

Britain

Labour Shortage Index

14 12 10

1975 2001

1990

8

1985

1995

6 1980

4 2 0 2

4

6

8

10

12

Unemployment Rate (%)

Netherlands 2.5 2001

Vacancy Rate (%)

2

1990

1.5 1980

1975

1

1995 1985

0.5 0 2

4

6

8

10

12

Unemployment Rate (%)

Source: de Koning et al., 2004. p. 47 – 8.

Figure 5: Three Countries with low Unemployment – (During the 1990s Unemployment at Given Vacancies Fell) 2 Supplement 57 – 2006

18

Richard Layard

Similar policies have been adopted in Britain and the Netherlands, with similar results. (In the Netherlands the high vacancies in the boom of 2000 were not inflationary, due to a policy of wage restraint.) In each of the three countries discussed so far, the policies have included: more active help with job search; stricter conditionality; more active sanctions for non-compliance; more use of job subsidies; and, after a period of time, the guarantee of activation, along with the corresponding responsibility to comply. In addition, the policies depend on a unified administration of benefits and active labor market assistance, which ensures that benefits act as an aid to job search, and not as a deterrent. In Britain, these policy changes were introduced gradually throughout the 1990s, first by a Conservative government (with a somewhat punitive slant), and then by a Labour government (with a more positive slant). The contrast between these successful experiences and those of France and Germany is striking, and provides social scientists with a rare natural experiment which can confirm or disprove their predictions. It, indeed, confirms all the predictions of the 1990 view of mainstream labor economists. During the 1990s, neither France nor Germany introduced any of the abovementioned reforms. The result was as predicted (see Figure 6). In the European boom of 1997 to 2001, unemployment fell only 2 percentage points in Germany and 3 in France, but vacancies rose to record levels, despite high unemployment. Clearly there had been a failure to mobilize the unemployed. In both countries, public opinion has been unwilling to accept the idea that the behaviour of the unemployed could have any affect on the number of jobs. This has been due partly to inappropriate theories – of the kind criticized at the beginning of this paper – and partly to a compassionate regard for the unemployed. But respect requires understanding as well as compassion. And unemployment is actually quite difficult to understand using traditional theories. For if unemployment makes people miserable, then, according to most scientific theories, the unemployed must be doing all they can to escape from it. But modern psychology explains that, although unemployment causes great unhappiness, unemployed people often reject opportunities which would make them happier. This is explained by the fact that a) people adapt, to some extent, to almost anything and fear further change (Frederick and Loewenstein, 1999), and b) individuals do not, as a rule, do well at predicting how they would actually feel in changed circumstances (Loewenstein and Schkade, 1999). In the interests of the unemployed themselves, employment services should introduce them to something that will make them feel better. That is the justification for a policy that is both tough and tender: it is on the side of the unemployed, but against unemployment as any form of solution for them.

Full Employment for Europe

19

France 25

1990

Labour Shortage Index

2001

20 15 1980

10

1975 1985

1995

5 0 2

4

6

8

10

12

10

12

Unemployment Rate (%)

West Germany 2

Vacancy Rate (%)

2001

1.5 1995

1980 1990

1 1975

0.5

1985

0 2

4

6

8

Unemployment Rate (%)

Source: de Koning et al. (2004) p. 47 – 8.

Figure 6: Two Countries with high Unemployment (During the 1990s Unemployment at Given Vacancies Did not Fall)

Unfortunately, such arguments are still not widely understood in France.7 In Germany the political leaders now understand these points quite well, and the Hartz reforms passed by the German legislature are an important and courageous step in the right direction. They include a unified set of benefits (with a means test after a year), stricter conditions about job search, more job opportunities through temporary work agencies, and the “activation principle,” applied to all people un7 Notable exceptions are a) the report, written in 2000 – 2001 for M. Jospin by his Council of Economic Analysis, and b) the example set by the negotiators of the Patronat and the unions (especially Mme Nicole Notat, Director General of VIGEO) in 2000 – 2001, who were unfortunately unable to secure effective government backing for their proposals.

2*

20

Richard Layard

der 25. In addition, the administration of benefits is to be integrated with active labor market assistance. But, as is well known, it is one thing to pass a law and quite another to implement it on the ground. Unless the workers in the Labor Ministry and the Employment Agency are convinced that these measures make sense, they will not be properly implemented. There is still a long way to go in establishing that consensus. 5. Regional Wage Flexibility However, mobilizing the unemployed through the Employment Service is not the answer everywhere. It cannot significantly increase employment in those parts of a country where wages are too high relative to productivity. This is the case in eastern Germany, in southern Italy and southern Spain. In those regions, trade unions, often with strong political support, have tried to hold wages close to the levels in the more productive parts of the country. As a result, unit labor costs are higher in the low-productivity regions, giving employers every incentive to generate their vacancies elsewhere. Thus, a mismatch arises between the pattern of vacancies and unemployment, with a disproportionate share of vacancies appearing in the high-productivity regions and a disproportionate share of unemployment in the low-productivity regions. Such a mismatch inevitably shifts the VU relationship in Figure 3 to the right and increases the equilibrium unemployment rate. The only remedy is to lower the relative unit-labor costs in the high-unemployment regions. In principle, this could be achieved either by raising relative productivity, or by lowering relative wages. The problem with the former approach is that employers are simply not willing to make the necessary investment until relative wages are low enough to raise relative rates of return on capital. (This situation is particularly acute in the former East Germany, because it is so uncompetitive compared with the neighbouring East European countries and with the former West Germany.) Thus the burden of adjustment falls on wages. In some cases, there is the possibility of achieving a relative wage adjustment by negotiation, or through the political process. But it is most unlikely to be achieved without some greater decentralization of wages. In Britain, this was achieved in the 1980s, as a result of the weakened power of the trade unions. As a consequence, local wages became more responsive to local unemployment rates, and relative wages in the north of England fell by 12 percent in the 1980s. This had the effect, in the 1990s, of halting the age-old drift of jobs from the north to the south of England. But each country must find its own way towards a more flexible structure of regional wages.

Full Employment for Europe

21

6. Transition Countries The analysis so far has concentrated on the countries of Western Europe; but, in fact, it applies equally to the countries of the former communist bloc. In the early days of the transition, it was widely believed that high unemployment would be a temporary phenomenon, while the workforce was redeployed, and that it would subsequently decline. However a number of writers observed quite quickly that unemployment in those countries could be explained by the same factors as in Western Europe and would be unlikely to change unless those factors also changed (Burda, 1994; Layard and Richter, 1995). Thus, with the exception of Hungary, unemployment in Eastern Europe has remained high (see Table 4). In fact, the redeployment of labor has occurred largely because people moved from one job to another, while inflow rates to unemployment were not particularly high. Therefore, the reason for high unemployment, as in Western Europe, was the low rate of outflow from unemployment. Thus, it can be seen that the lessons discussed earlier in this paper apply equally to the transition countries: high unemployment will only be reduced by having more flexible relative wages across regions, and better policies towards the unemployed. Table 4 Unemployment Rates (percent) in some Transition Countries

Poland Hungary Czech Republic Slovak Republic

1993

2003

16.4 14.5 3.5 14.4

19.2 5.9 7.8 17.4

Source: EBRD Transition Reports.

7. Conclusion Unemployment, and not productivity, is the key economic problem facing Europe. Since it is a major source of misery, it requires urgent attention, but that attention must be based on the evidence of what works. This means reform in two difficult areas: how unemployed people are treated (tough and tender works best); and greater flexibility in wage differentials between regions. It took courage and understanding to tackle inflation, now a thing of the past. In the end politicians in every country will find similar courage to tackle high unemployment in ways that work, so that it too will become a thing of the past.

22

Richard Layard

References Blanchard, O. (2004): “The economic future of Europe.” The Journal of Economic Perspectives 18(4): 3 – 26. Burda, M. (1994): “Unemployment, labour markets and structural changes in Eastern Europe.” Economic Policy 16, 101 – 138. de Koning, J. / Layard, R. / Nickell, S. / Westergaard-Nielsen, N. (2004): “Policies for full employment.” UK Department for Work and Pensions. Online at www.dwp.gov.uk. Frederick, S. / Loewenstein, G. (1999): “Hedonic adaptation.” D. Kahneman, E. Diener and N. Schwarz (eds.), Well-Being: The Foundations of Hedonic Psychology. New York: Russell Sage Foundation, 302 – 329. Helliwell, J. (2003a): “How’s life? Combining individual and national variables to explain subjective well-being.” Economic Modelling 20: 331 – 60. – (2003b): “Well-being and social capital: does suicide pose a puzzle?” University of British Columbia. Mimeo. Layard, R. (2005): Happiness: Lessons from a New Science. New York: Penguin Press. Layard, R. / Nickell, S. / Jackman, R. (1991): Unemployment: Macroeconomic Performance and the Labour Market. Oxford: Oxford University Press. – (2005): Unemployment: Macroeconomic Performance and the Labour Market. 2nd ed. Oxford: Oxford University Press. Layard, R. / Richter, A. (1995): “How much unemployment is needed for restructuring? The Russian experience.” Economics of Transition 3(1): 39 – 58. Loewenstein, G. / Schkade, D. (1999): “Wouldn’t it be nice? Predicting future feelings.” D. Kahneman, E. Diener and N. Schwarz (eds.), Well-Being: The Foundations of Hedonic Psychology. New York: Russell Sage Foundation, 85 – 108. Mitchell, B. (1992): International Historical Statistics: Europe 1750 – 1988. Basingstoke: Macmillan. Nickell, S. / Layard, R. (1999): “Labor market institutions and economic performance.” O. Ashenfelter and D. Card (eds.), Handbook of Labor Economics 3C. Amsterdam: Elsevier, 3029 – 3084. OECD (2001): Labour market policies and the public employment service. Paris: Conference. United Kingdom (2005): National Statistics: Labour Market Trends, Table A.11. HM Stationery Office. March.

Equality and Efficiency: Policy for Globalisation By Justin van de Ven and Martin Weale*

Abstract Policy responses to increasing wage inequality in advanced countries are considered. Protection is argued to be a means of diverting resources to appease particular interest groups at net cost to society. A social security system is shown to damp the effects of increased inequality at the cost of aggravated unemployment. Reforms linking benefits more firmly to work are shown to reduce unemployment and, in the presence of learning from experience, to raise productivity and reduce overall inequality. JEL Classifications: F160, J230, J240 Keywords: Social Security Reform, Employment-linked Benefits, Wage Inequality, Globalisation

1. Introduction Globalisation is a process with many facets, many of which are generally welcomed while others can be a source of considerable economic and political concern. Many people see benefits in falling transport costs and the increased choice which results from rising volumes of international trade. Fischer (2003) regarded globalisation as a challenge mainly for the international economic environment and for developing countries. He identified five key issues. 1. Implementing the right policies, 2. Delivering on Trade and Aid, 3. Making the International Financial System less Crisis-prone, 4. Dealing with Migration, 5. Improving Governance. Although it is obvious that the advanced countries are substantial actors in some of these and the key actors in others, it is noteworthy that the effect of globalisation on the domestic labour market in advanced countries does not feature on his list. * National Institute of Economic and Social Research, 2, Dean Trench Street, London SW1P 3HE. We are most grateful to James Sefton for developing the modelling framework employed here.

24

Justin van de Ven and Martin Weale

Perhaps someone in the United States does not see that any policy questions are raised in the public mind. Nevertheless in Europe and perhaps increasingly in the United States there are considerable concerns about the effects of access to cheap imports resulting in “unfair” competition with established domestic industries and the feeling that, as a result, wages for unskilled labour are being driven down or that unemployment is being driven up. These concerns are understandable. Change always creates uncertainty and a changing economic environment is often unpopular. In England we remember the Luddites who rioted against the advanced technology of the textile industry in the 18th century because they saw it as a threat to their jobs and way of life. In the late 1960s the unionised dock workers in London refused to handle containers because these posed a threat to their continuing employment. It is little surprise that England did not sustain the sort of cottage textile industry the Luddites sought to preserve or that the London docks are no longer of any economic significance. They have been replaced by non-unionised docks further down the Thames and on the East Coast, while the City of London’s satellite, Canary Wharf, stands on one of the former docks. International trade in goods is an alternative to international trade in factors of production. The developed countries are capital rich and labour scarce while the opposite is true of the developing countries such as China and India. We are not willing to accept the movements of labour which would balance out capital / labour ratios internationally but trade in goods has a similar effect on the wage / rental ratio. Factor prices are equalised if both developed and developing countries produce similar goods and provided also that labour efficiency levels are similar in both regions. If labour efficiency is higher in the developed than in the developing region, then the wage / rental rate may also be higher but the effects of globalisation will still be to depress it. This argument can be taken a step further by making a distinction between skilled and unskilled labour, recognising that, at least in the short run, the developing countries have a particular abundance of unskilled labour rather than labour in general. In such circumstances as developing countries start to produce goods which compete with those from developed countries factor-price equalisation is likely to lead to downward pressure on the price of unskilled labour. One might expect the United Kingdom to be more affected than Germany given that the poor quality of intermediate level education in the United Kingdom has been documented for many years. But in any case it is reasonable to expect developing countries to have some influence on wages in the developed world (Freeman, 1995). Developed countries can avoid the problem of factor price equalisation by stopping producing the goods which are also produced in the developing regions. Unskilled labour can nevertheless be employed in the non-traded service sector provided that rates of pay are low enough for there to be a demand for the services which they might provide. With institutional obstacles to low wages, such as the United Kingdom’s minimum wage, it is more likely that low-skilled workers will

Equality and Efficiency: Policy for Globalisation

25

tend to be unemployed. Even without a minimum wage, institutions such as unemployment benefit may set a practical floor to the wage rate and thus limit the employability of low-skill workers. A transatlantic summary of the matter is given by Lucas (1988) who remarked, the question why people choose not to work is not very different from the question why they choose to work for one employer rather than another. If people are paid not to work then some people will choose not to work. It should be noted that, while it is generally accepted that the position of unskilled labour in the advanced countries has worsened in the last twenty years or so, it is by no means clear that globalisation is responsible. Many authors have argued that the responsibility lies predominantly with skill-biased technical change rather than with globalisation per se (see Wood, 2002). But we wish to discuss policies to address the problems arising from the worsening position of unskilled labour and, at least in the context we are considering, it does not matter whether the cause of the worsening is globalisation or skill-biased technical change. I wish to consider first the use of protection to alleviate the symptoms and secondly to consider how, depending on its design, the social security system can be an engine for aggravating rather the mitigating the effects of these sorts of changes for unskilled labour. Our illustrations are with respect to the Untied Kingdom. However, I think the points they raise are relevant to all advanced countries.

2. Protection It is fair to say that there remains in both Europe and in North America a substantial body of opinion which sees protection as a solution to the labour market issues raised by or at least attributed to globalisation; in doing so it ignores the lessons of history. Kindleberger (1973) gives a good account of the tariff war which developed in the early 1930s. Obviously history would not repeat itself exactly. The scope for the new exporting countries such as China to retaliate must be limited by the magnitude of their trade surpluses. Thus it is conceivable that the developed world could achieve a terms of trade gains by means of protection – resulting in a genuine increase in its income. This is the standard optimal tariff argument. But as the world moves to a more balanced pattern of international trade it is hard to see such a situation being sustained and a tariff war on the scale of the early 1930s would probably then re-emerge. The domestic pressures for protection are easy to understand. Although, in the absence of sufficiently favourable terms of trade movements, protection results in a lower overall level of national income as compared to free trade, the StolperSamuleson theorem demonstrates that it raises the price of the factor in which the goods whose import is restricted are intensive. As a corollary it must reduce the relative price of other factors. Some factors gain and others lose. But typically industries seeking protection represent a band of coherent and well-organised

26

Justin van de Ven and Martin Weale

beneficiaries while the losers, although larger in number may not understand that they lose out from such a policy. Thus the argument for protection can be summarised by the Austrian prayer to St Florian. “Lieber heiliger Florian. Beschütz unseres Haus und sünd die andere an.” When tariffs are imposed at least the revenue accrues to the country imposing the tariff and the terms of trade effect means that aggregate domestic income may be raised. However, a particularly insidious form of protection is given by voluntary trade restrictions. Here producers in industrialising countries agree to limit their exports to developed countries. As a result they are able to raise their prices, so the tariff rent, which would otherwise accrue to the importing country is instead garnered by the exporter. This of course makes the arrangement much less controversial than a tariff but ensures that, with no favourable movement in the terms of trade, the protected country always loses out. Many years ago Cable and Weale (1982) looked at this using the Cambridge Growth Project Model. We found that with a tariff depressing imports of textiles and clothing into the United Kingdom by just over 30 % after ten years, the tariff led to real household income being depressed by 0.2 % while the voluntary export restraint led to a fall of 0.7 %. The simulations were carried out with taxes and the exchange rate adjusted to restore budget and external balance. Employment in the textile industry was raised by about 120,000 people in both cases and profits were also increased. So employment was raised to the benefit of the insiders in the industry, but at a cost to the consumer and a reduction in national income. In the last few years we have started to see the benefits of much cheaper clothing becoming available as a result of the end of the Multi-Fibre Agreement and it is perhaps not surprising that this has created difficulties for Italy whose textile industries have turned out not to be very competitive at world prices. It will be an important test for the European Union to resist domestic pressures for renegotiation of voluntary export restraints. Protection can certainly offer short-term relief to industries and employees affected by globalisation but, whether delivered by tariffs or voluntary export restraint, it is hard to see that it is solution to the effects of globalisation in depressing wages of low-skilled workers or leading to their unemployment. Theory suggests that, except in the most favourable circumstances where no retaliation results, it is likely to be little more than the appeasement of sectional interests. Rejecting protection then, the question arises what other solutions are appropriate. How is it possible to maintain, or not to depart too far from economic efficiency while at the same time ensuring that low-skill workers are protected, at least to some extent, from the effects of globalisation on their earning potential? Perhaps an alternative to consider is to explore the role of the social security system and to consider its effectiveness in alleviating the labour market problems which may be caused by globalisation.

Equality and Efficiency: Policy for Globalisation

27

3. Social Security and the Labour Market The basic issue and problem associated with social security is well known at least by economists. Flat rate taxes can be collected and subsidies paid without distorting incentives and without leading to economic inefficiency. If a citizen’s income – as a flat rate social credit is now often known – is paid to people whether they work or not, then although some satisficers will decide that they can afford not to go out to work, the social security arrangement will not offer a strong disincentive to work. But such an arrangement is very expensive and the taxes needed to pay for it are themselves bound to be distortionary. Social credit can be made cheaper by the introduction of a taper – withdrawing benefits from people as their incomes rise. But this is of course equivalent to a high marginal rate of taxation and promotes equality at the expense of economic inefficiency. In the United Kingdom effective tax rates faced by poor people are much higher than those faced by rich people with a significant number paying over 70p in the pound. When the framework for the United Kingdom’s Welfare State was set out (Beveridge, 1942) an attempt was made to keep the cost of the social security system within bounds by limiting them to people whose incomes were low because of unfortunate circumstances – unemployment, illness and old age. But we have since learned that the step from non-employment to employment is also often a matter of choice, recognising that people will be unlikely to take up work after a period of unemployment if their income falls or rises only slightly as a result. Thus since the mid 1980s and in particular in the last ten years there has been a move to make it much harder for people to draw benefits for long periods without working. Britain’s unemployment is now considerably lower than it was twenty years ago despite the pressures on unskilled labour arising from globalisation or skill-biased technical change. Indeed, because we have higher labour force participation income per head is now higher in the United Kingdom than in Germany and quite possibly even than in West Germany. A similar approach has been used elsewhere. Denmark has also succeeded in delivering a low unemployment rate despite apparently generous benefit arrangements. People who draw them are put under considerable pressure to find work so as to offset the disincentive effects of social security. In Germany the Hartz reforms are similar in spirit although Boss and Elendner (2005) doubt whether there will have a substantial effect. Immervoll et al. (2006) shed some light on the importance of linking social security payments to employment using a combination of a simple model of labour supply and the tax and benefit system as represented in country-specific microsimulation models. The difficulty with such ad hoc models is of course well known; they do not represent in any way the intertemporal consumption and labour supply decisions that households face and as a consequence they cannot lead to satisfactory conclusions about the welfare effects of different policies. Neverthe-

28

Justin van de Ven and Martin Weale

less, Immervoll et al. confirm the superiority of an arrangement like the United Kingdom’s working family tax credit over one where income rather than employment status is the main determinant of benefit entitlement. The effect of benefit withdrawal as a disincentive to work is much reduced by the arrangement. But the analysis provided by Immervoll and his colleagues is unlikely to be the whole story. If one considers the labour / leisure choice of an individual over the whole life cycle behaviour seems, on the face of it, most peculiar. Most men take most of their leisure late in life (e.g. they retire from the work-force) – although perhaps the fashion for gap years after graduation is indicative that the incentive to work is also fraying at the start of working life. While many women take career breaks in order to look after young children, they too tend to return to work, often full time after child care becomes easier, and then concentrate leisure late in life. Unless the earning power of old people is much below that of young people, economic theory suggests that people would prefer to take their leisure spread more evenly over their lives. Why don’t people work three days a week up to age eighty instead of five days a week up to age sixty or sixty-five and then not at all beyond that? At the National Institute we have been working on a modelling framework to explore the two key marginal decisions in economics, work / leisure and consumption / saving (Sefton, van de Ven and Weale, 2005a, b). We simulate the behaviour of a synthetic panel of households through the life-cycle. Each household faces uncertain income and an uncertain life-span, and makes its consumption and labour supply decisions optimally in the face of this. Our work so far has focused on pension policy and its interaction with the retirement decision. But the model can also be used to explore labour supply decisions of younger people and their interaction with social security. The approach has the attraction of being robust to the Lucas critique; economic behaviour is a direct consequence of optimising decisions. Our work suggests that, although intertemporal substitution does take place, the observed profile of rates of pay is not in itself adequate to explain the pattern of labour supply over the life cycle. On the other hand if, as Mincer (1974) and Weiss (1986) suggest, people’s earning power increases with experience, then the advantages of concentrating labour supply become very apparent; we find a degree of learning from experience is necessary to calibrate our model in a manner consistent with consensus views about intertemporal substitution. If this is the case, then the influence of the social security system on the incentive to supply labour will interact with the effects of experience to generate the distribution of individual earning power. A social security system which discourages work or encourages only casual and perhaps intermittent black economy work will be likely to lead to a distribution of market wages which is more unequal than that arising in an economy where “the caring hand” does not cripple, to use Herr Snower’s phrase. In much the same way, a social security system which dis-

Equality and Efficiency: Policy for Globalisation

29

courages retirement saving increases wealth inequality – because it discourages saving by relatively poor people (Sefton, Dutta and Weale, 1999). The social security system may also have a depressing effect on “true” productivity, i.e., the productivity which would be observed if everyone chose to work; despite this, it may raise observed productivity by filtering out of the labour force the least productive members. Obviously these effects are likely to be accentuated if a poorly-designed social security system interacts with an exogenous widening of the inequality of earning power, whether as a result of skill-biased technical progress or because of the factor price equalisation pressures arising from globalisation on the incomes of low earners. We use the National Institute model to indicate the effects of increased inequality and to show how they interact with the social security system. We identify how social security alleviates and aggravates inequality at the same time. Inequality in our model is driven by the initial distribution of earning power (human capital) at age twenty, when people are assumed to enter the labour force, and the dynamics of the income process after that. We represent the increased inequality by raising the standard deviation of log earning power at age twenty from 0.4 to 0.5, adjusting the mean value so as to keep mean earning power unchanged. The Gini coefficient is raised from 0.22 to 0.27. The impact of this change on the distribution of earning power at age twenty is shown in figure 1. In the stochastic income process which we use to represent the evolution of earning power (by which we mean the amount people would earn gross of taxes but before any social security receipts if fully employed), there is only a small degree of mean reversion, so this increased earning inequality persists throughout working life. The social security system as we have modelled it, and this is meant to be a representation of the UK system as far as is possible, does not protect poor people entirely from the increased inequality. Nevertheless the fall in average incomes after taxes and benefits of people in the bottom quintile of life-time incomes when aged 30 – 39 is only 51/2 % compared to a fall in earning power of 13 %. By age 55 – 59 both earning power and income fall by 6 %. Figure 2 shows these changes which result from greater inequality such as might result from globalisation or skill-biased technical progress. With no change to the social security system non-employment among the group in the lowest quintile of life-time earnings increases sharply, from 19 % to 28 % at age 30 – 39. The proportion of people aged 55 – 59 in the same income group who are retired increases from 46 % to 51 %. There is also an increase in the proportion of retired people in the top quintile, from 27 % to 28 % reflecting the fact that, with increased inequality and the same mean income, their incomes have risen. This raises the number of people who feel that they can afford to retire. This is shown in figure 3.

30

Justin van de Ven and Martin Weale

% of Population

45 40 35 30 25 20 15 10 5 0 0

50

100 150 200 250 300 350 400 450 500 £ per week

Figure 1: Inequality at Age 20 Before and After an Increase in Inequality

280

£ per week

260 240 220 200 180 160 140 120 100 Age 30-39 Earning Power

Age 30-39 Income

Age 55-59 Earning Power

Age 55-59 Income

Figure 2: Social Security Mitigages the Impact of Rising Inequality: Incomes of the Lowest Life-time Quintile

What might one hope for from reform of social security? As I mentioned earlier a key policy in the United Kingdom has been to link benefit receipts to work. We represent a policy of this type by reducing the benefits paid to people who do not work by £15 per week. This matches the reduced earning power of the twenty-year olds in the bottom income quintile. Obviously it considerably increases the incentive to work because the benefits accruing to people who do have jobs are not changed. The proportion of the population not working in the bottom quintile aged 30 – 39 falls to 6 % instead of rising to 28 %. And only 27 % of the population aged

Equality and Efficiency: Policy for Globalisation

31

55 – 59 retire instead of 51 % with the social security system unchanged. This is also shown in figure 3.

Non-employment Rate

60% 50% 40% 30% 20% 10% 0% Age 30-39

Age 55-59

Figure 3: The Effect of Social Security on Labour Force Participation by the Lowest Life-time Income Quintile

£ per week

Figure 4 shows the effect of social security reform on the income and earning power of people in the same life-time income quintile. Because incentives to work are stronger the population of low earners has higher earning power. Both the incomes of those in the bottom quintile and, at least for people aged 55 – 59, their earning power, is actually higher than before the “increase” in inequality. As a consequence the mean income of the bottom life-time income quintile rises as a proportion of the overall mean income in the age group and, as a result income inequality is reduced. 280 260 240 220 200 180 160 140 120 100 Age 30-39 Age 30-39 Age 55-59 Age 55-59 Income Income Earning Earning Power Power

Figure 4: The Effect of Employment-linked Benefits on Earning Power and Income of the Lowest Life-time Income Quintile

32

Justin van de Ven and Martin Weale

Figure 5 explores the effect of social security reform on average productivity of people aged 30 – 39 and 55 – 59. While the increase in inequality reduces this, a greater-focus on employment-linked social security mitigates the fall.

Change compared to Current Position

Obviously, if the results are believed, such a reform is desirable with or without the effects of globalisation, but the results do demonstrate that we should look at our own institutions, as well as exogenous factors, in influencing inequality.

0.0% -0.2% -0.4% -0.6% -0.8% -1.0% -1.2% -1.4% -1.6% -1.8% Age 30-39

Age 50-59

Figure 5: Employment-linked Benefits Raise Productivity of the Lowest Life-time Income Quintile

4. Conclusions Of course it can be objected that these results are nothing more than simulations from our model and the world may not turn out to be like that. The fact that the model is calibrated to the data we have is no real protection against that. And it is also true that a reform as stark as that described would not be easy to implement. We should also stress that we are not in any sense claiming that the social security system is the only route by which “helpful” policies can be a source of unemployment. Snower and Merkl (2006) show that it can interact with employment protection legislation to create an even more damaging mix – a finding which is easy to accept. Nevertheless we hope to have demonstrated how, in a framework which most economists regard as plausible, the social security system interacts with changes to inequality such as might arise from globalisation or skill-biased technical change and that some of the possible adverse effects of globalisation on the labour market can be mitigated by reviewing the working of the social security system. The tide of globalisation cannot be turned back but economists can certainly advise on how policy changes may affect its consequences.

Equality and Efficiency: Policy for Globalisation

33

References Beveridge, W. H. (1942): Social Insurance and Allied Services. Cmnd 6404. HMSO. Boss, A. / Elendner, T. (2005): “Incentives to Work: the Case of Germany”. Kieler arbeitspapier Nr 1237. http: // www.uni-kiel.de / ifw / pub / kap / 2005 / kap1237.pdf. Cable, V. / Weale, M. R. (1982): “Trade and Aid Policy Analysis: Use of the Cambridge Growth Project Model”. ODI Review, pp 50 – 70. Fischer, S. (2003): “Globalisation and its Challenges”. American Economic Review. Vol 93. No 2, pp 1 – 30. Freeman, R. B. (1995): “Are your Wages set in Beijing?”. Journal of Economic Perspectives. Vol 9, pp 15 – 32. Immervoll, H. / Kleven, H. J. / Kreiner, C. T. / Saez, E. (2006): “Welfare Reform in Europe: A Micro-simulation Analysis”, Economic Journal. Forthcoming. Kindleberger, C. P. (1973): The World in Depression. 1929 – 1939. Allen Lane, London Mincer, J. (1974): Schooling, Experience and Earnings. NBER New York. Sefton, J. / Dutta, J. / Weale, M. R. (1998): “Pension Finance in a Calibrated Model of Saving and Income Distirbution for the UK”. National Institute Economic Review. No 166, pp 97 – 107. Sefton, J. / van de Ven, J. / Weale, M. R. (2005a): “Means Testing and Retirement Choices in Europe: a Comparison of the British and Danish Systems” Fiscal Studies. Vol 26, pp 83 – 118. – (2005b): “The Effects of Means-testing Pensions on Savings and Retirement”. NIESR Discussion Paper 265. http: // www.niesr.ac.uk / pubs / dps / dp265.pdf. Snower, D. J. / Merkl, C. (2006: “The Caring Hand the Cripples: The East German Labour Market after Re-unification”. Kiel Working Paper No 1263. Willis, R. J. (1986): “Wage Determinants: a Survey and Re-interpretation of Human Capital Earnings Functions”. In Handbook of Labour Economics. Vol 1. O. Ashenfelter and R. Layard (eds.). Wood, A. (2002): “Globalization and Wage Inequalities: a Synthesis of Three Theories”. Weltwirtschafliches Archiv. Band 138, pp 54 – 82.

3 Supplement 57 – 2006

The Effect of Globalization on Aggregate Labour Demand in EU countries By Christian Dreger* and Stefan Kooths** Abstract Due to advances in economic integration especially in Europe, labour markets need to be more flexible to ensure a smooth transmission of shocks. A higher degree of flexibility might cause a shift in labour demand. For example, a rise in the output and real wage elasticities in absolute value would indicate a stronger response to the changing economic environment. Whether or not aggregated labour demand has become increasingly unstable in recent years, is explored in this paper. Several interesting conclusions are derived. Cointegration can be established between the variables entering a standard labour demand function. The long run vector is in line with economic theory and similar across EU countries. The error correction models are broadly robust to a wide range of specification tests. There is no reliable evidence that globalization indicators such as trade openness has affected the cointegration parameters. Hence, the presumption that aggregated labour demand has become more fragile in the course of globalization is not supported by the data. JEL Classifications: C22, E24 Keywords: Labour demand, trade openness, globalization

1. Introduction The globalization of factor and product markets has increased rapidly in recent years. Especially in Europe, the speed of integration has accelerated due to the introduction of the euro area and the EU eastern enlargement. Advances in the integration improves competition on product and labour markets and allow economic agents, private households and firms, to undertake price arbitrage. Because prices are forced to converge to marginal costs, factors and products are used in a more competitive and productive way. As incentives to research and innovate are stimulated, a higher path of economic growth could be achieved in the globalized environment. * Corresponding author. German Institute for Economic Research (DIW Berlin) and Institute for the Study of Labour (IZA Bonn), email: [email protected]. ** German Institute for Economic Research (DIW Berlin), email: [email protected]. The authors thank Herbert Buscher and the other participants of the ARGE conference for their helpful comments and suggestions. The usual disclaimer applies. 3*

36

Christian Dreger and Stefan Kooths

In the process of economic integration, national economic policies have become less important. Monetary policy or nominal exchange rate adjustments are no longer available for euro area member states, and fiscal policies have to take international commitments into account, outlined for example in the Stability and Growth Pact. The currency union not only requires a high degree of labour market flexibility, but serves also as a promoter in this regard (Frankel and Rose, 1998). According to the new economic geography, trade integration may contribute to higher specialisation pattern of countries, implying that the probability for asymmetric (country specific) shocks can increase (Krugman, 1991, Fujita, Krugman and Venables, 1999). In order to prevent long lasting disequilibria, labour markets play a more vital role to ensure a smooth transmission of shocks. Flexibility of labour markets can be measured in several ways, see Nickell, Layard and Jackman (1991) and Pissarides (1998). For example, specific indicators such as the long term unemployment rate can provide insights. The shorter the average duration of unemployment, the less persistent the impacts of economic shocks. In fact, the long term unemployment rates have decreased over the second half of the 1990s for the first time in most EU countries possibly due to the implementation of labour market reforms, with the notable exception of Germany. Generally, more market friendly institutions can improve the adjustment of labour markets in response to shocks. Blanchard and Wolfers (2000) have emphasized the relevance of the institutional setting to explain the labour market performance in Europe. Policies that aim to increase job security or reduce the losses from unemployment lead to less flexible labour markets (Ljungqvist and Sargent, 1998). According to Blanchard and Giavazzi (2003) an increase in product market competition lowers the rents available for redistribution in wage bargaining. Declining rents in product market rents limit the scope of distribution of incomes between firms and workers. As bargained wages tend to be more moderate, output and employment growth will be encouraged. A similar point can be made for trade integration. In general, the globalization of product and labour markets can induce a rise in labour demand elasticities in absolute value through a scale effect because of the increased competition, and via a substitution effect caused by expanding the firms production possibilities to cover additional inputs, see Hammermesh (1993) and Rauch and Trindade (2000). In this sense, trade integration might have affected the long run elasticities of employment to output and the real wage. Being the short side of the market, labour demand is crucial to determine the employment performance in the EU economies, see Bean (1994). Whether or not the relationship has become increasingly unstable in recent years, is the topic of this study. There is rather clear-cut evidence that globalization has crowded out the demand for low skilled labour in high wage countries, see for example Wood (1995), Greenaway and Nelson (2000) and Jean (2000). In contrast, this paper examines the effect on aggregated labour demand. As the variables in the labour demand function are usually nonstationary, a cointegration analysis is the appropriate way to proceed. Several interesting findings

The Effect of Globalization on Aggregate Labour Demand

37

are provided by the analysis. Cointegration can be established between the variables entering a standard labour demand function. The long run vector is in line with economic theory and similar across EU countries. The error correction models are broadly robust to a wide range of specification tests. There is no reliable evidence that globalization indicators such as trade openness has affected the cointegration parameters so far. Hence, the presumption that aggregated labour demand has become more fragile in the course of globalization is not supported by the data. The paper is organized as follows. First, a standard labor demand equation is obtained from cost minimization behaviour of firms (section 2). In section 3, the data used in the analysis are discussed. Robust evidence with respect to the cointegration property and the error correction adjustment is provided. The impact of trade openness on the long run elasticities is examined in section 4. Finally, section 5 offers a conclusion. 2. Labour Demand Behaviour According to standard economic theory, profit maximizing firms are faced by output demand and factor prices, which are both exogeneous to the firms decision, see Hammermesh (1993). As a result of duality, optimal behavior can be inferred from analysis of the cost function. Cost minimizing labor and capital input quantities are dependent on output demand and obtained by taking the partial derivatives of the total cost function with respect to the prices factor (Shephard’s lemma). Using a log linear approximation, labor demand L can be stated as …1†

Lt ˆ 0 ‡ 1 Yt ‡ 2 wt ; 1 > 0; 2 < 0

where Y is output, and w the real wage. Prices are measured in real terms, implying that the output price moves in tandem with nominal factor prices. The parameters 1 and 2 denote the elasticities of labor demand to output and real wages, respectively. Higher demand for goods will raise labour input, while an increase in real wages is expected to have a dampening effect. Because of imperfections such as institutional or cost restrictions, adjustment to economic conditions may not be instanteneous. Actual employment only partially reacts …2†


Lt ˆ …Lt

Lt 1 †  2 …0; 1†

1

at the level desired by firms. The higher the degree of persistence , the lower the employment response in the short run. By substituting the labour demand function into (2), an error correction mechanism …3†


Lt ˆ

…Lt

1

0

1 Yt

1

2 wt 1 †

is implied, which can be enhanced by a more complex dynamic structure. Note that the long run equilibrium (
L ˆ 0) can be already inferred from analysis of (1).

38

Christian Dreger and Stefan Kooths

After replacing labor demand by actual employment, this equation can be interpreted as a cointegration relationship.

3. Cointegration and Error Correction Analysis Data for the EU14 countries are annual and cover the 1973 – 2004 period. Due to data availability problems, the new EU member states and Greece have been excluded from the analysis. Starting with the year of the first oil crisis (1973), unemployment has become the major economic problem in the EU countries. All series have been taken from the AMECO database provided by the EU Commission. In particular, employment refer to the number of persons employed in the total economy, output is GDP at 1995 prices, and nominal wages are proxied by compensation per employee. To obtain the series at constant prices, nominal wages have been deflated using the GDP deflator (1995 = 100). Openness of the economy is considered as a proxy for advances in the process of globalization. It is conventionally measured as the share of foreign trade, i.e. the ratio of the exports plus imports to overall GDP, see Edwards (1998). Exports and imports include goods and services and are deflated by the export and import price deflators, respectively (1995 = 100). Table 1 ADF Statistics for Variables Entering Labour Demand

Austria Belgium Denmark Finland France Germany Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK

Employment

Output

Real wage

–2.95 / –3.23 –1.19 / –4.13 –3.13 / –3.45 –3.69 / –3.37 –2.49 / –2.91 –2.20 / –3.12 –1.20 / –2.11 –2.51 / –3.19 –1.61 / –2.26 –1.67 / –5.02 –2.51 / –3.82 –2.09 / –2.16 –2.99 / –2.57 –3.71 / –4.24

–2.20 / –3.76 –3.76 / –4.15 –4.22 / –3.42 -5.50 / -5.98 –1.17 / –4.46 –2.57 / –3.91 –1.70 / –3.58 –1.67 / –4.95 –3.75 / –3.35 –1.89 / –3.40 –0.77 / –2.45 –1.80 / –5.19 –3.43 / –5.74 –2.96 / –3.90

–1.72 / –5.25 –2.65 / –4.04 –1.73 / –4.39 –3.16 / -4.57 –1.19 / –4.09 –0.87 / –3.57 –4.64 / –3.43 –1.91 / –3.83 –2.08 / –7.08 –0.04 / –3.71 –0.77 / –2.00 –0.95 / –3.88 –4.51 / –6.44 –1.99 / –4.16

Note: Sample period 1973 – 2004. The lag length is determined by the Schwarz criterion, where a maximum lag of 4 years is allowed. The left (right) hand side entry denote the ADF statistic in the level (first difference) model. The regressions are specified with a constant and linear time trend for the variables in levels, and with a constant for the first differences. The 5 percent critical values are –3.59 for the level and –2.97 for the first difference models. An impulse dummy (1991 = 1) is present in the German regressions due to the unification.

The Effect of Globalization on Aggregate Labour Demand

39

The integration properties of the time series involved are investigated by the ADF test, see table 1. The lag length of the regressions is determined by the Schwarz criterion. The variables involved are usually integrated of order 1 in their level representation, but stationary in first differences. There are some exceptions from this rule: for example, output appears to be trend stationary especially in a number of smaller economies, like Denmark and Finland, and a second unit root might be present in the employment series in four cases (Ireland, Luxembourg, Spain and Sweden). However, the evidence is rather unique especially for the large EU members, including Germany, France, Italy and the UK. Overall, labour demand need to be examined by means of a cointegration analysis, and the Johansen (1995) trace statistics for the entire sample and subperiods are reported in table 2. The lag length for the VARs is determined by the Schwarz criterion, where a maximum lag of 4 years has been fixed. All models include an unrestricted constant to allow for deterministic trends in the level representation. Cointegration can be detected for all countries considered. Note that the evidence does not emerge due to a particular choice of the sample. In fact, the cointegration finding is robust even if subperiods are considered. Hence, the increase in globalization in recent years did not distort the existence of a long run relationship between the variables entering a conventional labour demand function. Table 2 Trace Statistics for a Standard Labour Demand Relationship

Austria Belgium Denmark Finland France Germany Ireland Italy Luxembourg Netherlands Portugal Spain Sweden UK

Lag

1973 – 2004

1973 – 1990

1973 – 1995

1973 – 2000

2 1 3 2 1 1 2 1 1 3 2 1 4 1

35.79 (0.009) 28.69 (0.067) 30.65 (0.040) 33.04 (0.020) 59.21 (0.000) 35.21 (0.011) 31.81 (0.029) 53.91 (0.000) 58.79 (0.000) 33.44 (0.018) 47.38 (0.000) 32.31 (0.025) 41.76 (0.001) 41.69 (0.001)

39.12 (0.003) 30.75 (0.039) 48.15 (0.000) 20.56 (0.386) 56.63 (0.000) 33.34 (0.019) 36.77 (0.007) 37.54 (0.005) 41.55 (0.001) 45.34 (0.000) 54.37 (0.000) 37.76 (0.005) 44.29 (0.001) 42.04 (0.001)

33.62 (0.017) 33.45 (0.018) 42.54 (0.001) 54.16 (0.000) 58.80 (0.000) 36.95 (0.006) 32.83 (0.022) 34.90 (0.012) 43.79 (0.001) 31.57 (0.031) 64.96 (0.000) 32.76 (0.022) 52.66 (0.000) 38.43 (0.004)

31.11 (0.035) 28.41 (0.072) 33.02 (0.021) 33.75 (0.017) 64.44 (0.000) 34.82 (0.012) 29.23 (0.058) 46.17 (0.000) 52.02 (0.000) 30.47 (0.042) 53.50 (0.000) 32.05 (0.027) 60.75 (0.000) 44.82 (0.001)

Note: Entire sample period 1973 – 2004 and several subperiods. The lag length for the VAR is determined for the whole sample period by the Schwarz criterion, where a maximum lag of 4 years is allowed. The models are estimated with an unrestricted constant. An impulse dummy (1991 = 1) is present in the German models due to the unification. Numbers behind the trace statistics denote p-values, according to MacKinnon, Haug and Michelis (1999).

40

Christian Dreger and Stefan Kooths

In principle, the cointegration vector can also be derived from the Johansen procedure. However, apart from multicollinearity problems and imprecise estimates in a VAR setting, the long run relationships and feedback mechanisms cannot be uniquely identified. Furthermore, parameter estimates in one equation might be affected by misspecifications in other equations. For these reasons, the long run vector is estimated using a single equation approach, see Hendry (1995). In particular, the Stock and Watson (1993) dynamic OLS method is applied, where the long run is estimated jointly with the short run dynamics in juse step. As a by-product, this strategy allows a further robustification of the cointegration evidence by the Banerjee, Dolado and Mestre (1998) test. Here, the significance of the feedback parameter is examined. At the initial stage of the estimation process, the contemporaneous and the first two lags of the changes of all variables and a constant are included in addition to the error correction term. Then, the variables with the lowest and insignificant t-values have been eliminated successively. The final prefered models are shown in table 3. The negative coefficient of the feedback coefficient is significant in almost all cases, according to the Banerjee, Dolado and Mestre critical values –3.57 (0.05),
–3.20 (0.10) . Hence the cointegration result is confirmed and employment reacts in response to shocks to restore the long run equilibrium. The variables in the long run vector are well signed. The hypothesis that the long run output and real wage elasticity are equal and of opposite sign, cannot be rejected for the vast majority of countries, including all large EU member states. The results even imply that a standard Cobb-Douglas technology might be a suitable approximation for the production process in the EU economy. In that case, the output and real wage elasticity are equal to 1 in absolute value. For example, the long run vector for Germany is Lt ˆ

0:29 Yt 0:23

0:20 wt ˆ 1:26Yt 0:23

0:87wt

where constants have been omitted. Given the range of uncertainty around the point estimates of the long run parameters, the cointegrating relationship is also consistent with the Cobb-Douglas elasticities. In addition to the error correction term, the employment dynamics are determined by the contemporaneous growth rates of GDP and real wages. Moreover, lagged employment changes appear to be significant rather often. With the exception of Portugal, the empirical fit of the equations is quite high. Standard specification tests are largely supportive for the error correction models, see table 3. LM is a Lagrange Multiplier test for autocorrelation in the residuals up to order 1 and 2. The p-values show, that usually no problems with autocorrelated residuals occur. ARCH is a Lagrange multiplier test for conditional heteroskedasticity. Again, the residuals do not exhibit such kind of behaviour. Furthermore, they are distributed as normal, as indicated by the Jarque-Bera test. For three countries (Germany, Netherlands and Sweden) the RESET test indicates some misspecification of the model. Hence, the linear specification could be eventually improved.

1.76 (0.43)

4.33 (1.37)

2.77 (0.63)

1.93 (0.50)

1.37 (0.47)

Spain

Sweden

UK

2.59 (0.89)

Ireland

Portugal

0.94 (0.29)

Germany

Netherlands

1.01 (0.05)

France

2.35 (0.65)

2.39 (0.41)

Finland

1.12 (0.21)

2.11 (0.70)

Denmark

Luxembourg

1.52 (0.37)

Belgium

Italy

2.93 (0.57)

Austria

Constant

–0.23 (0.06)

–0.25 (0.06)

-0.43 (0.08)

–0.54 (0.16)

–0.26 (0.07)

–0.19 (0.04)

–0.28 (0.07)

–0.38 (0.12)

–0.23 (0.07)

–0.15 (0.05)

–0.33 (0.06)

–0.30 (0.10)

–0.23 (0.06)

–0.42 (0.08)

L… 1†

0.27 (0.06)

0.08 (0.04)

0.40 (0.08)

0.12 (0.09)

0.23 (0.08)

0.08 (0.01)

0.15 (0.03)

0.25 (0.08)

0.29 (0.10)

0.13 (0.04)

0.31 (0.06)

0.15 (0.05)

0.18 (0.04)

0.17 (0.04)

Y… 1†

–0.29 (0.06)

–0.09 (0.05)

–0.43 (0.08)

–0.19 (0.11)

–0.23 (0.07)

–0.10 (0.02)

–0.18 (0.04)

–0.28 (0.10)

–0.20 (0.07)

–0.16 (0.04)

–0.39 (0.07)

–0.14 (0.07)

–0.17 (0.03)

–0.09 (0.03)

W… 1†

0.49 (0.07)

0.32 (0.07)

0.51 (0.11)

0.13 (0.09)

0.27 (0.08)

0.06 (0.02)

0.31 (0.09)

0.49 (0.08)

0.39 (0.05)

0.32 (0.04)

0.36 (0.07)

0.26 (0.04)

0.18 (0.03)


…Y†




0.04 (0.02)

0.04 (0.02)


y… j†

Table 3: Error-correction Models for Labour Demand

–0.46 (0.07)

–0.30 (0.08)

–0.48 (0.10)

-0.24 (0.12)

–0.12 (0.04)

–0.33 (0.11)

–0.33 (0.10)

–0.35 (0.08)

–0.44 (0.05)

–0.28 (0.07)

–0.25 (0.04)


…W†




0.06 (0.03)

0.07 (0.03)


W… j†

The Effect of Globalization on Aggregate Labour Demand 41

0.80

0.93

0.76

0.62 (0.22)

Netherlands

Portugal

0.74

0.76 (0.12)

0.30 (0.11)

Sweden

UK

0.31 (0.58)

0.00 (0.98)

2.81 (0.11)

0.63 (0.44)

0.02 (0.88)

2.27 (0.14)

0.32 (0.58)

0.08 (0.78)

2.31 (0.14)

3.29 (0.08)

2.73 (0.11)

0.53 (0.48)

0.02 (0.88)

0.02 (0.89)

LM…1†

0.58 (0.57)

2.35 (0.12)

1.36 (0.28)

1.38 (0.28)

0.03 (0.97)

1.16 (0.33)

1.28 (0.30)

0.05 (0.95)

1.19 (0.32)

2.36 (0.12)

1.40 (0.27)

0.53 (0.48)

0.06 (0.94)

0.89 (0.43)

LM…2†

0.04 (0.84)

0.81 (0.38)

0.02 (0.90)

2.95 (0.10)

0.20 (0.66)

1.97 (0.17)

1.84 (0.19)

0.46 (0.50)

0.41 (0.53)

0.07 (0.25)

0.01 (0.93)

0.47 (0.63)

0.64 (0.43)

0.12 (0.74)

ARCH…1†

0.12 (0.88)

1.52 (0.24)

0.06 (0.94)

1.50 (0.24)

0.41 (0.67)

5.20 (0.02)

0.75 (0.48)

0.27 (0.77)

0.89 (0.42)

0.07 (0.79)

0.38 (0.69)

0.00 (0.98)

0.61 (0.55)

1.37 (0.27)

ARCH…2†

0.52 (0.77)

1.37 (0.50)

4.83 (0.09)

0.13 (0.94)

0.01 (0.99)

1.22 (0.54)

0.86 (0.65)

0.29 (0.87)

1.66 (0.44)

0.87 (0.65)

0.79 (0.67)

0.80 (0.67)

1.05 (0.55)

0.90 (0.64)

JB

0.04 (0.83)

5.77 (0.03)

1.73 (0.20)

0.40 (0.54)

7.25 (0.02)

1.13 (0.30)

5.46 (0.29)

0.11 (0.74)

7.47 (0.01)

0.80 (0.38)

1.23 (0.28)

2.51 (0.13)

1.44 (0.24)

0.73 (0.40)

RESET…1†

Note: Sample period 1973 – 2004. Numbers in parantheses after coefficient estimates denote standard errors. R2 ˆ R squared adjusted, LM = Lagrange multiplier test for no autocorrelation in the residuals, ARCH = Lagrange multiplier test against conditional heteroscedasticity, JB = Jarque Bera test, RESET = Ramsey test. Numbers in parantheses after test statistics denote p-values.

0.85

0.84

Spain

0.30

0.73

0.80

0.63 (0.10)

Luxembourg

0.61

Italy

0.74

0.43 (0.11)

France

0.45 (0.14)

0.17 (0.11)

Finland

Ireland

0.46 (0.12)

Denmark

0.82

0.80

0.98

0.29 (0.11)

R2

Germany

0.43 (0.10)




Belgium


L… j†

Austria

Continued Table 3

42 Christian Dreger and Stefan Kooths

The Effect of Globalization on Aggregate Labour Demand

43

4. Trade Openness and Long Run Elasticities Although the empirical fit of standard labour demand equations is quite satisfactory, the increased openness to foreign trade might have affected the long run elasticities. In particular, the elasticities in absolute value are expected to increase in the process of globalization. As a degrees of freedom problem is apparent in a country specific analysis, the proposition is tested efficiently within a panel setting. Do develop the panel structure, recursive estimates of the long run output and real wage elasticities have been obtained by country specific error correction models described in table 3. Initially, all the equations are run over the 1973 – 1990 period, and the first set of elasticities (corresponding to 1990) is computed. The period is then prolonged by 1, the equations are re-estimated and this process is repeated until the end of the sample is reached. For illustration purposes, time varying output and real wage elasticities are displayed in figure 1 for some EU member states. Although there might be a slight upward trend in absolute value, the changes are hardly significant. The time series of the elasticities are treated as endogeneous in the panel framework. In addition to the country specific fixed effects, the only explanatory variable is openness to foreign trade, specified either in levels or first differences. As openness might act with some delay, the first lag is allowed to enter the regressions. The prefered specifications are exhibited in table 4. Globalization appears to be significant in the level models. But compared to the theoretical arguments outlined above, its sign is reversed. Both elasticities are expected to decrease in absolute value in response to a raise in integration, implying that employment will be less affected by the economic conditions. For example, if trade integration, i.e. the ratio of the sum of exports and imports to GDP increases by 1 percentage point, the elasticities will fall by approximately 0.002 points in absolute value. This may point to the relevance of political measures aiming to hamper a faster adjustment in labour markets. On the other hand, the result should not be overvalued. The evidence is supported only in the level, but not in the first difference specification, where globalization is insignificant. More important, the elasticities might indeed be constant, as fluctuations in these measures are largely artificial. Overall, the presumption that aggregated labour demand has become more fragile in the course of globalization is not supported by the data. The findings are broadly in line with a number of other studies, which have been conducted on the sectoral level, like Slaughter (2001) for the US economy and Bruno, Falzoni and Helg (2004) for a number of OECD economies. In this respect, further analysis of the effects of globalization on labour markets is clearly required.

44

Christian Dreger and Stefan Kooths A Output Elasticity

1,6 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 France

Germany

Italy

United Kingdom

B Real Wage Elasticity

0,0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 -0,2 -0,4 -0,6 -0,8 -1,0 -1,2 -1,4 -1,6 France

Germany

Italy

United Kingdom

Note: Recursive point estimates of output and real wage elasticities using the error correction models in table 3. Due to a structural break problem, the series for France has been smoothed over the 1999 – 2002 period, where the elasticities have been replaced by averages.

Figure 1: Time Varying Elasticities of Employment in Large EU Countries

The Effect of Globalization on Aggregate Labour Demand

45

Table 4 The Impact on Trade Integration on Labour Demand Elasticities A Output Elasticity

Constant TO… 1†
TO R2

Levels

First differences

0.89 (0.06) –0.23 (0.07)

0.68 (0.01) 0.35 (0.27) 0.82

0.83

B Real Wage Elasticity

Constant TO… 1†
TO R2

Levels

First differences

–0.90 (0.08) 0.23 (0.09)

–0.69 (0.02)

0.79

–0.39 (0.34) 0.79

Note: Sample period 1990 – 2004, panel regressions. Endogeneous: Output or real wage elasticities, recursively estimated by one step error correction models, exogeneous: openness to foreign trade (TO), either measured in levels or first differences. The constant refers to the mean of the country-specific fixed effcts. Standard errors in parantheses. R2 ˆ R squared adjusted.

5. Conclusions In this paper it is explored whether globalization have had an impact on macroeconomic labour demand behaviour so far. Several interesting conclusions are derived. Cointegration can be established between the variables entering a standard labour demand function. The long run vector is in line with economic theory and similar across EU countries. The error correction models are broadly robust to a wide range of specification tests. There is no reliable evidence that globalization indicators such as trade openness has affected the cointegration parameters. Hence, the presumption that aggregated labour demand has become more fragile in the course of globalization is not supported by the data.

References Banerjee, A. / Dolado, J. J. / Mestre, R. (1998): “Error correction mechanism for cointegration in a single equation framework,” Journal of time series analysis 19, 267 – 283. Bean, C. R. (1994): “European unemployment. A survey,” Journal of the Economic Literature 32, 573 – 619.

46

Christian Dreger and Stefan Kooths

Blanchard, O. / Giavazzi, F. (2003): “Macroeconomic effects of regulation and Deregulation in Goods and Labor Markets,” Quarterly Journal of Economics 118, 879 – 907. Blanchard, O. / Wolfers, J. (2000): “The role of shocks and institutions in the rise of European unemployment: the aggregate evidence,” Economic Journal 110, Conference Papers, C1 – C33. Bruno, G. S. F. / Falzoni, A. M. / Helg, R. (2004): “Measuring the effect of globalization on labour demand elasticity. An empirical application to OECD Countries,” Bocconi University Milan, Working Paper 153. Davidson, C. / Matusz, S. J. (2000): “Globalization and labour market adjustment: How fast and at what cost?,” Oxford Review of Economic Policy 16, 42 – 56. Edwards, S. (1998): “Openness, productivity and growth. What do we really know?,” Economic Journal 108, 383 – 398. Frankel, J. / Rose, A. (1998): “The endogeneity of the optimum currency area criteria”, Economic Journal 108, 1009 – 1025. Fujita, M. / Krugman, P. R. / Venables, A. J. (1999): “The spatial economy,” MIT Press, New York. Greenaway, D. / Nelson, D. (2000): “The assessment. Globalization and labour market adjustment,” Oxford Review of Economic Policy 16, 1 – 11. Hammermesh, D. S. (1993): “Labour demand,” Princeton, Princeton University Press. Hendry, D. F. (1995): “Dynamic Econometrics,” Oxford, Oxford University Press. Jean, S. (2000): “The effect of international trade on labour-demand elasticities: intersectoral matters,” Review of International Economics 8, 504 – 516. Johansen, S. (1995): “Likelihood based inference in cointegrated vector autoregressive models,” Oxford University Press, Oxford. Krugman, P. R. (1991): “Geography and Trade”, MIT Press, New York. Layard, R. / Nickell, S. / Jackman, R. (1991): “Unemployment. Macroeconomic performance and the labour market,” Oxford University Press, New York. Ljungqvist, L. / Sargent, T. (1998): “The European unemployment dilemma,” Journal of Political Economy 106, 514 – 550. MacKinnon, G. / Haug, A. / Michelis, L. (1999): “Numerical distribution functions of likelihood ratio tests for cointegration,” Journal of Applied Econometrics 14, 563 – 577. Pissarides, C. A. (1998): “The impact of employment tax cuts on unemployment and wages: The role of unemployment benefits and tax structure,” European Economic Review 42, 155 – 183. Rauch, J. E. / Trindade, V. (2003): “Information, international substitutability and globalisation,” American Economic Review 93, 775 – 791. Slaughter, M. J., (2001): “International trade and labor-demand elasticities,” Journal of International Economics, 54, 27 – 56. Stock, J. / Watson, M. W. (1993): “A simpler estimator of cointegrating vectors in higher order integrated systems,” Econometrica 61, 783 – 820. Wood, A. (1995): “How trade hurt unskilled workers,” Journal of Economic Perspectives 9, 57 – 80.

The Effect of Globalization on Aggregate Labour Demand in EU Countries – Comment By Herbert S. Buscher

1. Intro Listening to the news, nearly daily we hear that firms are planning to close plants and move production abroad or that part of the production will take place in a foreign country in the near future. And sometimes arguments for this decision are given: to produce at lower cost, to increase revenue, to avoid protection measures and the like. Thus, there is a permanent threat on the national or local economy to lose jobs, to increase unemployment, to accept wage cuts, to work in a more flexible way and so on. Thus, there seems to be a widespread fear in public opinion that globalization may destroy workplaces at home and that destruction may be of a large scale. Reading the paper by Christian Dreger and Stefan Kooths, these fears do not seem to be justified on empirical grounds, at least at the macro level.

2. How to Model the Influence of Globalization and Trade Liberalization on Labour Demand? There are two obvious approaches to deal with globalization and trade liberalization on specific domestic markets. The first approach starts from a theoretical model and derives testable implications of the theory with respect to the influence of globalization on domestic labour demand. Given that in most cases several theoretical explanations are available, one is usually restricted to a specific theory. The second approach does not refer to a special theoretical model and leaves open the question how foreign shocks are transmitted into national markets. Instead, the specification chosen here is more practically oriented in the sense that several measures thought of suitable to capture the impact of globalization on employment are tested. In a certain sense this approach seems to be more general (and less informative) than testing a specific theoretical content. Of course, this does not mean that the second approach is atheoretical und thus purely empirically determined. On the contrary, the authors explicitly refer to several publications to justify this route of research, see e. g. Slaughter (2001) and Jean (2000).

48

Herbert S. Buscher

To measure the possible impact of globalization on the national labour markets the authors estimated a standard labor demand equation over the whole sample period using an error-correction approach including the long-run relationship (cointegration) and checked for the adequateness of the specifications in economic as well as econometric terms. The restriction to labour demand follows from their argument that due to the high unemployment in European countries it is labour demand that determines the level as well as the structure of total employment. …1†

log…Lt † ˆ a0 ‡ a1  log…Wt =Pt † ‡ a2  log…Yt † ‡ ut :

In equation (1), log…L† is the log of total employment in persons, …W=P† is the log of real wage, log…Y† is real GDP in logs, u is an error term, t indicates time, and ai ; i ˆ 1; 2, are the elasticities of labour demand with respect to the real wage and to GDP. Country indices have been dropped for ease of exposition. Note that the authors only used total employment in persons and no other variables such as hours worked or the number of employee or the number of employee in different sectors and so on. Therefore, the results obtained from equation (1) provide the reader with information about the employment level, but not about the structure of employment, i.e. different sectors, different skill levels, firm sizes etc. Given the results of the various specification and misspecification tests, the authors conclude that the above equation (1) – using an ECM specification and cointegration – does not reveal any serious deficiencies in econometric terms. Consequently, the authors could stop here in their research activities. The reason is that given equation (1) is properly specified, then  no additional explanatory variables should enter the regression significantly (alternatively: there should be an omitted variables bias in the estimates if an important explanatory variable is missing), and  the estimated elasticities in equation (1) should be stable over time (in a statistical sense).

But here the authors follow a different route by proceeding to their second step in the estimation procedure. In a second step recursive estimation techniques are applied to obtain time varying elasticities of labor demand with respect to the real wage and to GDP for the period 1990 to 2004. For a theoretical justification see again e. g. Slaughter (2001) and Jean (2000). There are several arguments to proceed along this way. One argument relates to GDP. If it is true that due to globalization and a higher degree of trade liberalization foreign trade becomes a more important part in the composition of GDP, then it might well be that the elasticity of labour demand with respect to GDP may change over time. And the same can be expected for the real wage. But then it would perhaps be a better strategy to estimate equation (1), perform all the tests, re-estimate equation (1) using recursive or moving estimation techniques, looking at the CUSUM and CUSUM of squares test statistics, performing tests on the re-

The Effect of Globalization on Aggregate Labour Demand – Comment

49

cursive coefficient estimates under the restrictions of the cointegration properties of the time series etc. Furthermore, although there are no tests available to test for the constancy of recursively estimated coefficients, approximate test statistics could be obtained using bootstrapping procedures to construct the necessary distribution and to derive the appropriate test statistic. But I agree that this will be by no means an easy task. Finally, these estimates are used as dependent variables in auxiliary regressions with “openness of an economy” as an indicator to test for a possible impact of globalization on labor demand. The authors conclude that only weak empirical evidence could be found that globalization exerted a measurable impact on the elasticities of otherwise conventionally specified labor demand equations for the selected EU countries. What is right and what is wrong? Or are both messages compatible with each other? Or did globalization took place over the last fifteen years, but with much smaller effects on the national labor markets than we expected? Could the European economies compensate for the loss of industrial workplaces by expanding or creating new job opportunities in the service sector? Or is the level of aggregation too high to show possible effects of globalization on labor demand and, therefore, a more disaggregated level, say at a firm level, might be better suited to capture possible effects of globalization on labor demand?

3. How Did the Authors Account for Globalization? In the paper the authors do not explicitly refer to a theoretical framework the empirical approach can be derived from, say the Heckscher-Ohlin theorem or the New Economic Geography as proposed by Krugman (1991) and others. Instead, they start from trade integration and argue that globalization of product and labor markets can induce a rise in labor demand elasticities (in absolute value) through a scale effect as well as due to a substitution effect. The first effect arises because competition increases, and the second effect is originated in the presumption that globalization allows firms to expand their production possibilities to cover additional inputs (see p. 3). An additional explanation focuses on segmentation, which in the current context means that the production of one good is divided into several pieces and production of these pieces takes place in different countries, depending on relative cost advantages as well as on the availability of high skilled and low skilled labor. Translated into the current problem, the EU-countries are high wage countries as compared with Eastern European countries or even with Asian countries. Thus, one would expect that production of labor intensive parts will take place in low cost countries, whereas the part of the production process which requires high 4 Supplement 57 – 2006

50

Herbert S. Buscher

skills and latest knowledge about technology will take place in capital-intensive and thus high wage countries. If this explanation holds, we would expect a loss of low skilled work places in the EU-countries and therefore, a shift in labor demand in favor of high skilled workers. This effect may be compensated by creating new jobs in the service sector, leaving total employment (measured in persons) unchanged. To account for this effect, the estimated labor demand equations should be broken down at a sector level to gain insights of possible shifts across sectors. Rübel (2004) in his chapter on globalization and autonomy of national economic policies argues that the process of globalization especially puts small open economies under pressure in the sense that they are forced to undergo substantial structural changes which especially hit their labor markets. And a large part of these changes affects the composition of labor in favor of high skilled or at least skilled jobs. This means that most of the unskilled jobs, at least in most parts of the tradable sector, will move to more labor-intensive countries where production can proceed at lower cost. If this argument holds, one should observe a (steady) shift in the demand for labor from low skilled to high skilled labor. For the empirical analysis this means the labor demand functions estimated by Dreger and Kooth should at least account for different skill structures among employees to capture this possible effect of globalization.

4. How Would we Set up the Test Equations? Let us start with the hypothesis that globalization exerts an impact on national labor demand equations. Furthermore, the impact may be in a direct way in the sense that due to the changing environment additional explanatory variables capturing the influence of globalization enter the demand for labor equation. Alternatively, the impact may happen in an indirect way by changing the behavior of firms through changes in the estimated elasticities. Or let it be a mixture of both. The simplest way to proceed is presumably to augment the standard specification by additional explanatory variables thought of capturing the impact of openness and international trade. This leads to an extended labor demand equation as given in (2), where the vector X contains the new explanatory variables. …2†

log…Lt † ˆ a0 ‡ a1  log…Wt =Pt † ‡ a2  log…Yt † ‡ a3  Xt ‡ ut :

In the above equations L is the amount of labor demanded in an economy, …W=P† is the real wage, Y is real GDP, u is an error terms with the usual properties and a1 und a2 are the elasticities of labor demand with respect to the real wage and to GDP. What would we expect? If the vector X contains important variables to explain the variation in log…L†, then equation (1) should be misspecified (omitted variable bias) and properly conducted specification tests should indicate this misspecification.

The Effect of Globalization on Aggregate Labour Demand – Comment

51

In their paper the authors do not attempt to estimate an equation like (2), but instead they estimate equation (1), perform several (mis-)specification tests and, because none of these tests indicate any serious deficiencies, conclude that equation (1) is properly specified. With respect to the above mentioned hypothesis this means that no direct impact of globalization on national labor demands is detectable. Therefore, their next step consists of testing for indirect effects. To proceed along this way, the long-term labor demand equations are estimated recursively, starting in 1972 to 1990 as the base period, and recursive estimates are obtained from 1990 on until 2004. I think, the idea behind this approach is that globalization and increased international trade comes along with enforced competition, making firms more responsiveness to changes in wages and fluctuations in aggregate demand. But if this argument holds, then we should observe non constant coefficient estimates or structural breaks in equation (1). Therefore, one should estimate equation (1), accounting for co-integration and short-run adjustments, and perform tests on stability or structural breaks, using co-integration analysis techniques. In addition, equation (1) can be estimated recursively and CUSUM- and CUSUM-of-Squarestest can be used to test for the stability of the estimated relationship. Instead, the authors decided to take a different route. Let b1 and b2 the recursive estimates for the real wage and GDP variables, respectively, the test equations used by the authors are estimated using panel regressions with fixed country effects …3a†

b1t ˆ f …openness† ‡ e1t

…3b†

b2t ˆ g…openness† ‡ e2t ;

t ˆ 1990; 1991; . . . ; 2004 ;

and “openness” is the selected indicator to measure the impact of globalization on labor demand. The procedure is repeated by running equations (3a) and (3b) again, using first differences of the openness measure. The results are mixed in the sense that for the level specification “openness” shows up as a significant variable, whereas for the specification in first difference it does not. Perhaps a direct way would be to test for constancy over time of the recursively estimated elasticities. But this is in no way an easy task because no straightforward test statistic is available up to now. Therefore, it is necessary to use bootstrap procedures to approximate the distribution of the recursive coefficients and to draw inference on the constancy or non-constancy of the estimates.

5. Modeling the Effects of Globalization on Labor Demand Despite of a conventionally specified labor demand equation the authors apply recursive coefficient estimates (recursively estimated elasticities) and regress the explanatory variable “openness of an economy” on these elasticities for the period 1990 to 2004. Taking an indicator like openness suggests that the authors expect 4*

52

Herbert S. Buscher

that globalization shows an impact on labor demand via the “segmentation approach”. Part of a production proceeds abroad and is imported by the final goods country where the single parts are build together and the finished good is then exported to other countries. As a consequence, import and export shares will rise and therefore the indicator “openness of an economy” should account for this development. Taking “openness” as an indicator might be well suited to analyze how and where value added occurs, but it might be an inadequate indicator to capture labor market effects directly. Although it is by no means an easy task to find proper indicators at the macro level, additional variables related to foreign trade such as FDIB4s, should be tested Shifts in the demand for labor from industry to services It is common believe that in the course of globalization high wage countries will lose low wage workplaces in the industrial sector, but will create new workplaces for qualified and high skilled workers in industry and other sectors, namely services. Furthermore, one might think that due to growing export activities new workplaces may even be created in the industrial sector. And it is also true that in most EU-countries total employment either remained unchanged or even increased during the nineties and in the first years of the new century. Comparing the number of employees from 1990 to 2004 in the EU-15 countries, it increased in all countries with Germany and Sweden being the exceptions. And looking at employment for the same time period, employment only decreased slightly in Sweden. Taking Germany as an example, employment in industry dropped from 10.6 mill. in 1991 to 7.6 mill. in 2004, but at the same time employment in the service sector increased from 22.97 mill. to 27.7 mill persons. As these figures show, globalization must not necessarily effect total employment negatively in an open economy, but it seems more likely that sector-specific effects will take place, shifting employment from one sector into an other sector, changing the decomposition of low skilled and high skilled workers and so on. But these effects cannot be captured at a highly aggregated macro level. Therefore, it seems necessary to supplement studies at the macro level by appropriate studies at lower aggregation level and to bring the results of this research effort together in a concise manner to improve our knowledge about globalization and local markets.

6. What Conclusions and Suggestions Can Be Drawn from the Paper? The paper is a promising attempt to capture the impact of globalization on domestic labor demand equations. Furthermore, at the European level, there are only few chances to get more disaggregated data sets for all countries to run the regres-

The Effect of Globalization on Aggregate Labour Demand – Comment

53

sions in more detail. In this sense, the authors used most of the available information at this level. But nevertheless, due to their conclusion that no statistically significant impact on estimated labor demand elasticities could be detected, they should augment their results  by rerunning the regressions using hours worked as the dependent variable in the labor demand equation instead of total employment,  by rerunning the regressions using employment in the private sector only (excluded the public sector) and employment in industry only,  by rerunning the regressions using full-time employment equivalences to account for considerably changed employment patterns across countries as well as over time.

If these three additional sets of regressions produce similar results, more confidence will be obtained with respect to the effects of globalization on national labor markets. These regressions can be performed at the macro level with available data sets for at least most of the European countries included in the study. Furthermore, if the authors want to abstain from taking as many European countries in their study as possible, restricting themselves to a subset of countries might be useful to conduct an analysis at a more disaggregated level including those countries for which the respective data sets are available. And finally, thoughts should be devoted to how globalization could affect labor demand, either directly or indirectly, going beyond the segmentation approach and presenting additional empirical evidence about the validity of alternative explanations of the effects of globalization and international trade on national labor markets. But I agree that until today not all theoretical explanations are easily accessible to empirical testing. Concluding, the paper shows that there are still of lot of unanswered questions how globalization effects national economies or single markets. Further research is necessary to shed more light on these relationships and to improve our understanding of how world markets behave and how this will be transmitted into local markets. In this sense, the paper points in the right direction, but it also shows clearly that empirical research on this topic is just at its beginning.

The Impact of FDI on the Skill Structure in German Manufacturing By Robert Jäckle*

Abstract This paper tests whether foreign direct investments (FDI) of German manufacturing multinationals (MNE) have raised domestic skill intensity between 1996 and 2001. Using a sample of 1,557 firms, the results show that foreign activities of German manufacturing MNEs carry higher average wages on the home market. I interpret this as evidence indicating that part of the skill upgrading in German manufacturing is associated with the increasing job export to foreign locations. Other things equal, a rise in overall affiliate employment relative to domestic employment by 10 percentage points is associated with a rise in skill intensity at the parent firm by 0.1 % to 0.3 %. When distinguishing between different host regions, I find investments in industrialised countries consistent with the horizontal FDI motive, whereas investments in developing countries are driven by vertical production strategies. In the case of transition countries results are inconclusive. JEL Classifications: F21, F23, J31 Keywords: Multinational enterprises; Skill mix; Globalisation

1. Introduction In Germany (as in many other countries) an increase of multinational activities towards the end of the 20th century occurred. Between 1989 and 2001 the number of domestically owned multinational enterprises (MNE) rose from 6,762 to 8,857.1 At the same time, German firms built up or acquired another 15,196 subsidiaries * Ifo Institute for Economic Research, Poschingerstr. 5, 81679 München, Germany. Tel.: +49 (89) 9224-1603. E-mail: [email protected]. Many thanks to Georg Wamser, Christian Holzner, Sascha O. Becker, and Marc Muendler for helpful remarks. I also gratefully notice Matthew Scogin, who thoroughly reviewed the paper. Heinz Herrmann, Alexander Lipponer, Fred Ramb, and Thomas Wenger are thanked for access to and ongoing support with the BuBa MIDI and USTAN data. I gratefully acknowledge financial support from the VolkswagenStiftung under its grant initiative Global Structures and Their Governance, administrative and financial support from the Ifo Institute and the Deutsche Bundesbank. All remaining deficiencies are my responsibility. 1 See Becker, Jäckle and Muendler (2005c).

56

Robert Jäckle

abroad, totalling 33,527 foreign affiliates in 2001. The workforce at these operations amounted to 1.720 million employees in 1989 and rose to 4.549 million workers in 2001. By the end of the corresponding time period, German firms produced 1,279.49 billion Euros’ worth of goods abroad. Trade theory suggest that outsourcing is associated with changes in relative labour demand. Horizontal FDI, i.e. replicating domestic operations at different locations, could either keep labour demand unaffected or shift it towards the more skilled. The effects of vertical FDI, on the other hand, depend on whether the host country is skill-abundant relative to the domestic country or vice versa. FDI flows to nations, which are abundant with low-skilled workers compared to Germany, may bring skill upgrading at the parent firm, whereas FDI to countries with a highly qualified workforce might be accompanied by a decrease in the skill level at home. On the firm level this translates to a changing share of the non-production wage bill and changing average wages. Given those theoretical considerations, this paper investigates whether the international diversification strategy of German manufacturing MNEs has influenced the domestic skill mix between 1996 and 2001. For that purpose, I employ three different foreign activity measures for German-headquartered multinationals: The ratios of (aggregated) foreign affiliate to domestic employment, output, and capital. Using these FDI proxies, augmented demand functions for high-skilled labour are estimated. The main finding is that foreign activities of German manufacturing MNEs carry higher average wages at domestic operations. I interpret this as evidence indicating that part of the skill upgrading in German manufacturing is associated with the international production strategy of German firms. Other things equal, a rise in overall affiliate employment relative to domestic employment by 10 percentage points is accompanied by an increase in skill intensity at the parent firm by 0.1 % to 0.3 %. When distinguishing between different host regions, I find investments in industrialised countries consistent with the horizontal FDI motive, whereas investments in developing countries are driven by vertical production strategies. In the case of transition countries results are inconclusive, a distinction between the two motives is not possible. The remainder of this paper is organised as follows: the starting point is a discussion of differences between horizontal and vertical investment strategies and their impact on the domestic skill structure, followed by a brief summary of the existing literature in section 3, the next section gives an overview of the data and provides several descriptives, then I look at specification issues and discuss the estimation results in section 5, and finally section 6 concludes the paper.

The Impact of FDI on the Skill Structure in German Manufacturing

57

2. Theoretical Considerations The goal of this paper is to investigate the effects of multinational activities on the skill structure at their domestic locations. Theoretical answers to this question are not clear cut. Instead, the outcomes are influenced by what motives lie behind a firm’s decision to start up or acquire an affiliate in a certain region of the world. In this respect, the literature distinguishes between two types of FDI – vertical and horizontal FDI.2 Since models based on one theory or the other come to different conclusions, the task to examine whether foreign investments cause skill upgrading or downgrading must be addressed empirically. 2.1 Vertical FDI

Cost saving efforts are the most important driving force toward companies accomplishing vertical FDI.3 A firm decides to geographically fragment its production in separate stages, as a means to profit from differences in relative factor prices between the home and the host country. These benefits need to carry transport costs for the re-export of final or intermediate goods and fixed costs for starting up new production facilities abroad. Markusen (2002) summarises in his book that ‘‘for vertical firms, location advantages arise when trade costs are low, stages of production differ in factor intensities, and countries differ significantly in relative factor endowments.” Therefore, the effect of vertical FDI on the skill structure at home mostly depends on whether the domestic country is abundant in skilled labour relative to a large proportion of unskilled workers in the foreign country, or vice versa. In the first case, which is most likely for investments in low-income countries, one would expect that knowledge based assets are still produced at the parent firm, whereas final and intermediate production stages are accomplished at the affiliate operation. Hence, FDI flows to countries, which are relatively abundant with low-skilled workers, may come along with skill upgrading at the parent firm. In the latter case, which mostly refers to investments in high-income countries, theory suggests a decrease in the skill level at home. For Germany, a country which is relatively abundant with skilled labour, it is often argued that investments in low wage countries (especially those located in Central and Eastern Europe (CEE)) may put pressure on domestic (low-skilled) wages. Hence, according to the vertical FDI theory it seems straightforward that an investment strategy, which fragments production such that labour intensive stages are located in CEE countries, Asian-Pacific countries, or other developing countries, increases the skill intensity in German manufacturing. 2 Markusen, Venables, Konan and Zhang (1996) and Carr, Markusen and Maskus (2001) combine the two models to the so called “knowledge-capital” model. Through its three defining assumptions fragmentation, skilled-labour intensity, and jointness the model allows for vertical and horizontal activities in one common framework. 3 Among the first authors who described vertical multinationals were Helpman (1984, 1985), and Helpman and Krugman (1985).

58

Robert Jäckle

Marin (2004), on the other hand, finds evidence that Austrian and German companies take advantage of cheap high-skilled workers in Central and Eastern Europe. She argues that German firms locate headquarter activities like research and development (R&D) in low-wage countries. One may refer to this as some kind of inverted vertical FDI, i.e. instead of taking advantage of the abundant low skilled labour German firms (additionally) replace parts of their knowledge based assets at home with cheap (high-skilled) labour in transition economies. In this case, one would expect that, ceteris paribus, the skill level at the parent firm decreases. Reaffirming the above, it is a priori not clear whether the vertical division of the production process upgrades or downgrades the skill intensity in Germany. In the end, it depends on which of the two opposing forces – the relocation of lowor high-skilled labour – is stronger. If the first effect dominates one would expect vertical FDI to positively influence the skill structure. If, on the other hand, indirect vertical fragmentation turns out to be most important, there might be no effect or even a downgrade of the skill intensity.

2.2 Horizontal FDI

If a firm conducts horizontal FDI, aspirations for better market access and the proximity-concentration trade-off play a decisive role. Companies invest abroad if the costs of accessing new markets and transportation are higher than the expenditures of starting up a new firm and the loss due to a reduction of scale economies, when producing the same good across different markets. In this case, Markusen (2002) states that ‘‘for horizontal firms, location advantages arise when the host country market is large and when trade costs (broadly defined) are moderate to high.” Head and Ries (2002) bring an additional point into the discussion. They argue that horizontal FDI can be both a pure replication of downstream activities at different locations or only a replication of final goods production (in their terminology ‘‘branching”), where upstream stages of production, like design and marketing, are located at home. In the latter case, at least part of the firms’ (skill intensive) knowledge based assets stay at the parent location. If production in foreign countries is completely independent from local factors (replication) there exists only an indirect effect on the domestic firm through possible changes in the scale of the parent operation. These changes may occur if production at foreign affiliates substitutes for exports to the according markets. Head and Ries (2002) show that a scale-decrease in the domestic location may come along with either skill-upgrading or downgrading. However, in the extreme case where the demand for high skilled employees is completely independent of the output produced, i.e. knowledge based inputs require only a certain amount of high-skilled workers (independent of output), the skill intensity at the domestic operation increases if domestic production is reduced. Yet, in an empirical specification where one controls for the size of the parent firm (compare section 5) the

The Impact of FDI on the Skill Structure in German Manufacturing

59

expected impact of outward FDI on the skill structure with respect to the theory of horizontal replication should not be measured as significantly different from zero. In the branching case, where knowledge capital produced at home serves as a common input for production in subsidiaries all over the world and horizontal FDI is meant to build up production capacities for final goods only, controlling for the amount of output in a regression approach does not reduce the effect on the parent skill level to zero. The argument at this juncture is that in a case of accessing new markets (where no goods were exported to so far) the overall increase in worldwide production corresponds to an expansion of knowledge-based input factors at home, whereas domestic output stays constant. Therefore, the skill intensity at the parent firm will be enhanced. If, on the other hand, branching-investments act as a substitute for exports the scale of domestic production will be reduced. Given that the reduction at home is fully compensated by new capacities abroad, worldwide productions requires still the same input of knowledge capital. Therefore, even after controlling for the size of the domestic operation one should find a positive effect of branching on the skill level at the parent firm. For Germany, one might distinguish between vertical and horizontal FDI by discriminating between the different foreign locations German firms choose. High income regions especially in Western Europe and North America, may serve as locations where mainly horizontal FDI is conducted. In the case of developing and transition countries, however, the distinction is less clear. The motive for locating affiliates in these regions might be driven by both horizontal and vertical aspects. Since developing and transition countries are mostly abundant in low-skilled labour, a substantial part of investments made there may replace relatively expansive unskilled employees at home, thus leading to a higher skill intensity at the domestic location. To the contrary, many of these countries already serve as important markets for final goods. This is true, for instance, in the emerging markets of China and India, but also for many of the transition countries in Central and Eastern Europe. 3. Related Literature A number of recent studies examine the consequences of the international division of production for labour markets in industrialised countries. Among the first empirical papers written in this field are Feenstra and Hanson (1996, 1999). Their focus lies on the influence of globally integrated production on the observed wage divergence in the US between 1979 and 1990. The authors find evidence that outsourcing activities (defined as imports of intermediate inputs from their own and foreign affiliates) of American firms contributed significantly to the relative increase of high skilled wages in the considered time period. Instead of defining broad outsourcing measures, Slaughter (2000) solely concentrates on the question whether production transfers through foreign direct investments have contributed to the increased demand for high skilled workers in US manufacturing. In the descriptive part of his paper he shows that between 1997 and 1994 US multinationals

60

Robert Jäckle

extended foreign activities relative to domestic ones. However, his regression results suggest that in a hypothetical situation without increased MNE transfers the observed skill upgrading would have been almost the same. The results of Feenstra and Hanson (1999) along with those of Slaughter (2000) can be seen as evidence that for the US the wage gap is mainly effected by trade at arm’s length, sub-contracting, and licensing instead of direct investments. For Sweden, Hansson (2001) shows that transfers of production stages from Swedish MNEs to non-OECD countries positively effect relative wages of skilled workers. In a study of Japan, Head and Ries (2002) investigate the impact of overseas employment of Japanese enterprises on the skill structure at the domestic location. They use data that is aggregated on the sector level as well as micro-data at the level of the parent firm. Their results suggest that expanding the work force in low-income countries brings an upgrade of the skill level at parent operations in Japan. Geishecker and Görg (2004) are among the first to ask who the “Winners and Losers” from outsourcing in Germany are. They estimate wage equations separately for three (low, medium, and high) skill groups and augment the different specifications using foreign activity measures as defined by Feenstra and Hanson (1996, 1999). Their results suggest that workers in the lowest skill category lose from outsourcing and that – at least in some specifications – high-skilled workers gain through higher wages. Becker, Ekholm and Muendler (2006) use novel linked employer-employee data to study the skill demand at German multinationals. Their study covers all sectors of the economy and distinguishes between several skill groups. In another study of Germany, Jäckle (2006) looks amongst other performance measures at average wages of firms both prior to and after they have gone multinational (switching). He shows that compared to non-switchers average wages have already been superior in the run up to become a MNE, and secondly, that after switching average wages grow faster at newly founded multinationals than at purely domestic firms. 4. Data and Descriptives In the study at hand, I use data from the Deutsche Bundesbank’s (BuBa) MIDI database (Microdatabase Directinvestment) at the level of German manufacturing parents and their (aggregated) foreign affiliate activities between 1996 and 2001.4 From the MIDI data set information on ownership-weighted foreign employment, fixed assets, and turnover is gathered. All financial variables referring to affiliate operations are converted into Euro and deflated to real values at year end 1998 4 Before 1999 German firms were obliged by law to report their foreign affiliates’ balance sheet data if they controlled more than 20 percent of their foreign affiliates equity and the affiliates balance sheet total was at least 1 million DM. After 1998 either of the following two criteria had to be satisfied: (i) the parent controls at least 10 percent of equity and the balance sheet total is at least 5 million EUR; or (ii) the parent controls at least 50 percent of equity and the balance sheet is at least 0.5 million EUR (see Lipponer (2003)).

The Impact of FDI on the Skill Structure in German Manufacturing

61

employing a purchasing-power related method (for further details see Becker, Ekholm, Jäckle and Muendler (2005a, 2005b)). Information on parent specific variables is obtained from the BuBa USTAN (Unternehmensbilanzstatistik) data set, which was string matched by name to companies in the BuBa MIDI data base.5 Every firm in Germany that draws a bill of exchange is required by law to report its balance sheet to BuBa, which collects this information in its USTAN data base when the bill of exchange is rediscounted (see Deutsche Bundesbank (1998)). The variables extracted from USTAN are employment, turnover, fixed assets, overall labour costs, and intermediate input goods. All financial figures except intermediate input goods are deflated to 1998 real values using the German CPI (from the IMF’s International Financial Statistics). Intermediate input goods are converted to real values using the the intermediate input goods deflator from the OECD’s Main Economic Indicators. The value added is constructed as the difference between real turnover and real intermediate input goods. Both the MIDI as well as the USTAN data sets are available in the form of an unbalanced panel. Over the whole sample period, a total of 1,557 different manufacturing firms were matched. Table 6 in the appendix includes industry definitions and reports the panel attrition. In table 1 the development of the data sets between 1996 and 2001 is depicted. The first line reports the overall number of firms in German manufacturing with at least one foreign affiliate. The second row includes the according number of foreign affiliates. These figures show that both the number of parent firms (+8.5 %) as well as the number of their corresponding affiliates (+12.4 %) rose in the time period under consideration. A comparison with lines three and four, which include the total number of matched firms in each year, allows an evaluation of the matching algorithm. The merge process yields a matching quote between 40 percent in 1996 and 29 percent in 2001 for parent firms and a coverage between 54 percent (1996) and 42 percent (2001) for their foreign affiliates. Table 2 depicts the worldwide (domestic and foreign) usage of employment and capital of German manufacturing firms between 1996 and 2001. The table additionally focuses on the output produced with these inputs and on average labour costs incurred per employee.6 Table 3 presents the regional pattern of the MNEs’ aggregated foreign activities for three broad country groups. The groupings are: Industrialised countries (IND), transition countries (TRANS), and developing coun5 The string matching routine automatically chose firms with an equality of at least 50 percent of all letters included in their firm names. All potential matches were manually overseen before they were accepted as being the same company. In Becker, Ekholm, Jäckle and Muendler (2005a) (appendices A and B), the string matching procedure is described in more detail. 6 While affiliate numbers are aggregated values based an ownership-weighted firm-level observations from the MIDI data set (MIDI includes the overall population of German firms conducting outward FDI), parent values are based an aggregated numbers for the whole manufacturing sector (including also non-MNEs) provided by the national account series of the German Federal Statistical Office.

62

Robert Jäckle

tries (DEV) (for definitions see table 8 in the appendix). All tables also include the corresponding figures for the matched sample of MIDI and USTAN, used in the econometric analysis of section 5. Table 1 Number of FDI Firms and Foreign Affiliates in German Manufacturing

Parent firms total Foreign affiliates total Parent firms matched Foreign affiliates matched

1996

1997

1998

1999

2000

2001

(1)

(2)

(3)

(4)

(5)

(6)

2,596 10,403 1,034 5,639

2,710 10,935 1,047 5,787

2,782 11,432 993 5,860

2,782 11,375 969 5,506

2,849 11,828 929 5,558

2,817 11,689 811 4,949

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations.

Excluding self-employed persons, in 2001 the absolute value of employment in the German manufacturing sector (10.417 million) was 6.9 % lower than in 1996 (11.194 million). At the same time, overall employment at foreign affiliates increased by 19 % from 1.293 million workers in 1996 to about 1.538 million employees in 2001. The decline of parent employment in the matched sample (–21.4 %) virtually equals the decrease in sample size between 1996 and 2001 (–21.6 %) (see table 1). Hence, one may conclude that the employment reduction at manufacturing MNEs in Germany was less strong than in the whole sector. Assuming that absolute home market employment at manufacturing MNEs stayed constant, relative multinational-wide employment shifted towards foreign operations. These numbers therefore provide descriptive evidence, confirming the public opinion, that an increasing number of jobs at foreign affiliates substitute for parent employment in Germany.7 While the number of workers in German manufacturing declined, the figures in table 2 show that – at the same time – real average wages increased.8 Looking at 7 Using a translog cost function approach, Becker, Ekholm, Jäckle and Muendler (2005b) find that affiliate employment tends to be a substitute for parent employment. In a follow up study, Muendler and Becker (2006) show how multinational labour demand responds to wage differentials at the extensive margin, when a multinational enterprise expands, and at the intensive margin, when an MNE operates existing affiliates. They derive conditions to infer elasticities of labor substitution at both margins, controlling for location selectivity. Their results show that with every percentage increase in German wages, German MNEs allocate 1,600 manufacturing jobs to Eastern Europe at the extensive margin and 3,900 jobs overall. 8 The skill proxy used in the regressions of section 5 is the average wage paid by each firm. It is constructed using the USTAN variables wage bill and employment (see appendix A). To constructed comparable average wages for the overall sector (from the national account

The Impact of FDI on the Skill Structure in German Manufacturing

63

the matched sample, the development towards higher average wages per firm is even stronger. However, it is unfortunate that there is a relatively strong drop in average wages in 2001, which does not correspond to the sector-wide development, and might be due to the decreasing sample size in the last period. In sum, the increase in average wages by 3.7 % together with the declining employment (–6.9 %) in German manufacturing could be seen an initial piece of descriptive evidence for skill-upgrading during the time period under consideration. Table 2 Employment, Turnover, and Capital of German Manufacturing 1996

1997

1998

1999

2000

2001

(1)

(2)

(3)

(4)

(5)

(6)

Total observations Foreign affiliates Employment Fixed assets Turnover

1,292.616 78.248 299.733

1,343.883 77.422 319.484

1,531.898 96.532 420.904

1,504.524 100.297 461.512

1,562.200 124.638 473.793

1,538.254 129.285 449.038

Parent firms Employment Fixed assets Avg. wages Turnover

11,194 1,472.648 35,491 1,368.982

10,903 1,463.241 35,419 1,399.725

10,814 1,473.63 35,728 1,447.08

10,652 1,476.097 36,134 1,480.97

10,591 1,435.436 36,844 1,536.808

10,417 1,435.443 36,788 1,540.342

Matched sample Foreign affiliates Employment Fixed assets Turnover Parent firms Employment Fixed assets Avg. wages Turnover

836.398 60.083 219.683

786.018 49.292 198.565

794.413 50.700 198.983

757.393 49.860 189.167

780.564 53.837 204.036

741.339 50.100 191.468

1,376.887 127.128 45,330 292.789

1,316.960 137.454 45,602 307.974

1,272.875 145.456 48,248 318.963

1,196.742 144.761 49,140 304.930

1,205.114 163.804 52,269 333.855

1,082.388 156.093 47,654 310.724

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations, and Federal Statistical Office Germany, Fachserie 18 / Reihe 1.4. Foreign variables are ownership-weighted. All financial variables are deflated to unity at year end 1998 and, with the exception of average labour costs, measured in billions of Euros. Employment figures are in thousands.

series of the German Federal Statistical Office), 1 used sector-wide labour costs (including social security contributions paid by the employer) and divided those by the number of workers (without self-employed).

64

Robert Jäckle

Apart from variables related to the firms’ workforce, table 2 also focuses on capital usage and output. It shows that domestic firms reduced fixed assets (– 2.5 %), while output simultaneously rose by 12.5 %. At their foreign affiliates, German manufacturing MNEs strongly expanded both the use of capital (65.2 %) and the absolute value of output (49.8 %). Since affiliate output expansion was proportionally larger, the multinational-wide distribution of turnover shifted towards affiliates. Furthermore, when looking at the capital distribution, it becomes clear that investments abroad must have been relatively larger than investments at home. When looking at regional FDI patterns (see table 3), in 2001 industrialised countries are still the most important host region. At that time they accounted for 58.1 % (64 % in 1996) of affiliate employment, 85.9 % (83.9 %) of affiliate fixed assets, and 84.1 % (85.1 %) of the output German manufacturing firms produce abroad. In Germany, most of the public attention focuses on outward FDI to transition countries located in Central and Eastern Europe (CEE). In terms of employment numbers this country group raised its relative importance. German manufacturing MNEs increased their foreign employment share in transition countries from 10.7 % in 1996 to 17.5 % in 2001. Furthermore, with respect to capital and output the corresponding figures amount to 2.5 % and 2.8 % in 1996 and 2.8 % and accordingly 3.1 % in 2001. Finally, in developing host countries the relative share of foreign activities in 2001 was slightly lower than in 1996 (employment and fixed assets) or kept almost constant during the sample period (turnover).

5. Specification Issues and Econometric Results The figures in section 4 provide descriptive evidence that increasing foreign activities of German manufacturing firms in the time period between 1996 and 2001 were associated with higher average wages and, at the same time, decreasing employment numbers. However, to infer that these trends reflect within-firm shifts of labour demand from low-skilled employees towards the more high-skilled, one needs to turn to regression analysis.

5.1 Specification

The existing literature tries to understand, what explains changes in the skill structure, in a translog cost function framework. Throughout the different studies, the authors assume that capital inputs are a quasi-fixed factor and that firms / industries minimise their costs with respect to low- and high-skilled workers. Berman, Bound and Hanson (1994) use this approach to derive the corresponding (production and non-production) share equations and investigate whether laboursaving technological change shifts demand away from low-skilled employees.

The Impact of FDI on the Skill Structure in German Manufacturing

65

Table 3 Foreign Employment, Turnover, and Capital of German Manufacturing MNEs, Differentiated by World Regions 1996

1997

1998

1999

2000

2001

(1)

(2)

(3)

(4)

(5)

(6)

Total observations Industrialised countries Employment 827.727 Fixed assets 65.643 Turnover 255.155

825.508 64.191 269.034

957.416 80.542 362.389

914.149 82.215 401.026

909.902 106.233 403.075

893.121 111.098 377.692

Transition countries Employment 137.904 Fixed assets 1.987 Turnover 7.374

171.250 2.295 9.709

208.797 3.041 12.414

224.931 3.258 13.922

261.306 3.606 15.566

269.212 3.579 13.869

Developing countries Employment 326.985 Fixed assets 10.617 Turnover 37.205

347.125 10.937 40.741

365.685 12.949 46.101

365.444 14.825 46.564

390.991 14.800 55.151

375.922 14.608 57.477

Matched sample Industrialised countries Employment 541.002 Fixed assets 50.871 Turnover 185.107

488.567 40.176 164.583

484.489 40.501 164.215

444.784 38.593 153.703

433.839 42.176 160.938

415.411 39.510 150.571

76.881 1.389 5.630

91.220 1.678 6.864

106.542 2.209 8.897

111.508 2.231 9.801

136.257 2.384 10.818

134.454 2.197 8.778

Developing countries Employment 218.516 Fixed assets 7.823 Turnover 28.946

206.231 7.438 27.117

203.382 7.990 25.871

201.101 9.035 25.663

210.468 9.277 32.279

191.473 8.392 32.119

Transition countries Employment Fixed assets Turnover

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations. The variables are ownership-weighted. All financial variables are measured in billions EUR and deflated to unity at year end 1998. Employment figures are in thousands.

Feenstra and Hanson (1996, 1999) augment the Berman et al. (1994) specification and additionally include outsourcing proxies in their equations. Slaughter (2000) uses the cost function approach to explore what the impact of FDI on the skill structure in US manufacturing is. All of the above studies have in common that they use data on the industry or sector level. Unlike those, Head and Ries (2002) are the first who look at FDI-induced skill changes at the firm level. 5 Supplement 57 – 2006

66

Robert Jäckle

In the paper at hand, I follow in their footsteps and look at the firms’ demand equations for high skilled labour: …1†

H L SH j;t ˆ 0 ‡ 1 log wj;t ‡ 2 log wj;t ‡ k log

Kj;t ‡ v log Vj;t ‡ m MNEj;t ; Vj;t

H where SH j;t is the share of skilled workers at firm j in year t, wj;t are skilled wages, L wj;t are low-skilled wages, Kj;t are fixed assets, Vj;t is the real value-added output, and MNEj;t is a variable that measures the multinational activity of firm j in year t.

The parameter k accounts for the impact of capital intensity on the skill structure. It is positive if capital and skilled employees exhibit a complementary relationship. The value-added regressors v accounts for the size of the firm. Most importantly, the variable MNEj;t controls for the variation in SH j;t that is due to changes of the firms’ FDI-activities. I follow Slaughter (2000) and employ three different activity measures: affiliate employment, turnover, and fixed assets. I compute these measures by building the sum over all affiliates that belong to the same parent firm. To avoid double counting, if one affiliate is owned by more than one German parent, each variable is weighted with its parent firm’s ownership share.9 Using these absolute measures to account for foreign activities does not allow to control for general equilibrium shocks, that might change domestic and foreign performance. A worldwide recession, for example, could hit both parent and affiliate employment. Therefore, I construct the regressors MNEj;t as the ratio of (aggregated) foreign affiliate to domestic employment, output, and capital.10 Employing the two non-labour measures turnover and fixed assets allows to test, whether investing aborad accounts for more than just transferring jobs. Since FDI might additionally cause affiliate capital and output expansion, using these measures enables me to test for their impact on the skill mix in German manufacturing. The dependent variable in my regressions is the log of average wages (normalised by average sectoral wages) paid by each parent firm. The mean wage is meant to proxy the skill intensity of production and is computed as real labour costs over employment. The normalisation controls for productivity gains that are common to all workers. Appendix A derives conditions, under which average wages are a good approximation for the high-skill share of a firm’s workforce. While the rough estimates in Appendix A suggest that the log of average wages over average lowskilled wages can serve as an indicator for the skill intensity in German manufac9 The ownership variable also allows the restriction of the estimation sample in section 5.2 to majority owned affiliates. The use of the restricted sample did not significantly change the regression outcome. Modified estimation results are available an request. 10 In contrast to the paper at hand, Head and Ries (2002) use the share of a MNE’s total work force that is located offshore to quantify the firm’s international activities. For comparison reasons, I also constructed the regressors MNEj;t as affiliate activity divided by worldwide (domestic + foreign) activity. The use of these different measures, when estimating equation 1, did not alter results in any important way.

The Impact of FDI on the Skill Structure in German Manufacturing

67

turing MNEs, two problems arise at this point. First, since the USTAN data set does not include any information about the skill composition of the firms’ workforce, I have to rely on average yearly industry-wages from the German Statistical Office to normalise average wages.11 Second, as Head and Ries (2002) point out, average wages on the firm level might be an indicator for efficiency wages. Thus, as far as the efficiency mark-ups between firms differ, part of the variation in average wages might be explained independently of the employees’ skill levels. Burdett and Mortensen (1998) show that wage differentials and firm size are positively correlated, i.e. larger firms pay efficiency wages to reduce employment fluctuations. Therefore, including firm size as an explanatory variable should help to control for the efficiency-part of the wage variation. Additionally, the inclusion of firm fixedeffects in most specifications might also account for this kind of variation. Finally, since the USTAN data set does not include information about high- and low-skilled wages, I follow Berman et al. (1994), assuming that quality-adjusted high- and low-skilled labour does not vary over firms. Under this assumption, relative wages for the two skill groups are constant, and replacing the wage regressors in equation 1 by time dummies is a valid option. It is thus the case that equation 1 takes the following estimatable form: …2†

log

 j;t w  Lj;t w

! ˆ 0 ‡ yeart ‡ k log

Kj;t ‡ v log Vj;t ‡ m MNEj;t ‡ cj ‡ j;t ; Vj;t

 j;t is the average wage paid by firm j in year t, w  Lj;t is the average low skilled where w wage, yeart are yearly time dummies, cj is an unobserved firm specific factor, that does not change over time, and j;t is a normally distributed error term with mean zero. The decisive parameter in equation 2 is m . It measures the impact of FDI on the skill mix in German manufacturing. Firm fixed effects (cj ) are included to account for unobservable variables, which are correlated with the key variable MNEj;t – the standard omitted variable bias problem – and do not vary over time. For the short time period between 1996 and 2001, good examples in this respect are latent management skills and objectives. To compare, I also include pooled estimation results (without latent constant effects) in tables 4 and 5. Finally, in all specifications the variance estimators are calculated using the Huber/White/sandwich adjustment.

11 In other specifications, I also used average yearly wages of low-skilled workers from the German Socio-Economic Panel (GSOEP) to normalise average labour costs. I classify the workers’ educational attainment as low-skilled if they have either no school degree at all, no school degree plus vocational training, or a lower school degree without vocational training. However, the different denominator of the dependent variable did not alter results in any important way. Estimation outcomes, using these modifications, are available an request.

5*

68

Robert Jäckle 5.2 Results

Table 4 provides estimation results for equation 2. Columns (1) through (3) report specifications using pooled regressions, columns (4) through (6) include parameter estimates for the fixed effects model. For manufacturing firms in Germany, equation 1 explains between 7 % and 17 % (26 %) of the variance in log normalised average wages in the pooled sample (fixed effects model). Including relative affiliate employment instead of affiliate fixed assets or output increases the R2 from a mere 7 % to 17 % (26 %). This can be seen as the first piece of evidence, demonstrating that job creation at foreign subsidiaries better explains the variation in domestic skill composition than the expansion of affiliate output and capital. Corresponding to results obtained by other researchers, higher levels of output significantly increase average wages across all specifications. Burdett and Mortensen (1998) argue that wage differentials and firm size depend on each other, i.e. larger firms pay efficiency wages to reduce employment fluctuations. Hence, estimation results indicate that the output variable controls for this type of scale effects. Furthermore, a positive wage-scale relationship could indicate that successful firms (in terms of output) pass on part of their gains to the workforce (e.g. in the form of higher bonuses) and herewith increase average wages. The pooled regression model exhibits significant, positive effects of capital intensity, i.e. an expansion of capital usage comes along with higher average wages. However, estimated coefficients in the fixed effects model differ greatly from those without unobserved constant factors. As opposed to the pooled regression model, the inclusion of cj reveals a significant, negative influence of capital on the skill intensity. This means that estimated parameters in the pooled specifications might be severely upward biased. The substitutional relationship between capital and skill in German manufacturing stands in contrast to Slaughter’s US sector-level study, which finds the two input factors to be complementary. Head and Ries (2002), on the other hand, show that for Japanese MNEs – similar to German firms – greater capital investments are negatively correlated with average wages. Across all specifications, the F-test of joint significance of coefficients on the time dummies refers to regressions including a full set of yearly binary variables. Positive, significant parameter estimates on year dummies show that skill upgrading in German manufacturing cannot be fully explained by capital, output, and MNE activities. The coefficients are largest (and highly significant) in the years 2000 and 2001, which might be a response to the higher panel attrition at the end of the sample period (compare also Tables 1 and 6). Turning to the effects of multinational activities on the skill intensity, the central result is that FDI, especially when proxied by foreign employment, is one of the driving forces behind skill-upgrading in German manufacturing. The figures in table 4 indicate, that (i) in the pooled regression approach multinational activity measures, reflecting affiliate employment and capital, are significant and positive,

The Impact of FDI on the Skill Structure in German Manufacturing

69

Table 4 Multinational Activities and Average Wages in German Manufacturing, Results pooled

Activity measure MNE acivity. Log capital / value-added Log value-added Dummy 1997 Dummy 1998 Dummy 1999 Dummy 2000 Dummy 2001 Constant N N manufac. firms R2 F-test time dummies

pooled

pooled

within

(3)

(4)

within

within

(1)

(2)

employmt. .009

turnover .0007

(.002)***

(.0005)

(.0002)***

(.004)***

(.0003)

(.001)

.024

.042

.043

–.038

–.031

–.031

(.007)***

(.008)***

(.008)***

(.011)***

(.012)***

(.012)***

.048

.041

.041

.162

.125

.109

(.003)***

(.004)***

(.004)***

(.035)***

(.041)***

(.041)***

.007(.007)

fixed ass. employmt. .001 .026

(5)

(6)

turnover .0003

fixed ass. –.0001

.0005

.004

.006

–.002

.007

(.011)

(.012)

(.012)

(.006)

(.007)

.021

.021

.023

.006

.019

.019

(.012)*

(.012)*

(.012)*

(.007)

(.008)**

(.008)**

.010

.016

.017

.004

.017

.019

(.012)

(.012)

(.012)

(.007)

(.008)**

(.008)**

.023

.037

.042

.023

.043

.046

(.012)*

(.014)***

(.014)***

(.008)***

(.011)***

(.011)***

.024

.031

.031

.023

.045

.046

(.011)**

(.012)**

(.012)**

(.008)***

(.010)***

(.010)***

–

–

–

–.488

–.347

–.340

(.063)***

(.070)***

(.070)***

5458 1480 .166

5185 1394 .077

5210 1404 .074

5458 1480 .261

5185 1394 .074

5210 1404 .066

1.853*

2.52**

2.761**

3.681***

5.164***

5.497***

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations. Pooled and firm fixed effects estimation. Standard errors are in parenthesis, where * denotes significance at the 10 %, ** at the 5 %, and *** at the 1 % percent level. Standard errors are estimated using the Huber / White / sandwich adjustment.

and (ii) when including constant unobserved effects, only the coefficient on affiliate jobs positively affects skill intensity at the parent firm. Since the dependent variable exhibits a logarithmic form, m can be interpreted as growth rate of normalised average wages. Other things equal, rising overall affiliate employment relative to domestic employment by 10 percentage points is accompanied by an increase in skill intensity at the parent firm between 0.09 % in the pooled regression and 0.26 % in the fixed effects model. Table 5 includes results, where multinational activity measures are separated into three broad country groups. The groupings are: Industrialised countries (IND), transition countries (TRANS), and developing countries (DEV) (for definitions see table 8 in the appendix). Estimation results with respect to domestic capital invest-

70

Robert Jäckle

ment and value added lie within the range of the corresponding parameters depicted in table 4. Again, higher levels of output contribute significantly to skill upgrading, and a larger capital intensity comes along with lower skill levels at parent firms (fixed effects model). The same holds true for yearly time dummies. As in the specification with only one common foreign activity measure, I find significantly positive coefficients on many of the binary time variables and F-tests of joint significance refer to regressions including a full set of year dummies. Table 5 Multinational Activities and Average Wages in German Manufacturing, Results Differentiated by World Regions pooled

Activity measure MNE act. IND MNE act. TRANS MNE act. DEV Log capital / value-added Log value added Dummy 1997 Dummy 1998 Dummy 1999 Dummy 2000 Dummy 2001 Constant N N manufac. firms R2 F-test time dummies

pooled

pooled

within

(3)

(4)

(1)

(2)

employmt. .010

turnover .0003

(.004)**

(.0002)*

(.003)*

.006

.066

–.010

(.006)

(.021)***

.010

within

within

(5)

(6)

turnover .0001

fixed ass. .0003

(.009)**

(.0001)

(.0009)

.010

.064

–.004

(.006)

(.006)

(.011)***

(.008)

.272

.003

.029

.240

.0003

(.006)*

(.072)***

(.004)

(.004)***

(.202)

(.009)

.024

.028

.044

–.037

–.037

–.031

(.007)***

(.007)***

(.008)***

(.011)***

(.011)***

(.012)***

.048

.040

.041

.161

.141

.109

(.003)***

(.004)***

(.004)***

(.036)***

(.041)***

(.041)***

fixed ass. employmt. .006 .025

.0004

.004

.007

–.001

.007

.007

(.011)

(.012)

(.012)

(.006)

(.007)

(.007

.021

.019

.023

.007

.016

.019

(.012)*

(.012)

(.012)*

(.007)

(.008)*

(.008)**

.010

.015

.018

.006

.013

.019

(.012)

(.012)

(.012)

(.007)

(.009)

(.008)**

.023

.039

.042

.024

.036

.046

(.012)*

(.013)**

(.014)***

(.008)***

(.011)***

(.011)***

.024

.031

.031

.025

.038

.046

(.011)**

(.012)**

(.012)**

(.008)***

(.011)***

(.010)***

–

–

–

–.488

–.342

–.347

(.065)***

(.069)***

(.069)***

5458 1480 .167

5185 1394 .141

5210 1404 .075

5458 1480 .267

5185 1394 .111

5210 1404 .066

1.878*

2.357**

2.824**

3.96***

4.07***

5.557***

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations. Pooled and firm fixed effects estimation. Standard errors are in parenthesis, where * denotes significance at the 10 %, ** at the 5 %, and *** at the 1 % percent level. Standard errors are estimated using the Huber / White / sandwich adjustment.

The Impact of FDI on the Skill Structure in German Manufacturing

71

As for MNE activities in industrialised countries, the pooled regression approach suggests that higher affiliate employment, turnover, and capital brings rising average wages in German manufacturing. In the fixed effects model, only increasing affiliate employment is associated with higher skill intensities at the parent operation. Assuming that developed regions attract FDI of the horizontal type, estimation results suggest that most of the subsidiaries located in these countries replicate final good production at home. At the same time, upstream activities might stay on the domestic market, and a positive effect on the skill level at parent firms may occur. Therefore, at this juncture, evidence for the branching-hypothesis of Head and Ries (2002) (see section 2.2) is found. When turning to transition countries, for both the pooled and the fixed effects model affiliate output is the only foreign activity measure that significantly affects domestic skill intensity. Rising affiliate turnover is associated with increasing average wages at home. These findings could indicate that (i) FDI in these countries is of the horizontal branching-type and mainly driven by market access motives, or / and (ii) German manufacturing firms vertically divide their production process, locating final goods production in transition countries, and re-export finished goods. Finally, when focusing on developing countries, skill-upgrading in German manufacturing can be explained by job and output transfers in the case of pooled OLS, and when including unobserved constant effects only affiliate employment positively affects domestic wages. In the latter case, a 10 percentage points increase in the workforce in developing countries is associated with an 0.29 % skill increase at the parent operation. Since developing countries are relatively abundant with lowskilled labour compared to Germany, my findings clearly confirm that German manufacturing MNEs exploit relative factor price differences (= vertical FDI) when investing at these locations.

6. Conclusions This paper analyses the impact of the international diversification strategy of German manufacturing MNEs on the domestic skill mix between 1996 and 2001. The descriptive figures in section 4, which show increasing foreign activities, decreasing home employment, and rising average wages, suggest a shift in labour demand towards the more skilled. To confirm that these trends actually reflect within-firm changes of the skill structure of manufacturing firms, a translog cost function approach is employed and demand functions for high-skilled labour are estimated. The main finding of this paper is that foreign activities of German manufacturing MNEs are positively correlated with higher average wages at domestic operations. I interpret this as evidence that part of the skill upgrading in German manufacturing is associated with the rising job export to foreign locations. Other things equal, a rise of overall affiliate employment relative to domestic employment by

72

Robert Jäckle

10 percentage points is accompanied by an increase in skill intensity at the parent firm by 0.1 % to 0.3 %. When distinguishing between different host regions, I find investments in industrialised countries consistent with the horizontal FDI motive, whereas investments in developing countries are driven by vertical production strategies. In the case of transition countries results are inconclusive, a distinction between the two motives is not possible.

References Becker, Sascha O. / Ekholm, Karolina / Muendler, Marc-Andreas (2006): “Skill Demand at Multinationals: Evidence from Linked Employer-Employee Data,” mimeo. Becker, Sascha O. / Ekholm, Karolina / Jäckle, Robert / Muendler, Marc-Andreas (2005): “Location Choice and Employment Decisions: A Comparison of German and Swedish Multinationals,” CESifo Working Paper, January 2005, http://1374.ssrn.com/abstract=648081 1374. ssrn. com/abstract=648081. – (2005): “Location Choice and Employment Decisions: A Comparison of German and Swedish Multinationals,” Review of World Economics, 141 (4), 693 – 731. Becker, Sascha O. / Jäckle, Robert / Muendler, Marc-Andreas (2005): “Kehren deutsche Firmen ihrer Heimat den Rücken? Ausländische Direktinvestitionen deutscher Unternehmen,” ifo Schnelldienst, January 2005, 58 (1), 23 – 33. Berman, Eli / Bound, John / Griliches, Zvi (1994): “Changes in the Demand for Skilled Labor within U.S. Manufacturing: Evidence from the Annual Survey of Manufacturers,” The Quarterly Journal of Economics, May 1994, 109 (2), 367 – 397. Burdett, Kenneth / Mortensen, Dale T. (1998): “Wage Differentials, Employer Size, and Unemployment,” International Economic Review, 39 (2), 257 – 273. Carr, David L. / Markusen, James R. / Maskus, Keith E. (2001): “Estimating the KnowledgeCapital Model of the Multinational Enterprise,” American Economic Review, June 2001, 91 (3), 693 – 708. Deutsche Bundesbank (1998): “The Methodological Basis of the Deutsche Bundesbank’s Corporate Balance Sheet Statistics,” Monthly Report 10. Feenstra, Robert C. / Hanson, Gordon H. (1996): “Globalization, Outsourcing, and Wage Inequality,” American Economic Review, 86 (2), 240 – 245. – (1999): “The Impact of Outsourcing and High-Technology Capital an Wages: Estimates for the United States, 1979 – 1990,” Quarterly Journal of Economics, August 1999, 114 (3), 907 – 940. Geishecker, Ingo / Görg, Holger (2004): “Winners and Losers: Fragmentation, Trade and Wages Revisited,” IZA Discussion Paper, January 2004, 982. Hansson, Pär (2001): “Skill Upgrading and Production Transfer within Swedish Multiationals in the 1990s,” FIEF Trade Union Institute for Economic Research Stockholm Working Paper, January 2001, 166.

The Impact of FDI on the Skill Structure in German Manufacturing

73

Head, Keith / Ries, John (2002): “Offshore Production and Skill Upgrading by Japanese Manufacturing Firms,” Journal of International Economics, October 2002, 58 (1), 81 -1 05. Helpman, Elhanan (1984): “A Simple Theory of International Trade with Multinational Corporations,” Journal of Political Economy, March 1984, 92, 451 – 471. – (1985): “Multinational Corporations and Trade Structure,” Review of Economic Studies, 52, 443 – 458. Helpman, Elhanan / Krugman, Paul (1985): Market Structure and Foreign Trade, Cambridge: MIT Press. Jäckle, Robert (2006): “Going Multinational: What are the effects an home market performance?,” Bundesbank Discussion Paper Series 1, No 03. Lipponer, Alexander (2003): “A “New” Micro Database for German FDI,” in: Heinz Herrmann and Robert Lipsey (eds.), Foreign Direct Investment in the Real and Financial Sector of Industrial Countries, Berlin: Springer, 215 – 244. Marin, Dalia (2004): “A Nation of Poets and Thinkers-Less so with Eastern Enlargement? Austria and Germany,” CEPR Discussion Paper, 4358. Markusen, James R. (2002): Multinational Firms and the Theory of International Trade, Cambridge (Massachusetts) / London (England): The MIT Press. Markusen, James R. / Venables, Anthony J. / Konan, Denisee E. / Zhang, Kevin H. (1996): “A unified treatment of horizontal direct investment, vertical direct investment, and the pattern of Tade in goods and Services,” NBER Working Paper, 5696. Muendler, Marc-Andreas / Becker, Sascha O. (2006): “Margins of Multinational Labor Substitution,” IZA DP No. 2131, 2006, University of California, San Diego. Slaughter, Matthew J. (2000): “Production Transfer within Multinational Enterprises and American Wages,” Journal of International Economics, April 2000, 50 (2), 449 – 472.

Appendix A. Are Average Wages a Good Proxy for the Skill Intensity in German Manufacturing? The study at hand proxies the skill intensity in German manufacturing using average wages paid by each firm on the home market. It is constructed as the wage bill divided by employment, where both variables are available in the USTAN data set. By splitting up the overall wage bill into the sum of earnings of high- and low-skilled employees one obtains: PNj …3†

j ˆ w

iˆ1 wij ˆ Nj

PNjL iˆ1

wLij ‡

PNj iˆNjL ‡1

NjL ‡ NjH

wH ij

ˆ

H w  Lj NjL ‡ w H j Nj

NjL ‡ NjH

;

 j is the average wage paid by firm j, Nj is the overall number of workers employed by where w  Lj and w H firm j, NjL and NjH are low- and high-skilled employment, and w j denote average wages for low- and high-skilled employees at firm j, respectively. Some further transformation of equation 3 yields:

74 …4†

Robert Jäckle  Lj w

j ˆ w 1

H w H j Nj H H H  wH wj Nj ‡ j Nj



w H j

w  Lj H  wj

ˆ

1

 Lj w ; Ij  Pj

where Ij is the skill intensity at firm j, and Pj is wage premium of skilled over unskilled  Lj and taking logs results in: employees. Dividing both sides of equation 4 by w …5†

log

j w  Lj w

! ˆ

log…1

Ij  Pj †  Ij  Pj :

The second equality holds only for small values of the product of skill intensities and wage premia. Equation 5 states that the log of average payments to a firm’s domestic workforce over the average low-skilled wage is roughly proportional to the skill intensity at company j, given the skill premium on wages is constant in the period under consideration. According to calculations using the German Socio-Economic Panel (GSOEP) the premium for higher education as opposed to basic education (Pj ) in Germany between 1996 and 2001 remains roughly constant at 41 %.12 At the same time, the average skill-intensity in Germany is between 33% and 41%.13 Therefore, the respective values of log…1 Ij  Pj † lie in between 0.15 and 0.18 and the approximated values of Ij  Pj are 0.14 and 0.17, respectively.

12 I classify the workers’ educational attainment as low-skilled if they have either no educational degree at all, no school degree plus vocational training, or a lower school degree without vocational training. High-skill employment refers to persons with a high-school degree plus additional vocational training, higher technical college, or a university degree. To calculate average wages, fulltime gross earnings (incl. 13th month salary, vacation and Christmas bonus) of the respective skill groups are employed. Since I made use of the samples A – F of the GSOEP a weighting scheme is used to overcome the problem of different sampling probabilities when inferring average values of the target population. 13 A rough estimate of the above numbers can be obtained using the overall wage sum and the wage sum of unskilled employees from the GSOEP. Again, all observations are weighted according to their sample probabilities.

The Impact of FDI on the Skill Structure in German Manufacturing

75

Table 6 Industry Definitions and Panel Attrition

Food products and beverages Tobacco products Textiles Wearing apparel; dressing of furniture Tanning and dressing of leather Wood and cork (no furniture) Pulp, paper and paper products Publishing and printing Coke, petroleum, nuclear fuel Chemicals Rubber and plastic Other non-metallic Basic metals Fabricated metal Machinery and equipment Office machinery and computer Electrical machinery Communication equipment Medical and precision instruments Motor vehicles, trailers Other transport equipment Furniture Recycling total

1996

1997

1998

1999

2000

2001

total

(1)

(2)

(3)

(4)

(5)

(6)

(7)

53  3a) 42

54 3 42

43 3 42

40 3 39

34 3 35

34 3 35

258 10 235

38

30

28

31

27

22

176

8

10

11

9

7

5

50

10

14

13

12

9

10

68

19 13

20 13

19 16

16 18

15 18

12 13

101 91

5 103 60 43 45 101 265

5 110 65 45 39 105 266

5 98 62 40 31 97 261

4 93 65 34 32 93 247

3 94 60 29 39 90 241

3 80 50 27 37 78 213

23 578 362 218 223 564 1493

8 58 17

7 56 16

7 55 17

3 62 23

4 57 24

3 48 22

31 336 119

64 42 5 33 0

60 46 5 36 3

63 46 6 30 3

66 47 5 26 3

62 50 5 24 3

47 45 4 23 3

362 276 30 172 7

1.034

1.047

993

969

929

811

5.783b)

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations. a) Data protection guidelines of the Deutsche Bundesbank oblige researchers to hide figures where the number of observed firms is smaller than three. b) 5,783 observations from 1,557 unique firms between 1996 an 2001.

76

Robert Jäckle Table 7 Summary Statistics N

Mean

(1)

(2)

(3)

(4)

(5)

Dependent variable Log average wage normalised

5670

.366

.318

.074

.634

Regressors MNE activities employment MNE activities turnover MNE activities fixed assets Log capital/value added Log value added

5606 5312 5430 5596 5596

1.620 7.108 .573 –.776 17.421

11.572 196.442 6.686 .891 1.415

.023 .033 .010 –1.826 15.862

1.570 1.050 .961 .229 19.185

Regressors regional specification MNE activities IND Employment Turnover Fixed assets

5606 5312 5430

.937 6.798 .279

7.807 194.036 2.360

0 0 0

.854 .937 .565

MNE activities TRANS Employment Turnover Fixed assets

5606 5312 5430

.355 .040 .142

3.743 .766 1.541

0 0 0

.333 .055 .259

MNE activities DEV Employment Turnover Fixed assets

5606 5312 5430

.328 .270 .151

4.191 9.232 2.983

0 0 0

.194 .085 .190

Std. dev. 10 % pctl. 90 % pctl.

Source: USTAN and MIDI, Deutsche Bundesbank 1996 – 2001, own calculations. All summary statistics are on firm-year level. Regressors are ownwership-weighted. The dependent variable is normalized using industry-year wages form the German Statistical Office. Foreign activity variables are measured relative to domestic activities.

The Impact of FDI on the Skill Structure in German Manufacturing Table 8 Regional Definitions IND

TRANS DEV

Western European countries (EU 15 plus Norway and Switzerland) and Overseas Industrialised countries (Canada, Japan. USA, Australia, New Zealand, as well as Iceland and Greenland) Central and Eastern European countries (accession countries and candidates for EU membership) Developing countries (Asia-Pacific, Hong Kong, South Korea, Singapore, Taiwan, China, Mongolia, and North Korea; Russia and Central Asian economies) and other developing countries (South Asia (India / Pakistan), Africa, Latin America, the Middle East; including dominions of Western Europe countries and the United States)

77

The Impact of FDI on the Skill Structure in German Manufacturing – Comment By Alexander Lipponer*

1. Introduction The paper examines the impact of foreign direct investment (FDI) by German manufacturing enterprises on the domestic labour market. It focuses on the changes to the domestic parent company’s skill intensity which are triggered by FDI. In so doing, it makes an important contribution to the body of work analysing the implications of FDI. Until now, there has been a bulk of research on FDI determinants (for an overview see e.g., Blonigen 2006) but it still remains a particular dearth of empirical research into the effects of FDI, particularly on the domestic labour markets, especially as regards Germany. The author concludes that increased FDI activity by (German) firms results in higher skill intensity in the home country. In a panel estimation, the number of staff employed by the foreign affiliates, in particular, has a significant effect in the manner described. At the same time, this estimation also has the highest explanatory power (26%). In addition to the number of staff at the foreign affiliates, this paper also draws on their turnover and fixed assets as a measure of foreign activity.

2. Body of Data The paper is based on two Bundesbank micro databases – MIDI and USTAN – supplemented by sectoral wage data from the Federal Statistical Office. The Bundesbank MIDI data relates to German FDI abroad and FDI in Germany and is based on the statistics pertaining to international capital links. It is a comprehensive census on the basis of the Foreign Trade and Payments Act (Außenwirtschaftsgesetz) which mostly comprises individual data from the affiliates’ balance sheets between 1989 and 2004. However, the data up to 1995 are anonymised in such a way that a panel data set is only available from 1996 onwards. For this reason the study begins with that year. It should be noted that, prior to 2004, the number of staff at * Department of Economics, Deutsche Bundesbank, Frankfurt am Main, Germany, Tel.: +49 69 95 66 – 22 15, E-Mail: [email protected].

80

Alexander Lipponer

foreign branches was optional information. Where omitted, the figure was estimated based on the turnover data (affecting approximately 5 – 10 % of the reports). Indicators for parent firms were added to MIDI only recently. Hence, the data on domestic investors required for this study had to be obtained from another source (in this case USTAN). USTAN is the Bundesbank’s database of corporate balance sheet statistics. It contains data from 1971 onwards and, prior to monetary union, was principally used to check firms’ creditworthiness for rediscount business purposes. Until 1997, the database is comprehensive. After that date, however, the number of enterprises recorded drops significantly: to less than half the 1997 level in 2001 and to virtually zero thereafter. This is also likely to account for the decline in data matches found during cross-checking, as shown in table 1 of the paper, despite a simultaneous rise in the number of foreign direct investors. Hence, there is a meaningful overlap of the data – restricted to corporations – between 1996 and 2000.

3. Estimation It is not possible to measure the skill intensity SH of firm j at time t directly using the available data. In the literature, a proxy (the average wage relative to the wage of low-skilled workers) is suggested for such an eventuality. As this is not provided by the data either, the estimations use the ratio of the average wage to the average sectoral wage … ws;t †: …1†

SH j;t 

 j;t w  j;t w  :  Lj;t w  s;t w

Owing to a lack of data, the analysis further assumes that quality-adjusted high and low-skilled labour does not vary across firms. This assumption means that these variables can be replaced with time fixed effects in the estimation equation. Ultimately, the equation to be estimated is capital

firm fixed effects

     j;t Kj;t w ˆ 0 ‡ t t ‡ k log ‡ v log Vj;t ‡ m MNEj;t ‡ cj ‡ "j;t log  s;t Vj;t w |‚‚‚‚‚‚‚‚‚{z‚‚‚‚‚‚‚‚‚} see above

time fixed effects

value added

measure for the multinational activity

and leads to the following key results (for the panel estimator and employment abroad as the MNE variable):

The Impact of FDI on the Skill Structure in German Manufacturing – Comment

81

Table 1 Key Results of Panel Regressions employment 0.026*** .. .

employment . . . . . . in industrialised countries . . . in transition countries . . . in developing countries .. .

0.025*** 0.010 0.029*** .. .

Firms’ normalised average wages and hence, (in accordance with the hypothesis) the proportion of high-skilled labour to total labour input increases significantly as a result of higher FDI activity. From this, the author concludes that increased activity abroad causes the staff structure to shift to the detriment of low-skilled employees. This appears to be more or less unconnected with the specific target country or the motivation behind the FDI decisions. FDI in transition countries, which has aroused considerable public debate, is alone in appearing to have no significant effect on the domestic skill structure.

4. Remarks All in all, the observation period (1996 – 2001) appears decidedly short for the subject at hand. Many adjustments most certainly have already taken place in previous years or manifest themselves only over a longer period. Of all variables, it is the measure for the multinational activity (MNE) with the most pronounced effects in the panel estimation (employment) which is subject at least to a degree of uncertainty (see above). The impact on the estimation results is unclear. Potentially, other factors may be driving skill intensity and thus average wages. R&D activity may be important in this context and should therefore be taken into account. The omission of factors like this could create an omitted variable bias, which even the fixed effects introduced in the panel estimation may be unable to offset. Furthermore, average wages may not be tied too close to skill intensity and at least the second part of equation (1) might not be a reliable approximation. Do the matching results of the coefficients for industrialised and developing countries justify using different explanatory paradigms in the study? The positive coefficient is justified with horizontal FDI for industrialised countries and with vertical FDI in the case of developing countries. The results do not contradict the explanatory paradigms, though the connection is in no way causal. By contrast, the author states “. . . my findings clearly confirm . . .”. The findings may be correct, yet cannot be proven on the basis of this analysis. 6 Supplement 57 – 2006

82

Alexander Lipponer

Using a control group of firms which are not active abroad (source: USTAN), it would be possible to study whether they respond differently. How are they affected by productivity gains? How has their skill structure changed? Without the analysis of such a control group the possible existence of a selection bias in the data can not be rejected. Recent literature suggests that it is the larger, productive firms that engage in FDI (e.g., Helpman et al. 2004). Another possible selection bias in the data may stem from the aforementioned decline in the number of USTAN reports. This could, perhaps, be eliminated by linking the data with commercial databases of corporate balance sheet statistics. The number of employees abroad, particularly in the case of M&As, gives no direct information about the relocation or establishment of production sites. M&As do not generally represent any direct real economic effects, merely a transfer of ownership. In the event of a takeover, for example, the number of employees abroad rises without having a simultaneous impact on domestic employees. That may, by the way even in the case of greenfield investment, happen later, with a certain time-lag. It therefore seems to be necessary to account for lagged reactions in the estimation. In many cases, those enterprises subject to reporting requirements in MIDI are not the same as the manufacturing parent firms in Germany. Instead, they are rather “small” holding subsidiaries or Special Purpose Entities (SPEs) within the group, which manage and therefore report FDI. When cross-checking data, the domestic group structures ought to be taken into account. If they are not, then, as in the study at hand, in many cases, the indicators used for domestic (parent) companies required to report – and the changes in these indicators – cannot provide meaningful data on the adjustment measures at domestic production sites.

References Blonigen, B. (2006): “A Review of the Empirical Literature on FDI Determinants”, Atlantic Economic Journal, forthcoming. Helpman, E. / Melitz, M. J. / Yeaple, S. R. (2004): “Exports vs FDI with Heterogeneous Firms”, American Economic Review 94, 1, 300 – 316.

Relocation, Offshoring and Labour Market Repercussions: The Case of the German Automobile Industry in Central Europe By Peter Nunnenkamp*

Abstract The paper raises the proposition that Central Europe’s integration into the international division of labour has added significantly to competitive pressure in the German automobile industry. Based on production and trade data, we trace two dimensions of competitive pressure: relocation of assembly operations and offshoring of automotive parts production. The knowledge-capital model of multinational enterprises provides the analytical basis for the discussion of labour market repercussions. Vertical foreign direct investment in Central Europe may have helped the relatively favourable employment and earnings record of the German automobile industry, compared to other manufacturing industries. Yet recent industrial disputes can be attributed, though not exclusively, to the emergence of Central Europe as an attractive location for assembly operations and autoparts production. Employment and wages diverged considerably within the German automobile industry. Relative to skilled workers, the labour market situation of less skilled workers deteriorated significantly. JEL Classifications: F14, F23, L62 Keywords: vertical FDI, trade in intermediates, relative wages, employment

1. Introduction The automobile industry is widely regarded as an export champion in Germany. Only France and Japan exported more automobiles than Germany in 2004 (VDA a, 2005: 362). The contribution of exports of road vehicles (SITC 78) to Germany’s total exports of manufactures increased from 18.5 percent in 1993 to 22 percent in 2004 (OECD 2005). Moreover, employment and earnings opportunities have tradi* The Kiel Institute for the World Economy, Kiel, Germany. E-mail address: [email protected]. This paper is a substantially revised and updated version of Nunnenkamp (2005). Comments by Joergen Ulff-Moeller Nielsen and Konrad Pawlik from the Aarhus School of Business are gratefully acknowledged. The author also would like to thank the discussant, PD Dr. André Jungmittag, and the participants of the 69th Annual ARGE Meeting for critical comments and useful suggestions. The usual disclaimer applies. 6*

84

Peter Nunnenkamp

tionally been favourable in the German automobile industry, compared to the manufacturing average (Spatz and Nunnenkamp 2002a, 2002b). Yet the automobile industry offers a particularly interesting example to evaluate the fiercer competition from Central European countries as well as the production, trade and labour market repercussions in traditional locations such as Germany. The industry witnessed a series of industrial disputes in recent years, most of which were triggered by the threat of German companies to relocate production to cheaper locations, notably in Central Europe. For example, Opel, the German subsidiary of General Motors, decided in mid-2004 to locate part of its Zafira production in Gliwice, Poland, even though the assembly line at the company’s headquarter in Rüsselsheim had considerable spare capacity. The decision was based on a comparative analysis that revealed strong competitive advantages of the former location. Low-wage competition from the neighbouring Czech Republic notwithstanding, BMW decided to build its new production site in Saxony. However, Bosch, a major supplier of autoparts, revealed plans to relocate further 800 workplaces from France to the Czech Republic. Continental, a producer of tyres, announced in fall 2005 to close down production lines in Hanover, even though workers had agreed a few months earlier to longer working hours in order to reduce labour costs. Against this backdrop, the paper raises the proposition that Central Europe’s integration into the international division of labour has added significantly to competitive pressure in the German automobile industry, including the production of autoparts, even though this industry is relatively skill and technology intensive and represents a traditional stronghold of advanced countries.1 According to the knowledge-capital model of multinational enterprises (Carr et al. 2001), the labour market repercussions can be expected to depend on the type of foreign direct investment (FDI) (Section II). Wage inequality or unemployment of less qualified workers in Germany are supposed to increase if vertical FDI, which involves the relocation of relatively labour intensive stages of production to lower-income countries, plays a major role with regard to the automobile industry’s engagement in Central Europe. The evidence on the relocation of assembly operations and offshoring of autoparts production, presented in Sections III and IV, suggests that this is indeed the case. The labour market effects of fiercer competitive pressure are assessed in Section V. We argue that the recent controversy on whether the automobile industry exemplifies the case of Germany degenerating into a bazaar economy misses an important point, namely diverging employment and earnings trends within this industry. Section VI concludes.

1 Vickery (1996) and Weiß (2000) show that the development and manufacturing of automobiles requires increasing R&D and involves significant fixed costs.

Relocation, Offshoring and Labour Market Repercussions

85

2. Analytical Background and Earlier Findings Based on standard theoretical models on the distributional effects of the liberalization of trade with, and foreign direct investment (FDI) in lower-income countries, the integration of Central Europe into international production and sourcing networks can be expected to negatively affect the labour market situation of relatively low skilled workers in high-income countries such as Germany (Spatz and Nunnenkamp 2002b: 477).2 In a recent survey on trade and wages, Feenstra and Hanson (2003) argue that trade in intermediate inputs is a potentially important explanation for the increase in the wage gap between skilled and unskilled workers in advanced economies. The literature on the motives and effects of FDI offers further insights which help analyse the labour market repercussions of automobile production in countries with relatively low per-capita income. Marin et al. (2003) and Marin (2004) suspect that the wage and employment effects of outward FDI by economically advanced countries in lower-income countries depend on the type of FDI:  Companies undertaking horizontal FDI produce the same goods and services in their home country and in the host countries.3 This type of FDI is often motivated by trade barriers, transportation costs and other transaction costs that discourage exports (Carr et al. 2001). FDI is a means to avoid such costs. Horizontal FDI is driven by market considerations. That is why this type of FDI is also known as market-seeking FDI (UNCTAD 1998: 91).  Companies undertaking vertical FDI fragment the production process geographically and locate specific stages of the value chain in countries offering the relevant cost advantages.4 This type of FDI is motivated by cost considerations. Investors make use of varying factor endowments and differences in factor prices across countries (Zhang and Markusen 1999). FDI of this type is also known as efficiency-seeking FDI (UNCTAD 1998: 91).

According to Marin et al. (2003), wage inequality or unemployment in economically advanced economies is likely to increase if outward FDI is of the vertical type. This is because the investor relocates the relatively labour intensive stages of production to lower-income countries, thereby reducing the demand for unskilled workers in the home country.5 Unless unskilled workers agree to lower relative wages, they will face deteriorating employment opportunities. By contrast, these authors do not expect horizontal FDI to have effects on wage inequality or employment opportunities in the advanced country. See also the literature given there. For an early model of horizontal FDI, see Markusen (1984); more recent models include Markusen and Venables (1998, 2000). 4 For an early model of vertical FDI, see Helpman (1984); see also Helpman and Krugman (1985). 5 Feenstra and Hanson (2003) argue along similar lines. However, they consider foreign outsourcing which extends beyond FDI-related offshoring and includes arm’s-length trade in intermediate goods. 2 3

86

Peter Nunnenkamp

In particular the so-called knowledge-capital model of multinational enterprises (Carr et al. 2001) offers several arguments to suspect that the engagement of German automobile companies in Central Europe is largely vertical in nature. In many respects, this engagement resembles the vertical production networks of US-based automobile companies with their affiliates in NAFTA partner countries, notably Mexico (Hanson et al. 2005). First, the motive for horizontal FDI to avoid high trade and transaction costs associated with exporting from the German home base should be of minor importance for serving Central European markets.6 These markets are fairly close to the home base of German investors (i.e., transportation costs are relatively low), and the protection of these markets is rather weak as trade costs resulting from import barriers have been removed since various countries prepared for EU membership.7 Second, markets for (new) automobiles in Central Europe are small compared to the German home market.8 This limits the potential to exploit (plant-level) economies of scale in assembly operations located in Central European countries, which, in turn, should reduce the incentive to engage in horizontal FDI (Carr et al. 2001). Third, different factor endowments and factor price differentials between Germany and Central Europe, in combination with low trade costs and geographical proximity, provide incentives to undertake vertical FDI. Central Europe tends to be better endowed of relatively skilled labour than many developing countries. According to Zhang and Markusen (1999: 237), the case for vertical FDI no longer exists if “countries become extremely different”, i.e., sufficiently skilled labour being so scarce in the potential host country that multinational companies will find it difficult to hire local staff such as technicians and administrative employees. Likewise, vertical FDI is supposed to depend on the host country meeting minimum standards with regard to power supply, transport and telecommunication infrastructure as well as legal institutions. In contrast to many developing countries, it can be assumed that Central European countries fulfil these basic requirements for vertical FDI to take place. Nevertheless, the labour market repercussions of the engagement of the German automobile industry in Central Europe are open to question for both analytical and empirical reasons. The differentiation between horizontal and vertical FDI is not as clear-cut as it might appear at first sight.9 On the one hand, the labour market 6 In Helpman’s (1984) model of vertical FDI, trade costs were assumed to be zero. As noted by Carr et al. (2001), this assumption, in combination with plant-level economies of scale, removes the motive for horizontal FDI. 7 Prospective EU member countries had abolished import duties on cars imported from the EU by 2001 (van Tulder 2004: 106). 8 Even in Poland, i.e., the largest Central European market for automobiles, first registrations of passenger cars in 2001 – 2004 did not exceed one tenth of first registrations in Germany (VDA b). 9 According to Hanson et al. (2005: 666), different affiliates “fall in a continuum with pure horizontal FDI at one extreme and pure vertical FDI at the other extreme.” Ekholm et al.

Relocation, Offshoring and Labour Market Repercussions

87

implications of vertical FDI depend on whether the cost reduction associated with such a strategy results in an overall expansion of the investing company, including complementary operations at home (Becker et al. 2005). On the other hand, FDI appears to be horizontal if automobile companies produce the same final good, namely finished cars, at home and abroad. Yet, under certain circumstances, this engagement may be motivated by cost considerations and can have labour market repercussions at home. For instance, this may be the case if the automobile company produces higher quality cars at home, but relocates the production of models serving the lower segment of the market to countries offering cost advantages in the assembly of such cars. Even FDI undertaken for the assembly of cars that are similar to those assembled at home can be considered vertical as long as technology intensive and human-capital intensive activities such as the design and development of cars are concentrated in the relatively skill-abundant home country.10 Empirical studies portray an ambiguous picture with respect to the type of FDI in Central Europe and possible labour market repercussions.11 Earlier surveys typically suggest that market considerations are prominent motives for FDI in Central Europe (e.g., Lansbury et al. 1996: 104). Bechert and Cellarius (2004) note that “the great majority” of local employees of German subsidiaries in this region “are also involved in production that is intended for local markets.” As concerns the automobile industry, Sturgeon and Florida (1999: 53) find “a large measure of convergence toward building vehicles where they are sold.” Buch et al. (2005) show that the market size of host countries has a relatively large impact on German FDI in the automobile industry. The estimation results of Carstensen and Toubal (2003) indicate that both horizontal and vertical FDI exists in Central Europe.12 Marin et al. (2003) show that the affiliates of German companies in the machinery and transport equipment sector of Central and Eastern European countries deliver almost 40 percent of production to their German parents, which, according to (2003) model so-called export-platform FDI which has both horizontal and vertical features. The model of Markusen and Venables (2005) encompasses both market-serving and exportplatform motives for fragmentation of production. While it mainly depends on trade costs whether countries engage in market-serving or export-platform activity, it is mainly determined by factor endowments whether countries specialize in production of components or assembly of final goods. 10 Carr et al. (2001) derive the motive for vertical FDI, i.e., locating knowledge intensive activities such as R&D where skilled labour is relatively cheap and production where unskilled labour is relatively cheap, from two assumptions: (i) knowledge intensive activities can be geographically separated from production and supplied to production facilities at low cost, and (ii) production requires less skills than activities such as R&D. 11 This also applies to the earlier literature on the labour market effects of US FDI in lower-income countries (notably in Mexico in the context of NAFTA). This literature is shortly reviewed in Nunnenkamp (2006). See also Blomström et al. (1997) as well as Braconier and Ekholm (2000) on Swedish FDI, and Federico and Minerva (2005) on Italian FDI. 12 These authors find a robust and strong impact of the market potential of host countries on FDI. At the same time, relative unit labour costs are shown to exert a significant influence on FDI.

88

Peter Nunnenkamp

Hanson et al. (2001), is a clear indication of vertical FDI. The incentives for vertical specialization are stressed by Marin (2004), who finds that German direct investors can reduce unit labour costs by about 70 percent in several Central European countries. Few studies have assessed the repercussions of FDI-related relocation and offshoring on German labour markets.13 Becker et al. (2005) estimate a translog cost function to assess how outward FDI affects employment at home. With respect to German companies in Central and Eastern Europe, it turns out that a one percent wage reduction at existing affiliates in this region reduces employment in German parent companies, though only by about 0.04 percent. By contrast, Konings and Murphy (2001) reject the hypothesis that FDI by European direct investors, about 30 percent of which were based in Germany, has contributed to a relocation of domestic jobs to Central and Eastern Europe. By estimating the labour demand function of German parent companies, Marin (2004) even finds that a 10 percent wage decline for affiliates located in Central European EU-accession countries increases employment at home by 1.6 percent. Domestic job creation is attributed to cost savings and, thus, improved competitiveness that parent companies achieved through FDI-related offshoring. Apart from ambiguous findings, the aforementioned studies offer an incomplete picture of possible labour market repercussions of German FDI in Central Europe. In addition to the effects on employment in the parent companies, offshoring may affect employment in German companies which traditionally served as input suppliers of these parent companies. This suggests to assess labour market effects at the industry level, rather than only at the company level. Furthermore, vertical FDI in Central Europe may not only affect employment at home but also the wages paid there. Finally, none of the studies captures the distributional effects that can be expected to result from vertical FDI. More precisely, the question whether less skilled workers in Germany suffered deteriorating relative employment and wage prospects due to relocation and FDI-related offshoring to Central Europe is still unsettled. Data constraints typically prevent an adequate differentiation between skill groups. However, as shown below for the automobile industry, industry-specific case studies may offer at least tentative insights into the distributional effects of vertical FDI. Before returning to this issue in Section V, we present some stylised facts on the relocation of automobile assembly (Section III) and offshoring of autoparts production (Section IV) that are supposed to reveal the type of FDI undertaken by the German automobile industry in Central Europe. 13 The survey on the labour market implications of global production sharing by Feenstra and Hanson (2003) clearly reveals that the relevant literature is strongly concentrated on the case of the United States. Hardly any references are made to the case of Germany. At the same time, Feenstra and Hanson stress that more research is needed on outsourcing in Central and Eastern Europe, which should be of particular relevance to German companies.

Relocation, Offshoring and Labour Market Repercussions

89

3. Relocation of Car Assembly to Central Europe FDI by the German automobile industry, including autoparts, in Central and Eastern Europe has gained considerable momentum. FDI stocks soared fivefold since 1995 to A 80 billion in 2003 (Deutsche Bundesbank 2005). Since the late 1990s, Central and Eastern Europe has hosted higher FDI stocks than Latin America, which had traditionally been the preferred investment location of the German automobile industry outside the advanced OECD area. Hungary, the Czech Republic and Poland accounted for 80 percent of FDI stocks in the region in 2003.14 FDI stocks held by the German automobile industry in Hungary exceeded those in China, even though automobile multinationals consider China to be the most promising market and are eager to build or acquire production capacities there. As noted before, it is fairly difficult to clearly distinguish between horizontal and vertical FDI. Yet there are several indications that the activities of the German automobile industry in Central Europe are not restricted to horizontal FDI. The regime change in Central Europe, the region’s opening up to world markets and the accession of various countries to the EU not only promised new markets and export opportunities for German automobile producers, but also offered profitable investment opportunities.15 In contrast to China and Latin America, Central Europe has emerged as an important export platform for German automobile producers. Production and trade data for passenger cars (units) reveal that the character of German FDI in Central Europe differs from that in other low-income locations (Table 1). In China and Latin America (proxied by the most important locations, Brazil and Mexico), car production of German companies developed independently from trade. German car exports to these markets hardly existed before companies invested there; exports were no reasonable option because of high import barriers. As a consequence, labour market repercussions in Germany resulting from exports being replaced by foreign production are highly unlikely. At the same time, German car imports from China and Latin America remained marginal throughout the period under consideration, largely because production in China and Brazil lacked international competitiveness and transportation costs are high. Hence, production in Germany could hardly be affected negatively by rising imports from these locations. 14 The Slovak Republic, most likely, accounts for much of the rest, mainly because of Volkswagen’s engagement in this country. However, the Slovak Republic is not listed as an individual host country in the Bundesbank statistics (Deutsche Bundesbank 2005). 15 Humphrey and Memedovic (2003: 34) reckon: “The initial attraction for extending production networks from Western Europe to the peripheral regions was a combination of access to growing markets and reducing costs through the development of low-cost production sites.” For a similar line of reasoning with regard to Volkswagen’s acquisition of Skoda, see Pries (1999). The survey results of Dichtl and Hardock (1997) reveal that labour costs played an important role in motivating the first waves of relocation to Central Europe. Van Tulder and Ruigrok (1998) as well as van Tulder (2004) point out that European car manufacturers pursued different strategies: Some companies rated Central Europe primarily as a market, some as a production site, and some aimed at both.

90

Peter Nunnenkamp Table 1 Passenger Cars: Foreign Production by German Companies and German Imports and Exports, 1990 – 2004 (1000 units)

1990 1996 2002 2004 a)

Central Europea) prod. imp. exp.

prod.

China imp.

exp.

0.0 240.1 782.4 817.3

0.0 226.4 437.6 575.5

0.0 0.0 0.3 5.6

2.8 4.0 22.9 44.2b)

11.0 68.9 261.6 233.3

6.6 126.2 100.7 114.8

Brazil and Mexico prod. imp. exp. 425.8 735.0 799.8 748.9

1.3 6.9 24.0 42.6

1.1 11.9 33.9 22.9

Czech Rep., Hungary, Poland and Slovak Rep. – b) 2003.

Source: VDA (a).

A different situation prevails with regard to Central Europe. Almost one third of car production by German companies in the Czech Republic, Hungary, Poland and the Slovak Republic was destined for the German market in recent years. German car imports from these countries have multiplied since the early 1990s. Van Tulder and Ruigrok (1998: 10) expect this development to have labour market repercussions in Germany, as “the (threat of) reimportations puts the domestic bargaining arena under pressure.” Labour market repercussions resulting from car imports may be moderate for the time being. The ratio of imports from the four Central European locations to passenger car production in Germany increased substantially from 1.5 percent in 1995 / 96, but the ratio did not exceed the 5 percent mark in 2003 / 04. Moreover, it might be questioned that assembly operations in Germany were affected significantly since producers such as Volkswagen used production sites in Central Europe to complement their existing product range. Consequently, substitution effects may be minor even though the focus on relatively cheap lower-end cars may be characterized as vertical FDI (Section II). However, German car exports to the four Central European countries increased by much less than German imports. As a result, Germany reported a considerable import surplus in recent years (Table 1). Furthermore, the import surplus is no longer in terms of units only, as observed in the 1990s by van Tulder and Ruigrok (1998). In value terms, the German trade balance for motor vehicles (including chassis) vis-à-vis the four Central European countries switched from an export surplus of A 250 million in 1995 to a deficit of almost A 3 billion in 2004 (VDA a). The development over time of German exports of passenger cars to Central Europe provides further clues to substitution effects. Exports peaked in 1996 and remained almost flat thereafter, i.e., exactly when production by German automobile companies in the region soared from about 0.2 million units per annum to 0.8 million units in 2004 (Nunnenkamp 2005: Figure 2). It is almost impossible to decide how exports to Central Europe would have developed if German companies were

Relocation, Offshoring and Labour Market Repercussions

91

not engaged in assembly operations in this region. Yet, it is striking that the Czech Republic, Hungary, Poland and the Slovak Republic together absorbed less German exports of passenger cars than neighbouring Austria in 2002 (Nunnenkamp 2004). Additional substitution effects may have occurred in third markets if German companies exported assembled cars from Central European production locations to markets other than the German home market.16 To summarize, the import and export patterns associated with assembly operations of German automobile companies in Central Europe suggest that production in this region is more likely to affect domestic production and, thus, labour markets than the assembly operations of German companies in other host countries with relatively low per-capita income. In contrast to Latin America and China, production locations in Central Europe allowed for an internationally competitive assembly of automobiles close to European core markets.

4. Offshoring Production of Autoparts Substitution effects at the level of assembled cars are not the only transmission mechanism through which the emergence of Central Europe as an important player in the automobile industry may have repercussions on German labour markets. Additional labour market effects can be expected to result from the offshoring of automobile parts production to Central Europe through vertical FDI by car assemblers and so-called follow sourcing by parts suppliers.17 This is even though Kleinert (2003) does not find strong support for the offshoring hypothesis in time-series data on German FDI. As noted by this author, the finding that it is not so much outward FDI by German companies but rather FDI by foreign companies in Germany which drives German imports of intermediate goods may disguise that the importance of offshoring differs strongly between industries and host countries. Horizontal FDI probably accounts for the bulk of outward FDI by German companies that enters the analysis of Kleinert (2003),18 while vertical FDI seems to play a more prominent role in the case of FDI by the German automobile industry in Central Europe. Humphrey and Memedovic (2003) argue that changes in the strategies of automobile multinationals, particularly the integration of lower-income countries into corporate strategies, may be most obvious in car assembly, while even more signif16 Due to data constraints, it is not possible to assess the extent to which such exports replaced exports from Germany and, thus, affected domestic production of passenger cars. 17 Follow sourcing results from the preference of car assemblers to use the same suppliers in various locations. According to Kinkel (2004), there is strong pressure on parts suppliers to locate in the vicinity of their customers. 18 Note that about 85 percent of total German FDI stocks were located in industrialized countries in 2003 (Deutsche Bundesbank 2005). Horizontal FDI is likely to dominate in these host countries due to strong similarities in factor endowments.

92

Peter Nunnenkamp

icant changes were taking place in the production of parts and components. At the same time, these authors reckon that the key driving force in the restructuring of the Central European automobile industry was the creation of production networks and a closer division of labour with Western Europe. Van Tulder and Ruigrok (1998) and van Tulder (2004) show that several automobile multinationals, including German ones, aimed at a vertical division of labour with Central Europe and have integrated host countries in this region into international sourcing networks. For instance, Audi’s and Opel’s investments in Hungary were mainly to supply parts and components (notably engines) to Germany. In addition, Central European governments requested foreign car assemblers to help establish an advanced local industry of parts suppliers. For these reasons, companies such as Volkswagen developed local supplier bases in Central European host countries “through a mixture of encouraging follow sourcing by major transnational companies in components and the upgrading of existing local suppliers” (Humphrey and Memedovic 2003: 13). Various important component suppliers are located close to the assembly lines of German car assemblers in Central Europe. All member firms of the Association of the German Automobile Industry (Verband der deutschen Automobilindustrie, VDA) employ about 160000 workers in Central European countries that joined the EU recently; about 100000 of these workers are employed by autoparts suppliers (VDA 2004: 37). Offshoring does not necessarily result in one-way trade in autoparts from Central Europe to Germany. The fragmentation of the value chain by car assemblers and follow sourcing by parts suppliers through outward FDI may indeed help sustain employment in Germany, as the analysis of Kleinert (2003) suggests. This is because the host countries tend to import assembled cars from where FDI originates, and also the imports of intermediate goods originate predominantly from the home base of foreign investors. For instance, the export-oriented production of engines by German companies in Hungary relies heavily on inputs imported from Germany (Humphrey and Memedovic 2003). Nevertheless, the significance of trade in autoparts between Central Europe and Germany is likely to have added to labour market pressure in Germany. In the remainder of this section, we provide a short summary of trade patterns with the four major host countries of German automobile companies in Central Europe (the Czech Republic, Hungary, Poland and the Slovak Republic) by aggregating the most relevant items, i.e., engines and parts thereof as well as other parts and accessories.19 Central Europe has increasingly become integrated into the production networks of German automobile companies. This can be shown by relating the sum of German exports and imports of engines as well as other autoparts and accessories 19 For a more detailed picture of trade in autoparts between Central European countries and Germany, see VDA (2004: 59 – 68).

Relocation, Offshoring and Labour Market Repercussions

93

(in constant A as of 2000) to / from the four Central European countries to the volume of domestic automobile production in Germany. By this measure, trade in autoparts soared from less than A 400 per unit in 1995 to more than A 2800 in 2004.20 The integration of Central European countries through trade in autoparts is most advanced for Hungary, followed by the Czech Republic and Poland. The ranking of the four Central European countries is the same with regard to their share in total German imports of engines and other autoparts in 2004 (Figure 1). Taken together, Hungary, the Czech Republic, Poland and the Slovak Republic accounted for 28 percent of German imports; their share has increased fivefold since 1995.21 percent 14 1995 12

11.7

2004

10 7.9 8

7.0

6 4

2.9 1.7

1.5

2

0.5

0.7

0 Czech Rep.

Hungary

Poland

Slovak Rep.

Source: VDA (a).

Figure 1: Share of Central European Countries in Germany’s Total Imports of Engines and Other Autoparts,a) 1995 and 2004 (percent) a)

Car bodies not included.

Furthermore, similar to trade in assembled cars (Section III), it is no longer true what van Tulder and Ruigrok (1998) observed in the late 1990s, namely that Germany has a bilateral trade surplus with all major Central European countries. Rather, the German trade balance turned significantly negative vis-à-vis the Czech Republic and Hungary (Figure 2). The trade surplus vis-à-vis the Slovak Republic would shrink from A 0.9 billion to A 0.2 billion if car bodies were subsumed under autoparts. 20 In 2004, German imports of autoparts accounted for 52 percent of total trade in autoparts (German exports plus imports) per unit of domestic automobile production. 21 Note that steeply increasing imports of autoparts from Central Europe represent additional offshoring by the German automobile industry, rather than trade diversion to the detriment of other low-cost locations such as Spain and Latin America. As shown in Nunnenkamp (2004), German imports of autoparts from other locations continued to increase (in real terms) when imports from Central Europe gathered momentum.

94

Peter Nunnenkamp 1000

899

1995 2004

500 166

153

62 0

-55

-117 -500

-448

-1000

-925

-1500 Czech Rep.

Hungary

Poland

Slovak Rep.

Source: VDA (a).

Figure 2: Trade Balancea) for Engines and Other Autoparts: Germany vis-à-vis Central and Eastern European Countries, 1995 and 2004 (million A) a)

German exports minus German imports. Car bodies not included.

5. Labour Market Implications The labour market implications of the relocation of assembly lines and the offshoring of parts production to Central Europe are heavily disputed. The notion of Germany degenerating into a bazaar economy has been coined by Sinn (2004). Accordingly, companies use offshoring to overcome the competitive disadvantages at home. This is considered the reason why real value added of the German industry increased by only 5 percent between 1995 and 2003 and industrial employment decreased by 10 percent, even though industrial production increased by 15 percent. Sinn (2004) explicitly refers to the automobile industry to substantiate the argument that German companies remain competitive in international markets only because of “their Eastern European hinterland.” The export of Audi passenger cars whose engines are produced in Hungary is presented as an example of German sales of “high-quality products that were not produced in the country.”22 One may add that also the assembly of automobiles is increasingly taking place in Central Europe, as shown in Section III. Most interestingly, the opponents of this view, too, refer to the automobile industry when stressing positive labour market effects of international production networks and offshoring. For example, Klodt (2004) argues that employment losses are concentrated in industries that have failed to make use of offshoring, whereas high outward FDI and imports of autoparts are supposed to have helped a 22 Applying the classification of Markusen and Venables (2005), the German automobile industry would represent the case of exporting assembled goods produced with imported components, which are assumed by these authors to be capital intensive relative to assembly.

Relocation, Offshoring and Labour Market Repercussions

95

significant increase in employment in the German automobile industry since 1995.23 Bechert and Cellarius (2004) point to “numerous positive examples of outsourcing processes” that went along with rising employment at the companies’ German home base. The latter observation tends to support the reasoning of Becker et al. (2005), according to whom vertical FDI may add to employment at home if cost reduction through offshoring supports an overall expansion of the company (Section II). The evidence presented in the following qualifies both of these seemingly opposing views. In contrast to the decline in overall industrial employment referred to by Sinn (2004), employment in the German automobile industry has increased considerably since 1995 (Figure 3). Furthermore, the average wage earned in the automobile industry was about 25 percent higher than the average wage earned in the German manufacturing sector in 2004 / 05 (VDA a). Earlier wage comparisons reported by Spatz and Nunnenkamp (2002a: 67) suggest that the wage differential in favour of the automobile industry remained fairly stable over the last two decades. The earnings and employment situation does not appear to have suffered from relocation and offshoring if the German automobile industry as a whole is compared to other manufacturing industries. Yet, Sinn (2004) has a point when stressing the gap between production and value-added trends. Figure 3 reveals that gross production of the automobile industry (motor vehicles and parts) almost doubled in 1995 – 2004, whereas value added increased by less than 40 percent. Consequently, the share of value added in gross production declined by 9 percentage points to 24 percent in 2004. 250 192

200

138

150

117

120

employment

wages per employee

100 72 50

0 gross production value

gross value added

value-added share in prod.

Source: VDA (b).

Figure 3: Production, Value Added, Employment and Wages in the German Automobile Industry, 2004 (1995 = 100) 23

For a similar line of reasoning, see VDA (2004: 9).

96

Peter Nunnenkamp

Both sides of the debate tend to ignore diverging developments within the German automobile industry. Spatz and Nunnenkamp (2002a, 2002b) argue that the inter-industry perspective, i.e., comparing the automobile industry with other manufacturing industries, needs to be complemented in two respects in order to fully account for the labour market repercussions of relocation and offshoring. First, the differentiation of the automobile industry into assembly operations (including engines) and the production of parts and accessories reveals striking intra-industry differences:24  The gap between production and value-added trends widened dramatically in assembly operations, but less so in autoparts production. Comparing 1995 and 2004, the share of value added in production declined by more than 10 percentage points (to 21 percent) for assembly operations, compared to a decline by 7 percentage points (to 33 percent) for autoparts production.  Employment growth was by far higher in parts production than in assembly operations (36 versus 10 percent when comparing 1995 and 2004). This may have been helped by relatively low wages in the former segment of the automobile industry. However, the wage gap did not widen during the period under consideration.25 Successful adjustment to competitive pressure from lower-income countries, including Central European countries, through specialization in parts production appears to be another factor explaining the relatively favourable performance of this segment of the automobile industry (Spatz and Nunnenkamp 2002a, Nunnenkamp 2004).26

Second, and more importantly, employment and income trends diverge between specific groups of employees. Skill-specific employment and wage effects of relocation and offshoring may be captured by applying the conventional assumption that non-production workers are better qualified than production workers (Feenstra and Hanson 2003: 147). In addition, we consider three categories (so-called Leistungsgruppen) of production workers in the German automobile industry to assess skill-specific employment and earnings trends for better paid production workers 24 Details (covering the period 1995 – 2004) are not shown here, but are available upon request. 25 Wages paid in parts production amounted to 80 – 85 percent of wages paid in assembly operations in 1995 – 2004. 26 In this context, it may be noted that Sinn’s (2004) reference to Audi engines as exemplifying the trend towards a German bazaar economy is misleading. Trade in engines and engine parts expanded in both directions; the production of engines in Hungary relied heavily on the supply of parts from Germany. In other words, the differentiation by Markusen and Venables (2005) between (i) countries exporting assembled goods produced with imported components and (ii) countries importing assembled goods and exporting components is not easily applicable to the German automobile industry. The classification is complicated in two ways. First, case (i) may apply at the level of final goods, while case (ii) may apply once trade in intermediates is analysed at lower levels of aggregation. Second, case (i) may apply for the production of some final goods, while case (ii) may apply for other final goods in the same industry or even within the same company as the examples of Audi and Skoda in the VW Group suggest.

Relocation, Offshoring and Labour Market Repercussions

97

with more demanding tasks (Leistungsgruppe 1) and production workers with lower pay and less demanding tasks (Leistungsgruppen 2 and 3). Both approaches to account for skill-specific employment and wage effects reveal that the benefits which relocation and offshoring may offer to the automobile industry as a whole are not equally distributed within the industry. According to Figure 4, the employment share of production workers has declined by 6.6 percentage points since 1992. Employment losses in the early 1990s rested almost exclusively on production workers, whereas the subsequent recovery of overall employment in the German automobile industry benefited non-production workers overproportionally. Compared to the change in employment shares, the earnings of production workers have declined only marginally relative to the earnings of non-production workers since 1992. It is debatable whether the labour market situation of production workers deteriorated mainly because German automobile companies discovered Central European countries as competitive suppliers of assembled cars and autoparts. Spatz and Nunnenkamp (2002a) compared longer-term labour market trends for production workers and non-production workers and found that the employment and earnings opportunities of the former deteriorated in the 1980s already. If the “Eastern European hinterland” (Sinn 2004) had a larger effect on production workers than, for example, the previous integration of Spain into the networks of German automobile companies, changes in employment shares and relative earnings should have been particularly pronounced since the mid-1990s, when both the assembly of automobiles in Central Europe and trade in autoparts gathered momentum (Sections III and IV). This is hardly the case. Our second measure has some limitations, too. Throughout the period under consideration, more than half of production workers are grouped into category 1 and are, thus, considered highly skilled. On the other hand, few production workers are grouped into category 3.27 More importantly, the comparison of employment shares and relative wages over time may be affected by changes in the classification of production workers into different skill categories.28 This may raise doubts as to whether categories 1 – 3 adequately reflect skill differentials which we regard as the critical criterion to assess intra-industry distributional effects. Nevertheless, Figure 5 tends to support the view that the emergence of Central Europe as an attractive production location and a competitive trading partner has affected the labour market situation of specific categories of production workers:  Less skilled workers (categories 2+3) accounted for a declining share in overall employment of production workers. Moreover, the decline was more proTherefore, we combine categories 2 and 3 for calculating employment shares in Figure 5. For example, a significant share of workers previously classified into category 2 appears to have been reclassified into category 1 in 1998. A shift in the opposite direction occurred two years later (Figure 5). 27 28

7 Supplement 57 – 2006

98

Peter Nunnenkamp

nounced when the assembly of automobiles in Central Europe and trade in autoparts developed most dynamically. The employment share of less skilled workers fell by about five percentage points when comparing 1985 and 1995, but by more than nine percentage points when comparing 1995 and 2005.  Relative wages of less skilled workers declined only modestly (when comparing category 3 with category 1) or even improved (when comparing category 2 with category 1) until the mid-1990s. Subsequently relative wages dropped significantly, though no longer in most recent years. Employment share of production Employment production workersa)

workersa

percent 90 80

75.6

76.2

73.8

72.6

69.6

70

Relative earnings of production

b

b) Relative earnings of production workers workers

65.5

62.4

59.4

60.0

60.5

1996

2000

2004

60 50 40 30 20 10 0 1988

1992

1996

2000

2004

1988

1992

a) Percent of total employment (production plus non-production workers). – b) Per-capita annual earnings of production workers in percent of per-capita annual earnings of non-production workers.

Source: VDA (a).

Figure 4: Production versus Non-production Workers in the German Automobile Industry, 1988–2004

All in all, the evidence suggests that the seemingly opposing views on relocation and offshoring, mentioned at the beginning of this section, are not inconsistent with each other. Rather, they refer to two sides of the same coin. On the one hand, the automobile industry as a whole still compares favourably with other manufacturing industries in terms of employment and income opportunities. On the other hand, relocation and offshoring have resulted in distributional effects within the German automobile industry. The rising human-capital intensity of automobile production in Germany, reflected in the structure of employment, and declining relative wages of less skilled workers are longer-term phenomena that cannot be attributed exclusively to the emergence of Central Europe as an attractive location for assembly operations and autoparts production. Especially for less skilled production workers, however, the competition from Central Europe has added to pressure on relative wages and employment opportunities.

Relocation, Offshoring and Labour Market Repercussions

99 percent

percent 120

120

110

110

100

100

90

c)

relative wage: LG2/LG1

80

90 80

70

70

relative wage: LG3/LG1d)

60

60

50

50

40

40

30

30 employment share: LG2+LG3b)

20

20

10

10

0

0 1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

Source: VDA (a).

Figure 5: More Skilled versus Less Skilled Production Workers in the German Automobile Industry, 1985 – 2005a) a) As of January. – b) Share of categories (Leistungsgruppen) 2 plus 3 in total employment of production workers. – c) Category (Leistungsgruppe) 2 in percent of category 1. – d) Category (Leistungsgruppe) 3 in percent of category 1.

6. Summary and Conclusions The economic transformation of Central European countries has added significantly to competitive pressure in the automobile industry, even though this industry is relatively technology and human-capital intensive and, thus, represents a traditional manufacturing stronghold of advanced countries such as Germany. Substitution effects may be limited for the time being, but the relocation of assembly operations is likely to have affected domestic production by inducing higher imports of finished cars from Central Europe and, possibly, also by reducing the growth of exports from the German home base of automobile companies. At the same time, competitive pressure is due to offshoring and intensive trade in automotive parts between Germany and Central Europe. The finding that the German automobile industry still compares favourably with other manufacturing industries in terms of employment and income opportunities supports the view that relocation and offshoring are important means for German companies to remain competitive. Hence, the integration of Central Europe into the international division of labour is also in the interest of the workers employed in the German automobile industry. However, the benefits to be derived from relocation and offshoring are not equally distributed within the industry. Especially for low skilled production workers, the competition from Central Europe has intensified pressure on relative wages and has impaired employment opportunities. 7*

100

Peter Nunnenkamp

Competitive pressure and, thus, the need of the German automobile industry to adjust is unlikely to subside. Major automobile producers have announced plans to establish additional production facilities in Central and Eastern Europe.29 Parts suppliers have little choice but to follow the assemblers. Vertical FDI strategies will continue to be attractive. In 2003, labour costs in the four major Central European host countries of the German automobile industry amounted to only about one sixth of labour costs in Germany (VDA 2004: 23); and Havas (2000: 241) argued already in 2000 that “the productivity gap has almost been closed.” Recent agreements on wage restraint and longer working hours with assemblers and parts suppliers, reflecting the weakening bargaining position of trade unions in the German automobile industry, will narrow the gap in unit labour costs to some extent. But it will probably take long until “the catching-up process (of Central Europe) will have a tendency to increase investments by horizontal multinationals and depress investments by vertical multinationals” (Carstensen and Toubal 2003: 17).30 For the German automobile industry as a whole, there is no reasonable alternative to exploit the potential of cost savings through relocation and offshoring. Employment and wage prospects at home will depend on innovation and specialization according to comparative cost advantages. However, innovation and specialization offer little relief to low skilled workers unless they succeed to improve their level of qualification. Wage restraint and differentiation may provide part of the solution to the extent that it buys time for skill upgrading. By contrast, it would be counterproductive if economic policymakers and trade unions in Germany attempted to “protect” low skilled workers, either by insisting on EU harmonization of corporate tax rates and social standards, or by demanding effective minimum wages. Minimum wages would strengthen the incentives to relocate and offshore production to lower-income countries, including in Central Europe. As a result, the employment prospects of low skilled automobile workers would deteriorate further. EU harmonization may help contain distributional conflicts in German manufacturing to some extent in the short run, by reducing the cost advantages of new member states in Central Europe. In the longer run, however, the international competitiveness of German producers would suffer if they were constrained in cutting costs through relocation and offshoring. For industries such as automobile production, facing fierce competition on a worldwide scale, this would mean that employment and income prospects deteriorate for the overall workforce, independently of skill levels.

29 Heymann (2004) expected that production capacity in Central and Eastern Europe would double until 2006 / 07, which would add to the overcapacity problem the automobile industry is facing on an international scale. 30 In a similar vein, Heymann (2004) posits that Central Europe will enjoy a lasting competitive advantage in labour costs; see also VDA (2004: 25 – 26).

Relocation, Offshoring and Labour Market Repercussions

101

References Bechert, S. / Cellarius, G. (2004): “Outsourcing Offers a Chance for the Development of Strongly Competitive European Enterprises,” ifo Schnelldienst 57 (7), 7 – 10. Becker, S. O. / Ekholm, K. / Jäckle, R. / Mündler, M.-A. (2005): “Location Choice and Employment Decisions: A Comparison of German and Swedish Multinationals,” Working Paper 1243, Kiel Institute for the World Economy, Kiel. Blomström, M. / Fors, G. / Lipsey, R. E. (1997): “Foreign Direct Investment and Employment: Home Country Experience in the United States and Sweden,” Economic Journal 107, 1787 – 1797. Braconier, H. / Ekholm, K. (2000): “Swedish Multinationals and Competition from High- and Low-Wage Locations,” Review of International Economics 8, 448 – 461. Buch, C. / Kleinert, J. / Lipponer, A. / Toubal, F. (2005): “Determinants and Effects of Foreign Direct Investment: Evidence from German Firm-level Data,” Economic Policy 2005, 51 – 110. Carr, D. L. / Markusen, J. R. / Maskus, K. E. (2001): “Estimating the Knowledge-Capital Model of the Multinational Enterprise,” American Economic Review 91, 693 – 708. Carstensen, K. / Toubal, F. (2003): “Foreign Direct Investment in Central and Eastern European Countries: A Dynamic Panel Analysis,” Working Paper 1143, Kiel Institute for the World Economy, Kiel. Deutsche Bundesbank (2005): Kapitalverflechtung mit dem Ausland, Statistische Sonderveröffentlichung 10, Frankfurt am Main. Dichtl, E. / Hardock, P. (1997): Auslandsfertigung und Produktionsverlagerung von Unternehmen des Fahrzeugbaus: Ergebnisse einer empirischen Studie, Mannheim, University of Mannheim. Ekholm, K. / Forslid, R. / Markusen, J. R. (2003): “Export-Platform Foreign Direct Investment,” NBER Working Paper 9517, National Bureau of Economic Research, Cambridge (MA). Federico, S. / Minerva, G. A. (2005): “Fear of Relocation? Assessing the Impact of Italy’s FDI on Local Employment,” paper presented at the Euroframe Conference “Trade, FDI and Relocation: Challenges for the EU?,” http: // www.euroframe.org / fileadmin / user_upload / euroframe / docs / 2005 / session6 / eurof05_federico.pdf. Feenstra, R. C. / Hanson, G. H. (2003) : “Global Production Sharing and Rising Inequality: A Survey of Trade and Wages,” in Handbook of International Trade, edited by E.K. Choi and J. Harrigan, Malden (MA), Blackwell, 146 – 185. Hanson, G. H. / Mataloni, R. J. / Slaughter, M. J. (2001): “Expansion Strategies of U.S. Multinational Firms,” NBER Working Paper 8433, National Bureau of Economic Research, Cambridge (MA). – (2005): “Vertical Production Networks in Multinational Firms,” Review of Economics and Statistics 87, 664 – 678. Havas, A. (2000): “Changing Patterns of Inter- and Intra-regional Division of Labour: Central Europe’s Long and Winding Road,” in Global Strategies, Local Realities: The Auto Industry in Emerging Markets, edited by J. Humphrey, Y. Lecler, and M. Salerno, Basingstoke, Macmillan, 234 – 262.

102

Peter Nunnenkamp

Helpman, E. (1984): “A Simple Theory of International Trade with Multinational Corporations,” Journal of Political Economy 92, 451–471. Helpman, E. / Krugman, P. (1985): Market Structure and Foreign Trade, Cambridge, MA, MIT Press. Heymann, E. (2004): Perspektiven Automobilindustrie, Bauwirtschaft, Baumaschinenbau, Frankfurt a.M., Deutsche Bank Research. Humphrey, J. / Memedovic, O. (2003): The Global Automotive Industry Value Chain: What Prospects for Upgrading by Developing Countries, Vienna, UNIDO. Kinkel, S. (2004): Automobilzulieferer in der Klemme: Stand und Entwicklung der Auslandsproduktion bei deutschen Automobilzulieferern, Karlsruhe, Fraunhofer Institut für Systemtechnik und Innovationsforschung. Kleinert, J. (2003): “Growing Trade in Intermediate Goods: Outsourcing, Global Sourcing, or Increasing Importance of MNE Networks?,” Review of International Economics 11, 464– 482. Klodt, H. (2004): “Deutschland: Eine Basar-Ökonomie?,” http: // www.uni-kiel.de / ifw / staff / klodt / basaroekonomie.pdf. Konings, J. / Murphy, A. (2001): “Do Multinational Enterprises Substitute Parent Jobs for Foreign Ones? Evidence from European Firm-Level Panel Data,” Discussion Paper 2972, Centre for Economic Policy Research, London, www.cepr.org / pubs / dps / DP2972.asp. Lansbury, M. / Pain, N. / Smidkova, K. (1996): “Foreign Direct Investment in Central Europe since 1990: An Econometric Study,” National Institute Economic Review 156, 104 – 114. Marin, D. (2004): “A Nation of Poets and Thinkers – Less So With Eastern Enlargement? Austria and Germany,” Discussion Paper 4358, Centre for Economic Policy Research, London, www.cepr.org / pubs / dps / DP4358.asp. Marin, D. / Lorentowicz, A. / Raubold, A. (2003): “Ownership, Capital or Outsourcing: What Drives German Investment to Eastern Europe?,” in Foreign Direct Investment in the Real and Financial Sector of Industrial Countries, edited by H. Herrmann, Berlin, Springer, 147–163. Markusen, J. R. (1984): “Multinationals, Multi-plant Economies, and the Gains from Trade,” Journal of International Economics 16, 205 – 226. Markusen, J. R. / Venables, A. J. (1998): “Multinational Firms and the New Trade Theory,” Journal of International Economics 46, 183 – 203. – (2000): “The Theory of Endowment, Intra-industry and Multinational Trade,” Journal of International Economics 52, 209 – 234. – (2005): “Interacting Factor Endowments and Trade Costs: A Multi-Country, Multi-Good Approach to Trade Theory,” University of Colorado, Boulder, mimeo. Nunnenkamp, P. (2004): “Automobilstandort Deutschland unter Wettbewerbsdruck,” ifo Schnelldienst 57 (7), 28 – 36. – (2005): “The German Automobile Industry and Central Europe’s Integration into the International Division of Labour: Foreign Production, Intra-Industry Trade, and Labour Market Repercussions,” Papeles del Este. Revista electrónica 9, http: // www.ucm.es / BUCM / cee / papeles / 09 / pape0404220004a.pdf.

Relocation, Offshoring and Labour Market Repercussions

103

– (2006): “Was von ausländischen Direktinvestitionen zu erwarten ist: Unbegründete Ängste in den Heimatländern, übertriebene Hoffnungen in den Gastländern?,” Zeitschrift für Wirtschaftspolitik 55, 20 – 44. OECD (2005): International Trade by Commodities Statistics. Paris. Pries, L. (1999): Auf dem Weg zu global operierenden Konzernen? BMW, Daimler-Benz und Volkswagen: Die drei Großen der deutschen Automobilindustrie, Munich, Rainer Hampp Verlag. Sinn, H.-W. (2004): “Bazaar Economy,” ifo-Viewpoint 50, Munich, http: // www.cesifo.de. Spatz, J. / Nunnenkamp, P. (2002a): Globalisierung der Automobilindustrie: Wettbewerbsdruck, Arbeitsmarkteffekte und Anpassungsreaktionen, Kiel Studies 317, Berlin, Springer. – (2002b): “Globalization of the Automobile Industry: Traditional Locations under Pressure?,” Aussenwirtschaft 57, 469 – 493. Sturgeon, T. J. / Florida, R. L. (1999): The World That Changed the Machine: Globalization and Jobs in the Automotive Industry, Cambridge (MA), Massachusetts Institute of Technology. UNCTAD (1998): World Investment Report 1998, New York, United Nations. van Tulder, R. (2004): “Peripheral Regionalism: The Consequences of Integrating Central and Eastern Europe in the European Automobile Space,” in Cars: Carriers of Regionalism?, edited by J. Carillo, Y. Lung, and R. van Tulder, Basingstoke, Palgrave, 96 – 112. van Tulder, R. / Ruigrok, W. (1998): “European Cross-National Production Networks in the Auto Industry: Eastern Europe as the Low End of European Car Complex,” paper presented at the Berkeley Roundtable on the International Economy, University of California, http: // repositories.cdlib.org / brie / BRIEWP121. VDA (a) (various issues): Tatsachen und Zahlen aus der Kraftverkehrswirtschaft, Frankfurt am Main, Verband der Automobilindustrie. – (b) (various issues): International Auto Statistics, Frankfurt am Main, Verband der Automobilindustrie. – (2004): Die deutsche Automobilindustrie in der erweiterten EU: Motor der Integration, Frankfurt am Main, Verband der Automobilindustrie. Vickery, G. (1996): “Globalisation in the Automobile Industry,” in Globalisation of Industry: Overview and Sector Reports, Paris, OECD, 153 – 205. Weiß, J.-P. (2000): “Die deutsche Automobilindustrie im internationalen Wettbewerb,” DIWWochenbericht 12 / 00, Berlin, http: // www.diw-berlin.de / diwwbd / 00-12-2.html. Zhang, K.H. / Markusen, J. R. (1999): “Vertical Multinationals and Host-country Characteristics,” Journal of Development Economics 59, 233 – 252.

Relocation, Offshoring and Labour Market Repercussions: The Case of the German Automobile Industry in Central Europe – Comment By Andre Jungmittag*

The paper of Peter Nunnenkamp considers the question whether the relocation of car assembly and offshoring of automobile parts production to Central Europe have consequences for the employment situation of the German automobile industry. This question and the question of the characteristics of the consequences is investigated by means of a confrontation of descriptive statistics of foreign direct investment (FDI) and foreign trade, respectively, as well as of the employment situation in the automobile industry. Based on the descriptive findings and because the German automobile industry still compares favourably with other manufacturing industries in terms of employment and income opportunities, Nunnenkamp concludes that – in a nutshell – relocation and offshoring are important means for German companies to remain competitive. However, particularly for low skilled production workers, the competition from Central Europe has intensified the pressure on relative wages and impaired employment opportunities. My comments on the paper are divided in three parts. Firstly, I would like to place the role of vertical FDI within the larger framework of international production fragmentation and I will present some figures as well as measures which can be used to assess the degree of production fragmentation of the European automobile industry. Secondly, I will discuss the problem that the process of production fragmentation is superimposed by the general tendency towards de-industrialisation and by the impact of a high R&D intensity and innovativeness of the automobile industry, which both have rather similar effects on the employment situation as the former. Thirdly, I will postulate that these overlappings call for a econometric analysis and I will give some hints how such an analysis can be tackled.

* Chair for Macroeconomic Theory and Policy, University of Wuppertal, Gaußstraße 20, 42119 Wuppertal, Germany. [email protected].

106

Andre Jungmittag

1. International Production Fragmentation and Vertical FDI Fragmentation of production means the splitting-up of the value added chain allowing for a more in-depth specialisation. The reason therefore is that different stages of production correspond to different production functions so that a country may have a comparative advantage in one stage of production and a comparative disadvantage in other stages. One important driver is vertical foreign direct investment by large multinational companies, mainly based in the US, Japan and some European countries. An alternative mode of production fragmentation is international outsourcing. 35

35

30

30

25

25

20

20

15

15

10

10

5

5

0

0 95

96

97

98

99

00

01

02

03

95

96

97

98

99

00

01

02

03

Source: Eurostat.

Figure 1: Shares of Domestic Value added in Total Production of the Automobile Industries of selected European Countries

First of all, increasing international production fragmentation implies that the share of domestic value added in total production decreases. The first panel of figure 1 shows that the UK as well as Germany have the highest value-added-toproduction ratios within the group of the large European automobile manufacturers, and their shares are rather stable since 1996 and 1999, respectively. The shares of Italy and particularly France are clearly lower and have strong downward trends. The second panel of figure shows the same shares for some Central European countries, which are important destinations for relocations and offshoring, and for Spain as a benchmark. It is obvious that the Czech Republic and Hungary have rather similar shares like Spain, which is a rather popular host country for FDI of the German automobile industry. Their shares, however, are clearly lower than the share of Germany, but on a similar level like the share of Italy. The share of Poland is rather volatile, so that we cannot say whether it is actually higher than those of the two former countries, while that of Slovakia is clearly lower. Although shares of domestic value added to total production give a first impression of the degree of production fragmentation, an in-depth analysis requires the

Relocation, Offshoring and Labour Market Repercussions – Comment

107

use of some further indicators. One candidate is the vertical specialisation index suggested by Hummels / Ishii / Yi (2001), namely VSij ˆ

imported intermediatesij  exportsij gross outputij

for sector i in country j. For the calculation of this indicator the identification of SITC categories for parts, components and final products of the automobile industry, proposed by Kaminiski / Ng (2001), can be used. Moreover, Feenstra / Hanson (1999) as well as Hijzen / Görg / Hine (2005) used as a measure of outsourcing the ratio between imports of intermediates and value added. They distinguished between broad outsourcing capturing all imported intermediates within an industry and narrow outsourcing capturing only imported intermediates in a given industry from the same industry. The latter used import-use matrices of input-output tables to calculate the indices of narrow outsourcing.

2. Overlapping of De-industrialisation, Innovativeness and Production Fragmentation All developed countries experienced a massive de-industrialisation, which has not grind to a halt up to now. This development clearly superimposes the process of international production fragmentation, since it has a rather similar impact on the labour markets, namely on the development on total employment in the sectors of the manufacturing industry, the ratio of skilled to unskilled workers and the relative wages of these two groups. A further overlapping of the effects of production fragmentation and vertical FDI is caused by the high R&D intensity and innovativeness of the automobile industry. One third of the R&D expenditures of the whole business sector in Germany fall to the automobile industry (cf. RWI, 2005). The long-term positive development of employment in this sector is clearly a reflection of its R&D activities and innovativeness. Furthermore, the demand in Germany in the last decades was open to numerous expensive product innovations. Actually, Germany is a lead market for automobiles. However, these facts also imply an increase of the human capital intensity of the automobile industry and thus an increase of the demand for highly skilled workers. To put it in a different way, the development of employment and wages in the automobile industry might also be caused by de-industrialisation and / or a technical bias or by all three developments. Therefore, descriptive statistics might not be sufficient to discriminate between these effects.

108

Andre Jungmittag

3. Econometric Analysis is Needed To capture the effects of the just mentioned three developments in an econometric model, the basic idea of Rowthorn / Ramaswamy (1999), which only investigated the links between growth, foreign trade with low wage countries and deindustrialisation of the whole manufacturing sector in 18 OECD countries, can be taken up, augmented and refined for the automobile industry. Let relative labour productivity in a sector i (in our case: the automobile industry) in a country j (i.e. labour productivity in this sector compared to the labour productivity of the entire economy) depend on per capita income Yi as well as several further variables Zij . To these Zij belong domestic sector influences like the investment ratio and as an indicator of innovativeness the R&D intensity (or other indicators like patent intensity), but also indicators for production fragmentation. Thus, the equation for relative labour productivity in sector i is (the error terms are left out for the sake of a simpler notation): …1†

ln RELPRODijt ˆ 0i ‡ 1i ln Yjt ‡

K X

ki Zkijt :

kˆ2

Furthermore, let the relative price (price for this sector relative to the price of total domestic output) depend on relative labour productivity and the variables Zij , thus …2†

ln RELPRICEijt ˆ 0i ‡ 1i ln RELPRODijt ‡

K X

ki Zkijt :

kˆ2

Finally, the value added share of the sector considered in total GDP Sij can be modelled by the following equation: …3†

ln Sijt ˆ 0i ‡ 1i ln Yjt ‡ 2i …ln Yjt †2 ‡ 3i ln RELPRICEijt ‡

K X

ki Zkijt ;

kˆ4

where – due to the general de-industrialisation – a negative sign is expected for

2i . Then the following identity holds for the employment share Eij of the sector considered in total employment: …4†

ln Eijt ˆ ln Sijt

ln RELPRODijt :

Since RELPROD and therefore RELPRICE too only depend on Yi and the Zij , the two equations for value added and employment shares also can be written as

Relocation, Offshoring and Labour Market Repercussions – Comment …5†

ln Sijt ˆ 0i ‡ 1i ln Yjt ‡ 2i …ln Yjt †2 ‡

K X

109

ki Zkijt

kˆ3

and …6†

ln Eijt ˆ "0i ‡ "1i ln Yjt ‡ "2i …ln Yjt †2 ‡

K X

"ki Zkijt :

kˆ3

Since the data situation is surely problematic if only one sector and country is considered, a panel data approach should be applied, either across several countries or several sectors. Here, the heterogeneity of countries or sectors has to be taken into account, which can be done by including fixed country or sector effects. Furthermore, general trends concerning all countries or sectors considered can be captured by fixed time effects. After estimating the model and testing, whether the influences of production fragmentation and innovativeness on the value added and employment shares are statistically significant, decompositions of the changes of the value added and employment shares can be undertaken. Thereby, we have to differentiate between “normal” growth effects, which might lead to a reduction of the shares due to increasing per capita incomes, investment effects, production fragmentation and innovativeness effects. If the latter two effects cause higher labour productivities due to the competitive pressure in the low wages sphere, on the one hand, an increased reduction of the value added shares of these sectors can be expected indirectly via the relative prices, and, on the other hand, an increased reduction of the employment shares directly via the relative labour productivities. This process is associated with a weakening of de-industrialisation in those sectors, which produce skill and / or capital intensive. It would be naturally desirable to model the differentiation between unskilled and skilled workers directly. One approach can be found in Helg / Tajoli (2003), but also in some other analyses cited there. They propose to estimate the ratio Qijt of skilled to unskilled workers by K X
Qijt ˆ 0i ‡ 1i Yijt ‡ 2i Kijt =Yijt ‡ 3i …Wsk =Wunsk † ‡ ki Zkijt ; kˆ4

where Kijt is the net capital stock of the sector and W is the wage rate. The latter is probably endogenous, which has to be taken into account by an adequate choice of the estimation method, whereby a direct modelling of the wage differentials would be desirable, too. Although I am aware that such an analysis already is a rather challenging task in the face of the data situation, it has to be mentioned finally that production fragmentation and vertical FDI could not only improve the situation and the competitiveness of the particular firm or sector, but also have indirect effects on employment. Thus, e.g. RWI (2005) presented calculations that in 2000 1,024,000 jobs in

110

Andre Jungmittag

other sectors were created due to the demand for cars (of them 219,000 in business related services and 315,600 in other services). What qualifications are required for these new jobs? Summary Summing up, the paper of Peter Nunnenkamp addresses the important question of the domestic labour market effects of the relocation of car assembly and offshoring of automobile parts production to Central Europe by the German automobile industry and derives a set of plausible hypotheses, which, however, need further testing. Particularly, the discrimination between the effects of production fragmentation, general de-industrialisation and the high R&D intensity and innovativeness of the automobile industry is necessary, which only can be achieved by an in-depth econometric analysis. In a further step the indirect employment effects also should be taken into account.

Zusammenfassung Der Beitrag von Peter Nunnenkamp behandelt die wichtige Frage, welche Auswirkungen Standortverlagerungen und Offshoring nach Zentraleuropa auf die Beschäftigungssituation im bundesdeutschen Automobilsektor haben. Er leitet dabei eine Reihe von plausiblen Hypothesen ab, die aber der weiteren Überprüfung bedürfen. Insbesondere bedarf es einer Diskriminierung zwischen den Arbeitsmarkteffekten der Produktionsfragmentierung, allgemeine Deindustrialisierung sowie der hohen FuE-Intensität und Innovativität des Automobilsektors, die nur durch eine stringente ökonometrische Analyse erreicht werden kann. In einem weiteren Schritt sollten auch die indirekten Beschäftigungseffekte einbezogen werden. References Feenstra, R. C. / Hanson, G. H. (1999): The Impact of Outsourcing and High-technology Capital on Wages: Estimates for the United States, 1979 – 1990, Quarterly Journal of Economics 114, 907 – 941. Helg, R. / Tajoli, L. (2003): Patterns of International Fragmentation of Production and Implications for Labor Markets, Flowenla Discussion Paper No. 14, HWWA, Hamburg. Hijzen, A. / Görg, H. / Hine, R. C. (2005): International Outsourcing and the Skill Structure of Labour Demand in the United Kingdom, The Economic Journal 115, 860 – 878. Hummels, D. / Ishii, J. / Yi, K. (2001): The Nature and Growth of Vertical Specialisation in World Trade, Journal of International Economics 54, 75 – 96. Kaminski, B. / Ng, F. (2001): Trade and Production Fragmentation: Central European Economies in EU Networks of Production and Marketing, Worldbank Policy Research Working Paper Series No. 2611, Washington D.C.

Relocation, Offshoring and Labour Market Repercussions – Comment

111

Rowthorn, R. / Ramaswamy, R. (1999): Growth, Trade, and Deindustrialization, IMF Staff Papers 46, 18 – 41. RWI (2005): Beschäftigungswirkungen von Forschung und Innovationen, Forschungsvorhaben im Auftrag des Bundesministeriums für Wirtschaft und Arbeit, Endbericht, Essen.

Basic Income Reform in Germany: Better Gradualism than Cold Turkey By Alexander Spermann*

Abstract This paper advocates the cautious and constitutional evolution of existing basic income schemes (“unemployment benefit II”) and Targeted Negative Income Tax (“Einstiegsgeld”) into a means-tested combi-wage model for the future long-term unemployed (gradualism strategy). The paper argues that, with regard to existing unemployment benefit II claimants, stronger financial incentives should be offered on a time restricted basis by largely disregarding (up to the relative poverty line) earnings from “mini”, “midi” and part-time jobs – with the aim of providing current unemployment II claimants with a powerful incentive to work at least 15 hours a week and thus to relinquish their unemployed status. Bearing in mind the uncertain employment impact and the related fiscal risks, the paper advises against additional financial incentives by reducing support levels from one day to the next (cold turkey strategy). JEL Classifications: I 38, J 22 Keywords: Labor Market Reform, Employee Subsidy, Workfare, long-term unemployment

1. Introduction After more than ten years of debate on wage subsidies (combi-wage) the year 2006 may finally bring the breakthrough: The ‘grand coalition’ is planning a legislative initiative on the combi-wage for the autumn of 2006. The government will be able to draw on the lessons learned from a large number of pilot schemes which have been run in recent years in Germany (cf. Kaltenborn 2001, 2005 and Dietz et al. 2005). The relevant debate has taken on board numerous proposals originating from the academic world, business and industry associations and the political parties – the FDP’s suggestion of a citizen’s wage, the ifo Institute’s Combi-Wage Model (cf. Sinn et al. 2002, 2006) and the Magdeburger Alternative (cf. Schöb / Weimann 2005) are the concepts on which most debate has focused, although the respective pros and cons of each are not examined in this paper (cf. Spermann 2006 for a more detailed discussion). Instead, this paper presents a reform proposal which is based on practical experience with Targeted Negative Income Tax (TNIT) * Centre for European Economic Research (ZEW), Mannheim, [email protected]. 8 Supplement 57 – 2006

114

Alexander Spermann

in Germany. Initially, however, the paper discusses the key problems on the labor market and the dynamic relationship between the duty to work and the financial incentives built into the design of a basic income support system (Chapter 2). Chapter 3 presents a simplified diagram, supported by recent basic income statistics, of the distorting incentive effects associated with the unemployment trap which contribute to prolonging periods of joblessness. The four-component model of fair basic income support based on insurance, TNIT, poverty gap concept, and workfare components, is presented in Chapter 4. Chapter 5 discusses the potential employment and fiscal effects while Chapter 6 elaborates the advantages of a gradualist compared with a cold turkey strategy. Chapter 7 provides a brief conclusion. 2. Key Problems on the Labor Market and the Design of Basic Income Support Dramatic Increase in Numbers of Long-Term and Low-Skilled Unemployed The main problems on the German labor market are long-term unemployment and joblessness among the low skilled, although both problems do of course overlap (cf. Sachverständigenrat – the German Council of Economic Experts – 2005). Despite the Hartz reforms, the long-term unemployment rate has leapt up in recent years and – as defined by the OECD – reached 51.8 % in 2004 (cf. OECD 2005). According to figures published by the Federal Labor Institute (Bundesagentur für Arbeit), there has been a continuous increase in the number of long-term unemployed since 2001. The unprecedentedly high figures for 2005 were over 1.8 million long-term jobless (cf. Federal Labor Institute 2005). The trend among the low skilled with without professional qualifications in Germany has been one of consistently rising unemployment since the 1970s; in western Germany in 2004 over 20 % of this group was unemployed, in eastern Germany over 50 % (cf. Reinberg / Hummel 2005). The situation is truly dramatic. Reform is more urgently required than ever. Dramatic Surge in the Number of People Receiving Basic Income Support The “basic jobseeker’s allowance” (so-called unemployment benefit II = Arbeitslosengeld II = ALG II) was introduced throughout Germany under the German Social Code SGB II at the beginning of 2005. This entailed the merging of social assistance (Sozialhilfe) for claimants who are fit to work with unemployment assistance (Arbeitslosenhilfe) into a single new means-tested and tax-financed transfer system. In March 2004, the official administrative bodies anticipated that the new basic allowance would be paid to around 3 million claimants. However, in March 2006 the actual figures for claimants of the basic allowance under SGB II is over 7 million – more than twice as many as originally expected. The explanation for this development basically lies in two politically determined definitions which were not current in quite the same form prior to the reform. On the one hand, ‘capacity to work’ was defined in such broad terms that even drug

Basic Income Reform in Germany

115

addicts who were very unlikely to find a job were, for example, defined as available for work. On the other, the term low income household (Bedarfsgemeinschaft or a household entitled to receive benefits) was also defined in a way which created incentives for young adults to move out of their parental homes to form new low income households – with the result that in the space of a very short time the number of low income households rose to almost four million. In March 2006, 5.2 million people were on ALG II and 1.85 million people – mainly children – were on social allowances (Sozialgeld) so that, in total, over 7 million people were in receipt of basic income support under SGB II (cf. Federal Labor Institute 2006a, b). Unemployed Claimants of Basic Income Support and of Combi-Wages From a labor market policy perspective, two subgroups of ALG II claimants are of particular significance. The almost three million registered unemployed ALG II claimants must be considered separately. A person is defined as unemployed if he or she works fewer than 15 hours a week. On the other hand, there are also over 900,000 people in work who also receive basic income support (cf. Federal Labor Institute 2006c). Unfortunately these subgroups overlap as the available statistics do not specify how many working claimants of ALG II are no longer registered unemployed because they are working over the threshold value of 15 hours / week. A more detailed statistical breakdown is required to distinguish these two subgroups. Failure to Enforce the Duty to Seek and Accept Work In Germany claimants of ALG II are subject to a duty to seek and accept work. With the introduction of the Job Center and the tightening up of suitability criteria requiring claimants to accept work under the Hartz laws, the duty to accept suitable jobs is much tougher than in the past, even if considerable regional differences persist. Even though the basic jobseeker’s allowance has now been in force for over a year, Job Centers are still in the process of being set up. Case management in the working parties comprising the Federal Labor Institute and local authorities (ARGEn) is – apart from one or two exceptions – nothing like as intensive as that undertaken by the local labor offices themselves. A popular lament at the time this paper was written is that “in den ARGen noch viel im Argen liegt” (or “it’s no party in the working parties”), particularly because the envisaged ratio of case managers to unemployed claimants has not yet been achieved. The situation is exacerbated by IT problems which eat up case manager’s scare client contact time owing to the sheer volume of data which needs to be entered and corrected manually. However, in reality the duty to seek and accept work cannot be enforced if intensive case management is not available. But even if these transitional difficulties were to be solved in the near future, the absence of an automatic mechanism for reducing benefits and regular involvement of officially appointed doctors means that imposing benefit sanctions for those re8*

116

Alexander Spermann

fusing a job is likely to remain an enduring problem. Although case managers are able to resort to “one-euro jobs” (in effect additional jobs) as a means of verifying a claimant’s willingness to accept work, claimants are always entitled to lodge objections to, or bring an action against, any benefit reductions whilst claiming legal aid to cover all their court costs. As a result, case managers are currently having to bear a huge increase in administrative work and consequently have less time available to support other clients. Is it Enough to Impose Compulsory Activation Requirements on Claimants Without Providing Financial Incentives? At first glance a consistently enforced duty on claimants receiving jobseeker’s allowance to seek and accept work might appear to make additional financial incentives in the form of improved additional earnings options superfluous. However, compulsory activation is seriously hampered by the near impossibility of precisely assessing capacity to work in hourly categories (e.g. 5, 10, 15, 20 etc. hours). If we assume that officially appointed doctors are able to dictate a minimum hour working load (e.g. three hours a day) at reasonable costs, the more stringent the state is in urging compliance with compulsory activation policies, the more likely it is that people will accept work (participation decision). However – and this is the decisive argument in favor of additional financial incentives – the amount of time which individuals choose to spend beyond this administered threshold depends critically on the disregard rules in the transfer system (working hours decision). What is more, in a system in which work obligations are perfectly enforced and consistent benefit sanctions are imposed on those refusing to accept a job, there is a very strong incentive to acquire incapacity to work status in order to continue receiving benefits. The lessons learned by military draft boards have amply demonstrated how imaginative people can be in this respect. Nonetheless, it is still worthwhile thinking about the existing financial incentives for claimants receiving jobseeker’s allowance and the way they respond to these rules, as well as about alternative arrangements. Enforcing the duty to work is a necessary but not a sufficient condition for tackling key labor market problems.

3. Distorting Incentives Effects and the Unemployment Trap in 2006 The status quo of ALG II claimants in the year 2006 is illustrated more clearly by considering a simplified gross / net earnings diagram (Figure 1) drawing on the disregard rules which have been in force since October 1, 2005. The average basic income support for a single person in western Germany is assumed to be A 575. This amount is composed of a basic allowance of A 345 and an amount to cover the combined costs of housing and heating of A 230. Income earned by claimants is subject to a disregard of A 100. The benefit reduction rate

Basic Income Reform in Germany

117

on earnings between A 100 and A 800 is 80 % …t ˆ 0:8†, while claimants earning between A 800 and A 1200 have 90 % of their additional earnings deducted from their benefit entitlement …t ˆ 0:9†. The figures in this simplified example, which takes account of social security contributions and income tax, demonstrate that supplementary ALG II is paid to single people with gross earnings of up to A 1,100. The case in which own earnings are (almost) entirely swallowed up by a reduction in benefit is referred to as the unemployment trap. Given the way the benefit withdrawal tables for single households work, it would hardly be worthwhile for a claimant on jobseeker’s allowance to accept a mini job (A 400) from which the claimant would only take home an extra A 160. Net earnings 45°

t=0.8

t=0.9

€845

Net earnings line for those paying social security contributions and Y* taxes

B=€575

Gross earnings €100

€400

€800

€1,100

Figure 1: Status quo, single Claimant, Western Germany

The unemployment trap is even more egregious for a family with two children (western Germany). The basic income support is significantly higher at A 1,471 and is comprised of the basic allowance for two adults (A 311 each), for two children under 14 (A 207 for each child) and the costs of housing and heating amounting, in average, to A 435. The additional earnings options are calculated as above, although the 90 % withdrawal rule applies on earnings up to A 1,500 for low income household with children. Over and above this gross earnings level, additional income is withdrawn in full …t ˆ 1†. Claimants are entitled to supplementary ALG II if their gross earnings are below around A 2,400 …Y  †. This clearly demonstrates how difficult it is for families in particular to escape from the transfer system, even those that make very great efforts to do so. If both partners were to accept full-time jobs on a wage of A 5 an hour their joint earnings for working two 40-hour weeks would be around A 1,600, which means that even

118

Alexander Spermann

in this case of full-time employment there would still be no alternative but to continue paying the family supplementary ALG II unless the hourly wage were to rise significantly. Net earnings 45°

t=0.8

t=0.9

t=1 Y*

€1,780

Net earnings line for those paying social security contributions and taxes

B=€1,471

Gross earnings €100 €400

€800 €

€1,200 €

€1,600 €

€2,000 €

€2,400 €

Figure 2: Status Qou, Family (2 Children), Western Germany

Empirical Relevance of the Unemployment Trap Differentiated empirical evaluations of the gainful employment of ALG II claimants are now available for September 2005 – a period in which disregard rules which were considerably more restrictive than those considered above were in force. The available figures show that around 20 % of ALG II claimants – 906,000 people in 844,000 low income households – were in gainful employment (cf. Federal Labor Institute 2006c). Two phenomenon or behavioral responses are apparent: (a) “Stealth” employment Directors of security departments and job centers have provided anecdotal evidence of “stealth” employment when the pressure on people to accept work increases. Marginal employment is accepted, for example, in order to meet the legal requirement to seek and accept work and to avoid a reduction in benefit entitlement. This phenomenon is also substantiated by the latest evaluations performed by the Federal Labor Institute: more than half of all claimants work below the marginal earnings threshold of A 400 Euro – with a remarkably high frequency of jobs paying between A 100 and A 200 (gross earnings).

Basic Income Reform in Germany

119

(b) The working and the hidden poor Low skilled people, who are not only in marginal employment, but whose net earnings only just cover their basic needs, are often referred to as the working poor. The hidden poor are those do not claim their full entitlement to state transfer payments because of perceived stigma or because they are inadequately informed. Prior to the introduction of ALG II, several hundred thousand people were estimated to be living in hidden poverty because they failed to submit claims to the social security offices. Job centers appear to be less stigmatized than social security offices. One indication of this is the increase in the number of people entitled to submit supplementary claims. The number of claimants in gainful employment doubled, for example, from almost 470,000 previous social and unemployment assistance claimants to 906,000 ALG II recipients. However, this must be qualified by noting that no information was available about the number of hours worked or the distribution of gross earnings among people previously receiving social and unemployment assistance. Basic income reform must therefore take account of the fact that claimants in “stealth” employment are able to settle down permanently in the transfer system without having to fear having their benefits cut – and that they need to be enticed away from the transfer system with incentives. The number of claimants may also rise dramatically if claims are made for people in densely populated gross earnings categories. Accordingly, a reform of this nature would have to be implemented with great caution – not least because of the lack of comprehensive empirical knowledge and the imponderables concerning the response of those affected.

4. Proposed Solution: The Four-Component Model of Fair Basic Income Support: a Means-Tested Combi-Wage Model for the Long-Term Unemployed This paper advocates the cautious and constitutional evolution of existing unemployment benefit instruments (ALG II) and Targeted Negative Income Tax (“Einstiegsgeld”) under Paragraph 29 of the German Social Code (SGB II) into a means-tested combi-wage model for the future long-term unemployed (gradualism strategy). The four component model is a combi-wage model within the existing basic income support system. ALG II is the equivalent of an indefinite entitlement to income supplement; Targeted Negative Income Tax (TNIT or “Einstiegsgeld”) on the other hand is time-restricted income supplement. As far as the three million existing unemployed ALG II claimants are concerned, this paper only argues for an additional, time-restricted income supplement (TNIT) up to the poverty line for each household type coupled with a stringently enforced duty to seek and accept work. In other words, the model represents a gradualist carrot-and-stick strategy intended to make low-paid jobs more attractive to the long-term unemployed.

120

Alexander Spermann

The four component model consists of 1. degressive unemployment benefit (ALG I) reduced stepwise over 12 months (insurance component), 2. a time-restricted income supplement up to the poverty line (TNIT component), 3. an indefinite entitlement to ALG II equivalent to a subsistence minimum (poverty gap concept component), and 4. an absorbing community or voluntary work element (workfare components). In order to properly understand this proposal, it is important to distinguish between a number of different threshold values – as shown in Table 1. Table 1 Threshold Values For the Four Component Model Single Social / cultural minimum income B Subsistence minimum income (after halving of ALG II) B00 Break-even income Y  Relative poverty line (60 % of median income based on OECD formula)

Family with 2 children

A 575

A 1,471

A 402.50 A 840

A 1,160 A 1,781

A 938

A 1,970

(1) Social / cultural minimum income B The social / cultural minimum income referred to above corresponds with the requirements of the household type in the status quo and is composed of the basic allowance (ALG II), social allowances for children, and the combined cost of housing and heating. (2) Subsistence minimum income B00 The subsistence minimum is not currently defined in the constitution. According to rulings on the Federal Social Assistance Act (BSHG) made in the social courts it would, however, be fair to assume that a 50 % reduction in standard benefits in the case of a persistent refusal to accept work would not be overturned by the courts. A standard benefit for a single person of A 345 (ALG II) could potentially be cut by A 172.50, for example. In the case of standard benefits of A 622 for both adults in a four-person family, it would be possible to cut benefits by A 311. The children’s social allowance and the reimbursement of housing and heating costs would remain in place. In some cases benefits in kind would be granted instead of monetary benefits in order to ensure that the needs of children are met. (3) Break-even income Y  This income is the amount at which no further transfers are paid. To the right of Y  only taxes and fiscal charges are paid, while to the left of Y  transfer payments

Basic Income Reform in Germany

121

are made (although taxes and fiscal charges may, in certain circumstances, also be paid); Y  only lies on the 45º line if no taxes and fiscal charges need to be paid. (4) Relative poverty line The relative poverty line is identical with the OECD definition found in the German government’s report on the rich-poor divide (cf. Bundesregierung 2005). The OECD poverty line for singles is 60 % of median income or A 938 in 2004. For a family this value is equivalent to a value of 1 for the first adult, 0.5 for the second adult and the figure 0.3 for each child in the household, adding up to A 1970 (2.1  A 938) for two-child families. Component 1: Degressive Unemployment Benefit Reduced Stepwise Over 12 Months The first component is aimed at bringing about a change in the benefits side of the unemployment insurance equation. Instead of the current practice of limiting payment of unemployment benefits to 12 months and paying the benefit as a constant proportion of the claimant’s last earned net income, the benefit could be paid at progressively lower levels over time (cf. the German Council of Economic Experts 2003). This would mean that the loss in income in the first few months of unemployment would be significantly lower than is currently the case, enabling search activities to be maximized. Within a period of one year, unemployment benefit would sink substantially as the total value of payments is to remain unchanged. At the same time, additional earnings up to the relative poverty line would continue to disregarded – in contrast to the current situation in which only A 165 is disregarded and amounts over and above this are deducted in full from the insurance benefit. Component 2: Degressive TNIT Empirical studies based on experiments with control groups demonstrate the positive employment effects of time-restricted income supplements paid to the longterm unemployed (cf. Spermann / Strotmann 2005). International experience with time-restricted supplements – in Canada for example – have proved to be successful, as evaluation studies with control groups have shown (cf. Michalopoulos et al. 2005). In principle TNIT is already part of the current basic jobseeker’s allowance system and in fact 17,600 relevant cases were registered in 2005 (cf. Federal Labor Institute 2006b). Internal Federal Labor Institute guidelines also envisage the use of these back-to-work bonuses for self-employed people in particular (the “me incorporated [Ich-AG]” instrument and transitional benefits under the Social Code (SGB III) are not available to ALG II claimants). As a result 86 % of financial support was dedicated to self-employed recipients and only 14 % to the acceptance

122

Alexander Spermann

of employment subject to compulsory social insurance contributions. Practice differs markedly from region to region, however. The state government of Saxony-Anhalt, for example – a region with a very high level of unemployment – has made massive use of these instruments with the outcome that around 60 % of unemployed claimants receiving this form of support switched, with the aid of TNIT, into employment subject to compulsory social insurance (cf. Federal Labor Institute 2006b). Whether or not supplementary financial support is granted continues to be at the discretion of case managers and the strategies adopted by individual job centers. Experiences with TNIT from pilot studies run in Baden-Württemberg and Hesse demonstrate, however, that case managers in the agencies often regard incentive instruments with considerable skepticism. They tend to adopt the argument that claimants are subject to a duty to seek and accept work and that there is therefore no need for any additional incentives. Differentiated analyses of administrative data at the national level are not yet available (studies in this direction will be launched by the IAB in the future). The basic idea underlying TNIT is discussed in detail in Spermann (2001). In the current institutional framework, ALG II claimants are the target group for this supplementary payment. The continued development of TNIT links current timerestricted disregards of claimants’ earned income with the relative poverty line for each type of household and a gradual and automatic reduction in ALG II (not of the social allowance) to a subsistence minimum income which conforms with constitutional requirements. The social / cultural minimum income for a single person is the current ALG II level of A 575 (basic rate plus housing costs). The subsistence minimum income is A 402 (50 % of the basic rate plus housing costs). The relative poverty line as defined by the OECD and as used in the German government’s report on the rich-poor divide is A 938 for a single person or 60 % of median income. The following figures (3a – c) illustrate the two components of this proposal drawing on the example of a single person: After one year of unemployment, unemployed claimants receive basic jobseeker’s allowance under SGB II if they successfully pass a tough means test. During the first month in which they are in receipt of benefits they are entitled to claim TNIT. Figure 3a demonstrates that in this way the income derived from a mini job is disregarded so that the poverty line of A938 can be reached. In other words, a mini job becomes just as attractive as moonlighting – if only for a limited period of time. In contrast to the situation today, a mini job is financially less attractive to claimants – with the outcome that only around 8 % of those with mini jobs were previously unemployed (cf. Fertig et al. 2004). We may therefore anticipate that a combination of more stringently applied work requirements (the stick) and financial incentives (the carrot) will induce the long-term unemployed to step up their efforts to acquire mini jobs.

Basic Income Reform in Germany

123

However, unless it is accompanied by other measures this incentive may well lead to new “stealth” employment – i.e. ALG II plus a mini job. It is for this reason that the TNIT must be degressive over time. This can be achieved by making automatic and gradual reductions in ALG II down to a constitutionally permissible subsistence minimum income. Figures 3b and 3c illustrate this additional incentive mechanism which pushes the long-term unemployed into earning more than they would in a mini job. The message is clear: unless they are more active in their job searching, they will have less disposable income. If, after the time-restricted TNIT income supplement has expired, the claimant’s earnings are still only just sufficient to cover basic needs, the third component of the model comes into play. The period for which TNIT is granted may vary according to the type of household concerned. The larger the household is, the longer the period during which TNIT is granted given that a higher net income is required in order to permanently escape the unemployment trap. Net earnings 45° Time-restricted TNIT €938  YY*

Net earnings line for those paying social security contributions and taxes

B=€575

Gross earnings €100

€400

€800

€1,100

Figure 3(a): 2. Component – Month 1, Single Claimant, Western Germany

Component 3: Indefinite Basic Income Support Based on the Poverty Gap Concept The third component provides for indefinitely payable basic income support at a subsistence minimum level – for future long-term unemployed claimants. This component is based on Milton Friedman’s poverty gap concept (1962 / 68). In the long run this component also does away with the unemployment trap. In contrast to the present system, there is an ongoing financial incentive to work more even after TNIT has expired. It is true that, at a benefit reduction rate of around 66%,

124

Alexander Spermann

Net earnings 45° Time restricted TNIT €938  Y*

Y

Net earnings line for those paying social security contributions and taxes

B‘=A500 B‘=€500

Gross earnings €100

€400

€800

€1,100

Figure 3(b): 2. Component – Month 6, Single Claimant, Western Germany

the marginal burden for claimants is relatively high – however, this could only be reduced further by either cutting ALG II levels even further (B00 drops further) and / or lifting the transfer limit Y  so that it applies to higher gross earnings. Neither of these options would be advisable. A reduction beneath the subsistence minimum would conflict with constitutional requirements, and an increase in the transfer limit would entail an enormous additional fiscal burden. Similar scenarios can be mapped for larger households – although for space reasons these cannot be dealt with in this paper. Three points must, however, be underlined. First: The reduction to a subsistence minimum income level only relates to standard benefits, not however to the social allowance (cf. Table 1). Second: The TNIT for this type of household would push people towards accepting midi jobs (up to A 800) or part-time employment – they would then be able to work their way out poverty under their own steam and, with the help of additional earnings, reach the relative poverty line of A 1,970. Rising hourly wage rates (productivity effect) and more working hours (working hours effect) would, in certain circumstances, allow both partners to earn a gross income which would enable them to live permanently outside the transfer system. They would in any case remain in the working process so that their human capital would not be eroded by long-term unemployment and their social networks would be maintained. Third: Even after TNIT expires – three years after a person becomes unemployed – there would still be a financial incentive to earn more even if the benefit reduction rate is relatively high at 71 %. This again underlines the crucial need to enforce the duty to seek and accept work. In a basic income support system which

Basic Income Reform in Germany

125

provides relatively generous income support, it will not be possible to offer permanent financial incentives (carrots) which will be able to compete at all with moonlighting, and it is therefore essential that the model is backed up by the duty to seek and accept work (sticks). Net earnings

Fig. 4: Component 3, Single Claimants, Western Germany 45°

Y*

€845 t=0.66

Net earnings line for those paying social security contributions and taxes

B‘‘=402 €

Gross earnings €100

€400

€800

€1,100

Figure 4: Component 3, Single Claimant, Western Germany

Component 4: Community Work in Exceptional Cases (Workfare Component) Any reform of basic income support aimed at improving financial incentives and tightening up the legal requirement to seek and accept work must offer a solution to the dilemma of how people who cannot find work on the labor market, despite their ability to demonstrate that they have made strenuous efforts, should be treated. This is where community work can play an important role – although as the exception rather than the rule. Community work becomes relevant as soon as ALG II begins to fall. A social / cultural minimum income is intended to be secured with the help of community work – with working hours varying according to type of household – not, however, an income at the poverty line level. This ensures that it is always more worthwhile to accept private jobs than to undertake community work – in contrast to the current system in which one-euro jobs can prove to be financially more attractive than mini jobs, for example. Community work would continue to be used to assess claimant’s willingness to accept work. This instrument has proved effective in a) deterring people from making claims who are not in fact willing to accept work and b) ensuring that cuts in the benefits received by claimants who refuse work are legally watertight. On the

126

Alexander Spermann

other hand, it became apparent in 2005 that some ALG II claimants do not regard one-euro jobs as a sanction, but that these jobs are regarded as a welcome employment opportunity (cf. Koch et al. 2005). Community work thus fulfils a dual role as an instrument to check up on a claimant’s willingness to accept work and to provide desirable occupations for the long-term unemployed who are unable to find employment on the open labor market.

5. Employment Effects and Fiscal Impact Field studies with control groups in Germany have demonstrated that time-restricted income supplements – such as TNIT – increase the probability of employment (cf. Spermann / Strotmann 2005). Indefinitely payable income supplements based on the poverty gap concept – in other words a reduction in benefit levels and lower benefit reduction rates – have positive employment effects, as simulation studies performed by the ZEW with the general equilibrium model show (Boeters et al. 2003). Positive employment effects are even more probable if a reform of basic income support is backed up by labor law reforms (shorter dismissal protection periods and lower back-to-work wages for the long-term unemployed) and collective bargaining measures (larger wage spread). Positive employment effects would bring fiscal savings with them. The author of this paper can only speculate however on the extent and sustainability of these employment effects and the resulting fiscal impact. Neither microsimulation studies nor general equilibrium models are capable of producing reliable results because ultimately both are based on the behavioral responses of people in a world in which a legal low-wage sector is, to all intents and purposes, very small or even non-existent. Empirical elasticities on which such models draw, are based on observations in wage segments which exist today – it goes without saying that they cannot be based on the responses of people in an, as yet, non-existent low-wage sector. The significance of non-cognitive abilities in filling jobs in a low-wage service sector continues to be underestimated (cf. Hieming et al. 2005). For this reason well-founded skepticism is warranted as far as model calculations are concerned which promise major employment gains and high levels of fiscal savings. What is more, the available empirical ex ante microsimulations for the low-wage sector are only able to capture partial effects and to provide educated guesses about where the impact will be felt. Microsimulations of mini jobs are not, for example, able to capture the basic impact of sideline employment, as this type of model only allows for the simulation of changes in main occupations (cf. Arntz /Feil / Spermann 2003 and Steiner / Wrohlich 2004). Impressive though this is, it nonetheless only covers part of the overall picture. Empirical studies with existing simulation models are, however, able to demonstrate under what circumstances neutral or negative employment effects and additional fiscal burdens may arise. The introduction of a minimum wage would be

Basic Income Reform in Germany

127

counterproductive – existing jobs below the agreed minimum wage would be destroyed, new jobs in the low-wage sector would not be created at all. A combi-wage reform which expanded the population of entitled claimants would be fiscally extremely risky: additional fiscal burdens would be guaranteed, but the additional employment effects in an inflexible labor market would be highly uncertain. High minimum wages and an across-the-board combi-wage for the low-skilled under existing legal and collective bargaining restraints would prove a fiasco.

6. The Advantages of the Gradualism Strategy Compared with the Cold Turkey Strategy Bearing in mind the many imponderables and the lack of empirical clarity regarding the level of employment effects, as well as the resulting fiscal impact, this paper argues in favor of a gradual strategy – in contrast to a cold turkey strategy which promises an immediate reduction in benefit levels and substantially higher employment of up to 3.2 million people in work, plus fiscal savings of up to 21 billion euros (cf. Sinn et al. 2006). Fairness Considerations For reasons of fairness, the combi-wage reform coupled with an automatic trimming down of basic income support to a subsistence minimum should only be focused on future ALG II claimants who would be able to adjust their behavior accordingly in good time. As far as existing ALG II claimants are concerned – those who have previously been in employment for many years and who have slipped into long-term unemployment as a result of structural changes – fairness would dictate a different approach. Current ALG II claimants have received the message over decades of their working lives that, should they become unemployed, they would be able to rely on a system of unemployment insurance which would provide them with a transfer income coupled to their previous take-home pay. All the economic decisions taken by this group have also been taken in this framework. The merging of unemployment assistance and social assistance at social assistance levels has already produced several hundred thousand reform losers (cf. Rudolph / Blos 2005). It would be inappropriate to further reduce the benefits payable to this group of people, who would then be forced to engage in community work in order to maintain the level of their benefit entitlements. Existing ALG II claimants could, however, be at least motivated to take up mini, midi and part-time jobs if adjustments were made to the additional earnings regulations. Where these mini jobs entail more than 15-hours of work a week, these ALG II claimants would no longer be counted as unemployed, even if they continued to receive supplementary ALG II. A realistic intermediary target for a combi-wage

128

Alexander Spermann

reform for existing ALG II might be the loss of “unemployed” status. As approximately 2.8 million ALG II claimants are registered unemployed, this would represent a major step in the right direction, and one that would substantially reduce the fiscal burden. It should prove easier for people to find fulltime employment if they already have a job than would be the case if they were to seek work as an unemployed applicant. Reduced Scope of Community Work A combination of time-restricted and permanent income supplements would almost certainly reduce the scale of community work undertaken in comparison with a cold turkey strategy. On the one hand, the long-term unemployed may move out the basic income support system within the time limits as a result of rising wages and higher hourly wage rates – they are able to find their way into the labor market and, in certain circumstances, break out of the basic income system without being dependent on community work. On the other hand, benefits received by future ALG II claimants would only gradually be reduced to a subsistence minimum so that more time would be available to seek a job in the private labor market. Avoiding the Production of Working Poor A further element of the four component model is the link between time restricted income supplements and the poverty line. This can enable the long-term unemployed to work their own way out of poverty – and out of the category of the working poor – within a specified timeframe. A substantial number of long-term unemployed people may be expected to remain in employment once they have managed the leap into the private labor market. The lessons learned from field experiments with TNIT in Germany suggest that people tend to get used to higher incomes – around 60 % of those receiving TNIT support remained in employment even after the scheme had expired (cf. Dann et al. 2002).

7. Conclusion A cautious change of system to a combi-wage system would appear to be an appropriate way of enabling people and institutions to adapt – in particular with regard to labor law and collectively agreed structures (gradualism strategy). A distinction must, however, be made between future long-term unemployed people and existing long-term unemployed claimants. Those who become unemployed in the future should receive a clear message from the state that their state income supplement will drop to a subsistence minimum wage level and that individual effort combined with generous disregard rules make a life above the poverty line possible. The status of a member of the working poor can be avoided by individual effort. The four component model describes such a system consisting of time-re-

Basic Income Reform in Germany

129

stricted and indefinitely payable income supplements with an automatic reduction in basic income support to the level of a subsistence minimum income. At the same time, current unemployed ALG II claimants should – out of fairness considerations – only be offered generous time-restricted income supplements (TNIT) to motivate them to work at least 15 hours a week and thus to lose their status as unemployed. The danger inherent in a fast change of system (cold turkey strategy) in which benefits for all ALG II claimants are immediately and substantially reduced is that the employment effects promised by its protagonists will not materialize in anything like the near future because the behavioral responses in the low-wage sector have been incorrectly understood and non-cognitive abilities crucial for successful job matches are ignored. If employment effects fail to materialize, fiscal savings above and beyond those arising from cuts in ALG II alone would not be realized. This would entail the establishment of community work opportunities on a major scale which would again generate very high fiscal costs. The net fiscal effect is uncertain. The logical conclusion is: better gradualism than cold turkey.

References Arntz, M. / Feil, M. / Spermann, A. (2003): Die Arbeitsangebotseffekte der neuen Mini- und Midi jobs – eine ex-ante Evaluation, Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 36, 271 – 290. Boeters, S. / Schnabel, R. / Gürtzgen, N. (2003): Reforming Social Security in Germany – An Applied General Equilibrium Analysis, ZEW Discussion Paper 03 – 70, ZEW, Mannheim. Dann, S. / Kirchmann, A. / Spermann, A. / Volkert, J. (2002): Einstiegsgeld in Baden-Württemberg, Schlussbericht, in: Baden-Württemberg Ministry of Socia Affairs (publisher), Stuttgart. Dietz, M. / Koch, S. / Walwei, U. (2006): Kombilohn, Ein Ansatz mit Haken und Ösen, IABKurzbericht Nr. 3 v. 1. 3. 2006, Nuremberg. Federal Government of the Federal Republic of Germany (2005): Lebenslagen in Deutschland, Der 2. Armuts- und Reichtumsbericht der Bundesregierung, Berlin. Federal Labor Institute (2005): Der Arbeits- und Ausbildungsmarkt in Deutschland, Monthly Reports for December and the year 2005, Nuremberg. – (2006a): Der Arbeits- und Ausbildungsmarkt in Deutschland, Monthly Report March, Nuremberg. – (2006b): Grundsicherung für Arbeitsuchende, Annual Report 2005, Nuremberg. – (2006c): Grundsicherung für Arbeitsuchende, Anrechenbare Einkommen und Erwerbstätigkeit, March, Nuremberg. Fertig, M. / Kluve, J. / Scheuer, M. (2004): Aspekte der Entwicklung der Mini jobs, Essen. Friedman, M. (1962): Capitalism and Freedom, Chicago. 9 Supplement 57 – 2006

130

Alexander Spermann

– (1968): The Case for the Negative Income Tax, in: Melvin, L. (publisher): Republican Papers, 202 – 220. Hieming, B. / Jaehrling, K. / Kalina, T. / Vanselow, A. / Weinkopf, C. (2005): Stellenbesetzungsprozesse im Bereich “einfacher” Dienstleistungen, Dokumentation Nr. 550, published by Federal Ministry of Economics and Labor, Berlin. Kaltenborn, B. (2001): Kombilöhne in Deutschland, Eine systematische Übersicht, IABWerkstattbericht Nr. 14, Nuremberg. Kaltenborn, B. et al. (2005): Evaluierung der arbeitsmarktpolitischen Sonderprogramme CAST und Mainzer Modell, Forschungsbericht Nr. 552, published by Federal Ministry of Economics and Labor, Berlin. Koch, S. / Stephan, G. / Walwei, U. (2005): Workfare: Möglichkeiten und Grenzen, IAB Discussion Paper No. 17 / 2005, Nuremberg. Michalopoulos, C. / Robins, P. K. / Card, D. (2005): When financial work incentives pay for themselves: evidence from a randomized social experiment for security claimants, Journal of Public Economics 89, 5 – 29. OECD (2005): Employment Outlook, Paris. Reinberg, A. / Hummel, M. (2005): Höhere Bildung schützt auch in der Krise vor Arbeitslosigkeit, IAB-Kurzbericht Nr. 9 / 2005, Nuremberg, 2005. Rudolph, H. / Blos, K. (2005): Schätzung der Auswirkungen des Hartz-IV-Gesetzes auf Arbeitslosenhilfe-Bezieher, IAB-Forschungsbericht Nr. 14 / 2005, Nuremberg. Sachverständigenrat für die gesamtwirtschaftliche Entwicklung (2005): Die Chance nutzen – Reformen mutig vorantreiben, Jahresgutachten 2005 / 06, Wiesbaden. Sachverständigenrat zur Begutachtung der gesamtwirtschaftlichen Entwicklung (2003): Jahresgutachten 2003 / 04: Staatsfinanzen konsolidieren – Steuersystem reformieren, Wiesbaden. Schöb, R. / Weimann, J. (2005): Arbeit ist machbar. Die Magdeburger Alternative: Eine sanfte Therapie für Deutschland, 4. Aufl., Dößel. Sinn, H.-W. et al. (2002): Aktivierende Sozialhilfe, Ein Weg zu mehr Beschäftigung und Wachstum, ifo-Schnelldienst Nr. 9, Munich. – (2006): Aktivierende Sozialhilfe 2006: Das Kombilohn-Modell des ifo Instituts, ifo schnelldienst Nr. 2, Munich. Spermann, A. (2001): Negative Einkommensteuer, Lohnsubventionen und Langzeitarbeitslosigkeit, Finanzwissenschaftliche Schriften Bd. 104, Frankfurt am Main. – (2006): Der Kombilohn: Chancen, Gefahren und Erfahrungen aus Modellversuchen, erscheint in: Wirtschaftspolitische Blätter. Spermann, A. / Strotmann, H. (2005): The Targeted Negative Income Tax (TNIT) in Germany: Evidence from a Quasi Experiment, ZEW Discussion Paper 05 – 68, ZEW, Mannheim. Steiner, V. / Wrohlich, K. (2004): Work Incentives and Labor Supply Effects of the ‘Mini jobs Reform’ in Germany, DIW Discussion Papers 438, Berlin.

Basic Income Reform in Germany: Better Gradualism than Cold Turkey – Comment By Steffen J. Roth*

Abstract This paper accentuates some critical points about wage subsidies in general and explains this criticism discussing the instruments of Alexander Spermann’s proposal of a basic income reform in Germany. First, there is not much evidence that the so-called unemployment trap is relevant for many unemployed. Second, if many unemployed are reluctant to work because of a lack of monetary work incentives we have to ask how these persons will react to time-restricted exemptions of earned income and to reductions of monetary work incentives for achieving higher earned incomes. In particular, this paper points out the medium-term and long-term adverse side-effects of wage subsidies. In total, these instruments of active labour market policy likely encourage a person with high wage claims not to accept low-income jobs. This would be the exact opposite of what proponents of wage subsidies wish to achieve. JEL Classifications: I 38, J 22, Z 13 Keywords: Labour Market Reform, Wage Subsidies, Social Norms

In another article, Alexander Spermann recently declared simplicity as one of three requirements for a successful concept of complementary wages (Spermann 2006). Whether this criterion still holds regarding Four-Component Model seems dubious. In principle, the proposal contains all instruments which are known from public discussion: be it very generous subsidies with high tax exemptions or a model of negative income tax with a deprivation rate of 66 or 71 per cent or compulsory community jobs for the maintenance of the actual transfer level (compulsory in this case meaning a transfer cut in case of a rejection of the community job by the unemployed person). At a first glance, this seems a bit indecisive. However, suggesting several instruments in combination is more likely the result of a profound analysis of the single components, their individual shortcomings and difficulties. Spermann tries to re* Institute for Economic Policy at the University of Cologne, Germany. E-mail address: [email protected]. The author gives his thanks to the Fritz Thyssen Foundation for the financial support of the research project “Not-intended Effects of the Labour Market Policy”. This comment has been written as a part of this research project. 9*

132

Steffen J. Roth

medy the respective problems of the single instruments in such a way that, after a short period, another strategy follows – which, however, has in turn to struggle with its respective difficulties. For a structured discussion, at first I intend to provide general thought-provoking impulses and criticism concerning complementary wages. Afterwards, the single elements of the Four-Component Model shall be discussed in more detail.

1. General Remarks on Wage Subsidies 1.1 Wage Subsidies Contradict Fundamental Principles of the Tax-Transfer-System

Whenever wage subsidies exceed a subsidy of low earned incomes up to the social basic income level, such measures are incompatible with the basic principles of the German tax-transfer-system. Most employees in Germany are obliged to pay taxes. As long as the resulting market income exceeds the tax exemption for covering a minimum living standard, citizens have to contribute to the financing of collective tasks, according to their economic capacity. This includes paying taxes to guarantee people of less economic capacity a dignified lifestyle. However, the society should not overdo this: if taxes swallow too much of the market income, they may reduce incentives to supply labour. If people reduce their working effort, this will result in lower taxes and, therefore, risk the ability of the state to finance its tasks. As such, it is significant to pay attention to that basic principle “Effort must pay”: The available income of (net) tax payers has to bear explicit noticeable reference to the original market income, with a focus on labour productivity as valued by the labour market. If employees have just enough earned income to finance their own livelihood and that of their family, they do not have to pay taxes. In this case, they will be exempt from the duty to have a share in the financing of collective tasks. If persons achieve less earned income than required by themselves and their relatives there is also no duty to contribute to the collective tasks financing. They are allowed to keep the earned income tax free. In addition, the community subsidises this income up to the specific amount, which, according to the effective social standards, enables a person to live a dignified lifestyle. However, this does not release these persons from the general obligation of keeping transfers as low as possible – they should still earn their living according to their own labour productivity. As such, transfers are based on the lack of the recipients’ labour productivity, that is on their indigence. People get transfers up to the level of what they themselves cannot achieve with their own productivity in order to cover their basic income level. So, transfer recipients, who achieve earned income but need additional

Basic Income Reform in Germany – Comment

133

support, receive a kind of complementary income: a combination of market income and complementary support by means of tax financed transfers. This complementary income is a part of the German tax-transfer-system and intended to be this way as an instrument of social policy. Transfer schemes which go beyond that, like those active employment policy models of wage subsidies, go against both aforesaid principles of income determination: against the principle of labour productivity and against that of indigence. These transfer schemes enable people to receive an income which is not justified by their productivity on the market. At the same time, this income is not justified by the indigence of the concerned people, because it is above the basic income level. This is not without consequences. Inevitably and directly, severely unequal treatments and redistribution effects that are difficult to justify will ensue. Conversely, some suggestions for complementary wages intend to cut transfers, if employable transfer recipients refuse to accept jobs offered. However, this also entails profound problems: it is difficult to fathom how to reduce a minimum subsistence level. Either the transfer level actually granted is higher than intended by society or it is not. In the first case a general reduction would have to be discussed. In the second case a principle of penalties on minimum transfers abandons the idea to guarantee every citizen the income covering a dignified lifestyle as deemed appropriate. In the latter case, we first would have to enquire more deeply about the intentions of the net payers of this system – but this would go beyond this article. It is enough to say at this point that a shortfall of a basic income level would be incompatible with a large part of the arguments put forward in the literature on distribution policy (Roth 2002).

1.2 Wage Subsidies Do Not Have an Unambiguously Positive Effect on Work Effort

Those who are able to work and have the opportunity of getting a reasonable job are morally and legally obliged to reduce the required aid or to end dependence from tax-financed transfers by taking every chance. Of course, this obligation has only a limited effect for reluctant transfer recipients. Controlling the efforts as well as penalties in case of such reluctant behaviour is difficult to implement in practice. In a system based only on indigence, there is no monetary incentive for transfer recipients to take up a job as long as a significantly higher wage than the subsistence level is unachievable, at least in a short term perspective. This situation of a lack of work incentives, often defined as “unemployment trap”, is well known and extensively discussed. Wage subsidies presume that this problem explains a major part of the unemployment in Germany. Therefore, the proponents of these subsidies try to enhance

134

Steffen J. Roth

the work incentives by increasing the disposable incomes. Actually, though, we neither know how many unemployed are affected by a lack of (monetary) work incentives, nor how high the monetary incentives should be in order to decidedly influence their choice to work or not. Additionally, wage subsidies are unable to enhance the monetary work incentives consistently at all income levels. In order to secure their financing, higher financial incentives for achieving low market incomes are accompanied by corresponding reductions of the financial incentives for achieving even higher market incomes – for example, through an extension of working hours, further education or the acceptance of hard and unpleasant jobs. In total, working incentives are not increased for the whole range of income levels by complementary wages, when compared with a pure transfer system based only on indigence. The effects on the employees’ decisions on the length of work, on the attempts for more qualifications, on working flexibility and on the general working efforts in the concerned income spans are difficult to predict. An adverse effect on work incentives through additional fiscal burden on higher incomes for financing the wage subsidies may have to be added to these effects. If the additional fiscal burden is financed through indirect taxes, however, this would simply reduce the disposable income also of the lower income recipients targeted by the wage subsidies, reducing their monetary incentives to work.

1.3 Wage Subsidies Often Include a Fiscal Risk

Additional fiscal burden would thwart the aim of more employment. Therefore, the proponents of complementary wages usually try to show a fiscal neutrality or even fiscal relief, as they anticipate savings on transfers. Evidently, this only makes sense, if reductions of the transfer requirements of the program participants cannot alternatively be achieved by lower costs (Eekhoff / Roth 2006). Spermann is correct in maintaining that subsidies become a fiscal risk, anyhow, once they are granted permanently and / or extend the initially targeted group of unemployed.

2. The Instruments In the Four-Component Model 2.1 The Cutting of the Transfer Claim

Spermann suggests to cut the tax-financed transfers to the unemployed (AlG II) to the “subsistence minimum income”, which he considers as unobjectionable under German constitutional law. He equates this level with a half of standard benefits plus accommodation costs. After what was discussed so far, this implies at first sight, that he does not consider as necessary the granting of a “social” or “cultural minimum income”, which is identified as the German basic income as we know it.

Basic Income Reform in Germany – Comment

135

Actually, however, the Four-Component-Model continues to guarantee the transfer recipients the present basic income, so long as they start a regular job or give an evidence of their willingness to work. In this, Spermann is not very radical. The requirement that everyone has to work to the best of his or her ability does not put at risk a dignified lifestyle for those who stick to the rules. Spermann even appears somewhat less strict when compared to transfer reductions as penalties in the current German social law. The gradual reduction of the transfers in the “second component” and the reduced unemployment transfer in the following permanent “third component” of a negative income tax merely means that Spermann demands less working willingness at the beginning of the AlG II-transfers than after the expiry of a “period of grace”. In the 13th month of unemployment, the author even grants the effective monthly transfer level of 575 EUR, if the transfer recipients still do not work. With the 14th month of unemployment and beyond, transfer recipients either have to work regularly or do a community service, if they want to continue receiving the cultural subsistence level. The demands of such work are limited: even in the 24th month of unemployment, for a single transfer recipient it is enough to earn a market income of only 173 EUR. In the phase of the negative income tax after the 25th month in unemployment, the same single has to earn 315 EUR per month in order to achieve the effective transfer level. Alternatively, in both phases the acceptance of a community service employment will suffice.1

2.2 The Willful Neglect of Claimants’ Earned Income

Transfer Claimants earning a market income above the before mentioned amounts do not have to accept being limited to the basic income level in the FourComponent Model. On the contrary, during the time of the “second component”, singles easily can achieve up to 938 EUR disposable income per month.2 This is 63 per cent more than the current effective basic income level. Within this period, the decrease of the basic transfers corresponds again to more and more demands for an earned income. Claimants who just start receiving the tax-financed transfers achieve a disposable income of 938 EUR at an earned income of 363 EUR gross per month. Those receiving already tax-financed transfers for 12 months have to earn 536 EUR per month on the market to reach a total disposable income of 938 EUR. 1 The transfer claims which Spermann intends for a sole earner of a four-person household, are a bit more demanding. At the end of the “Degressive Targeted Negative Income Tax”, this person has to earn 311 EUR on the market for the maintenance of the actually granted cultural subsistence level, in the following phase of the negative income tax it is, after all, almost 830 EUR. Alternatively, however, even at this point, the acceptance of a community work job is always sufficient. 2 This amount corresponds to the current boundary of “relative poverty”, according to Spermann. There is, however, no reason to include a statistical value of 60 per cent of the median income into the system of tax financed transfers.

136

Steffen J. Roth

This means a grave violation of the principles of basing labour market income on productivity and transfers on indigence for a wide range of income levels. Within this system, every claimant who is a single with a market income between 174 and 938 EUR per month receives more transfers than his or her productivity on the labour market yields, and with no reasonable consideration of his or her indigence. The unequal treatment taking place can be highlighted with the help of a concrete example: Suppose a person supported by the Four-Component Model accepts a job as an untrained assistant of a hairdresser. Regular full-time employees in this job earn about 800 EUR per month.3 The transfer recipient, participating in the programme who expands his of her working hours step-by-step over one year from only 18 to 27 hours per week, receives finally 938 EUR every month. This exceeds the income of the regular fully employed colleagues by more than 17 per cent.4 The result: the participants of Four-Component Model earn more than the regular fully employed persons in the same profession. It goes without saying that the fully employed colleagues at the hairdresser will have little understanding for this. Moreover, unequal treatment against less productive transfer recipients has to be discussed. This stems directly from the violation of the principle of indigence. According to that principle, less productive transfer recipients should get more support than those with a higher productivity. However, similar to most wage subsidy schemes, Spermann’s model achieves the opposite: a person who is lucky enough to achieve higher hourly rates gets the disposable income of 938 EUR per month with less work effort.5 As such, with the “second component” those with higher productivity have to work less for their transfers, not more. Spermann wilfully neglects earned income, temporarily. During the second component, he deducts no earned income on the transfers in a low-income area. For every single claimant who earns a minimum 536 EUR this means a disposable income of 938 EUR. In the case of a family with two children, this means 1,970 EUR for every claimant who earns a minimum 810 EUR. Spermann is generous in this aspect, because the Four-Component Model is intended to compete with illegal incomes from clandestine employment.6 3 The negotiated wage in North Rhine-Westphalia includes 793 EUR per month for simple hairdresser assistant activities (as of December 2005). Claims for complementary transfers could be raised. However, this would be not up to the standard of the relative poverty level of about 938 EUR, but only up to 845 EUR per month according to Spermann’s own calculations. 4 Or the other way round, for having 800 EUR per month at his or her disposal, the program participant needs less working hours: 11 hours per week at the beginning, rising up to 20 hours after 12 months of the tax-financed transfer. 5 In an even more paradox case, transfer recipients with a 12 EUR hourly rate comply with the conditions for support already at 7,5 or 11 hours per week. Spermann mentions, however, as a condition for support 15 minimum working hours per week. 6 Strictly speaking, even in this model the monetary work incentives of a regular and illegal job are the same only if they can provide the same hourly wages. This is not the case in

Basic Income Reform in Germany – Comment

137

However, this increase of monetary work incentives in lower income areas forces to accept a reduction of the monetary incentives to work for higher incomes. A single of the above example can only achieve a disposable income of more than 938 EUR when he or she has a gross market wage of around 1,300 EUR or above. Every EUR, earned between 536 and 1300 EUR, will be deducted from the transfer by 100 per cent. The family can only achieve a higher income than 1,970 EUR, if the sole earner has a gross income per month of approximately 2,700 EUR or above. Market incomes between 810 and 2,700 EUR do not at all increase the available family income. This contrasts with a pure indigence-based transfer: without any exempts from transfer reductions, no single would have any monetary incentive to achieve an income up to 575 EUR. However, the monetary incentives would be effective from this amount upwards, even after the 50 per cent actual tax and social security deductions. In the Four-Component Model, a single has a strong motivation to achieve an income of up to 536 EUR, but no monetary incentives at all to increase this income from 537 to 1,300 EUR. The so-called unemployment trap is merely moved to higher income levels, which would pose a considerable new trap for part-time or low productive activities.

2.3 The Component of the Negative Income Tax

Alexander Spermann realizes the problem of the new incentive trap. Thus, he redeems the second component of neglected earned income after 12 months through a negative income tax with a transfer reduction rate of 66 per cent. What a single gets then in this phase from each earned EUR is 34 Cent.7 However, the respective households drop significantly from their comparatively high level of disposable income. While the single with his or her own earned income of 536 EUR in the 12th month of the tax-financed transfer has a total disposable income of about 938 EUR, in the 13th month the disposable income shrinks to 650 EUR. In the four-person-family example with 810 EUR earned income this means falling from 1,970 EUR in the 24th month of tax-financed transfer to 1,466 EUR in the 25th month. If the competition with clandestine employment is relevant, the transition from the second component to the third component of the negative income tax will push many participants again into illicit work.

situations when employers prefer illegal activity to reduce social security contributions or bureaucracy costs. 7 Even here, a reduction of the monetary incentives for higher incomes can be expected compared to a pure indigence-based system. This is because the transfer recipient keeps only 34 per cent of additional income, whereas the normal tax and social security system would leave 50 per cent of the recipient’s additional income.

138

Steffen J. Roth 2.4 The Component of Community Work

With the fourth component Spermann finally acknowledges that, at least, in the current situation of an insufficiently developed low-wage sector there are people who, regardless of their real efforts, cannot get any employment on the primary market. For those people community services are envisioned. Unfortunately, Spermann does not describe these services in more detail. The participants of such activities receive transfers amounting to the present transfer level of basic income in Germany according to the tax-financed AlG II. If the arrangement of these measurements follows the usual definition of community work, we have to expect serious distortions of competition between nonprofit employers eligible for community work and their private competitors. If these jobs have to be “additional to regular jobs” in order to prevent distorted competition problems, a mere occupational therapy would be the result, which would cause more costs than create benefit. A job, appropriate to the capabilities of the participants and useful for society as a whole, would be almost excluded in this case. Finally, the process could overload the administration charged with organising the community services, since it will have to create such jobs for all persons without a regular employment. An all-embracing provision of community work jobs is inevitable, because within the Four-Component Model participants receive still the basic income and are not pushed back to the physical subsistence income level. After all, in the case of the four-person-family this would be 311 EUR difference per month. As soon as monetary improvement can be expected from accepting a job with community services, every unemployed interested has to be provided with the opportunity of practising such a job. Mr. Spermann, though, expects that such jobs will be only necessary in a few exceptional cases. He mistakenly argues that for the claimants it would always be worthwhile to accept a regular employment instead of the artificially organised job opportunities in community work. However, this is far from being clear-cut. For example, a sole earner of a four-person family with a full-time job at an hourly wage of 5 EUR would not achieve as much money as what is granted in the case of a community service job.8 If, furthermore, it could not be excluded that the community work jobs seem less exhausting and unpleasant or less time-intensive than a regular employment in the low-wage sector, a regular employment would be unattractive even at higher hourly wages.9 8 At 160 hours per month, a family, in line with allowances of the third component of Spermann’s model, gets 1,463 EUR. A community service job grants the same family 1,471 EUR. 9 It is confusing that Spermann explicitly suggests that the working hours required for these activities should vary between household types. In fact, these activities must include full-time working, as far as the working capacity of the respective person permits. And this is all the more valid, if one suspects that transfer recipients increase their incomes with clandestine employment.

Basic Income Reform in Germany – Comment

139

3. Wage Subsidies Have Grave Adverse Side-effects in the Medium and Long Term Spermann does not explicitly discuss the imminent medium-term and long-term risks that aggravate the problems to be solved by subsidising the labour market with complementary wages. Wage subsidies suppose that too high wage claims by the unemployed are the reason for the high unemployment in the low-wage sector. The assumption is that citizens are not willing to accept jobs remunerated according to their productivity and the scarcity of their work. As a short-run symptom therapy, complementary wages give in to this illegal attitude with the payment of additional subsidies which are not justified by the indigence of the recipients. This reaction is purely mechanical. There is neither a search for the true causes of the problem nor an analysis of the causes and possible solutions from a regulatory policy point of view. Undoubtedly, people react to incentives in economic theory. Limiting the analysis of work incentives to only immediate incentives of monetary nature is, however, misleading. Including more aspects, a decision about choosing an employment will have a more complex background. The intrinsic value system of the individuals plays a more significant role for the wage demanded. In turn, this system of social norms, ethical qualities and individualistic values will likely not remain untouched by labour market subsidies. Values, attitudes and preferences usually are considered as constant in economic models. For short term partial-analytic considerations this is a practicable and reasonable approach. However, it is not appropriate when we evaluate broad system reforms: in the medium and long term preferences of the individuals show reactions to the institutional regulations (Lindbeck 1995a, 1995b, 2004). Seen from this angle, it becomes more and more clear that the models of subsidising labour market policy obviously become counterproductive in the medium and long term. All labour market participants get the idea that society does not expect the unemployed to accept a job for the sake of self-aid and the reduction of claimed benefits. Evidently, even the legal entities do not ask the unemployed to accept jobs at the common wage level. Rather, they consider it their task to raise employees’ remuneration above the basic income level. Clearly, the claim attitude of the today’s transfer recipients reluctant to work would thus be affirmed and intensified. The current law would be inconsistent and not credible. With this, the adverse side-effects of wage subsidies inevitable lead to an aggravation of the situation. Wage subsidies increase wage claims which have been mentioned before as the cause of the problem. Conversely, the incentives for unemployed persons and less productive employees to earn higher incomes through qualification, further education, flexibility and efforts, would decrease in many cases.

140

Steffen J. Roth

4. Conclusion The Four-Component Model shows the well-known problems of a subsidising labour market policy. Due to contradictory incentive effects for employment, it is unlikely that it will achieve the intended increase of labour supply in the short term. Whether such an increase in job offers would actually lead to an increase in employment depends also on how much the subsidies would lead to fiscal burdens. Financing the subsidies could affect labour demand, thwarting the targeted employment. Spermann points to these uncertainties – he cautions against raising too many hopes and reminds of beneficiary potential from measures in labour market regulation and in the collective wage bargaining process. It may be unsatisfactory, but sometimes less could be more. In view of the short-term uncertainty about labour market and fiscal effects caused by this kind of complementary wages, the middle and long-term counterproductive effects stemming from a deteriorating claim mentality of those concerned, become more important. As such, we have to think twice whether wage subsidies are really a good idea. An alternative would be a concentration on consistently implementing the wage supplement already legally in place. This should be accompanied by a mixture of, on the one hand, more pressure on transfer recipients to contribute as much as they can themselves for their support, and, on the other hand, improved possibilities for useful activities through tapping fully the resources of the low-wage sector. Here, obviously, our evaluations concerning the practicability of requesting self aid are different. It should be explicitly pointed out, though, our unanimity in the appraisal of the current political discussion of labour market instruments for the low-wage sector in Germany: minimum wages would be counterproductive in view of employment politics. Subsidising complementary wages without the effort to make the labour market more flexible at the same time and without a stretching of the wage structure are a fiscal adventure. And introducing minimum wage and complementary wage programs at the same time is the worst of all imaginable alternatives. References Eekhoff, J. / Roth, S. J. (2006): “Vorsicht vor Kombilöhnen – Eine vergleichende Analyse ausgewählter Maßnahmen und Konzepte subventionierender Arbeitsmarktpolitik”, München. Download at www.iwp.uni-koeln.de. Lindbeck, A. (1995a): “Welfare-State Disincentives with Endogenous Habits and Norms”, Scandinavian Journal of Economics 97(4). – (1995b): “Hazardous Welfare-State Dynamics”, American Economic Review, Papers and Proceedings 85.

Basic Income Reform in Germany – Comment

141

– (2004): “Improving the Performance of the European Social Model – The Welfare State over the Life Cycle”, in: Building a Dynamic Europe, edited by J. Gaul, Cambridge University Press. Roth, S. J. (2002): “Beschäftigungsorientierte Sozialpolitik”, Untersuchungen zur Wirtschaftspolitik 125, Köln. Spermann, A. (2006): “Mehr Druck, mehr Anreiz”, DIE ZEIT Nr. 4 / 2006 vom 19. 1. 2006, p. 26.

Participants 1. Jakob Ache

Deutscher Bundestag

2. J. Ade

Deutscher Bundestag

3. Sule Akkoyunlu

DIW Berlin

4. Mandy Alkens

University of Potsdam

5. Vincenzo Andrietti

University of Pescara, Italy

6. Balve-Hauff

BMWi

7. Annette Bender

BMF

8. Andreas Bley

BVR, Berlin

9. Ulrich Blum

IWH, Halle

10. Janna Böhringer

University of Potsdam

11. Philipp Bohle

Deutscher Bundestag

12. Deborah Bowen

DIW Berlin / AEQ

13. Nicola Brandt

BMF

14. Rolf Bürkl

GfK, Nuremberg

15. Bugru

BMWi

16. Herbert S. Buscher

IWH, Halle

17. Martin T. Clemens

IZA, Bonn

18. Hermann Clement

Munich

19. Andreas Cors

BMWi

20. Berend Diekmann

BMWi

21. Christian Dreger

DIW Berlin

22. Deike Fuchs

BMF

23. Reinhard Heck

BMWi

24. Christoph Helberger

TU Berlin

25. Andrea Hergt

FU Berlin

26. Dörte Höppner

DIW Berlin

27. Jochen Homann

BMWi

28. Klaus Hüfner

Deutsche Unesco Commission

29. Robert Jäckle

ifo, Munich

144

Participants

30. André Jungmittag

Bergische Universität Wuppertal

31. Christiane Kardweil

FTD

32. Astrid Klesse

BMWi

33. Sabine Klinger

IAB, Nuremberg

34. Thomas Knaus

BMWi

35. Stefan Kooths

DIW Berlin

36. Andreas Krallmann

BMWi

37. Ines Krug

BMWi / Teles AG

38. Richard Layard

CEP / London School of Economics

39. Harald Legler

NIW, Hannover

40. Georg Licht

ZEW, Mannheim

41. Alexander Lipponer

Deutsche Bundesbank, Frankfurt am Main

42. Karolina Lyczywek

FU Berlin

43. Philipp Mende

British Embassy

44. Ralf Messer

ARGE / DIW Berlin

45. Martin Meurers

BMWi

46. Philipp Mohl

BMF

47. André Müller

BMF

48. Johannes Müller

BMWi

49. Karl-Heinz Nöhrbaß

BMF

50. Peter Nunnenkamp

IfW, Kiel

51. Frank Oskamp

IfW, Kiel

52. Wolfgang Quaisser

OEI, Munich

53. Guido Raddatz

Stiftung Marktwirtschaft

54. Martina Rengers

Federal Statistical Office

55. Javier Revilla Diez

NIW, Hannover

56. Alexander Romanski

Berliner Volksbank

57. Steffen J. Roth

Institut für Wirtschaftspolitik, University of Cologne

58. Karin Scheffel

BMWi

59. Claudius Schmidt-Faber

Berlin

60. Sibylle Schmerbach

HU Berlin

61. Danny Schulze

Europa-Universität Viadrina, Frankfurt / Oder

62. Ludwig Schuster

BMWi

63. Jeanette Schwamberger

BMF

64. Patrick Seidewitz

Canadian Embassy

Participants 65. Boris Siliverstovs

DIW Berlin

66. Dennis J. Snower

IfW, Kiel

67. Albrecht Söllner

Europa-Universität Viadrina, Frankfurt / Oder

68. Alexander Spermann

ZEW, Mannheim

69. Hildegard Stahmer

HWWA, Hamburg

70. Thomas Straubhaar

HWWA, Hamburg

71. Bettina Stuchtey

BMWi

72. Sebastian Thomasius

BMF

73. Carsten Ullrich

Canadian Embassy, Berlin

74. Ingrid Vogel

US Embassy, Berlin

75. Christian von Kienlin

BMWi

76. Martin Weale

NIESR, London

77. Wolfram Wilde

HU Berlin

78. Joachim Wuermeling

BMWi

79. Klaus Zimmermann

DIW Berlin

10 Supplement 57 – 2006

145