This Selected Issues paper on Euro Area Policies underlies global rebalancing of accounts. From a growth-accounting perspective, slower growth in the capital-labor ratio seems to be the main driver behind the deceleration in labor productivity. The increase in bilateral trade was accompanied by a large bilateral EU trade deficit. China’s market share seems to have increased mainly at the expense of other East Asian countries. EU trade with China increased at more than twice the rate of total EU external trade, and China became the EU’s second largest trading partner.

Abstract

This Selected Issues paper on Euro Area Policies underlies global rebalancing of accounts. From a growth-accounting perspective, slower growth in the capital-labor ratio seems to be the main driver behind the deceleration in labor productivity. The increase in bilateral trade was accompanied by a large bilateral EU trade deficit. China’s market share seems to have increased mainly at the expense of other East Asian countries. EU trade with China increased at more than twice the rate of total EU external trade, and China became the EU’s second largest trading partner.

I. Why is Productivity Growth in the Euro Area So Sluggish?1

Core Questions, Issues, and Findings

  • What are the main stylized facts regarding recent trends in euro-area labor productivity growth? Since the mid-1990s, the area’s labor productivity growth (output per hour) has declined markedly. Across sectors, the deceleration was concentrated in the traditional production sectors, i.e. sectors that are neither producers nor intensive users of information and communications technology (ICT). The deceleration was largely common across countries. From a growth accounting perspective, slower growth in the capital-labor ratio seems to be the main driver behind the deceleration in labor productivity. (¶12 to ¶26)

  • Why did the pace of euro-area productivity growth slow? The main factors seems to be sustained wage moderation and some progress on structural reforms since the early 1980s, which have induced firms to shift to more labor-intensive production, reversing earlier substitution policies in favor of capital. (¶27 to ¶39)

  • Can labor reabsorption in the euro area fully account for the divergence between euro-area and U.S. labor productivity growth since the mid-1990s? It does not. Unlike the United States, the euro area did not experience a productivity surge in ICT-using service sectors, mainly wholesale and retail trade and financial intermediation. (¶17)

  • Does the exceptionally low productivity growth during the recent protracted slowdown point to a further decline of underlying trend growth? It may be too early to tell. At least in part, the very low productivity growth since 2001 reflects cyclical factors, especially more labor hoarding than during previous cycles. (Appendix IV)

  • Given this chapter’s analysis, what should be the structural policy priorities? The Lisbon agenda and the Broad Economic Policy Guidelines (BEPGs) contain the right pointers. Reversing the area’s secular decline in labor utilization should be a priority to ensure fiscal sustainability and preserve the integrity of social protection systems. Moreover, lagging productivity growth in ICT-using service sectors points to the need to accelerate both product and labor market reforms. (See Chapter II for a detailed discussion of the Lisbon agenda).

A. Introduction

1. Labor productivity in the euro area seems to have risen a bit above U.S. levels in the mid-1990s, hinting at a full technological catch-up, but has lost some ground since then. Several analysts have pointed to a decline in total factor productivity (TFP) growth in the euro area as an important cause for the sluggish labor productivity since 1995.2 Others have highlighted the productivity surge in key high-tech sectors in the United States as crucial to the performance gap.3

2. Unlike previous studies, this chapter focuses on the euro area (rather than on the EU or individual countries) and argues that:

  • The bulk of the labor productivity deceleration in the euro area in the second half of the 1990s can be explained by slower capital deepening (slower growth in the capital-labor ratio), as opposed to slower TFP growth. The apparent slowdown in TFP growth obtained from productivity calculations using national accounts data for the euro area disappears once better, industry-level data for Germany are considered in the analysis. Therefore, the sluggishness in euro-area labor productivity in the second half of the 1990s should be more associated with the use of production inputs and not with negative technological or structural shocks.

  • The slower capital deepening in the euro area in the second half of the 1990s can be explained by structural wage-setting changes. These changes made labor cheaper, inducing firms to slow the process of capital accumulation and to hire more workers. To quantify the effect of these structural labor market changes on capital deepening, the chapter develops a simple model for evaluating how structural changes in wage setting affect labor productivity growth. Calculations based on econometric estimates using industry-level data for a subset of euro-area countries (France, Germany and the Netherlands) show that wage-setting shocks would have forced capital-labor ratios to decline in the second half of the 1990s. In the event, capital-labor ratios grew at a slower rate but did not decline, as other factors, including cheaper information and communications technology (ICT) equipment, partly offset the wage shock.

  • The productivity growth differential with respect to the United States since the mid-1990s can be explained by a faster labor productivity deceleration in traditional industries (i.e. industries that are neither producers nor intensive users of ICT) in the euro area and, to a lesser extent, by a surge in productivity growth in intensive-ICT-using sectors (mainly wholesale and retail trade and financial intermediation) in the United States. Productivity behavior in ICT-producing sectors (e.g. computers, semiconductors, and communication services) was similar in the two areas.

3. Looking ahead, policies to improve labor utilization in Europe should continue in the medium term as the Lisbon targets are pursued, which might dampen labor productivity growth through slower capital deepening. However, lower labor productivity growth is a temporary phenomenon that will fade away when the economy reaches a new equilibrium unemployment rate. In addition, the labor market reforms needed for the continuation of low wage growth and reductions in the unemployment rate should improve economic efficiency. Besides labor market reforms, further product market deregulation (particularly in wholesale and retail trade) would promote efficiency gains, and help to close the productivity growth gap with respect to the United States. Higher TFP growth could also be attained by letting markets better reward individual effort, which would raise risk-taking activities, R&D spending, and human capital accumulation.

4. The next section discusses labor productivity developments in the euro area and in the United States using aggregate national accounts data within a larger context of convergence in GDP per capita between the two regions. It serves as a motivation for the paper and presents a decomposition of labor productivity growth in the euro area and in the United States into the contributions of capital deepening and TFP growth. Section C presents calculations using the industry-level database from the Groningen Growth and Development Center (GGDC) for the 12 euro-area countries and the United States. These calculations document productivity developments among intensive users of ICT equipment, producers of ICT equipment, and more traditional industries. Then, the GGDC growth accounting database for France, Germany, the Netherlands, and the United States is used to provide a breakdown of labor productivity growth into the contributions of changes in ICT and non-ICT capital, labor quality, and TFP. Section D proposes a simple wage-bargaining model to illustrate how structural labor market changes would affect the adjustment path of labor productivity growth through changes in capital deepening. An econometric estimate for the effect of structural wage-setting changes on capital deepening and, therefore, labor productivity is provided. Section E concludes this chapter by using key results from the literature to highlight the effect of structural changes, including deregulation of product markets, on TFP growth. Appendix IV provides a simple way to integrate the conclusions of this chapter into an analysis of the degree of labor hoarding at the current cyclical juncture.

B. GDP Per Capita and Productivity Growth in the Euro Area and in the United States

5. The long-run pattern of declining GDP per capita growth in the euro area has a mirror image in declining trend rates of labor productivity growth. Trend GDP per capita growth in the euro area has been declining since the 1950s, finally bringing to a halt the convergence to U.S. levels in the 1970s (Figures 1 and 2). In the United States, labor productivity growth oscillated around 1½ percent for many years until it trended up in the second half of the 1990s, surpassing the euro-area figures for the first time (Figure 3 and Table 1).4 Increasing employment rates in the United States (Figure 4 and Table 1) widened this gap and GDP per capita growth in the second half of the 1990s was about 1 percentage point higher than in the euro area.

Figure 1.
Figure 1.

GDP per Capita Trend Growth

(5-year moving average, in percent)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Figure 2.
Figure 2.

PPP GDP per Capita in the Euro Area as Percentage of U.S. Value

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Sources: EC - AMECO database; OECD Productivity database; and staff calculations.
Figure 3.
Figure 3.

Labor Productivity Growth

(5-year moving average, in percent)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Figure 4.
Figure 4.

Employment Rates

(In percent of total population)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Sources: EC - AMECO database; OECD Productivity database; and staff calculations.
Table 1.

GDP Per Capita Growth

(Annual rates, in percent)

article image
Sources: EC-AMECO database; OECD productivity database; and staff calculations.

6. GDP per capita growth in the euro area, even if lower than in the United States, did increase in the second half of the 1990s, when a surge in employment rates offset a deceleration in labor productivity and continued declines in average hours of work (Figures 4 and 5, and Table 1). The opposite movements of employment rates and labor productivity during this period suggest that lower labor productivity growth in the euro area could be related to the reinsertion of unemployed individuals into jobs. On the other hand, the positive correlation between accelerating productivity and employment rates in the U.S. during the same period is consistent with increased technological growth and economic activity in an economy near its natural rate of unemployment.

Figure 5.
Figure 5.

Annual Hours Per Worker

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Sources: Staff calculations based on total hours from OECD and employment from EC -AMECO.

7. Breaking down labor productivity growth into the contribution of hours of work, capital, and TFP shows that a significant decline in capital deepening (a slower increase in the capital-labor ratio) explains a large part of the productivity deceleration in the euro area (Table 2). 5 However, the aggregate national accounts-based data used here also show that TFP growth declined in the euro area while sharply increasing in the United States. A note of caution should be introduced at this point. Cross-country comparisons using national accounts data could be compromised by different national methodologies in the calculation of investment flows, deflators (including the treatment of quality improvements in high-tech equipment), aggregation methods, and so on. In addition, changes in labor quality could bias the TFP measures shown in Table 2. While these are crucial issues, the chapter assumes them away for now but will return to them later.

Table 2.

Labor Productivity Growth

(Annual rates, in percent)

article image
Sources: EC-AMECO database; OECD productivity database; and staff calculations.

8. TFP growth picked up in the US in the second half of the 1990s but not in the euro area. In fact, euro-area TFP seems to have converged to U.S. rates for 1970-95. The boost in U.S. technological growth in the second half of the 1990s has been associated to the larger production of ICT equipment and more intensive ICT use. The cyclical decline in TFP growth during 2001-2003 was about the same in the two countries.

9. The reduced rate of capital deepening in the euro area in the second half of the 1990s can be associated with the reinsertion of unemployed workers into jobs because of reduced wage demands. That is consistent with the rate of capital growth declining only slightly while work hours growth surged in the euro area in the second half of the 1990s (Figure 6). In addition, real hourly compensation in the euro area in the second half of the 1990s grew significantly more slowly than in the U.S. for the first time since the series has been available (Figure 7). Overall, euro-area hourly compensation seems to follow a “boombust” pattern, but the downward trend in its growth rate is probably associated with labor market reforms and wage moderation agreements between social partners, that began in the 1980s and continued through the 1990s. This trend seems to have affected capital/labor growth already in the 1980s, and the first half of the 1990s represented only a pause. High unemployment rates could also have tamed wage demands, and, in Section D, the two effects will be isolated. In any case, labor cost developments were translated into a negative trend unit labor cost growth (total labor compensation divided by output, as in Figure 8).

Figure 6.
Figure 6.

Breaking Down Changes in the Capital-Labor Ratio

(Percent, annual rate)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Sources: EC-AMECO database for capital stock and employment; OECD for average hours of work; and staff calculations.
Figure 7.
Figure 7.

Real Hourly Compensation

(Percent changes)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Figure 8.
Figure 8.

Unit Labor Costs

(Percent changes)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Sources: EC - AMECO database; OECD Productivity database; and staff calculations.

C. New Evidence on Labor Productivity Growth in the Euro Area and in the United States Using Industry-Level Data

10. Observers have attributed the productivity acceleration in the United States in the 1990s to what has been dubbed the “new economy”—an acceleration in technical change in which rapid investment and use of ICT transformed business practices leading to new breakthroughs and the wider adoption and use of ICT. Oliner and Sichel (2000), and Jorgenson and Stiroh (1999 and 2000) first documented the surge in U.S. productivity growth using traditional growth accounting techniques. They show that the accumulation of ICT capital plus the growth in TFP in the computer and semiconductor industries accounted for over three-fourths of the labor productivity acceleration in the U.S. nonfarm business sector. Still, about one-third of the acceleration is accounted for by TFP growth in non-ICT sectors.

11. More recent work sheds light on differences between U.S. and European productivity developments, focusing on either a small sample of European countries (Jorgenson, 2003, who also provides evidence for Japan) or on the European Union as a whole (O’Mahony and van Ark (2003)). O’Mahony and van Ark (2003) also present country-specific calculations for labor productivity growth and document some of the cross-country disparities within the European Union. In this section, the focus is shifted to the euro area as a whole, and to comparisons with the United States.

Labor productivity growth by ICT classification and countries

12. The first industry database used provides information for 15 EU countries and the United States. The database was constructed by the GGDC departing from the OECD STAN database and national sources. It contains information on value added (real and nominal), employment, and hours of work for 56 industries in each of these countries. The database corrects several problems with the aggregate data used in Section B. Most important, the GGDC used information on quality changes in ICT equipment from the U.S. statistical agencies to correct data for all the other countries. All sector and country aggregations performed in this chapter use value-added weights at the industry level. For more information on the Industry Productivity Database, see Appendix I.

13. The industry data broadly confirm labor productivity developments described in the previous section, with one important difference: labor productivity growth does not decline as much in the second half of the 1990s as shown in Tables 1 and 2. According to the results in Table 3, labor productivity decelerated by 0.7 percentage point in the euro area in the second half of the 1990s, as opposed to the 1 percentage point indicated in the first two tables. Again, one could claim that labor productivity growth in the euro area converged to U.S. rates up to the mid-1990s (about 1.5 percent at an annual rate) but missed the technological shock observed in the United States thereafter.

Table 3.

Labor Productivity Growth by ICT Classification1

(In percent, at an annual rate)

article image
Sources: Industry Labor Productivity Database - EC and GGDC; and staff calculations.

Productivity is defined as real value added per hours worked. Detailed breakdown by ICT type listed in Appendix I.

Includes office machinery, telecommunications equipment, and scientific instruments.

Comprises communications, and computer and related activities.

Includes most transportation equipment (excludes motor vehicles), mechanical engineering, and printing and publishing.

Includes wholesale and retail trade, and financial intermediation.

Includes motor vehicles, chemicals, basic and fabricated metals.

Includes real estate activities and public services.

Includes agriculture, construction, and mining and quarrying.

14. The productivity deceleration in non-ICT industries accounts for most of the gap between the two regions, as this grouping covers over 60 percent of economic activity in each country (Table 3). While the euro area has always outperformed the U.S. in this category, labor productivity growth fell from 2.1 percent at an annual rate in the first half of the 1990s to 0.9 percent at an annual rate in the subsequent six years. In the U.S., the deceleration was of only 0.4 percentage point. This gap accounts for ½ percentage point of the ¾ percentage point difference between labor productivity growth in the euro area and in the U.S. at the end of the 1990s. The deceleration in productivity in this grouping accounts for virtually all of the deceleration in aggregate labor productivity in the euro area. The service industries in this category account for the majority of the discrepancy, not least because of their large weight in the economy.

15. The much faster acceleration in work hours in the non-ICT sector vis-à-vis the United States (1.5 percentage points in the euro area versus 0.5 percentage point in the United States) explains all of the relative labor productivity deceleration (1.2 percentage points in the euro area versus 0.4 percentage point in the United States) (Table 4). The acceleration in hours of work suggests that changes in the relative costs of capital and labor may be behind the sluggish productivity in the sector.

Table 4.

Acceleration in Total Work Hours1

(In percent, at an annual rate)

article image
Sources: Industry Labor Productivity Database - EC and GGDC; and staff calculations.

Detailed listing of all industries in each ICT category in Appendix III.

Includes office machinery, telecommunications equipment, and scientific instruments.

Comprises communications, and computer and related activities.

Includes most transportation equipment (excludes motor vehicles), mechanical engineering, and printing and publishing.

Includes wholesale and retail trade, and financial intermediation.

Includes motor vehicles, chemicals, basic and fabricated metals.

Includes real estate activities and public services.

Includes agriculture, construction, and mining and quarrying.

16. The euro area has seen large productivity increases in several high-tech industries. 6 Among ICT-producing industries, the euro area has always lagged behind the United States but not by much (Table 3, row 2). In addition, labor productivity growth in this category increased in the second half of the 1990s in both regions. Within ICT producers, the euro area lags in manufacturing but is an outstanding performer in services, where productivity growth jumped significantly in the second half of the 1990s while declining in the United States.

17. Turning to intensive ICT users (mainly wholesale and retail trade, and financial services), productivity growth declined in the euro area but surged in the United States in the second half of the 1990s. The large difference between the two regions in this category was caused by a productivity surge in service industries in the United States. Productivity growth among intensive ICT users in manufacturing in the euro area remained unchanged and much above U.S. rates. Lagging deregulation in product and labor markets, as described in chapter 2, is likely to have dampened efficiency gains in ICT-using service industries in the euro area. In addition, the much faster acceleration in work hours in the sector vis-à-vis the United States partly explains the lower productivity growth (Table 4).

18. The aggregate euro-area pattern masks important cross-country differences (Table 5). In ICT-using sectors, labor productivity growth increased between the first and the second half of the 1990s in several countries (Ireland, Netherlands, Portugal and Spain, although only Ireland had larger growth than the United States). However, the weight of the three largest euro-area countries (with some help from other smaller countries) forced down productivity growth in this category. The largest countries also imposed most of the productivity deceleration on the large non-ICT sector. Among them, Italy experienced the largest declines in productivity growth after 1995. Overall, Italy contributed about 40 percent of the 0.7 percentage point deceleration in labor productivity growth in the euro area in the second half of the 1990s (Table 6).

Table 5.

Labor Productivity Growth Across Countries

(In percent, at an annual rate)

article image
Sources: Industry Labor Productivity Database - EC and GGDC; and staff calculations.Notes: Productivity is defined as real value added per hours worked. Detailed breakdown by ICT type listed in Appendix III.
Table 6.

Labor Productivity Deceleration in the 1990s

article image
Source: Staff calculations based on aggregation shown in Table 5.

Demonstrating the importance of capital deepening and correct TFP calculation

19. The previous analysis of labor productivity developments is hampered by the lack of information on capital formation and changes in labor quality. The Growth Accounting Database put together by the GGDC close this gap. It provides information on growth in real value added, hours of work, ICT capital, non-ICT capital, labor quality, and TFP. Data availability determined its coverage—the database contains information for three euro-area countries (France, Germany and the Netherlands), the U.K. (not used here), and the United States—the end-point for the analysis (2000), and a somewhat more aggregated industry classification (26 industries) than provided by the Industry Productivity Database. All the methodological improvements introduced in the Industry Productivity Database, including the homogenization of treatment of quality changes in ICT equipment, are also present in the Growth Accounting Database. The method used to break down labor productivity growth into its main components corresponds to the traditional methodology discussed, for instance, in Oliner and Sichel (2000). The database is described in more detail in Appendix II and the breakdown of labor productivity growth follows equation (A.2). When comparing to the breakdown shown in Table 2, capital deepening has two components, ICT and non-ICT capital deepening, and changes in labor quality are measured separately instead of being included in TFP growth.

20. Turning to the components of labor productivity growth, the TFP growth shown in Table 2 is misleading: while German TFP accelerates continuously when carefully measured according to the GGDC, it declines sharply when using aggregate data (Table 7). Given the weight of Germany in the euro area’s aggregate (about 30 percent of total value added in the area) and considering the TFP calculations based on the detailed industry database as superior, TFP growth in the area would actually have been 0.35 percentage point higher than shown in Table 2—about the size of the deceleration in TFP shown in that table. If labor productivity growth in Table 2 were augmented by this amount, the deceleration in euro-area labor productivity would conform to the measurement based on the industry data shown in Table 4 (about 0.7 percentage point). The general profile of TFP growth in France and in the Netherlands is similar in both calculations.

Table 7.

Productivity Growth in Two Different Databases1

(In percent, at an annual rate)

article image
Sources: Growth Accounting Database - EC and GGDC; EC-AMECO and OECD; and staff calculations.

Productivity is defined as real value added per hours worked.

Total factor productivity (TFP) from the Growth Accounting Database calculated as a residual after taking into account the contribution of different types of capital deepening and labor quality changes. Calculations using AMECO and OECD data do not correct for quality changes in ICT equipment, changes in labor quality, and aggregation issues.

21. The contribution of ICT capital deepening to productivity growth increased significantly for all countries while the contribution of non-ICT capital deepening declined, becoming negative in France and zero in the Netherlands (Table 8). Labor quality growth contributed less to productivity growth in the Netherlands and in Germany, but not in France. Looking at the ICT-based breakdown, the contribution of non-ICT capital deepening declined in all groupings for all countries between the first and the second halves of the 1990s, while the contribution of ICT capital deepening increased. That is consistent with the widespread use of ICT equipment in these countries even in the face of large increases in labor usage. TFP grew differently depending on the country and the sector being analyzed.

Table 8.

Decomposition of Labor Productivity Growth in Three Euro Area Countries1

(In percent, at an annual rate)

article image
Sources: Growth Accounting Database - EC and GGDC; and staff calculations.

Productivity is defined as real value added per hours worked. Detailed breakdown by ICT type listed in Appendix III.

Capital deepening defined as changes in the capital to hours worked ratio.

Labor quality changes calculated by the ratio of hours weighted by wages of individuals with different educational backgrounds.

Total factor productivity (TFP) calculated as a residual.

Includes office machinery, telecommunications equipment, scientific instruments, communications, and computer and related activities.

Includes most transportation equipment, mechanical engineering, printing and publishing, wholesale and retail trade, and financial services.

Includes agriculture, construction, mining, motor vehicles, chemicals, basic and fabricated metals, real estate activities and public services.

22. A deceleration of capital deepening is the key factor behind gaps in labor productivity growth between the U.S. and an aggregate of France, Germany, and the Netherlands (called euro-3 in Table 9). The contribution of non-ICT capital deepening to labor productivity growth remained unchanged in the U.S. in the second half of the 1990s but declined markedly in the euro-3 aggregate. In addition, the contribution of ICT capital deepening to labor productivity growth increased by twice as much in the U.S. than in euro-3.

Table 9.

Decomposition of Labor Productivity Growth in Euro-3 and in the U.S.1

(In percent, at an annual rate)

article image
Source: Growth Accounting Database - EC and GGDC; and staff calculations.

Productivity is defined as real value added per hours worked. Detailed breakdown by ICT type listed in Appendix III.

Industry value-added weights used to aggregate data underlying Table 8.

Capital deepening defined as changes in the capital to hours worked ratio.

Labor quality changes calculated by the ratio of hours weighted by wages of individuals with different educational backgrounds.

Total factor productivity (TFP) calculated as a residual.

Includes office machinery, telecommunications equipment, scientific instruments, communications, and computer and related activities.

Includes most transportation equipment, mechanical engineering, printing and publishing, wholesale and retail trade, and financial services.

Includes agriculture, construction, mining, motor vehicles, chemicals, basic and fabricated metals, real estate activities and public services.

23. TFP growth rose by ¾ percentage point in the U.S. in the second half of the 1990s but remained lower than the rates posted in euro-3, which increased 1/3 percentage point during this period. The TFP growth differential in favor of the euro-3 aggregate contrasts with the message for the euro area as a whole shown in Table 2. Again, methodological problems with the aggregate data used in Table 2 likely overestimate the decline in TFP growth for the euro area as a whole, but the partial coverage of the euro-3 aggregate (in particular, the exclusion of Italy) may help to explain the more upbeat productivity scenario.

24. Looking at the ICT groupings, labor productivity in non-ICT industries decelerated much less in the United States than in the euro-3 aggregate. In addition, the productivity deceleration in the U.S. non-ICT sector was caused by a large decline in TFP growth that was partly offset by more capital deepening and faster improvements in labor quality. In contrast, in the euro-3 aggregate, TFP growth in the non-ICT sector remained nearly unchanged while declines in non-ICT capital deepening and labor quality growth accounted for the deceleration in labor productivity. Unlike the non-ICT grouping, labor quality growth in the euro-3 grouping increased in the ICT sectors in the second half of the 1990s. The United States posted larger increases in both TFP growth and capital deepening in ICT-producing and, more important, ICT-using industries than the euro-3 aggregate.

Summary of results from the sectoral productivity analysis

25. A much slower deceleration in labor productivity in non-ICT industries and a faster acceleration in ICT-using sectors accounted for the U.S. productivity growth lead over the euro area in the second half of the 1990s (Tables 10 and 11). Labor productivity acceleration in ICT-producing industries in the second half of the 1990s was faster in the euro area than in the United States but that had little effect on aggregate developments because of the small share of this sector in total value added (Table 10, row 2). The decline in labor productivity growth in the euro area is almost fully accounted for by the decline in labor productivity growth in non-ICT sectors (Table 10, third column, eighth row.)

Table 10.

Contributions to Aggregate Labor Productivity Acceleration1

(In percentage points, at an annual rate)

article image
Sources: Industry Labor Productivity Database - EC and GGDC; and staff calculations.

Productivity is defined as real value added per hours worked. Detailed breakdown by ICT type listed in Appendix III.

Includes office machinery, telecommunications equipment, and scientific instruments.

Comprises communications, and computer and related activities.

Includes most transportation equipment (excludes motor vehicles), mechanical engineering, and printing and publishing.

Includes wholesale and retail trade, and financial intermediation.

Includes motor vehicles, chemicals, basic and fabricated metals.

Includes real estate activities and public services.

Includes agriculture, construction, and mining and quarrying.

Table 11.

Contributions to Labor Productivity Acceleration1

(In percentage points, at an annual rate)

article image
Sources: Growth Accounting Database - EC and GGDC; and staff calculations.

Productivity is defined as real value added per hours worked. Detailed breakdown by ICT type listed in Appendix III.

Industry value-added weights used to aggregate data underlying Table 7.

Includes office machinery, telecommunications equipment, scientific instruments, communications, and computer and related activities.

Includes most transportation equipment, mechanical engineering, printing and publishing, wholesale and retail trade, and financial services.

Includes agriculture, construction, mining, motor vehicles, chemicals, basic and fabricated metals, real estate activities and public services.

26. Turning to a breakdown of aggregate labor productivity growth, the difference in performance vis-à-vis the United States can be accounted for by a decline in capital deepening and slower labor quality improvements observed in an aggregate of France, Germany and the Netherlands. These variables grew at a faster rate in the United States after 1995. TFP growth increased in the euro-3 aggregate in the second half of the 1990s but more slowly than in the United States. These variables are not readily available for the euro area as a whole but if generalized for the remaining 40 percent of the economy, they suggest that the decline in labor productivity growth in the second half of the 1990s discussed in Section B was not caused by slower technological growth (or at least not as much as suggested by the aggregate data used in Table 2). Slower capital deepening was the most important culprit.

D. Structural Labor Market Changes and Capital Deepening

27. The stylized facts produced so far can be mapped into an analytical framework relating structural labor market changes and productivity developments. Taking the results for the euro-3 aggregate as representative for the euro area as a whole, the actual reduction in labor productivity growth in the second half of the 1990s was rooted in the sharp declines in non-ICT capital deepening, which were the counterpart of the large increase in work hours in the period. Some studies suggest that this job-rich growth was caused in part by changes in the basic parameters of the wage-setting mechanism that shifted rightward a “labor-supply-like” relationship between real wages and the unemployment rate.7 Other studies claim that workers actually learned from the mistakes of the past after observing the consequences of excessive wage demands8, or that a set of factors could have conspired to generate lower wage growth in the 1990s.9 Among many factors, declines in unions’ bargaining power—maybe related to globalization—implicit contracts with governments—who provided services to workers in exchange for less wage demands—and targeted reductions in labor cost taxation are worth listing. Increased use of active labor market policies (mainly the policies directed toward increasing labor demand by private corporations) were also shown to have lowered wages for a given rate of unemployment and increased employment rates in a sample of OECD countries, including most euro-area economies.10 Finally, labor market reforms allowing a better use of temporary and part-time work in many euro-area countries could also have strengthened labor market competition and held wage growth down.

Benchmark model

28. Structural labor market changes such as the ones described in the previous paragraph are quite consistent with the stylized facts unearthed in Section C and a simple model captures the basic idea and provides a framework for the econometric analysis.

29. A short-run labor demand curve, as SLD in Figure 9, can be obtained under standard neoclassical assumptions. Following Blanchard (1997), assume the economy grows along a balanced path determined by the rate of labor-augmenting (Harrod-neutral) technological growth, ga. The curve SLD is derived by assuming that the production function combines labor and capital according to a constant-returns-to-scale technology, that capital is fixed in the short run and that firms maximize profits. The labor force is normalized to 1 and employment is N = 1-u (u is the unemployment rate). Wages are defined in efficiency units, i.e. as a ratio of the technology level, A.

Figure 9.
Figure 9.

Structural Labor Market Changes and Long-Run Adjustment

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

30. In the long run, capital varies and, assuming interest rates are determined abroad, the user cost of capital is exogenously given. In this case, labor cost in efficiency units is set to equalize the profit rate to the user cost of capital independently of the unemployment rate (LLD in Figure 9).

31. A “labor-supply-like” relationship can be modeled according to the right-to-manage model, in which firms and unions bargain over wages, given the short-run labor demand. A version of such a model, developed in Estevão and Nargis (2002), generates

WA*B*τ=f(m,u)fm>0andfu<0,(1)

where B stands for the income a worker would receive if unemployed, and τ stands for the ratio of the fiscal wedge on unemployment income to the fiscal wedge on labor income; m is a structural parameter determining the position of the wage curve and its steepness. Equation (1) represents a contract curve relating wages in efficiency units to the unemployment rate (the wage-setting curve, WS, in Figure 9). For a given rate of unemployment, wages depend on unemployment income (net of the relative tax wedge) and on the position of the wage curve, a function of m. Ceteris paribus, wage demands are higher the higher is unemployment income (which depends, among other things, on unemployment benefits replacement rates), as the outcome in case of disagreement (and the worker is unemployed) is less unattractive. On the other hand, when the unemployment rate increases, the probability of not finding a job also rises and wage demands are more subdued. Whenever workers’ bargaining power becomes weaker, or whenever workers value employment more, the parameter m decreases and wages are lower for a given rate of unemployment. Changes in the degree of labor market competition (e.g. because of reforms that allow better allocation of labor, like the deregulation of part-time and “temp” work in Spain and France in the 1990s), will also affect the position of the wage-setting relationship.

32. Wage-setting changes trigger an adjustment path where labor productivity growth declines at first, but then surges before returning to its original steady state. Point E in Figure 9 represents the long-run equilibrium in the labor market, where wages are such that the profit rate equals the worldwide user cost of capital. In this steady state, output, capital, and employment in efficiency units (AN) grow at ga percent. Under the hypothesis of a significant downward shift in the wage-setting curve—due, for instance, either to a general agreement for wage moderation, as in the Wassenaar agreement in the Netherlands in the 1980s, or to some labor market deregulation—wages will grow more slowly than technological progress and the unemployment rate will decline as the economy moves along a negatively sloped short-run labor demand curve and reaches the short-run equilibrium point E1. In this transition path, the rate of growth of the capital-labor ratio declines as labor grows faster than capital in efficiency units, K/A.

33. However, wage-setting changes in favor of cheaper labor for a given rate of unemployment will ultimately raise investment, as low wages raise profit rates to a level above the user cost of capital. In the longer run, the short-run labor demand will then shift outward, moving along the labor supply relationship, until the profit rate and the unit cost of capital are equal at point E2. Structural unemployment is lower than in E but wages in efficiency units are unchanged. While labor demand shifts, capital deepening speeds up as capital in efficiency units grows at a faster rate than labor.

34. During the transition path, technological growth remains unchanged, but the capital-labor ratio first decelerates and, then, accelerates, causing labor productivity growth to change as well. This adjustment pattern does not account for other possible effects from structural labor market changes on labor productivity growth. In particular, TFP growth is likely to benefit in the long run from labor market reforms as labor is allocated more efficiently. TFP growth may also suffer in the short run if labor quality is mismeasured and the newly hired unemployed are less efficient than currently employed workers. Changes in the sector composition of the labor force may also affect TFP aggregate productivity growth, although that seems to be a minor factor in explaining the disparities in productivity growth between the United States and the euro area.

Estimating the impact of wage moderation on capital deepening

35. The wage-setting relationship has been estimated in different ways, but, in general, empirical work has tended to prefer regressing the logarithm of wages on the logarithm of the unemployment rate. Therefore, empirical versions of equation (1) are in general written as

In[WtCPt*At]=ξt*γ-θ*ln(ut),(2)

where CPt represents consumer prices, ln(.) stands for the natural logarithm of a variable, and deviations from equilibrium levels of real hourly wages in efficiency units (ln(Wt/(CPt*At))) are modeled as ξt≠1. Therefore, in equilibrium at time 0, the wage-setting curve intercept is determined by γ, and structural shocks move the curve away from this value. Estimates of these changes can be obtained by assuming θ = 0.1, as has been estimated by Blanchflower and Oswald (1994) for many different countries.11

36. The large, negative wage-setting shocks of the 1970s, when workers resisted the efficiency shock from higher oil prices, were reversed in the 1980s and in the second half of the 1990s. This path is shown in Figure 10, which plots the accumulated wage-setting shocks for the euro area using aggregate data from the AMECO database and the OECD. By the end of the sample period, the wage-setting curve is roughly back at its position at the beginning of the 1970s, although there is some evidence of a small upward shift during the recent slowdown.

Figure 10.
Figure 10.

Accumulating Wage-Setting Shocks in the Euro Area

(Variable as defined in equation (7), 1970 = 100)

Citation: IMF Staff Country Reports 2004, 235; 10.5089/9781451812954.002.A001

Sources: EC-AMECO database and staff calculations. Labor cost data refer to hourly labor compensation.

37. In order to know the impact of wage-setting changes on capital deepening an elasticity estimate is needed. This estimate may be obtained by using the industry data discussed in Section C. This is a superior alternative to using the aggregate cross-country data because of the greater degrees of freedom, and the quality of TFP estimates and capital deepening obtained from the growth accounting database. Using these data, industry-specific measures of wage-setting shocks can be built as

ξijt*γij=ln(WijtCPjt*Aijt)+0.1*ln(ujt),(3)

where i stands for country, j for industry, and t for the time period. Consumer prices and the unemployment rate are measured at the country level. Industry-level technology, Aijt, gives the right norm for the wage increases industries could afford without weakening profit rates. Because wages are not available in the growth accounting database, hourly labor compensation is used instead.

38. The estimated equation is consistent with a simple relationship between the capital-labor ratio and the relative price of labor and capital, as implied by the neoclassical labor demand equation used in the model sketched above. Empirically, percent changes in the capital-labor ratio are modeled as a function of industry/country/year specific dummies and their interactions, represented by the linear function F(.), shocks in wage-setting (Δξijt) and in the user cost of capital (Δηijt), and residuals that are identically and independently distributed (εijt):

Δln(KijtLijt)=F(countryi,industryj,timet)+β*Δξijt-α*Δηijt+εijt.(8)

β is the parameter of interest here. Function F(.) captures a significant amount of variation in the data, including common industry shocks within a country (e.g. variations in central bank interest rate policy), common country shocks within an industry (e.g. industry-specific technological shocks), and time shocks in industry characteristics (e.g. changes in the composition of the labor force), among others. Because of a lack of information, the residual of the estimated regression includes industry-specific shocks in the user cost of capital, which are assumed to follow an AR(1) process but to be uncorrelated to wage-setting shocks. Information on total capital deepening was obtained by averaging the accumulation of ICT and of non-ICT capital, using the shares of ICT and non-ICT capital income in total capital income as weights.

39. Wage-setting shocks are estimated to affect capital deepening significantly in the panel data formed by France, Germany and the Netherlands, with an elasticity of 0.64 (Table 12). This elasticity can be used as representative of the euro area, since the estimation takes care of country-specific effects. Based on the evolution of wage-setting shocks as displayed in Figure 10, capital-labor ratios would have declined in the euro area in the absence of further shocks. The contribution of capital deepening to annual labor productivity growth would have been about -0.3 percentage point as opposed to the 0.4 percentage point shown in Table 2. Other factors, such as drops in the user cost of capital because of declining interest rates and ICT equipment prices, offset the strong push from these wage shocks for firms to substitute away from capital toward labor.

Table 12.

Elasticity of Capital Deepening to Wage-Setting Shocks1

article image
Sources: GGDC; AMECO database; and staff estimates.

Estimation uses industry-level data for France, Germany and the Netherlands. Standard errors are shown in parentheses and are corrected for AR(1) residuals.

stands for significant at the 5 percent level.

Wage-setting shocks measured as shown in equation (7). Consumer prices are measured by the implicit deflator for private consumption expenditures.

E. Additional Structural Changes to Boost TFP and Investment Growth

40. The same labor market reforms necessary to continuing reabsorbing people into jobs will probably ultimately increase TFP growth. These reforms should aim at increasing the incentives to work vis-à-vis receiving social benefits and correct incentives for human capital formation, with labor income better reflecting individual abilities and efforts. The increase in human capital accumulation and the better allocation of labor across alternative uses should boost TFP growth in the long term.

41. However, recent research has shown that other direct measures could be helpful in addressing the relatively weak TFP growth in Europe. The European Commission (2003) shows some evidence that the recent labor productivity differential between the U.S. and the European Union can be related to some fundamental structural differences at the individual country level, with five areas of significant quantitative importance: the level of product market regulation, the structure of financial markets, the degree of product market integration, the size of “knowledge” investment, and the aging of the labor force.

42. Turning to product market reforms, the analysis provided in this chapter points to the need for reforms in specific sectors. Notwithstanding considerable progress in product and financial market reforms (see, for instance, Chapter 2 and Blanchard (2004) for a recent positive evaluation) the gap in productivity growth in ICT-using services, which includes wholesale trade, retail trade, and financial intermediation, is worrisome. However, evidence from the McKinsey Global Institute research on productivity growth in France, Germany and the United States, does not clearly indicate which reforms should be implemented. Take the case of the retail food sector, for instance. McKinsey finds that labor productivity in that sector was actually 7 percent higher in France than in the U.S. in 2000. In addition, the degree of IT use in that sector was about the same in France, Germany and the United States in 1999, with the United States holding only a small lead. Blanchard (2004) suggests that barriers to firms’ entry and exit in the retail sector in Europe could be behind the productivity differentials. In fact, Foster, Haltiwanger and Krizan (2002) show that productivity growth in the U.S. retail trade sector in the 1990s can be attributed to the replacement of less productive by more productive establishments. In this sense, lowering barriers to and easing the regulatory burden on the creation of enterprises in Europe seem to be necessary.

43. The European Commission (2003) argues that, although it is important to address static efficiency problems, product market deregulation would not actually increase TFP growth in the long term. The document provides some simulations showing that even relatively rapid deregulation toward the U.S. levels would not lead to sufficiently large productivity gains over the next seven years to close the efficiency gap with the United States. The document stresses that any gains from deregulation in terms of technological catching-up or from privatizations of state monopolies should be interpreted more as static efficiency gains and not as the dynamic efficiency gains needed to expand the technological frontier.

44. However, product market reforms could positively affect those risk-taking activities that are the engine of technological progress. Furthermore, Chapter 2 finds some evidence linking product market reforms to future labor market reforms, which would not only improve labor market functioning, but also, depending on the type of labor market reforms, increase human capital accumulation—an engine of TFP growth.

45. The Commission’s work also suggests that long-run productivity gains from investments in both education and R&D would have a direct positive impact in TFP growth. With respect to R&D, the paper argues that the focus should not be on boosting R&D spending directly, but on creating the necessary conditions for promoting an endogenous increase in research spending. These could be obtained through two main channels: higher product market integration (e.g. through the completion of the single market program), and an investment environment that ensures the development of a more active market for risk capital.

46. Given the pattern of TFP growth in the three euro-area countries studied in detail in Section C, it is equally possible to argue that the euro area is only lagging the United States in terms of adoption of ICT technologies in some service sector industries. Although product market reforms and other structural changes would speed the diffusion of technology in the euro area, the diffusion will, nonetheless, happen. Evaluating such a hypothesis is outside the scope of this chapter and will be left to future research.

APPENDIX I The Industry Labor Productivity Database

47. The Industry Labor Productivity Database, put together by the Groninger Center for Development and Growth (GCDG), contains information on value added, employment, and hours worked in the 15 EU member states and the United States for 56 separate industries between 1979 and 2001. The point of departure for most countries was the new OECD STAN Database of national accounts. The STAN Database contains information on the most important national accounts variables from 1970 onward based on a common industrial classification. However, for a number of industries STAN does not contain sufficient detail. For example, the electrical engineering sector does not distinguish among semiconductors, telecommunications equipment, and radio and TV receivers. Wholesale trade and retail trade are aggregated in STAN, as are all industries within transport services as well as those within business services. To obtain a sufficiently detailed perspective on industry performance, the GGDC supplemented STAN with additional detail from annual production surveys, and service statistics. In addition, where necessary, more detailed national accounts were used from individual countries (e.g. in the case of Ireland). In general, the method employed was to use STAN aggregates as control totals and data from alternative sources to divide these totals into subindustries. The data series available from STAN are value added in current and constant prices (at basic prices), numbers of persons engaged (including self-employed), number of employees, total labor compensation, and, in a limited number of cases, working hours. Similar variables were available from survey statistics.12

48. Most important for this chapter, the Industry Labor Productivity Database homogenized the treatment of quality changes in computer and semiconductor prices across all countries. Following the work of Schreyer (2000 and 2002), the GGDC achieved international comparability in this area by using harmonized U.S. deflators for six ICT producing industries encompassing the production of computers, semiconductors, communications equipment and others, to correct value-added data for other countries. In the process, U.S. value-added deflators are corrected for differences in overall inflation between each country and the United States. In addition, the GGDC minimized the substitution bias in fixed-weight indices (like the Laspeyres) when calculating value-added at constant prices for higher levels of aggregation.

49. The GGDC used the Törnqvist method of aggregation to approximate an ideal Fisher price index, a procedure also followed here when calculating industry aggregates for the euro area and the United States. All the tables and results presented in this chapter for the euro area, the United States or the euro-3 aggregate uses value-added weights to get to (ICT-based) sectoral breakdowns.

APPENDIX II The Growth Accounting Database

50. The Growth Accounting Database from the GGDC provides information for three euro-area countries (France, Germany and the Netherlands), the United Kingdom (not used here), and the United States. The sample goes from 1980 to 2000, and it uses a somewhat more aggregated industry classification (26 industries) than provided by the Industry Productivity Database. The aggregations by the ICT taxonomy are based on a mapping between the listing in Appendix III and the 26 industries in the database. This was also the procedure used by O’Mahony and van Ark (2003) but it is possible that the mapping used here differs slightly from theirs, mainly in cataloguing some service industries as non-ICT users, as opposed to ICT users. All the methodological improvements presented by the Industry Productivity Database, including the homogenization of treatment of quality changes in ICT equipment, apply to this database. For more details, see the reference in footnote 13.13

51. The method used to break down labor productivity growth into several components assumes perfect markets and constant returns to scale so that the share of total capital is one minus the share of labor compensation in total value added—the same procedure used to break down the aggregate data in Section B. The database provides information on the labor share and the share of ICT capital income in total capital income. The assumption of constant returns to scale allows the share of each type of capital stock on value added to be recovered with this information.

52. The database also provides information on changes in labor quality calculated by first dividing total hours by skill level (education attainment), weighting the growth in each type by its wage share and subtracting total hours. The researchers divided, for each country, total hours worked into a number of different skill types. These types vary across country, but all include a high-skill category (degree and above) and a low-skill category (broadly equivalent to no high school graduation in the U.S.). Therefore, variations across countries in skill types are confined to intermediate categories. Second, capital input is measured using a Törnqvist capital service index, which comprises three assets for ICT—software, computers, and communications equipment—and three for non-ICT—non-ICT equipment, structures, and vehicles. Capital inputs are measured as service flows, and the share of each type in the value of capital is based on its user cost and not its acquisition cost.

53. To derive the productivity growth accounting equation, the GGDC assumed percent changes in output can be written as

Δy=αi*Δl+αl*Δq+αict*Δkict+αnict*Δknict+Δtfp,(A.1)

where αi represents the share of input i’s income in value added, Δ represents first differences, lower-case letters refer to the natural logarithm of each variable, y is real value added in a particular industry at time t (subscripts are omitted for simplicity), l is total hours of work, q is labor quality, kict and knict represent capital services of ICT and non-ICT equipment, respectively, and tfp is total factor productivity. Subtracting total hours from both sides of the above equation, and rearranging and employing constant returns to scale so that al + aict + anict = 1, gives a decomposition of average labor productivity growth as

Δp=αi*Δq+αict*(Δkict-Δl)+αnict*(Δknict-Δl)+Δtfp,(A.2)

where p is labor productivity, and the terms in parentheses are ICT and non-ICT capitalhours ratios.

APPENDIX III ICT Taxonomy14

1. ICT Producing - Manufacturing (ICTPM): Office machinery (30); Insulated wire (313); Electronic valves and tubes (321); Telecommunication equipment (322); Radio and television receivers (323); Scientific instruments (331).

2. ICT Producing – Services (ICTPS): Communications (64); Computer & related activities (72).

3. ICT Using – Manufacturing (ICTUM): Clothing (18); Printing & publishing (22); Mechanical engineering (29); Other electrical machinery & apparatus (31-313); Other instruments (33-331); Building and repairing of ships and boats (351); Aircraft and spacecraft (353); Railroad equipment and transport equipment not elsewhere classified (352+359); Furniture, miscellaneous manufacturing; recycling (36-37).

4. ICT Using – Services (ICTUS): Wholesale trade and commission trade, except for motor vehicles and motorcycles (51); Retail trade, except for motor vehicles and motorcycles; repair of personal and household goods (52); Financial intermediation, except insurance and pension funding (65); Insurance and pension funding, except compulsory social security (66); Activities auxiliary to financial intermediation (67); Renting of machinery & equipment (71); Research & development (73); Legal, technical & advertising (741-3).

5. Non-ICT Manufacturing (NICTM): Food, drink & tobacco (15-16); Textiles (17); Leather and footwear (19); Wood & products of wood and cork (20); Pulp, paper & paper products (21); Mineral oil refining, coke & nuclear fuel (23); Chemicals (24); Rubber & plastics (25); Nonmetallic mineral products (26); Basic metals (27); Fabricated metal products (28); Motor vehicles (34).

6. Non-ICT Services (NICTS): Sale, maintenance and repair of motor vehicles and motorcycles; retail sale of automotive fuel (50); Hotels & catering (55); Inland transport (60); Water transport (61); Air transport (62); Supporting and auxiliary transport activities; activities of travel agencies (63); Real estate activities (70); Other business activities, not elsewhere classified (749); Public administration and defense; compulsory social security (75); Education (80); Health and social work (85); Other community, social, and personal services (90-93); Private households with employed persons (95); Extraterritorial organizations and bodies (99).

7. Non-ICT Other (NICTO): Agriculture (01); Forestry (02); Fishing (05); Mining and quarrying (10-14); Electricity, gas, and water supply (40-41); Construction (45).

APPENDIX IV Labor Hoarding in the Recent Slowdown

54. The latest cyclical downturn did not trigger the area’s traditional pattern of labor shakeouts and upward-ratcheting unemployment rates. Employment has also been surprisingly resilient. This appendix will attempt to shed light on these recent events using aggregate data because the detailed growth accounting database stops in 2000 and does not cover the euro area as a whole.

55. The apparent change in employment behavior might be related to the increased job intensity of growth in the area in the 1990s—the mirror image of the slowdown in labor productivity growth—because of the more favorable wage-setting patterns documented elsewhere in this chapter. In this case, output can grow more slowly and still generate hirings. This type of interpretation is consistent with two main characteristics of the recent slowdown: annual GDP changes were never negative and employment increased marginally every year between 2001 and 2003 (Table 13).

Table 13.

Euro area: Output and Employment Changes during Previous Recessions

(Percent)

article image
Sources: AMECO database; and staff calculations.

Dating was done using quarterly data. Annual definition is an approximation of original dating.

IMF staff dating.

56. However, labor-hoarding intensity might also have picked up when compared to previous slowdowns. Actually, more labor hoarding could also be associated with the wage moderation observed in the 1990s and the resulting increased job-intensity of growth. As opposed to previous slowdowns, this moderation seemed to have raised the profitability of companies, which might have weakened firms’ justification for firing excess labor.

57. This appendix provides a benchmark for the extent of labor hoarding during the current slowdown. Labor hoarding can be defined as deviations from firms’ optimal labor utilization levels in the absence of firing and hiring costs; it is, by definition, a temporary phenomenon that disappears after firms learn the true nature of a certain economic shock (i.e. how permanent such a shock is).

58. Labor can be hoarded in three ways: firms may hire fewer hours from each employee (because costs of varying average hours of work when facing unexpected shocks are smaller than costs of firing and hiring people), they may allocate workers to nonproduction tasks (e.g. equipment maintenance and cleaning), or they may simply reduce the intensity of production even if average hours paid are unchanged. Of course, during a slowdown firms choose a combination of the three types of labor hoarding. In all cases, we should expect a reduction in TFP growth, measured as the Solow residual, because it captures how efficiently labor and capital are combined to produce a certain level of output.

59. Depending on how labor hoarding takes shape, the behavior of employment and hours of work during a full business cycle (slowdown and recovery) will differ:

  • Reductions in hours worked by each employee would keep employment roughly constant but total hours of work closer to its optimal value. Output would decline in tandem with total hours of work, although some short-run inefficiency in the matching of labor and capital could imply a small reduction in the Solow residual calculated using hours of work as a measure of labor input.

  • A reallocation of workers to nonproduction activities or reductions in production intensity would keep employment and hours of work unchanged but would cause a large decline in the Solow residual, as more hours of work (than in the case described in the first bullet) would be hired for the same decline in output.

60. Using this logic, a natural way to define the extent of labor hoarding is the following:

  • First, measure how efficiently employment and capital are combined during a recession.

  • Second, isolate the importance of reductions in average hours of work to explain the inefficient combination of employment and capital in the first measure.

  • Third, assess the importance of structural changes in the underlying behavior of production efficiency to isolate cyclical effects.

61. The table below shows changes in production efficiency, measured as the Solow residual, during CEPR-dated recessions. Efficiency changes in each recession (bolded dates in the table) are compared to the period extending from the initial cyclical recovery after the previous recession to the end of the current recession. This comparison gives some room for trend changes in the Solow residual while keeping the business cycle fluctuations nearly balanced during each comparison period. The first period begins in 1970 because average hours are not available for the 1960s.

Table 14.

Euro area: Solow Residual

(Percent at an annual rate within each period)

article image
Sources: AMECO database; OECD; European LFS; and staff calculations.

Dating was done using quarterly data. Annual definition is an approximation of original dating.

IMF staff dating.

OECD average hours of work refer to business sector. It takes national LFS as a basis but also uses information on payroll data and others.

Average hours of work from the European LFS refers to the first quarter of each year.

62. Labor hoarding was only slightly more intense during the current slowdown than in the previous two recessions: -0.82 percentage point versus -0.72 percentage point and -0.71 percentage point, as seen in the first column of Table 1. Moving across columns, reductions in average hours of work do account for part of the labor hoarding in the latest recession as the decline in efficiency growth is smaller once hours of work from the OECD are taken into account (-0.69 percentage point compared to -0.82 percentage point). That is in fact a change from previous recessions when calculations using hours of work data yielded the same or even larger swings in the Solow residual.

63. Using average hours from the European labor force survey does not change this picture significantly, although the ELFS data seems to be a bit more cyclical than the OECD average hours series. (In other words, changes in average hours of work account for a slightly larger share of the observed labor hoarding when using the ELFS: -0.82+0.58 = -0.24 percentage point, versus -0.82+0.69 = -0.13 percentage point. The same is true for the previous recession.)

64. There is an important caveat to the calculations presented in this section: as shown in the main body of this chapter, aggregate data are quite imperfect for the calculation of production efficiency. Solow residual calculations based on the aggregate data overestimate the decline in TFP growth after the mid-1990s and may taint the benchmark for comparing recent cyclical swings. However, the existence of any bias depends on how the mismeasurement of TFP growth affects the 2001-03 period, which cannot be assessed due to data availability.

65. In summary, this analysis suggests that the changed response of labor markets to the most recent downturn reflects two factors:

  • Lower underlying labor productivity growth, in part reflecting the reabsorption of labor owing to sustained wage moderation and some labor markets reforms, has shielded employment from the effects of lower output growth.

  • There has been some increase in labor hoarding (number of employees) compared with earlier cycles, likely reflecting somewhat the higher cyclicality of hours and the interplay between overall improvements in the profitability of companies and employment protection laws.

66. Looking forward, a large cyclical pickup in labor productivity, as the hoarded labor is directed toward actual production, will introduce disinflationary pressures in the economy through a deceleration in unit labor costs. Household income growth should increase as individuals work longer hours but will be somewhat dampened by slow hirings. The amount of excess labor will be increased by rises in labor force participation, and unemployment rates are expected to decline very slowly in the near term.

References

  • Blanchard, O., 1997, “The Medium Run,Brookings Papers on Economic Activity, 2, pp. 89158.

  • Blanchard, O., 2004, “The Economic Future of Europe,NBER Working Paper, n. 10310, November. Forthcoming in the Journal of Economic Perspectives.

    • Search Google Scholar
    • Export Citation
  • Blanchard, O., and Thomas Philippon, 2003, “The Decline of Rents and the Rise and Fall of European Unemployment,unpublished, MIT.

  • Blanchflower, David G., and Andrew J. Oswald, 1994, The Wage Curve, MIT Press.

  • Card, David, 1995, “The Wage Curve: A Review,Journal of Economic Literature, Vol. XXXIII, June, pp.785799.

  • Crépon, Bruno and Rozenn Dezplatz, 2001, “Une Nouvelle Évaluation des Effets des Allégements de Charges Sociales sur les Bas Salaires,Economie et Statistique, No. 348, pp. 324, August.

    • Search Google Scholar
    • Export Citation
  • Debrun, Xavier, 2003, “Unemployment and Labor Market Institutions: Why Reforms Pay Off,” World Economic Outlook, Chapter IV, Washington, IMF, April.

    • Search Google Scholar
    • Export Citation
  • Decressin, Jorg, Marcello Estevão, Philip Gerson and Christoph Klingen, 2001, “Job-Rich Growth in Europe,France, Germany, Italy and Spain: Selected Issues Paper, Chapter 2, IMF Country Report No. 01/203, pp. 3680, October.

    • Search Google Scholar
    • Export Citation
  • Estevão, Marcello, 2003a, “Do Active Labor Market Policies Increase Employment?IMF Working Paper, WP/03/234, Washington, IMF, September.

    • Search Google Scholar
    • Export Citation
  • Estevão, Marcello, 2003b, “Structural and Cyclical Labor Market Changes in Poland,” in Poland: Selected Issues, IMF Country Report No. 03/188, Chapter II, pp. 1444.

    • Search Google Scholar
    • Export Citation
  • Estevão, Marcello, and Nigar Nargis, 2002, “Wage Moderation in France,IMF Working Paper, WP/02/151, Washington, IMF, September.

  • European Commission, 2003, “Drivers of Productivity Growth: An Economy-Wide and Industry Level Perspective,The EU Economy 2003 Review, Jorgenson, Dale, 2003, “Information Technology and the G-7 Economies,unpublished, Harvard University.

    • Search Google Scholar
    • Export Citation
  • Foster, Lucia, John Haltiwanger, and C. J. Krizan, 2002, “The Link Between Aggregate and Micro Productivity Growth: Evidence from Retail Trade,NBER Working Paper, 9120, August.

    • Search Google Scholar
    • Export Citation
  • Jorgenson, Dale, 2003, “Information Technology and the G7 Economies,” unpublished, Harvard University.

  • Jorgenson, Dale, and Kevin Stiroh, 1999, “Information Technology and Growth.American Economic Review, 109115, May.

  • Jorgenson, Dale, and Kevin Stiroh, 2000, “Raising the Speed Limit: U.S. Economic Growth in the Information Age,Brookings Papers on Economic Activity, 1.

    • Search Google Scholar
    • Export Citation
  • Oliner, Steve and Dan Sichel, 2000, “The Resurgence of Growth in the late 1990s: Is Information Technology the Story?”, Journal of Economic Perspectives, vol. 14, no. 4, pp. 322.

    • Search Google Scholar
    • Export Citation
  • O’Mahony, Mary and Bart van Ark (editors), 2003, EU Productivity and Competitiveness: An Industry Perspective, European Commission, Brussels.

    • Search Google Scholar
    • Export Citation
  • Schreyer, P., 2000, “The Contribution of Information and Communication Technology to Output Growth: A Study of the G7 Countries,STI Working Papers, 2000/2, OECD, Paris.

    • Search Google Scholar
    • Export Citation
  • Schreyer, P., 2002, “Computer Price Indices and International Growth and Productivity Comparisons,Review of Income and Wealth, 48(1), pp. 1531, March.

    • Search Google Scholar
    • Export Citation
1

Prepared by Marcello Estevão.

4

Basic identity: Growth in GDP per capita = Growth in GDP per hours of work + Growth in employment as a ratio of total population + Growth in average hours of work per person. Data used in this section come primarily from the AMECO database, produced by the European Commission. Data on economywide average hours of work come from the new OECD productivity database.

5

Basic identity: TF^P=(Y^-L^)-(1-α)(K^-L^), where ^ denotes percent changes, Y is real value added, L is total hours of work (employment*average hours of work), K is the capital stock and α is the share of labor compensation in total domestic income.

6

Appendix IV provides a listing of industries by ICT classification according to work presented in O’Mahony and Van Ark (2003).

7

Decressin and others (2001) analyze macro data for the largest four euro-area countries and claim that wage moderation by unions was likely behind job-rich growth. Estevão and Nargis (2002) make the same claim for France after a detailed analysis.

9

Estevão and Nargis (2002) use household-level data for France to show that the trade-off between unemployment and real wages did improve in the 1990s. However, they caution that other factors beyond wage moderation could be behind the clear structural improvement in French labor markets.

11

Several papers since Blanchflower and Oswald (1994) show that there may be some variation around the -0.1 estimate. Card (1995), in particular, raises doubts about their basic specification and notices that elasticities for the United States could be smaller than their estimate. More recently, Estevão and Nigar (2002) use micro data from the French labor force survey and estimate a wage-setting elasticity of -0.1. This general result does not seem to be unique to more developed industrial economies: Estevão (2003b) estimates, also using micro data and different methods, an elasticity of about the same size (but a bit smaller) for Poland. Finally, Estevão (2003a) has estimated the same -0.1 elasticity using aggregate information for a panel of 15 OECD countries, suggesting that the results are not dependent on the use of household-level data.

12

All the data described here are explained in detail in “Data Sources and Methodology” by R. Inklaar and others, published as Chapter 7 in O’Mahony and Van Ark (2003).

13

The results for labor productivity growth using information from this database will differ from the ones using the Industry Productivity Database for many reasons. First, the tables using the Growth Accounting Database will stop with averages up to 2000. The addition of 2001 in the tables based on the Industry Productivity Database lowers productivity growth slightly in the last sample period. Second, the aggregation by ICT grouping will differ because there is not a perfect match between the classification put together for the 56 industries in the Industry Productivity Database and the 26 industries included in the Growth Accounting Database. Third, small differences can be attributed to approximations made in the aggregation process.

14

Original list can be found in O’Mahony and van Ark (2003).