Selected Euro-Area Countries: Rules-Based Fiscal Policy and Job-Rich Growth in France, Germany, Italy and Spain

Fiscal deficits and the public debt has grown throughout much of the postwar period in most industrialized countries under the pressure of rising public expenditure, a trend that has begun to reverse after 1992. A number of studies argue that fiscal consolidation in association with expenditure restraint, particularly reductions in primary current expenditure, has proved more durable historically. All in all, the fiscal consolidation essential to qualify for European Monetary Union is a major achievement but also a difficult process in the four countries (France, Germany, Italy, and Spain).

Abstract

Fiscal deficits and the public debt has grown throughout much of the postwar period in most industrialized countries under the pressure of rising public expenditure, a trend that has begun to reverse after 1992. A number of studies argue that fiscal consolidation in association with expenditure restraint, particularly reductions in primary current expenditure, has proved more durable historically. All in all, the fiscal consolidation essential to qualify for European Monetary Union is a major achievement but also a difficult process in the four countries (France, Germany, Italy, and Spain).

III. JOB-RICH GROWTH IN EUROPE43

A. Introduction

75. After years of relatively lackluster performance, employment in the larger euro-area countries accelerated in the second half of the 1990s. This paper examines the facts of employment generation in France, Germany, Italy and Spain, in an attempt to answer two questions: first, is the behavior of employment during the last economic upswing markedly different from that in previous expansions? Second, if so, what factors could explain this improved performance? The results of the analysis suggest that output growth has been significantly more employment intensive of late than during previous growth episodes, although in Germany almost all of the employment increase in the 1990s was in part-time jobs. This increased job-intensity of growth may largely be due to substantial reductions in labor costs. The bulk of the decline in labor costs reflects increased wage moderation on the part of employees. A stabilization and, in a few cases, reduction in nonwage components of labor costs, notably employer social security contributions, has also played a role, but only recently. A decline in replacement rates for unemployment benefits in some of the countries reviewed has also helped. A loosening of labor market regulation may have catalyzed the response of employment to wage moderation in some countries, while in others employment responded despite tighter regulations. The use of active labor market policies probably played only a modest role.

76. The analysis also suggests that while labor cost moderation has occurred on a national level—at least in France, Italy and Spain—there is little evidence that this moderation has been more pronounced in regions that have suffered from relatively high rates of unemployment. This finding may have negative implications for future employment growth and for equilibrium unemployment, as it raises concerns that job growth may run up against supply constraints well before full employment is achieved at the national level.

77. The paper is organized as follows. Section B reviews historical trends of employment, labor supply and unemployment, and presents some evidence that the employment intensity of output growth has increased recently. To this end, it employs a simple econometric test for a structural break in employment growth, which confirms that the elasticity of employment growth with respect to output growth was indeed significantly different in the 1990s from its value in previous decades in France, Italy and Spain but, when measured in full-time units of labor, not in Germany. Section C analyzes the sources of this increased employment intensity, distinguishing between changes in labor demand and labor supply, and focusing particularly on labor market deregulation and active labor market policies in promoting employment growth. Section D looks beyond wage moderation and the role of wage dispersion, Section E concludes.

B. Employment Growth in the 1990s From an Historical Perspective

78. Data on total employment and output growth confirm that in the larger euro-area countries, the expansion of the late 1990s has been significantly more employment intensive than were previous growth episodes. Specifically, the ratio of employment growth to output growth is substantially above the historical norm in all four major euro-area countries (Table III.1). Growth has been especially employment intensive in Spain, where the rate of employment growth has been close to that of output growth, but in all four countries the ratio of employment growth to output growth has been much larger recently than in previous expansions, averaging 0.6 in the late 1990s. This is three times the average achieved in all expansions from the early 1970s to the mid-1990s.

Table III.1.

Selected Countries: Employment Intensity During Upswings, 1970-20001/

article image
Sources: OECD Analytical Database; IMF WEO Database (for output gap); INSEE for (fulltime-equivalent employment in France); IAB (for worktime by employment groups in Germany); and IMF staff calculations.

Upswings are defined as years of shrinking output gap. Employment intensity is the ratio of employment to GDP growth.

Excluding 1991, the first year of data for reunited Germany.

Parttime jobs, self-employment, and work of unpaid family members converted into fulltime jobs of employees in proportion to relative hours worked.

79. The increase of employment intensity reflects developments in the business sector, as public sector employment acted as a drag on employment growth in all four countries. Public employment grew more slowly than business sector employment in France and Spain, and contracted in Germany and Italy during the expansion of the late 1990s. The ratio of business sector employment to output growth averaged 0.6 in the late 1990s, compared to barely 0.2, on average, during previous expansions (Table III.1).

80. Although part-time and temporary employment have both increased during the current upswing, employment on a full-time equivalent basis has also risen rapidly, except in Germany. Indeed, once employment is measured in full-time equivalent units, the recent expansion is no longer particularly job-rich in Germany (Table III.l), as part-time jobs accounted for much of the recent growth of employment there. In addition, while fixed-term or temporary contracts are used in all countries, only in Spain do they constitute a substantial share of total employment (Figure III.1). Total hours of work grew less than employment in the 1990s because of the downward trend in average hours of work. However, when compared to the recovery at the end of the 1980s, both hours of work and employment have increased at a stronger pace in the last four years, even though average GDP growth for the four countries was stronger in the earlier period. In France, the strength in hours of work, even after the implementation since 1996 of different laws aiming at the reduction of the standard workweek, is remarkable.44

Figure III.1.
Figure III.1.

Selected Countries: Atypical Work Contracts, 1983-2000

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Source: EUROSTAT NewCronos Database.

81. The services sector has been the main engine of employment generation in the 1990s, consistent with trends in output growth and with the relatively labor-intensive nature of many service sector jobs (Table III.2). Thus, part of Germany’s relatively lackluster employment growth performance in the late 1990s, at least when measured on a full-time equivalent basis, can be explained by slow growth in its services sector, where gains were confined to the financial sector and enterprise services. In addition, Germany made less headway than the other countries even when excluding eastern Germany, the construction sector, or the government sector. It does, however, still record the highest labor force participation rate and lowest unemployment rate among the countries reviewed (Figure III.2).

Table III.2.

Selected Countries: Fulltime-Equivalent Employment by Sector, 1991-2000

article image
Sources: IMF staff calculations based on data from INSEE, IAB. Banca d’ltalia, and Banco de España.

Cumulative during 1991-2000.

Figure III.2.
Figure III.2.

Selected Countries: Unemployment and Labor Force Participation, 1970-2000

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Source: OECD Analytical Database.

82. Rapid growth in employment has been accompanied by both significant declines in unemployment rates and increases in labor force participation. The unemployment rate has fallen most dramatically in Spain, from 23.7 percent in 1994 to 14.1 percent in 2000 (Figure III.2). However, declines in France and Germany were also significant: the rate of unemployment in France dropped from 12.2 percent in 1997 to 9.7 percent in 2000, while that in Germany fell from 9.5 percent to 7.8 percent over the same period. In Italy, progress has been more recent and, to date, less pronounced, with the rate of unemployment falling to 10.7 percent from a recent high of 11.9 percent in 1998. Labor force participation has risen considerably in Spain and Italy and more gradually in France (Figure III.2). In Germany, the participation rate had already increased steadily through the 1980s and jumped at the turn of the decade, largely, but not only, because of the incorporation in the data of eastern Germany where the participation rate of women is relatively high.

83. Recent declines in unemployment rates have done little to redress longstanding imbalances across demographic groups. Data on unemployment rates for women and the young show little evidence of convergence in recent years (Figure III.3). Only in Germany did the ratio of female to male unemployment rates decline noticeably since 1995, and in Spain the ratio has worsened considerably. Progress in addressing youth unemployment has been similarly slow since 1995, although all four countries have made progress relative to the early 1980s. Only in Germany does unemployment appear not to have an important gender or age component: in the remaining countries, female and youth unemployment rates substantially exceed those for male and older workers.

Figure III.3.
Figure III.3.

Selected Countries: Unemployment Dispersion, 1993-2000

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: IMF staff calculations; and EUROSTAT NewCronos database.1/ Unemployment rate for the 15-24 year old unemployed divided by the unemployment rate for the 25-64 year unemployed.2/ Coefficient of variation across NUTS level 1 regional unemployment rates.

84. The benefits of declines in unemployment rates have also been distributed unevenly across regions in most countries. Since the mid-1990s, the dispersion of regional unemployment rates (measured by the coefficient of variation of regional unemployment rates) has widened everywhere but in western Germany (Figure III.3).45 These regional disparities are particularly acute in Italy, and have worsened notably both there and in Spain. They have also plagued Germany since reunification.

85. The impressionistic findings on the relationship between employment and output growth discussed above can be analyzed and expressed more formally in a simple econometric exercise. For each country (for Germany, only the western part), the following regression was run over 1973-2000 (for France, only 1981-2000) to investigate the connection between employment and output:

dlt=β0+β1dyt+β2dyt1+β3D90dyt+β4D90dyt1+β5dnt1+β6dnt2+εt

where d denotes the difference operator, and l and y the natural logarithms of aggregate employment and real GDP, respectively. The variable D90 is a dummy with D90=l for 1990 and subsequent years and D90=0 otherwise.46 Accordingly, an estimate for β3 significantly different from zero suggests that the contemporaneous relation between employment and output differed significantly in the 1990s relative to earlier decades. The year 1990 was chosen as a break-point to obtain a reasonably long span over which to check for a change in the employment-output relation: alternatives would have been to identify a full cycle for each country or to concentrate on periods of cyclical expansions. But this would have introduced more judgment in the analysis and added little value to the exercise. Employment is expressed in full-time equivalent units, so as to adjust for changes in the share of part-time work. Furthermore, all variables are expressed in first differences: thus the equation captures the short-run relation between output and employment.

86. The econometric evidence suggests that growth has become more employment intensive during the 1990s in all countries but Germany.47 The table below shows the regression output: the standard test statistics suggest a good fit and reasonable error properties. As can be seen from the estimates for β1—which are all significant at a 5 percent level—a one percentage point increase in real GDP is associated with a contemporaneous (short run) rise in employment of between 0.39 percentage point (Italy) and 0.63 percentage point (Spain). However, the estimates for β3 suggest that during the 1990s the increase in employment arising from a one percentage point increase in real GDP is significantly larger in France, Italy, and Spain: the difference relative to the 1970s and 1980s amounts to between 0.15 percentage point (France) and 0.49 percentage point (Spain).48 By contrast, the estimates for β3 for western Germany are both small and wholly insignificant, suggesting no change in the employment-output relation.

87. Moreover, in Germany growth during the 1990s was less labor intensive than in any of the other countries. The sum of the estimates for β1 and β3 suggests that output growth in western Germany during the 1990s was less employment intensive than in any of the other countries (when measured in full-time equivalent units), by a margin of 30 percent (relative to France, which had the least employment intensive growth of the other three countries). To check for statistical significance, the regressions for France, Italy, and Spain were run by restricting the estimates for β1 and β3 to the values obtained for Germany. Wald tests reject the restrictions at the 5 percent significance level for all countries.

Regression of Employment on Real GDP Growth 1/

article image
Sources: OECD; national authorities; and IMF staff calculations.

Employment measured in full-time equivalent units.

Note: Standard errors in parentheses. A * indicates significance at a 5 percent level; a ** indicates significance at a 10 percent level.

88. The flip-side of the increased elasticity of labor input to changes in output in these countries was a decline in labor productivity growth in the 1990s. It is important to remark, though, that increased employment growth does not need to be associated with a deceleration in labor productivity: employment growth was strong in the United States in the 1990s when labor productivity also kept a healthy pace. In addition, labor productivity growth has been declining in Europe for decades although until the recent episode employment growth had been quite sluggish (Table III.3). Labor productivity can be decomposed into two components: capital deepening and total factor productivity (TFP) growth. Both have been responsible for the deceleration in productivity since the 1970s. During the latest recovery period, capital deepening was quite weak in all countries and, actually, significantly lower than in the 1980s; a sign of the employment-based growth of late. TFP growth during the latest recovery was also significantly weaker than in the 1980s, although only moderately so in the case of France. Workplace reorganizations following the introduction of the standard workweek reduction laws in France since 1996 may have supported TFP growth.49

Table III.3.

Selected Countries: Decomposition of Labor Productivity Growth, 1970-20001/

article image
Sources: IMF staff calculations; OECD Analytical Database, and unpublished data on hours worked.

Refers to the business sector.

Growth of output per hour worked.

Contribution of total factor productivity (TFP) growth to labor productivity growth.

Contribution of capital per hour worked to labor productivity growth.

Calculations exclude growth in 1991, the first year of data for reunited Germany.

89. In summary, the data confirm that output growth has been more employment intensive during the current expansion than in previous ones. The increased job-intensity of growth has been accompanied by a decline in labor productivity growth. Such a decline originated from reductions in capital deepening and TFP growth in all countries with the exception of France where TFP growth has actually gone up in the last few years. The increase in employment intensity has led to a significant decline in overall unemployment rates, but there is little evidence that unemployment rates are converging among demographic groups or across regions, which could suggest that wage dispersion remains insufficient. Maintaining rapid rates of employment growth, however, will increasingly require the entry into employment of individuals with relatively low initial productivity, a process that thus far appears to be inconsistent with present levels of wage dispersion.

C. Explaining the Pick-Up of Employment Growth

90. Understanding the factors responsible for more rapid employment growth in recent years requires disentangling changes in labor supply conditions from changes in labor demand. This section provides a framework for doing this. It argues that changes on the supply side of the labor market were primarily responsible for favorable employment developments of late. That is, a behavioral change has taken place on the part of workers and their representatives who appear to have been attaching more importance to the employment consequences of negotiated wage settlements than in the past. This phenomenon is referred to as wage moderation. Labor demand conditions also appear to have supported stronger employment growth but in a more passive way through a slowdown in the pace at which firms have been adopting labor-saving technology.

91. Employment growth driven by wage moderation fits well with the phenomenon of job-rich growth. As workers moderate their wage demands, firms have less reason for adopting labor-saving technology. As a consequence, output per worker grows less rapidly and the inverse, job-intensity, accelerates. Slower productivity growth would also be consistent with a change in the skill-mix of employment to the extent that new jobs would be filled by previously unemployed workers, which are disproportionately unskilled.

The analytical framework

92. In a partial equilibrium setting, employment and wages are determined by the intersection of a labor demand curve and a wage curve. The latter captures the wage setting behavior of workers or their representatives and substitutes. The intersection typically does not involve market clearing defined as zero unemployment or unemployment reduced to its frictional level. It does constitute an equilibrium, however, in the sense that market forces will drive employment and wages to the values determined by the intersection. In this framework, observed changes in employment and wages reflect movements of the intersecting point as labor demand and the wage curve shift over time. In turn, the position of the labor demand and wage curve is determined by technological, behavioral, and institutional parameters. The relationship between wage and employment movements would not be expected to be systematic. For example, a shift in the labor demand curve would move employment and wages in the same direction, but a shift in the wage curve would move them in opposite directions (see diagram below).50 In practice, both curves might shift at the same time with the outcome for wages and employment depending on the direction and size of the shifts and the slope of the two curves.

93. The labor demand curve is downward sloping under the standard assumptions of profit maximizing firms and a production technology where the additional output produced per unit of labor shrinks with expanding employment. A simple CES production function (equation (1)) provides a sufficiently rich analytical structure to illustrate the main factors that might be expected to shift labor demand:

Yt=[α(atLt)σ1σ+(1α)Ktσ1σ]σ1σ,(1)

with Y standing for output, L for labor, K for capital, σ for the elasticity of substitution between effective labor (aL) and capital, and a technology parameter a between 0 and 1. Smaller values of α are associated with more capital-intensive modes of production. Equating the marginal product of labor to real effective compensation, i.e., real effective product wage plus employers’ social security contributions at the rate te, to maximize profits, yields after rearranging and some approximation the following relationship:

logαte=log(ωt)+1σlog(atLtYt).(2)

In order to satisfy equation (2), employment must fall as real effective wages rise, which establishes that the labor demand curve is downward sloping. Equation (2) also shows that the technology parameter α and employers’ social security contributions shift the labor demand curve: for higher α (i.e., less capital-intensive production), or lower te, a given real effective wage is associated with more employment, meaning that the labor demand curve shifts out.

94. The supply side of the labor market is modeled by a wage curve that tries to capture how wages are determined in negotiations between employers’ and workers’ associations. It assumes that workers’ associations negotiate wages knowing that hiring decisions rest with employers and are based on profit considerations (Box III.1). Unemployment has a dampening effect on wage demands because it raises the probability of long spells of joblessness and hence the associated risks and costs to workers. This establishes the upward sloping shape of the wage curve.

95. The wage curve is subject to a variety of shift parameters that capture workers’ preferences and institutional arrangements, in particular unemployment benefits and labor taxes. As in the case of labor demand, the position of the wage curve changes as the underlying behavioral and institutional factors change. Key shift parameters are workers’ preferences for employment, as opposed to merely earnings, captured by the parameter γ, the level of unemployment benefits, B, and labor taxation, C. For example, if the value attached to employment increases, wage demands moderate and the wage curve shifts out. Likewise, the level of unemployment benefits matters as it is critical for workers’ well-being during spells of unemployment: lower benefits make unemployment more unpleasant, thus dampening wage demands and shifting the wage curve outward. Finally, the taxation of labor relative to the taxation of income when unemployed influences the position of the wage curve. Lowering the taxation of wages, but not of income when unemployed, increases the net income loss during unemployment spells, thus reducing wage demands and shifting the wage curve out.

Deriving the Wage Curve

The wage curve seeks to model the relation between wages and unemployment as the outcome of wage bargaining between employers and employees. Unlike in a setting with a conventional labor supply curve, workers cannot costlessly switch between jobs nor are they costlessly and instantaneously replaceable by other workers. In a depressed labor market, where it is hard to find alternative employment, workers will therefore settle for lower wages than in a booming labor market where jobs are easy to find. When bargaining is about wages, and hiring decisions rest with employers, the outcome is the solution to the maximization problem

MaxWiΩi=[Liγ(WiCA)]θΠis.tLi=L(Wi),fromfirm’sprofitmaximization.(3)

Li, Wi, and Πi represent, respectively, employment, wage, and profits of firm i. θ measures workers’ relative bargaining power and γ indicates how much workers care about aggregate employment. C is the consumer price index, Pc, adjusted for the tax wedge between earned wages and workers’ purchasing power. Defining fc as the consumption tax rate, td as the income tax rate, and tss as the social security tax rate, C can be written as

C=PC[1+tc(1td)(1tss)](4)

A represents workers’ reservation wage and equals

A=(1u)WC+uBCυ.(5)

u, W, B, and CU denote the unemployment rate, wages, the income received if unemployed, and the consumer price index adjusted for the tax wedge applicable to the income of the unemployed. The unemployment rate is a proxy for the probability of finding work elsewhere in case wage negotiations fail.

Solving the maximization problem (5) yields

WiC=mA,m>1.(6)

Negotiated wages, corrected for the tax wedge, turn out to be marked-up alternative income of workers. The markup factor is increasing in workers’ bargaining power. It is decreasing in the elasticity of labor and output demand, the labor intensity of production, and the value workers attach to employment as such.

Finally, assuming symmetry across workers and replacing A yields the wage curve

W=mu1m(1u)BCCU.(7)

It has four important features: (i) the wage level is inversely related to the unemployment rate; (ii) negotiated wages are higher when the level of income received during unemployment spells is higher; (iii) the tax wedge matters for wages only to the extent that it affects wage income and the income of the unemployed differently; and (iv) changes of workers’ preferences, technology and product market conditions alter the relationship between wages and the unemployment rate and are thus shift parameters of the curve.

96. Note that in the model, labor productivity increases are assumed to be passed through one-for-one into wage demands. Deviations from this behavioral benchmark consequently constitute changes of employees’ preferences. In other words, workers not seeking full compensation for productivity gains would be viewed as placing a greater emphasis on securing jobs for the unemployed. In the simple supply and demand diagram above, the wage curve is strictly drawn as a function of real effective product wages meaning that wages growing in line with labor productivity do not shift the curve and thus do not affect equilibrium unemployment. At least in the long run this is a reasonable assumption well covered by empirical evidence.51

Real effective wage trends

97. Real effective wages in the four countries increased substantially in the 1970s but since the early 1980s have been on a declining trend that was temporarily interrupted in the early 1990s, most severely in Germany (Figure III.4).52 The decline of real effective wages, which was most dramatic in Spain and in Italy, left them below their 1970-levels in all countries except for Germany. Data on real contractual wage growth, unadjusted for productivity growth, also show a substantial slowdown in the 1990s and confirm the relative country experience (Figure III.5).53 In Germany the declining trend of real effective wages set in a little later than in the other countries and was somewhat less pronounced. It was also most severely interrupted around 1990 as a consequence of reunification. For one thing, eastern Germany joining the statistical base at wages out of line with productivity in 1991 pushed up real effective wages. And in the ensuing years wage increases in eastern Germany were driven primarily by the desire to converge to western earnings levels rather than by productivity considerations. However, the reunification boom also triggered rapid wage growth in western Germany. Indeed, since 1992 real unit labor costs have increased faster in western Germany than in eastern Germany.

Figure III.4.
Figure III.4.

Selected Countries: Real Effective Wage in the Business Sector, 1970-2000

(1970=100)

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: IMF staff calculations; OECD Analytical Database and unpublished data on hours worked per employee in the business sector (France and Germany) and in the whole economy (Italy and Spain).
Figure III.5.
Figure III.5.

Selected Countries: Real Contractual Wages, 1985-2000

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: IMF staff calculations; national sources (nominal contractual wages); and OECD Analytical Database (GDP deflator).

98. The simultaneous decline of real effective wages and the unemployment rate in the second half of the 1990s suggests that developments were dominated by an outward shift of the wage curve. With the labor force expanding in this period, only an outward shift of the wage curve can produce declining wages coupled with falling unemployment as long as the labor supply curve is downward sloping and the wage curve is upward sloping. This is not to say that the labor demand curve must have remained unchanged; it might have also shifted out, but the effect of the wage curve shift must have dominated for otherwise real effective wages would not have fallen.

Shifts in labor demand

99. Pronounced swings of the share of wage income in national income over the last three decades suggest that the labor demand curve has moved around significantly, although it has perhaps been more stable in the 1990s. The right-hand side of equation (2) is approximately equal to the share of wage income, at least as long as is close to 1. Therefore, changes of the wage share in national income over time indicate that the technology parameter, α, and/or employers’ social security contributions must have changed over time. The wage share increased in the 1970s, declined in the 1980s, and stabilized in the 1990s, especially in the second half of the decade (Figure III.6). Over the 30-year period, the wage share has declined substantially in all countries and by as much as 30 percent in Spain. The stability of the wage share in the second half of the 1990s suggests that the position of the labor demand curve has been more stable recently so that observed changes of real effective wages have been likely driven by shifts of the wage curve.

Figure III.6.
Figure III.6.

Selected Countries: Labor and Capital in the Business Sector, 1970-2000

(1970=100)

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: IMF staff calculations; and OECD Analytical Database.

100. Both technology changes and rising social security contributions by employers contributed to the declining share of wage income. Technology changes are traced by the labor share of income, i.e., the share of wages plus employers’ social security contributions in national income.54 Like the wage share, the labor share also fell quite strongly over the last three decades, but not as much (Figure III.6). It declined most in Spain (by about 12 percent) and least in Germany (by about 6 percent). The excess drop of the wage share is due to rising social security contributions of employers. On average, about two thirds of the decline in labor share in can be attributed to technology factors and the remainder to increases in social security contributions. In recent times, however, employers’ social security contributions have been reduced in France (after 1993) as well as in Italy and Spain (after 1996).

101. Observed technology changes likely reflect the adoption of increasingly capital-intensive technologies, not least in response to rising wage costs in the 1970s and early 1980s. In the face of rising labor costs and limited substitutability between capital and labor in the short run the labor share of income initially rose. It declined again as more capital-intensive modes of production were adopted over the medium term, as reflected in the rising capital-output ratio during the 1970s and 1980s (Figure III.6). The stabilization of labor shares and the capital-output ratio in the second half of the 1990s suggests that this process is now petering out.

Shifts in the wage curve

102. This section examines more closely the prima facie evidence that an outward shift of the wage curve is responsible for recent job-rich growth. It first tries to quantify the extent to which wage demands have moderated and how this might have affected employment and the NAIRU. It then looks at the factors that could have curbed wage demand.

103. In order to isolate shifts in the wage curve, changes in wages need to be adjusted for inflation and productivity growth and the effects of unemployment on wage demands. The latter is key: observed changes in wages might be nothing but a response to developments in unemployment without any change of workers’ underlying stance at the negotiation table—i.e., a movement along the wage curve. Making the adjustment requires an estimate of the slope of the wage curve assuming, as argued above, the 1990s were a period of relatively stable labor demand conditions. In turn, the slope of the wage curve—or relatedly, the elasticity of wage costs (δ1) with respect to unemployment—can be estimated by using a simplified model based on equation (7) of Box III.1.

ω=δ0+δ1u(8)whereδ0=δ0(u*,m,C,Cu),δ1=δ1(u*,m,C,Cu)

Staff estimates (see Appendix) put this elasticity at -0.2, somewhat higher than estimates elsewhere in the literature, although not sufficiently different to alter the conclusions qualitatively.

104. The quantification of δ1 enables wages to be adjusted for unemployment rates, to derive a better measure of underlying wage demands (Figure III.7). As can be seen, unemployment-adjusted real effective wages rose sharply during 1975-85, most of all in Germany and least of all in Italy. Since the mid-1980s, however, they declined in all countries but Germany, where the events surrounding reunification pushed up wage demands.

Figure III.7.
Figure III.7.

Selected Countries: Shift of the Wage Curve, 1970-2000

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: IMF staff calculations; and OECD Analytical Database.

105. Information on the slope of the wage curve can also be used to calculate changes in the NAIRU. The vertical shift of the wage curve can be computed from the slope of the wage curve and observed unemployment-wage combinations. In the medium run this shift should be proportional to the change in the NAIRU or natural rate of unemployment. This is because over longer periods of time—that is, periods over which firms can adjust all factors of production including capital—labor demand should be independent of unemployment. In other words, the labor demand curve would be horizontal and equilibrium unemployment changes by as much as the wage curve shifts.55

106. In all countries, the calculated NAIRU reached high levels in the mid-1980s and then declined. Figure III.7 traces the NAIRU over time using the WEO estimate for 1985 as a starting point and the year-to-year shifts of the wage curve. Calculated NAIRUs rose again in the early 1990s, but since then have declined again in France, Italy, and Spain. In Germany, where the labor market had been in a much better position to begin with and where reunification exacerbated labor market conditions, the calculated NAIRU did not decline significantly. In almost all cases, the calculated NAIRUs for 2000 are below estimates in the WEO or those made by the OECD. This might mean that the WEO and OECD estimates do not yet reflect the full extent of wage moderation—although it could also indicate that estimates of the NAIRU for 1985 are too low.

Selected Countries: NAIRUs, 1985 and 2000

(In percent)

article image
Sources: IMF, World Economic Outlook; OECD, Analytical Database; and IMF staff calculations.

Assumes δ1=0.2.

Explaining wage curve shifts

107. The factors that could have curbed wage demands include: (i) a reduction of unemployment benefits; (ii) a lowering of labor taxation, relative to taxes applicable to income if unemployed; and (iii) behavioral changes of workers and their representatives toward increased emphasis on jobs. The last of these appears to have been by far the most important.

The role of unemployment benefits

108. Benefits are an important determinant of reservation wages and changes in benefit levels therefore shift the wage curve. Germany and Italy span the spectrum of generosity of unemployment benefits but once other income support and tax benefits are taken into account, the differences narrow (see tabulation below). In Germany, benefits can reach almost two thirds of wages earned prior to unemployment and generally last indefinitely. In Italy, by contrast, the replacement rate is 40 percent and duration is limited to 6 months in most cases. Both in Italy and Spain many unemployed are not eligible for benefits, mostly because they are first-time job seekers and because eligibility was tightened significantly in Spain.56 Accordingly, only a little over half of the unemployed receive assistance in Italy and Spain. In terms of trend, OECD information for replacement rates shows that gross replacement rates have declined since the beginning of the 1980s in Spain and in Germany, which may have contributed somewhat to curbing wage demands in these countries (Figure III.8). In France, replacement rates increased dramatically in the 1980s and stood at the top of the rank among the four countries during the 1990s. In Italy, average gross replacement rates increased in the 1990s from low levels due to more generous benefits in the first year of unemployment.

Figure III.8.
Figure III.8.

Selected Countries: Gross Unemployment Benefit Replacement Ratio, 1970-1997 1/

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Source: OECD tax and benefits database.1/ Data plotted refer to the simple average of replacement ratios applicable to different demographic groups and different unemployment durations.

109. On the whole changes of unemployment benefits are unlikely to have mattered in the recent pick-up of employment growth. For one, changes in the 1990s were rather modest in most countries. In Italy, the only country with a significant increase of generosity, the change should have dampened employment growth, not accelerated it. Moreover, relative employment performance of countries fails to even remotely match relative benefit changes.

Income Support for the Jobless

article image
Sources: Nickell (1997); and OECD (1999d).

Starting in 1994, replacement rates for unemployment compensation and assistance are, respectively, 67 percent and 57 percent. Duration of assistance for the former is one year and for the latter unlimited. Since 1994, a recipient who has not received unemployment compensation before can receive unemployment assistance only for one year and then needs to apply for social assistance. The duration is longer for older workers.

Ordinary unemployment compensation; replacement rates and duration are higher and longer under a scheme for workers laid off by large firms in industry (up to 80 percent for 4 years). These workers account for about 10 percent of all the unemployed receiving support. The over-50 years old get benefits for 9 months as of 2000.

Replacement rates can reach 70 percent; benefits last 0.5 to 6 years, depending on length of contribution.

Taxes are recorded with a negative sign and thus the share of unemployment assistance can exceed 100 percent.

Developments in tax wedges

110. The effect of taxes on employment is fundamentally an empirical issue. Within the confines of the model outlined in Box III. 1, changes in taxes might be expected to have little impact on employment unless they alter the difference between after-tax incomes for workers and the unemployed. It is debatable whether it is really relative taxation that matters, or whether the taxation of labor income plays a role independent of taxes on income if unemployed. It could be argued that, at least in the short run, cuts in labor taxation—more broadly defined to include employers’ social security contributions—stimulate employment even if taxes on unemployment benefits are cut proportionately. The empirical evidence in the vast literature on the link between taxes and unemployment is not clear cut: Blanchard and Katz (1997), for example, note that the cross-sectional evidence in Europe does not reveal a strong correlation between tax rates and unemployment rates, nor between changes in tax rates and unemployment rates. By contrast, Daveri and Tabellini (2000) argue that if wages are set by strong trade unions, an increase in labor taxes is shifted onto higher real compensation and results in higher unemployment.57

111. After-tax wages have moderated to an even larger extent than have pre-tax wages in the countries reviewed. Figure III.9 shows developments in the tax wedge computed by the OECD. This wedge between wages and take-home pay comprises income tax as well as employees’ social security contributions and is a very imprecise measure of C/Cu in equation (7). Tax wedges increased during the 1980s and the first half of the 1990s in most countries—the main exception was Spain—constraining growth of take-home pay and employment. The tax wedge has subsequently leveled off. Germany experienced a particularly sharp increase in the tax wedge in 1992-95 to pay for some of the costs of reunification.

Figure III.9.
Figure III.9.

Selected Countries: Labor Taxation, 1970-2000

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: OECD Economic Outlook Database and 2EAPWHIS.ivt Database; and IMF staff calculations.1/ Average of rates for single and married employees; wedge includes income tax, employees’ social security contributions and cash benefits (as a negative entry).2/ Total social security collections as percent of total employee compensation minus social security collections.

112. Social security contributions were cut in some countries in the second half of the 1990s owing largely to cuts in the employer-paid portion (Figure III.9). In France, these cuts which began in 1993 and were quite sharp in 1998 were targeted to low-skilled, low-paid labor, and indeed, the long-standing trend of falling employment in this segment of the labor market came to a halt around that time. Social security contributions were also cut sharply in Italy in 1998. In Germany and Spain, social security contributions were increased slightly, although targeted reductions in employer social security contributions in Spain may have contributed to employment growth for specific demographic groups.

113. Despite significant changes, labor taxation is unlikely to hold the key for explaining the shift of the wage curve in the 1990s. Changes of employers’ social security contributions played a major role in developments of the tax wedge and they affect the labor demand curve rather than the wage curve in the framework of this study. Moreover, the framework suggests that the relative taxation of labor income and benefits matters. Relative taxation likely changed little; in Germany, for example, benefits remain untaxed and are set at fixed fraction of the net earnings prior to unemployment, so that relative taxation does not change by definition.

Changes in worker preferences

114. The part of the wage curve shift that is neither due to changes in unemployment benefits nor to changes of labor taxation can be attributed to workers and their representatives placing more emphasis on job preservation and creation. In the 1990s, this phenomenon, called here wage moderation, was the dominant factor behind the outward shift in the wage curve.

115. Figure III.10 illustrates how wage moderation evolved over time. It simply plots unemployment-adjusted real effective wages while controlling for effects from changing unemployment benefits and labor taxation.58 Since the early to mid-1980s, wages have moderated significantly but increasing labor taxation offset much of the effect on labor costs (Figure III.10). Wages moderated by around 10 percent in Spain and France; by somewhat less in Italy; and actually increased in Germany (see bold line in Figure III.10). Labor taxation increased in all countries except for Spain, so that gross wages moderated by much less (see dotted line in Figure III.10).

Figure III.10.
Figure III.10.

Selected Countries: Gauging Underlying Wage Moderation, 1980-2000 1/

(1980=100)

Citation: IMF Staff Country Reports 2001, 203; 10.5089/9781451813005.002.A003

Sources: IMF staff calculations; OECD Analytical Database rod 2EAPWHIS.ivt Database.1/ All wage data refers to hourly wages.

116. These results—together with those obtained earlier on labor demand—would imply that much of the employment intensity of growth in the current expansion can be attributed to wage moderation on the part of labor. Workers’ wage demands grew sharply in the 1970s, and, boosted by large increases in the tax wedge, this led to a slowdown in employment growth. At the same time, labor demand declined because of the adoption of more capital-intensive technologies and rising payroll taxes. With a trend toward greater wage moderation since the mid-1980s, and in the face of a stabilization of the labor demand curve, employment has begun to rise significantly.

117. The above results are subject to the potential caveat that they are derived using average wages rather than disaggregated wage data. Average wages might not adequately reflect wage developments if the sectoral, demographic, or skill composition of employment changes over time. For example, if employment starts shifting toward the service sector, where wages tend to be lower, average wage growth will decelerate even if wage growth in each sector individually remains constant.59 Therefore, part of observed average wage restraint over the last 15 years might simply reflect compositional changes of employment caused by an accelerated expansion of the service sector, by higher growth of low-skilled employment, or by increased employment of low-paid demographic groups. However, the broad trends in the composition of employment across countries have been similar, with the service sector gaining at the expense of manufacturing and agriculture everywhere. This suggests that the conclusions regarding the relative degree of wage moderation across countries would still hold. Moreover, absolute estimates of wage moderation only change if the compositional change in the workforce is not adequately reflected in the productivity measure that is used to obtain real effective wages, an issue which lies beyond the scope of this paper.60

D. Beyond Wage Moderation

118. While wage moderation is playing a key role in creating jobs its effectiveness depends on how individual labor markets function at a more micro level. For example, across-the-board wage moderation is only a second-best strategy if labor market imbalances afflict predominantly certain regions and demographic groups. And in general, it is less effective where tight regulation or active policies inhibit the functioning of the labor market.

Wage dispersion

119. Wage dispersion measures differences in wage rates across skill groups, regions, or age groups. Wide differences in unemployment rates between various groups of workers, the so-called mismatch, are indicative of insufficient wage dispersion. Mismatch captures both transitory and persistent imbalances between the supply of and demand for labor across skill groups, regions, and age groups. Section B showed that mismatch is an issue in the countries reviewed: youth unemployment is between 2½ to 4 times higher than adult unemployment in all countries but Germany; unemployment among women is higher than among men; and unemployment rates differ considerably across regions within countries. Within the framework of overall wage moderation, more variation in wages across groups and regions would help to speed the narrowing of such unemployment differentials. However, overall, the evidence suggests that both wage dispersion across regions and skills has not increased significantly during the 1990s, except in eastern Germany, which started out with highly overpriced labor. Accordingly, there should be considerable room for reducing the NAIRU by allowing more wage dispersion.61

Regional wage dispersion

120. Regional relative unit labor costs are not negatively correlated with regional relative unemployment rates. In particular, unit labor costs levels in areas with higher unemployment are not lower than in areas with low unemployment, nor have they slowed sharply relative to such areas, except in eastern Germany. Despite progress, eastern Germany’s unit wage costs remain far above those prevailing in western Germany. Indeed, in Germany, relative wage costs of the high-unemployment region exceed those in Spain and Italy, where unit labor costs in the high unemployment regions are comparable to those in the low-unemployment regions (see tabulation below). France has no regional unemployment problems on the scale of those in the other countries.

121. Except in eastern Germany during 1991-95, relative unit labor costs respond little to relative change in unemployment rates. The wage curve of equation (8) can be used to gauge how labor costs should have evolved in the less prosperous regions relative to their more prosperous counterparts. The evolution of relative unemployment rates predicts a decline in relative unit labor costs between 1983-2000 of 27 percent in southern Italy relative to the center-north and of 5½ percent in Andalucia relative to the country as a whole. Actual relative unit labor costs hardly moved or even increased.62 The same holds for the 1990s. In contrast, relative unit labor costs in eastern Germany declined during 1991-2000 by some 20 percent, more than suggested by the evolution of relative unemployment rates (9 percent). Unfortunately the decline of relative unit costs was primarily the result of massive labor-shedding.

Regional Labor Market Developments

article image
Sources: EUROSTAT; national authorities; and IMF staff calculations.

Data on unit labor costs are only available through 1995 for Spain and 1998 for Italy. For Spain, unit labor costs were calculated for 1996-2000 with the help of regional data on contractual wages.

Unemployment across skill levels and demographic groups

122. With respect to skills or demographic groups, data limitations make it difficult to assess the extent to which wage moderation has occurred. In all countries but Germany, the unemployment rate of the young is considerably higher than that of the adult population (Figure III.3). The compression of wages across skills during the 1970s and 1980s is well documented in the literature.63

123. The evidence on the development of wage differentiation in the 1990s is mixed, but sea changes do not appear to have taken place. According to Atkinson, Glaude, and Olier (2001) wage dispersion in France remained relatively stable in the 1990s, indicating that wage compression across skill levels came to a halt. Interestingly, wage dispersion within low skill categories has lately increased. In the case of Italy, wage differentials have started to edge up in the first half of the 1990s, especially for male workers (Brandolini and Sestito, 1999). Germany was characterized by a relatively stable wage structure in the 1990s. While Prasad (2001) reports a marginal increase of the skill premium over the last decade, Schimmelpfenning (2000) finds that wage differentiation across skills continued to decline into the 1990s. In Spain, wage dispersion was stable at the low end of the income distribution but increased significantly in the upper part of the distribution during the 1980s (Bover, Bentalila, and Arellano (2001)).

Active labor market policies

124. Studies have confirmed that some active labor market policies—such as job search assistance, training programs and financial assistance for would-be entrepreneurs—can be effective in stimulating employment and reducing unemployment.64 However, studies have also found that, in general, the more broadly-defined the intended beneficiaries of an ALMP, the less effective these policies tend to be. For example, broadly-based employment subsidies may have little impact on employment relative to the amount of spending they involve either because of heavy deadweight losses (when firms accept subsidies to hire workers they would have been willing to contract in any case) or substitution effects (when firms simply replace nonsubsidized workers with subsidized ones without increasing total employment). Because the composition of spending on ALMPs is as important as the level, it is not surprising that there is little empirical evidence that high aggregate levels of ALMPs correspond to lower rates of unemployment or faster employment growth.65

125. Assessing the role of ALMPs in boosting employment during the 1990s would require time series data on spending and an evaluation of the evolution of policies over time. Due to lack of data, this section merely takes stock of these policies. Spending on ALMPs as a percentage of GDP varies from about 1¼ percent of GDP in France to ¾ percent of GDP in Spain (see tabulation below). The variation in spending is significantly greater once scaled for unemployment levels and worker productivity: spending on ALMPs per unemployed person in Germany was equal to about 12 percent of output per worker, roughly three times the level prevailing in Spain.

Selected Countries: Spending on Active Labor Market Policies

article image
Sources: OECD; national authorities; and IMF staff calculations.Note: Data for France and Italy are for 1996. Data for Germany are for 2000, and data for Spain are for 1999.

126. The composition of spending on ALMPs differs significantly across countries. In particular, spending in Italy in the late 1990s was heavily biased towards private sector employment subsidies, compared to France and especially Germany and Spain. By contrast, all three of these latter countries dedicated a significantly greater share of spending to direct job creation in the public or nonprofit sectors. None of the four countries devoted a particularly large share of spending to public employment services (which in the United States and the United Kingdom account for more than one-third of spending on ALMPs). This type of spending is reckoned to be among the most efficient forms of ALMPs. In addition, Italy devoted almost no spending to labor market training, although the bulk of the subsidies serves to sustain apprenticeship schemes and on-the-job training.

Selected Countries: Distribution of Spending on Active Labor Market Policies

(As percent of total)

article image
Source: OECD.

“Training” refers to labor market training; “subsidies” to subsidies for regular employment in the private sector; “job creation” to direct job creation in the public or nonprofit sectors; and “other” to youth measures, support for unemployed persons starting enterprises, and measures for the disabled. Data are for 1997 (France 1996).

127. Evidence suggests that considerable scope exists to improve the efficiency of ALMPs. In France, some active labor market programs have created a substantial number of jobs, although they do not seem effective in securing participants durable jobs in the private sector. Among them, the emploi jeunes program, which aims at ameliorating the problem of youth unemployment in France, created about 240,000 jobs between 1997 and 2000. In Germany, a study conducted for the budget committee of the parliament found that public employment programs have had a limited impact in reducing unemployment as such programs have not served as a useful means of moving workers into the primary labor market. In Italy, the targeting of policies could be improved significantly: while the south harbors two thirds of the unemployed, with the youth unemployment rate at over 50 percent, only about ½ of the positions supported under active policies are located there and less than a quarter of these positions support education or training. Also, some of the policies have been ineffective in finding lasting employment for participants: under a specific job creation scheme, young labor force participants are publicly employed for “socially useful” purposes. This scheme was introduced in 1998 and covered some 150,000 people, about 80 percent of whom were employed in the south, but as the scheme expired, participants found it very difficult to find durable jobs in the private sector. In Spain, studies conducted by INEM, the national employment service, have found that unemployed individuals completing training courses are only slightly more likely to find employment than individuals who do not receive training.

128. Overall, there seems to be little reason to suppose that the increased employment intensity of growth in recent years is substantially due to the use of ALMPs. While an extensive study of the changing impact of ALMPs in recent years is beyond the scope of this paper, a highly impressionistic examination of the evidence—along with some studies undertaken by country authorities—raises doubts. First, the country that has spent the most on these policies per unemployed population—Germany-has recorded the slowest rate of employment growth, while that which has spent the least—Spain—has recorded the most dramatic increase in employment growth. Second, the allocation of spending on ALMPs in all four countries would appear to be sub optimal, either in terms of services, regions, or both; in particular, a number of important ALMPs have failed to provide participants with lasting jobs.

Labor market deregulation

129. The revival of employment growth in Europe in the second half of the 1990s coincided with a loosening of stringent labor market regulations. As a proxy for labor market regulation in general, this paper examines the strictness of Employment Protection Legislation (EPL), which is monitored by the OECD on a regular basis. In the definition of the OECD, EPL encompasses regulations governing severance pay, redundancy procedures, special requirements for collective dismissals, fixed-term contracts, temporary work agencies, rules favoring disadvantaged groups, etc. It takes into account legislated regulation as well as regulation that is grounded in collective bargaining, court interpretations of legislative and contractual provisions, and private agreements. However, the EPL measure does not capture some potentially important aspects of labor market regulation such as the setup of the process of wage determination, working time restrictions, and governments’ role in wage fixing.

130. The effect of EPL on employment and unemployment has been the subject of extensive empirical and theoretical research. While economists generally tend to favor more flexible labor markets arrangements, the findings of the research on the effect of EPL are less clear cut. A recent empirical study by the OECD finds that stricter EPL tends to be associated with lower employment-to-population ratios, but detects no significant linkage between EPL and overall unemployment (OECD, 1999b). It also finds that stricter EPL tends to improve employment opportunities of prime-age men at the expense of the rest of the population. Moreover, stricter EPL tends to lower the turnover in the labor market, meaning that long-term unemployment accounts for a bigger share in total unemployment. Blanchard and Wolfers (2000) find that a measure of EPL interacted with shocks explains well the different unemployment performances across Europe over time.

131. In line with the general trend, Italy, Germany, and Spain made EPL more flexible over the last decade, but France made it somewhat more stringent. All four countries remain on the more restrictive side compared to the OECD average. Among the four countries, Germany is listed as the least restrictive followed by France, Spain, and Italy. In their relative ranking Germany and France have switched position since the late 1980s; there was no change in the relative position of Spain and Italy.

132. Much headway is attributable to the relaxation of regulations governing temporary employment. In Italy and Germany, it has become easier to renew fixed-term contracts and temporary work agencies were allowed broader scope of operation, while the maximum cumulated duration of temporary work contracts was increased. Moreover, the bias against part-time employment stemming from social security contributions was eliminated in Italy, leading to very strong growth in these positions. Recently, the trend toward more flexibility in Germany is likely to have come to a halt since the OECD took stock of EPL.

133. Spain is the only country to have introduced significantly more flexibility in the area of permanent contracts, while France stands out for having tightened regulation during the 1990s. In Spain, notification periods for dismissal were shortened and severance payments in the case of unfair dismissal reduced. The definition of fair dismissals was also broadened, although labor courts were slow to adopt this new definition. Moreover, 1997 saw the introduction of a new permanent contract with lower firing costs for special groups. Also, the liberalization of regulations governing fixed-term contracts predated the late 1980s, and therefore does not show up as an improvement during the 1990s.66 France’s EPLs became more stringent as a result of new limitations on fixed-term contracts and the operation of temporary work agencies. Significant other changes, such as the reduction of the standard workweek and reductions of employer’s social security contributions, are not captured by the OECD’s EPL measure.

134. EPL does not to appear to be a promising candidate for explaining cross-country differences in employment performance during the 1990s, although their loosening in Italy and Spain certainly supported employment generation. While Germany and Italy show the largest degree of EPL relaxation they also feature the worst employment growth performance. France, the only country that tightened EPL, nonetheless experienced rapid employment growth. As to Spain, important labor market deregulation predated the period covered by OECD EPL data, so that its role is potentially understated here. A number of variables could potentially be influenced by EPLs. The only variable that has the expected association is the ratio of employment to working age population, which is strictly falling with the strictness of EPLs. Of course, this is not to say that employment performance is unrelated to the strictness of EPLs. Rather, it means that other more important factors dominated the effects of cross-country differences in measured EPL reform.

E. Conclusions

135. Output growth has been significantly more employment intensive in the 1990s than in previous decades in France, Italy and Spain, but not in Germany if full-time equivalent units are used. This increased output elasticity of employment growth reflects primarily wage moderation. There is strong evidence of a downward shift in the wage curve: after controlling for the unemployment rate and changes in productivity, real wages grew sharply in the 1970s and early 1980s, but have since moderated everywhere. Put another way, for a given unemployment rate, workers today appear to be willing to accept a lower real effective wage than was the case in the past. The shift of the wage curve appears to reflect mainly a more job-conscious bargaining stance of workers. Labor demand also contributed to more job-intensive growth in the second half of the 1990s, possibly owing to less substitution of capital for labor as wage costs have moderated and to cuts in employer-paid social security contributions.

136. Other factors such as the use of active labor market policies, labor market deregulation, or benefit reform seem unlikely to be at the heart of the cross-country differences in employment performance and the increased job intensity of growth. More often than not these factors were not systematically associated with the relative employment performance of countries; some of the factors did not change much over time and are therefore unlikely to have had much of a significant quantitative impact. However, deregulation of the labor market, particularly of part-time and temporary employment, may have increased incentives for some segments in the population to return to the labor market. And, in France, the laws aiming at the reduction of the standard workweek to 35 hours seems to have had a positive effect on job creation so far.

137. The lack of improvement in employment growth in Germany can be traced to a smaller decline in productivity-adjusted labor costs since the mid-1980s. As Germany avoided the sort of excesses that plagued other continental labor markets in the 1970s and early 1980s, unemployment—despite less wage moderation—is still lower than in the other countries, although it remains high by historical standards. However, an overly rapid convergence of eastern toward western German wage levels has caused unemployment in the east to shoot up to nearly 20 percent. As a result, a structural, regional labor market problem has emerged that, in many ways, is comparable to the regional problems in Italy and Spain. Barring a major change in wage setting, the problem in the east may well turn out as long-lasting as the regional problems in the other countries. In addition, the increase in the tax wedge during the 1990s, which was related to the financing of reunification and was unique among the countries reviewed here, made wage moderation all the more painful in Germany.

138. A number of policy implications arise from this study:

  • First, and most critically, wage moderation does work. Much of the growth of employment in recent years can be traced to greater wage moderation. With unemployment rates in all four countries at historically high levels, wage moderation needs to continue.

  • Scope exists for continued labor cost reduction through cuts in the tax wedge. Some countries have recently embarked on moves in this direction, but given the sharp increase in the tax wedge that has occurred everywhere since the 1970s, there remains considerable room to be exploited in this regard.

  • There is a need for greater wage dispersion in addition to continued moderation. Within individual countries there is no evidence that wage moderation has been greater in regions with relatively high rates of unemployment, and substantial disparities in unemployment rates exist across either regions, genders, age groups, or all three in all countries.

  • Employment growth can be supported by less generous unemployment benefits. In the countries reviewed, net replacement rates remain very high. A reform of unemployment compensation schemes could contribute to further increasing the employment intensity of growth by reducing workers’ reservation wages. Unemployment compensation in each of the countries under review, except Italy, is relatively generous, with respect either to level, duration, or both.

  • Finally, there appears to be scope to further support employment growth through better use of active labor market policies.

Estimating the Wage Curve

139. The slope of wage curves has been estimated in numerous contributions to the literature and is estimated here with regional data of the four countries of this study. Specifically, the following equation based on equation (7) of Box III. 1 is estimated:

ω=δ0+δ1u(8)whereδ0=δ0(u*,m,C,Cu),δ1=δ1(u*,m,C,Cu)

The methodology is similar to that of Blanchflower and Oswald (1994). In fitting the equation, assuming labor-augmenting technological progress,67 real wages W must be adjusted for labor productivity growth. Typically this equation is fitted to the data with u standing for the natural logarithm of the unemployment rate; here the equation is also fitted with u standing for the raw unemployment rate.68

140. Blanchflower and Oswald (1995) investigate the relation between real wages and unemployment for various countries, including Italy among those reviewed here. Using broadly comparable microeconomic data, they estimate a cross (individual)-sectional earnings equation for each country, in which, together with the a set of control variables,69 the regional unemployment rate is entered as an explanatory variable. They demonstrate that there appears to be an empirical regularity in international pay and unemployment data, whereby estimates of the unemployment elasticity of pay (i.e., δ1) cluster around-0.1.

141. This paper estimates the wage curve with regional national accounts data on labor costs and labor force survey data on regional unemployment rates, using the following setup:

ωit=δi0+δ1ituit+δ2ωit1+ϑit
Wit=γi0+γ1uit+γ2Wit1+ρit

where ω denotes the unit labor cost,70 u the unemployment rate (or the natural logarithm thereof), and w real compensation per employee. These regressions are fitted to EUROSTAT NUTS2 regional data for a total of i=53 regions, including 17 regions in Spain, 16 in Germany, and 20 in Italy. Regional labor cost data for France are not available in EUROSTAT.71 The time series span 1983-95 for Spain, 1983-1998 for Italy, and 1991-1999 for Germany. The regression set-up allows for region-specific intercepts (fixed effects) and thus for region-specific equilibrium rates of unemployment or region-specific real effective wages: both theoretical arguments and empirical evidence support the existence of stable, region-specific natural rates of unemployment.72 And, by introducing a lagged dependent variable, it allows for equilibrium real wages adjusting gradually to shocks. Notice, though, that the regression assumes constant parameters across time and is therefore a simplified, dynamic version of the wage equation (7) in Box III. 1.

142. The model specification is similar to that used elsewhere in the literature. It resembles closely the models fitted to data by Layard, Nickell, and Jackman (1991) as well as by Blanchard and Katz (1997) for the U.K. regions and U.S. states.73 But unlike these studies, data on regional productivity and GDP deflators allow the computation of productivity adjusted real compensation, or unit labor costs: a wage curve should be estimated using productivity adjusted real compensation because there should be no relation between productivity and unemployment rates in the long run. To have comparable results, the regressions were also run using real compensation unadjusted for productivity. Moreover, all regressions were run by instrumenting for the dependent variable on the right-hand side with the same variable lagged twice, to address the endogeneity of the regressors.74

Regression of Regional Labor Costs on Unemployment Rates

article image
Sources: EUROSTAT; national authorities; and IMF staff calculations.Note: Standard errors in parentheses. A * indicates significance at a 5 percent level.

143. The results suggest a lower short-run but higher long-run elasticity of labor costs with respect to the unemployment rate than in Blanchflower and Oswald (1995). The short-run elasticity is estimated at about 0.05 in the ω (ULC) equation and 0.025 in the w (real compensation) equation or, respectively, at one half or one quarter of the 0.1 elasticity found by Blanchflower and Oswald. The long-run elasticities amount, respectively, to 0.25 and 0.18. Furthermore, the adjustment of labor costs (particularly of compensation) in response to shocks is slow, as evidenced by coefficient estimates on the lagged dependent variables which are between 0.8-0.9: following a shock, it takes about five years to complete half of the adjustment to the new equilibrium. Blanchflower and Oswald do not find a significant estimate for the lagged dependent variable, unlike here and in other studies.75 The results from the specification with the unemployment rate in levels rather than logarithms are virtually identical: they suggest a short-run semi-elasticity of 0.4 and a long-run semi-elasticity of 2: note that with an average rate of unemployment of 10 percent, this finding is equivalent to the raw elasticities of 0.05 of 0.2 found with the logarithm specification. For all further analysis, this paper works with the elasticity of 0.2, but key results are also checked with an elasticity of 0.1, as found by Blanchflower and Oswald.

References

  • Atkinson, T, M. Glaude and L. Olier (2001), Inégalités Économiques, Conseil d’Analyse Économique, La Documentation Française, Paris.

    • Search Google Scholar
    • Export Citation
  • Blanchard, O. and L. Katz (1992), Regional Evolutions, Brookings Papers on Economic Activity, (1), pp. 175.

  • Blanchard, O. and L. Katz (1997), “What We Know and Do Not Know About the Natural Rate of Unemployment,Journal of Economic Perspectives, Vol. 11 (1), pp. 5172.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Blanchard O. and J. Wolfers (2000), “The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence,The Economic Journal, No. 110, pp. C1C33.

    • Search Google Scholar
    • Export Citation
  • Blanchflower D. and A. Oswald (1994), The Wage Curve, MIT Press, Cambridge (MA).

  • Blanchflower D. and A. Oswald (1995), “International Wage Curves”, in Freeman and Katz (eds.), Differences and Changes in Wage Structures, University of Chicago Press, Chicago.

    • Search Google Scholar
    • Export Citation
  • Bover, O. S. Bentolila, and M. Arellano (2001), “The Distribution of Earnings in Spain During the 1980s: The Effects of Skill, Unemployment, and Union Power,CEPR Discussion Paper, No. 2770.

    • Search Google Scholar
    • Export Citation
  • Brandolini, A. and P. Sestito (1999), “Wage Dispersion, Low-Paid Workers and Household Poverty in Italy, 1977-95”, mimeo, Bank of Italy.

    • Search Google Scholar
    • Export Citation
  • Brunello, G. Lupi C., and P. Ordine (2000), “Regional Disparities and the Italian NAIRU,Oxford Economic Papers, Vol. 52, pp. 146177.

  • Chirinko R.S. (1980), The Real Wage Curve Over the Business Cycle, Review of Economics and Statistics, Vol. 67.

  • Cotis, J-Ph. R. Meary and N. Sobczak (1996), “Le Chomage d’Équilibre en France. Une Evaluation,Document de Travail de la Direction de la Prévision, no. 96-14.

    • Search Google Scholar
    • Export Citation
  • Daveri F. and G. Tabellini (2000), “Unemployment and Taxes,Economic Policy, London.

  • Decressin, J. and A. Fatás (1995), “Regional Labor Market Dynamics in Europe,European Economic Review, Vol 39, pp. 16271655.

  • Erickson, C. and A. Ichino (1995), “Wage Differentials in Italy,” in Freeman and Katz (eds.), Differences and Changes in Wage Structures, University of Chicago Press, Chicago.

    • Search Google Scholar
    • Export Citation
  • Estevão, M. and N. Nargis (2001), “Wage Moderation in France,IMF Working Paper.

  • Freeman R. and L. Katz (1995), Differences and Changes in Wage Structures, University of Chicago Press, Chicago.

  • Hsiao, C. (1986), Analysis of Panel Data, Economometric Society Monographs, Cambridge University Press, Cambridge.

  • Keane, M. Moffit, R. and D. Runkle (1988), “Real Wages over the Business Cycle: Estimating the Impact of Heterogeneity with Micro Data,Journal of Political Economy, Vol 96.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Keane, M. and E. Prasad (1993), “The Relation Between Skill Levels and the Cyclical Variability of Employment, Hours, and Wages,IMF Working Paper 93/44.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Layard R. Nickell S. and R. Jackman (1991), Unemployment, Oxford University Press, Oxford.

  • Lerais, F. (2001), “Une Croissance Plus Riche en Emplois,Premieres Syntheses, DARES, February, no. 07.1.

  • Marino M. and R. Rinaldi (2000), “Tassazione e Costo del Lavoro nei Paesi Industriali,Temi di Discussione No. 373, Bank of Italy, Rome.

    • Search Google Scholar
    • Export Citation
  • Mauro, P. E. Prasad, and A. Spilimbergo (1999), Perspectives on Regional Unemployment in Europe, IMF Occasional Paper No. 177, IMF: Washington D.C.

    • Search Google Scholar
    • Export Citation
  • Nickell, S. (1997), “Unemployment and Labor Market Rigidities: Europe versus North America,Journal of Economic Perspectives, Vol. 11 (3), pp. 5474.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • OECD (1994), The Jobs Study, OECD, Paris.

  • OECD (1999a), Benefit Systems and Work Incentives, OECD, Paris.

  • OECD (1999b), Employment Outlook, OECD, Paris.

  • Passeron, V. (2000), “Les 35 Heures, l’Emploi et les Salaires,Premières Synthèses, DARES, December, no. 50.2.

  • Pisani-Ferry, J. (2000), Plein Emploi, Conseil d’Analyse Économique, La Documentation Française, Paris.

  • Prasad, E. (2001), The Unbearable Stability of the German Wage Structure: Evidence and Interpretations, mimeo, International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Romer D. (1996), Advanced Macroeconomics, McGraw-Hill, New York.

  • Sachverständigenrat (2000), Chancen auf einen höheren Wachstumspfad, Jahresgutachten 2000/01, Metzler-Poeschel Stuttgart.

  • Scarpetta, S. (1996), “Assessing the Role of Labour Market Policies and Institutional Settings on Unemployment: A Cross-Country Study,OECD Economic Studies, Vol. 0, No.26, pp. 4398.

    • Search Google Scholar
    • Export Citation
  • Schimmelpfennig A. (2000), Structural Change of the Production Process and Unemployment in Germany, Kieler Studies 307, Mohr Siebeck, Tübingen.

    • Search Google Scholar
    • Export Citation
43

Prepared by Jörg Decressin, Marcello Estevão, Philip Gerson and Christoph Klingen.

44

In fact, Passeron (2000) presents evidence of the positive effect on employment growth from the 35-hour workweek laws enacted in 1996 and 1998 in France. Employment growth in France was also very strong in 2000, right after the introduction of a third law enforcing the 35-hour workweek in firms employing more than 20 individuals.

45

The coefficient of variation is defined as the standard error divided by the mean of regional unemployment rates defined according to EUROSTATs’ NUTS level 1.

46

An intercept dummy was introduced as well but turned out insignificant in all countries except Italy. For Italy, the intercept dummy suggests that employment growth was 1.3 percent lower during the 1990s upon holding constant current and lagged output growth and lagged employment growth. The slope dummy then indicates that a one percentage point increase in output was associated with an additional 0.68 percent employment growth during the 1990s. Since these estimations serve only descriptive purposes, not to provide a structural interpretation of the employment intensity of growth during the 1990s, only the results without the intercept dummy are reported.

47

If number of employees, instead of full-time equivalent units, are used to estimate the above equation, all countries, including Germany, show an increased sensitivity of employment to output growth in the 1990s.

48

Even though the result for Spain is highly significant, the relatively early onset of strong wage restraint, along with labor market deregulation in the mid-1980s, suggests that the structural break might have occurred prior to 1990.

50

In the diagram, wages (to) are corrected for both price and productivity levels. Employment (x - axis) is normalized by the size of the labor force and is equivalent to one minus the unemployment rate (u).

51

Blanchard and Katz (1992), Decressin and Fatàs (1995), and Mauro et al. (1998) provide evidence for stable, region-specific equilibrium unemployment, region-specific rates of labor productivity growth notwithstanding. In the formal model sketched in Box III.1, real wage demands, W, grow one-for-one with productivity if income of the unemployed, B, is linear in labor productivity. The latter is likely to be the case, as unemployment benefits are typically explicitly or implicitly linked to the prevailing wage level.

52

To derive real effective wages, the nominal gross hourly wage (i.e., wage costs minus employers’ social security contributions) is deflated by the GDP deflator and the index of labor augmenting technical progress. The growth rate of the latter can be estimated as TFP growth divided by the labor share of income. Real unit labor costs are a rough approximation of real effective product wages. The estimates refer to the business sector, rather than the whole economy, and are calculated from the Analytical Database of the OECD and unpublished OECD data on average hours worked. Real effective wages per worker exhibit a similar pattern and, in particular, show the same ranking of countries. Data for Germany refer to western Germany until 1990 and united Germany thereafter.

53

Calculated from national data on nominal wage increases stipulated in collective bargaining agreements and deflated by the consumer price index.

54

To see this, rearrange equation (2) by moving te to the right-hand side. This isolates the technology parameter a on the left-hand side and the right-hand side approximates the labor share of income for reasonably close to 1.

56

First-time job seekers are not eligible for unemployment benefits in Germany too but German youth unemployment is much lower than in the other countries.

57

The authors’ results suggest that a 10 percentage point increase in effective labor tax rates, through the effect on wages, can account for a 4 percentage point increase in European unemployment. Similarly, Marino and Rinaldi (2000) find that, in OECD countries, an increase in the effective labor tax rate by 1 percentage point raises labor costs by 1.1 percent in the first year and 1.4 percent in the long run. And, for Italy in particular, Brunello and others (2000) observe a cointegrating relationship between unemployment, the tax wedge, the real interest rate, and union power: according to their simulations, a reduction in the tax wedge from the level in 1996 to the one of the early 1980s would yield a 15 percent reduction in the unemployment rate, with about 70 percent of this reduction taking place after five years.

58

Variations in prices and labor productivity are assumed to capture changes of income if unemployed and variations of the tax wedge are used as a proxy for changes of labor taxation.

59

In the literature, composition changes of employment have been shown to seriously affect average wages over the business cycle. See for instance, Chirinko (1980), Keane, Moffitt, and Runkle (1988), or Keane and Prasad (1993).

60

Estevão and Nargis (2001) use individual level data to estimate shifts in the wage curve in France in the 1990s. They show that there was a 16 percent decline in average hourly wages adjusted for inflation, productivity growth, unemployment changes, and variations in the composition of the unemployment pool.

61

For the relation between the NAIRU, the mismatch, and the variance of unemployment rates see Layard, Nickell, and Jackman (1991).

62

Brunello, Lupi, and Ordine (2000) find that wages in southern Italy are a function of labor market conditions in northern Italy and thus do not respond to local labor market conditions.

64

For example, see Scarpetta (1996).

66

In 2001, firms were required to provide termination payments to workers upon the expiration of fixed term contracts, increasing the level of EPL.

67

This assumption ensures the existence of a well-specified, steady-state growth path, with an unchanged capital-output ratio. See, for example, Romer (1996).

68

When estimating wage curves from unemployment and wage data parameter estimates are typically distorted if wage curves shifts between observations. Across regions, however, wage curves should be more stable than labor demand curves, thus limiting the distortion when using regional data. Moreover, estimates derived here are simular to results found by the literature and all conclusions of this study are not very sensitive to the exact maginitude of parameters.

69

These variables include dummies for the industry and region in which a worker is employed; his or her gender, marital status, experience, schooling, rank, and union status. The countries include the United States, Canada, Austria, Italy, the Netherlands, Switzerland, and Norway.

70

Data to compute total factor productivity are only available for a subset of the regions considered here.

71

See Estevão and Nargis (2001) for estimates of the wage curve for France using individual-level data from 1990 to 2000.

72

See, for example, Blanchard and Katz (1992) for evidence on the United States, and Decressin and Fatás (1995), as well as Mauro and others (1998) for Europe.

73

Blanchard and Katz ran the regressions with nominal wages rather than real wages and did not adjust for productivity because data on prices and regional output were not available. Accordingly, they included time-specific fixed effects, effectively assuming that price and productivity developments are very similar across individual states.

74

See, for example, Hsiao (1986). The bias in the estimates as a result of introducing a lagged dependent variable in a set-up allowing for fixed effects declines rapidly with the time length of the sample.

75

Blanchard and Katz (1997), after adjusting for specific features of the micro data of Blanchflower and Oswald, also find significant and sizeable estimates for the coefficient on the lagged dependent variable; their results with micro data on individuals were then similar to those that they obtain with macro data. Estevão and Nargis (2001) also find a significant coefficient for lagged wages in wage curve estimates for France.

Selected Euro-Area Countries: Rules-Based Fiscal Policy and Job-Rich Growth in France, Germany, Italy and Spain
Author: International Monetary Fund