Chapter

2 Adjustment Dynamics in the German Labor Market

Editor(s):
Dennis Snower, and S. Henry
Published Date:
January 1997
Share
  • ShareShare
Show Summary Details

Unemployment IN (WEST) Germany remains low compared with that in other European countries. Nevertheless, as elsewhere, it has risen considerably since the 1970s, because employment growth has failed to keep pace with a steady rise in the labor force (Chart 1). Much of the literature has focused on the possible rise in the “equilibrium” rate of unemployment that may underlie this increase. This paper takes a rather different approach, focusing instead on adjustment dynamics in the labor market. Slow adjustment of the labor market means that the effects of temporary shocks will be persistent and the effects of permanent shocks will be delayed. An understanding of why adjustment is slow should point to policies that could speed it up—a particularly important question in Germany in light of the danger of persistent high unemployment in the new Länder.

Chart 1.West Germany: Unemployment and Employment

(In thousands)

Sources: Federal Bureau of Statistics, Volkswirtschaftliche Gesamtrechnungen; and Deutsche Bundesbank, Monthly Report.

This paper reports the results of a set of estimates based on data on the west German labor market for 1970–94.1 Broadly following the methodology in Karanassou and Snower (1994), a system of labor demand, labor supply, and wage determination equations is estimated, using current and, most important, lagged aggregate variables, such as employment, labor force, and wages, as well as policy indicators. The study singles out the effects of lagged variables, including the ways in which they may have changed over the last two and a half decades, as objects of attention. A number of sources of adjustment lags are identified and are related to specific features of labor market institutions and policies. In addition, the study pays particular attention to the way the effects of different lagged variables interact, and the response of the system to various types of shocks is simulated over time. As will be seen, the results—tentative as they are—suggest that unemployment may remain far from its long-run equilibrium for a long time, so that the long-run equilibrium itself may lose much of its relevance over any time horizon of practical significance.

Developments in Labor Market Institutions and Policies Since 1970

The German labor market is characterized by a centralized system of collective bargaining and by rather extensive social protection.2 Its institutions and policies have remained relatively stable throughout the postwar period, although some trend toward greater government activism in labor market matters can be discerned after 1970, when the Government proclaimed an unconditional “full employment guarantee” (Soltwedel (1988)).

The following paragraphs describe a number of important labor market institutions and policies as they stand at present, and note significant changes that have occurred since 1970. In each case an attempt is also made to identify the ways in which the institutions and policies described may affect the equilibrium rate of unemployment and the speed with which unemployment moves to its equilibrium.

The basic principles governing wage bargaining in Germany date back to the 1920s. The partners to the negotiations are industrywide employers’ federations and trade unions. Negotiations take place at the regional level, but national demands are publicized in advance of the wage round, and wage contracts within a sector are typically identical nationwide. Wage agreements specify a set of minimum tariff wages; minimum wages are established by law. The coverage of collective bargaining is very broad. Although the unionization rate is only about 40 percent (up only a little from about 35 percent in 1970), the majority of west German employers belong to employers’ federations and are obliged to pay at least the union-negotiated wage.3 Individual firms are free to pay wages higher than the tariff, and most do. However, in practice, perhaps in part because industrial action is prohibited in firm-level negotiations, increases in actual wages tend to follow the increase in the tariff, and “wage drift” from year to year is small.

In practice, the system operates more like a centralized one than these bare facts suggest. Not only is there little regional differentiation between wage contracts within sectors, but coordination mechanisms between sectors include the employers’ and unions’ umbrella organizations and a system of “pattern bargaining” whereby certain sectors—metalworking in particular (represented by the largest of the unions), followed by construction and the public sector—set the standard for each wage round.

There has been no significant change in either the operation or the coverage of the bargaining system since 1970. Something of a watershed was reached in 1984, with a nine-week strike by the metalworkers’ union, affecting about half a million workers by the end—the biggest and arguably most embittered strike in the history of west Germany (Streeck (1988)). Remarkably, the traditional consensual pattern of west German industrial relations survived this strike, and, unlike in other countries in the 1980s, there was no determined effort by employers or the Government to roll back union power.4

The generally consensual patterns evident in collective bargaining also prevail at firm and plant level. “Industrial democracy” in Germany has a long history, but was strengthened significantly with the 1972 Works Constitution Act (Betriebsverfassungsgesetz) and the 1976 Codetermination Act. These laws established the rights of Works Councils (Betriebsräte, representing the workforce) to approve decisions on most aspects of working practices and to be consulted or at least informed on other matters. They also strengthened worker representation on supervisory boards (Aufsichtsräte), which guide long-term company strategy.

The high degree of coverage of collective bargaining and the influence of employees over firm-level working practices suggest a potential role for “insider-outsider” effects in German labor markets.5 Both the unions and the firms’ own workforces represent primarily the interests of the employed, so that high unemployment may have only a weak tendency to correct itself through wage restraint. This effect may be magnified by centralized bargaining, which will tend to make all wages in the economy move together and, hence, may prevent wages in particularly weak segments of the labor market from adjusting to eliminate unemployment.6 Pattern bargaining in Germany may introduce a further element of persistence in the form of wage staggering, even though most of the agreements are concluded over a time span of only a few months: adjustment to a labor market shock may be delayed by the heavy influence exerted on contracts concluded later in the wage round by contracts concluded earlier in the round.7

A unique feature of the German labor market is the prominence of apprenticeship programs, which provide employment for over half of all 16- to 19-year-olds and which combine practical training with compulsory part-time education at vocational schools. Practical training has long been provided in private firms, but since 1969 it has been regulated by the Vocational Training Act (Berufsbildungsgesetz). In 1976, fears of free-rider problems led the Government to introduce fines for employers for a collective failure to create enough apprenticeships to absorb all those leaving school. Although the relevant law was struck down as unconstitutional in 1980, apprenticeship programs expanded considerably during the 1970s and 1980s. These programs are undoubtedly a major reason why Germany does not suffer from the widespread youth unemployment prevalent in other countries; in addition, they may contribute to reducing insider power because they “enfranchise” outsiders in the wage determination process. Another factor is Germany’s educational system, which helps keep the proportion of unskilled workers in the population low.

Germany places considerable emphasis on active labor market policies—an emphasis that has increased since 1970. The precursor of the Federal Labor Office was created as early as 1927, with responsibilities for job placement in addition to the compulsory unemployment insurance system, but the Labor Promotion Act (Arbeitsförderungsgesetz) of 1969 set the stage for an expansion of vocational training and counseling and job creation programs.

To the extent that the expansion of both apprenticeship programs and active labor market policies has mitigated skill mismatches, it would be expected to have reduced equilibrium unemployment (at least compared with what it might have been otherwise). At the same time, retraining and job creation programs may help smooth adjustment by ensuring that the unemployed remain in touch with the labor market.

Income maintenance for the unemployed is provided under the headings of unemployment benefit (Arbeitslosengeld, typically for the first year of unemployment), unemployment assistance (Arbeitslosenhilfe, thereafter and indefinitely, subject to means-testing), and social assistance (Sozialhilfe, available to all whose income is inadequate).8 Replacement ratios rose in 1975, but were reduced in 1984 and again in 1994 (Table 1). At the same time, from the beginning of the 1980s, eligibility requirements for unemployment insurance payments—in particular the minimum contribution period and the penalties for turning down job offers—were gradually tightened.

Table 1.Germany: Unemployment Insurance Statutory Replacement Ratios(In percent)
1970–741975–831984–931994
Unemployment benefit
With child(ren)63686867
Without child(ren)63686360
Unemployment assistance
With child(ren)53585857
Without child(ren)53585653
Source: German authorities.Note: Statutory replacement ratios were constant over the periods indicated. Replacement ratios apply to previous after-tax earnings, excluding the typical thirteenth-month wage; thus, as a proportion of total annual earnings, replacement ratios are a little lower.
Source: German authorities.Note: Statutory replacement ratios were constant over the periods indicated. Replacement ratios apply to previous after-tax earnings, excluding the typical thirteenth-month wage; thus, as a proportion of total annual earnings, replacement ratios are a little lower.

An exception to the trend of decreasing generosity of benefits from the mid-1980s occurred in 1986, when the period of eligibility of older workers for unemployment benefits was extended from one year to, depending on age and tenure, up to 32 months. Because unemployment in Germany is disproportionately concentrated among older workers, unlike in many other European countries, the significance of this change should not be underestimated: in 1994, 40 percent of the unemployed in west Germany were 45 or older (up from about 30 percent in the early 1980s), compared with 30 percent of those in dependent employment.

From an international perspective, Germany’s replacement ratios for unemployment insurance are not unusually high, at least in the first few years of unemployment. However, Germany is unusual in that it provides unemployment insurance benefits (albeit at a slightly reduced level, in the form of unemployment assistance) indefinitely.

From a theoretical point of view, the more generous the unemployment insurance, the less intense the job search by the unemployed and the downward pressure unemployment puts on wages. The result could be not only higher equilibrium unemployment, but also higher persistence in unemployment following adverse macroeconomic shocks.

The growing cost of both active labor market policies and income support for the unemployed, together with a more general expansion of government activities, has been reflected in rising taxation of labor. Like other major industrial countries, Germany has experienced an enormous rise in the “tax wedge” between net take-home pay and the total cost of labor to employers (Charts 2 and 3). The overall wedge has risen from just over 50 percent (of take-home pay) in 1970 to close to 90 percent in 1994.

Chart 2.West Germany: Labor Costs and Take-Home Pay

Source: Federal Bureau of Statistics, Volkswirtschaftliche Gesamtrechnungen

Chart 3.West Germany: Labor Taxes

Source: Mendoza. Razin, and Tesar (1993).

Note: All taxes on labor as a percentage of all payments for labor (including both social security contributions and payroll taxes).

Rising taxation of labor would be expected—for given take-home pay—to raise labor costs, reduce employment demand, and hence raise the equilibrium rate of unemployment. Of course, if rising taxation of labor is instead absorbed through a decline in take-home pay, employment demand would be unaffected, and measured unemployment might even fall to the extent that labor supply contracts in response to falling net wages. In either case, changes in the taxation of labor might be expected to affect both long-run equilibrium unemployment and the dynamics of unemployment (as it may take time for agents to adjust fully to the tax changes).

Employment protection in Germany is fairly severe. This may be one reason why the German economy operates with large amounts of overtime even in periods of recession (Franz and König (1986)), and why, as shown by Abraham and Houseman (1993), shocks to labor demand tend to result in changes in the number of hours worked rather than in the number of employees.

For strictness of protection against individual dismissals, Grubb and Wells (1993) rank Germany fifth among 11 countries of the Organization for Economic Cooperation and Development (OECD), directly behind 4 southern European countries (Portugal, Spain, Italy, and Greece). Perhaps even more important, much is left to the discretion of the courts, creating uncertainty for employers. Individual dismissals have to be “fair,” but criteria for fairness are only partly laid down in legislation. Similarly, individual redundancies can be implemented only if alternatives are “intolerable,” and the choice of redundant workers is subject to “social criteria,” but the definitions of these requirements have in practice been left up to the courts.9 In addition, large-scale dismissals must be accompanied by social plans, under which compensation is negotiated.10

Although there were no basic legal changes in employment protection during the 1970s and early 1980s, labor courts over this period increasingly developed the idea of employees’ “social property rights” in their jobs. There is some evidence that in practice average dismissal compensation payments rose considerably over the period (Soltwedel (1988)). Employment protection legislation was relaxed a little in 1985, with the passage of the Employment Promotion Act (Beschäftigungsförderangsgesetz). This law expanded the possibilities for fixed-term contracts, increased their maximum duration to 18–24 months from 6 months, and exempted new firms from the social plan requirement. Due to expire in 1995, its validity was recently extended to the year 2000.

While the effects of employment protection are the object of debate, the high costs and uncertainty involved in making workers redundant in Germany are likely to affect the dynamics of adjustment in the labor market, by making current employment depend more strongly on past employment. In addition, the relevance of social criteria to the dismissal decision is likely to magnify long-term unemployment and thus to increase the persistence of unemployment: those who have been long-term unemployed are relatively more likely to be protected by these criteria, and are therefore at still more of a disadvantage in attempting to find employment.

Aside from their direct effects on the speed of adjustment of employment, hiring and firing costs also strengthen the hand of insiders in wage negotiations, and would thereby be expected to delay the adjustment of wages to weak labor market conditions and to result in higher equilibrium unemployment rates overall.

The peculiarities of the German housing market—in particular a relatively high proportion of owner occupancy, together with an underdeveloped market in previously owned houses (owing in part to tax preferences for new housing)—are likely to have contributed substantially to regional unemployment differentials and, thus, to aggregate unemployment rates. The sharp increase in unemployment in Germany during the 1970s and the 1980s coincided with the structural decline of the coal, shipbuilding, and metallurgy industries located in the north of the country, and these areas remain characterized by especially high unemployment rates. However, a study of this phenomenon would require a model disaggregated by region, and the issue thus lies beyond the scope of this study.

In sum, German labor market policies and institutions exhibit a number of features that would be expected to raise equilibrium unemployment and reduce the speed with which unemployment responds to changing conditions. The most important of these features are relatively strict employment protection, high and rising labor taxation, and generous unemployment insurance (with a rising duration of benefits in the 1980s). The most important influences in the opposite direction are likely to be training policies, including the apprenticeship system.

Labor Market Model

The characteristics of the German labor market were analyzed by estimating a model of labor demand, labor supply, and wage determination that allows for a rich lag structure. Its adjustment dynamics were subsequently investigated by simulating the effects of temporary and permanent labor demand shocks.

Description of the Model

The system of equations was specified as follows: a labor demand equation, with employment as a function of total labor costs to the employer and GDP (the latter as a proxy for product demand); a wage bargaining equation, with real take-home pay as a function of productivity and the wedge between real total labor cost to employers and real take-home pay for employees (which consists of the tax and social security wedge between total labor costs and take-home pay, and of the discrepancy between consumer and producer prices); and a labor force equation, with labor force as a function of working-age population and unemployment (the latter as a proxy for the “discouraged-worker” effect). A number of other independent variables were tried in each equation, but proved insignificant.

The potential effects of labor market policies and institutions were analyzed in several ways in this model. First, policy variables were included in the system as exogenous, explanatory variables. Second, the stability of the relationships was analyzed by exploring potential changes in the estimated coefficients for different sample periods, corresponding to changes in labor market policies and institutions. The relative stability of these policies and institutions in Germany over the past twenty-five years sets limits on what either method can achieve. Nevertheless, it is possible to identify important ways in which labor market behavior has changed, ways that in themselves suggest possible policy responses. Finally, the equations themselves, and the effects of lagged variables in particular, point to various possible sources of sluggish adjustment, some of which may be amenable to policy changes.

Estimation

The estimation strategy (autoregressive distributed lag, or ARDL, strategy) was to estimate equations including a number of lags of the independent and dependent variables and to infer the long-run relationships implied by them.11 The general structure of the approach relies on (1) testing that the variables are all integrated of order 1,12 (2) testing for cointegration (also via the Johansen procedure),13 (3) OLS estimation, (4) misspecification tests and consideration of the possibility of endogeneity, and (5) 3SLS estimation. In principle, the result would be a structural model that is consistent with an underlying cointegrating vector autoregression (VAR), as shown in Karanassou and Snower (1994).

The long-run relationship in the ARDL model drops out of a general dynamic specification, which is estimated directly using a “general to specific” specification search to simplify the lag structure. In each case, the coefficients on the lagged variables provide key information on the speed with which the system adjusts.

The models were estimated through a variety of econometric techniques to counter potential problems, such as simultaneous equation bias. Quarterly data for 1970–94 were used.14 The results presented here are based on ordinary least squares.15 Lags of up to eight quarters were used in light of the strong autocorrelation of many of the data series, not only at the first but also at the higher orders.16

Results and Implications

Tables 2, 3, and 4 report the individual-equation estimates for the employment, wage, and labor force equations, respectively.

Table 2.Employment Equation
E = number of people employed
wg=real total labor costs (total labor income per worker, inclusive of taxes and social security payments by employers and employees, divided by the GDP’ deflator)
y = real GDP
Q1, Q2, Q3 = quarterly dummies
4 = four-quarter difference operator
Autoregressive distributed lag (ARDL) model
R2 = 0.98; DW = 1.91; serial correlation statistic = 1.51 [0.82]
Implied long-run relationship from the ARDL model
R2 = 0.63; DW = 1.91.
Note; All variables are in logarithms. t-statistics are in parentheses. The serial correlation statistic is a Lagrange multiplier statistic of the null hypothesis that the residuals are serially uncorrelated (up to the fourth order) and is distributed as a Χ2 with 4 degrees of freedom; the rejection level is reported in square brackets.
Note; All variables are in logarithms. t-statistics are in parentheses. The serial correlation statistic is a Lagrange multiplier statistic of the null hypothesis that the residuals are serially uncorrelated (up to the fourth order) and is distributed as a Χ2 with 4 degrees of freedom; the rejection level is reported in square brackets.
Table 3.Wage Bargaining Equation
wn = real net take-home pay (labor income per worker, exclusive of taxes and social security payments by employers and employees, divided by the CPI)
g = real GDP per person employed
rtw = real tax wedge, or ratio of real total labor cost to real net take-home pay
Q1, Q2, Q3 = quarterly dummies
4 = four-quarter difference operator
Autoregressive distributed lag (ARDL) model
R2 = 0.99; DW = 1.47; serial correlation statistic = 9.62 [0.47]
Implied long-run relationship from the ARDL model
R2=0.81; DW=1.47.
Note: All variables are in logarithms, t-statistics are in parentheses. The serial correlation statistic is a Lagrange multiplier statistic of the null hypothesis that the residuals are serially uncorrelated (up to the fourth order) and is distributed as a Χ2 with 4 degrees of freedom; the rejection level is reported in square brackets.
Note: All variables are in logarithms, t-statistics are in parentheses. The serial correlation statistic is a Lagrange multiplier statistic of the null hypothesis that the residuals are serially uncorrelated (up to the fourth order) and is distributed as a Χ2 with 4 degrees of freedom; the rejection level is reported in square brackets.
Table 4.Labor Force Equation
L = number of people in employment and unemployment
wn = real net take-home pay
U = unemployment race
N = working-age population
∆ = one-quarter difference operator
4 = four-quarter difference operator
Autoregressive distributed lag (ARDL) model
R2 = 0.87; DW = 1.75; serial correlation statistic = 10.5 [0.03]
Implied long-run relationship from the ARDL model
R2 = 0.92; DW = 1.75.
Note: All variables are in logarithms, t-statistics are in parentheses. The serial correlation statistic is a Lagrange multiplier statistic of the null hypothesis that the residuals are serially uncorrelated (up to the fourth order) and is distributed as a X2 with 4 degrees of freedom; the rejection level is reported in square brackets.
Note: All variables are in logarithms, t-statistics are in parentheses. The serial correlation statistic is a Lagrange multiplier statistic of the null hypothesis that the residuals are serially uncorrelated (up to the fourth order) and is distributed as a X2 with 4 degrees of freedom; the rejection level is reported in square brackets.

Employment Equation

Employment (E) is modeled as a function of total labor costs to the employer (wg) and GDP (y) (the latter as a proxy for product demand):17

The results for the employment equation are reported in Table 2.18

The signs of the coefficients are as expected if the equation represents a labor demand function: high labor costs reduce employment, while high output demand raises it.19 The long-run elasticity of employment with respect to the real total labor cost is estimated to be -0.2, using the implied long-run relationship from the distributed lag model.

The results suggest that there is considerable persistence in employment. The sum of the coefficients on lagged values of employment in the distributed lag form of the employment equation is very high (0.92), indicating that employment depends very strongly on its own past values. Strong persistence in employment may be related to the presence of considerable hiring and firing costs in Germany, as described in the section on labor market institutions and policies since 1970.

Recursive estimation of the employment equation yields no evidence of changes in the coefficients on the individual lagged employment terms, or in their sum, so that there is no evidence of changing persistence in employment. In particular, there is no sign of reduced persistence in employment around 1985, the date of the Employment Promotion Act. This finding accords with that of Abraham and Houseman (1993), who analyze the responsiveness of employment to output changes and find no evidence of instability in this relationship around 1985.

Wage Bargaining Equation

Real take-home pay (wn) is modeled as a function of productivity (g) and the wedge between real total labor cost to employers and real take-home pay for employees (rtw) (which is equal to the ratio of nominal total labor cost to nominal take-home pay divided by the ratio of the GDP deflator to the consumer price index (CPI)):20

The results for the wage bargaining equation are reported in Table 3.21 The coefficients on both the real tax wedge and productivity display the expected signs in the long-run relationship. Real take-home pay responds negatively to the real tax wedge and positively to productivity. In this model, employees bear the whole (indeed, more than the whole) increase in the tax wedge. Anecdotal evidence suggests that the trade unions have not recently sought to “shift forward” increases in the personal income tax burden, although they did in the early 1970s and, on the other hand, that increases in social security contributions tend to be accepted by employers and employees at the statutory levels, that is, 50 percent each.

Although real take-home pay is positively associated with productivity in the long run, the elasticity is estimated to be less than 1. Thus, increases in productivity have not been fully reflected in increases in real take-home pay over the sample period considered, as would be expected in the presence of a rising tax wedge. Chart 4 shows the gross and net labor shares over the sample period.

Chart 4.West Germany: Labor Share

Source: Federal Bureau of Statistics, Volkswirtschaftliche Gesamtrechnungen.

Note: Labor shares calculated as percent of GDP at factor cost.

Perhaps the most important result from the long-run relationships, and one that significantly affects the dynamics of the system, is the small role played by unemployment in wage bargaining. A priori, one would expect unemployment to feature prominently in the wage equation: higher unemployment should dampen wage claims, both by raising fears of unemployment among “insiders,” and by increasing competition for jobs from “outsiders.” It is, however, difficult to find evidence of such an effect in these regressions. In particular, the unemployment rate (or its change) does not enter significantly into the “levels” form of the wage bargaining equation. The first lag of unemployment was found to be marginally significant, but recursive estimation of this coefficient (Chart 5) suggests that the effect of unemployment on wage settlements diminished in the 1980s.

Chart 5.West Germany: The Effect of Unemployment on Wages

Note: Wage bargaining equation: coefficient of lagged unemployment rate and its two-standard-error bands based on recursive ordinary least squares.

Since the restraining effect of current and lagged unemployment on wages is important for equilibrating labor markets, the weakness of this effect in Germany is a significant finding. It points to hysteresis as a major danger and to insider power as a possible explanation. Certainly, strong worker influence over firms’ working practices and generous social protection—for all the other advantages that these institutions bring with them—must help to entrench insider power. In addition, as previously noted, there is some evidence that the effect of unemployment on wages has weakened, suggesting that insider power may have increased over time. With few marked changes in labor market policies and institutions during the period under consideration, it is difficult to pinpoint the sources of the increase. However, the general direction of policy changes toward more government involvement in the labor market may be relevant to this result, as may be the increased duration of unemployment benefits for older workers. Both of these would tend to strengthen insider power by reducing the search intensity of the unemployed and tempering insiders’ own fear of unemployment.

In addition, the finding of increasing insider power is consistent with the observed growing segmentation of the labor market between high- and low-skilled individuals in particular, with the latter becoming increasingly marginalized.22 It is widely recognized that skill-biased technological change and skill-biased international trade flows have in recent decades increased the demand for skilled labor in industrial countries relative to that for unskilled labor. These developments would tend to require a widening of skilled-unskilled wage differentials, whereas the centralized bargaining system in Germany has instead tended to compress wages.

High persistence in wages is suggested by the relatively high (0.75) sum of the coefficients on lagged wages in the wage equation. High persistence in wages may reflect in part staggered wage setting, which will tend to make current wages depend on past wages. Such staggering may be rather prevalent in Germany, with its systems of pattern bargaining.

Labor Force Equation

The labor force is modeled as a function of working-age population (N), unemployment (U), and real take-home pay (wn):

The labor force is positively and significantly associated with working-age population, with an elasticity of about 1 (Table 4). The labor force is also negatively associated with the unemployment rate, perhaps owing to the discouraged-worker effect While the elasticity is rather small, it is statistically significant at the conventional levels. There is no significant evidence of any effect of real take-home pay on the labor force, perhaps owing to offsetting income and substitution effects. It is also possible to speculate that the decision of whether to enter the labor force is determined by other factors, while wages may play a role in determining effort and the number of hours worked.

Labor supply displays considerable persistence and there is significant evidence that—holding working-age population and other factors constant—past increases in unemployment are negatively associated with increases in the labor force. There is also tentative evidence that changes in working-age population are associated with labor supply changes, although the elasticity is relatively low. Again, there is no evidence that wage changes play a significant part in changes in the labor force.

The labor force exhibits high persistence. The sum of the coefficients on lags of the labor force is 0.88. This suggests that entering and leaving the labor force involve large adjustment costs. The lasting discouraged-worker effects found in these equations suggest that unemployment will be associated not only with lower output today, but also with lower potential output tomorrow, and point to the risk that measured unemployment may not capture total welfare costs. The fact that unemployment enters into the labor force equation in lagged form suggests that it takes some time before it discourages workers from searching for a job, and points to the important role of long-term unemployment. Measures to keep the long-term unemployed in touch with the labor market may thus be particularly important.

Labor Market Dynamics

The importance of lagged effects in the individual equations of the labor market model has been highlighted. These lags, in addition, interact to create more complex dynamics in the system as a whole. To investigate these dynamics, the labor market system23 was simulated in the presence and absence of, first, a temporary adverse shock to labor demand and, second, a permanent adverse shock to labor demand. Slow adjustment is manifested in the former case in the form of a slow return of unemployment to what its level would have been in the absence of a shock (following Karanassou and Snower (1994), we call this “persistence” of unemployment), and in the latter case in the form of a slow adjustment of unemployment to its new equilibrium level (we call this “imperfect responsiveness” of unemployment).

The simulations show that the effects of a temporary shock are very persistent and that it takes considerable time to reach a new long-run equilibrium following a permanent shock. Each case is analyzed in turn below.

Following a 1 percent temporary shock to employment demand, it takes 20 years for the unemployment rate to stabilize within the vicinity of equilibrium (defined as a distance equivalent to one-fourth of the initial shock) (Charts 6 and 7). The results reflect strong and mutually reinforcing persistence in employment, wages, and the labor force. When the negative shock is applied to employment demand, employment falls, thus raising productivity. The rise in productivity raises wages, and this in turn reinforces the initial fall in employment. Also, as employment falls, unemployment rises, and in subsequent periods this puts downward pressure on the labor force, through the discouraged-worker effect. This mechanism slightly mitigates the effects of the fall in employment on the (measured) unemployment rate. The lag structure produces overadjustment, and unemployment oscillates toward its equilibrium level.

Chart 6.West Germany: Temporary Shock; Difference in the Unemployment Rate With and Without the Shock

(In percentage points)

Source: IMF staff estimates.

Chart 7.West Germany: Temporary Shock; Difference in Each Series With and Without the Shock

(In percentage points)

Source: IME staff estimates.

Following a permanent shock to employment demand, it takes 12 years for the unemployment rate to return to and stabilize in the vicinity of the new equilibrium (Charts 8 and 9). Following the permanent rise in the growth rate of employment demand, the same mechanisms as described in the case of a temporary shock raise productivity and wages and reduce the labor force; the overall effect is a rise in unemployment.

Chart 8.West Germany: Permanent Shock: Difference in the Unemployment Rate With and Without the Shock

(In percentage points)

Source: IMF staff estimates.

Note: Dotted line is the difference in long-run equilibrium with and without the shock.

Chart 9.West Germany: Permanent Shock; Difference in Each Series With and Without the Shock

(In percentage points)

Source: IMF staff estimates.

Note: Dotted line is the difference in long-run equilibrium with and without the shock.

Overall, the model suggests that there is considerable sluggishness in German labor markets. This conclusion is further supported by the intetnational comparisons reported in Chapter 1 of this volume. Based on the simulations reported above for Germany and similar ones for other European countries, Henry and Snower present the time taken to recover after temporary and permanent shocks, as well as aggregate measures of persistence and imperfect responsiveness.24 The time to recover following either a temporary or a permanent shock is longer in Germany than in any of the other countries. Similarly, the measure of persistence is higher in Germany than anywhere else, and the measure of imperfect responsiveness indicates that inertia was higher only in the United Kingdom.

These results may appear to be at variance with the traditional view of the German labor market as rather more “flexible” than those of other European countries—a view that is based on Germany’s relatively low unemployment rate.25 However, there is no necessary connection between low equilibrium unemployment and rapid labor market adjustment. Although the results of the present study need to be treated with caution, they suggest that while equilibrium unemployment may be low in (west) Germany, adjustment to that equilibrium may be at least as slow as in the rest of Europe.

Of course, because the model is by its nature symmetric (responses to rises in employment demand are the exact opposite of responses to falls in employment demand), both persistence in the face of a temporary shock and overshooting in the face of a permanent shock will work to lower unemployment if a shock is favorable, just as they work to raise unemployment if a shock is adverse. In practice, many of the effects identified as sources of sluggish adjustment are likely to have asymmetric effects. An exploration of these asymmetries is beyond the scope of this study. Nonetheless, the presence of considerable sluggishness in German labor markets is sufficient to raise serious concerns about the likely path of unemployment should Germany face serious adverse shocks in the future. To the extent that west German labor market behavior is replicated in east Germany, slow adjustment is also likely to keep unemployment in east Germany higher than it would otherwise be.

Conclusions

The estimation results suggest a number of tentative conclusions on the potential links between institutional features and adjustment dynamics in the German labor market, which may be relevant for the debate on possible policy implications.

The study finds that the German labor market exhibits considerable sluggishness in adjustment. The specific numerical results should be treated with caution, but they are at least suggestive that, in the present model, it takes one to two decades for the effects of a temporary shock to employment demand to disappear, and for employment to adjust to its new equilibrium level following a permanent shock. Adjustment to a temporary shock appears to be particularly slow. It may thus be that unemployment at any given point in time is determined much more by its own history than by its “long-run equilibrium.” Such slow adjustment—as in other European countries—may help to explain the rise in unemployment since the early 1980s, and does not bode well for the path of unemployment in the future, including in east Germany. In addition, there is tentative evidence from this and companion studies that the German labor market adjusts more slowly than those of other European countries. Thus, even though the (west) German unemployment rate is relatively low by European standards, structural reform of the labor market may be no less necessary in Germany than elsewhere in Europe. At the same time, strong persistence in unemployment strengthens the presumption that macroeconomic policies should be stability oriented. If inflation is allowed to get out of hand, the rise in unemployment necessary to get it back to acceptable levels may turn out to be very long lived.

The results show that there is substantial persistence in employment in Germany, perhaps related to relatively strict employment protection and to the fact that the courts have considerable discretion in the area of dismissals. Even so, there is no evidence of reduced persistence in employment around the time of the Employment Promotion Act of 1985, which somewhat loosened the restrictions on dismissals. Another factor underlying high persistence in employment and unemployment is a weak effect of unemployment on wage growth. This result may be related to Germany’s labor market institutions (widespread collective bargaining, industrial democracy, and generous social protection), which—for all their strengths in other dimensions—would be expected to prevent outsiders from exerting a major influence on wages. The indefinite duration of unemployment assistance, which distinguishes the German unemployment insurance system from that of most other countries, probably plays an important role in this respect.

The fact that collective bargaining is in practice highly centralized may also contribute to the weak effect of unemployment on wage growth, as it reduces the likelihood that wages will adjust to clear segments of the labor market (for example, in particular regions or sectors) that are subject to particularly high unemployment. Although centralized bargaining need not produce a relatively undifferentiated structure of wages across skill levels, in practice it tends to do so, reflecting unions’ objectives; thus, high unemployment among particular skill groups may also exert only weak downward pressure on wages overall.

To the extent that any effects of unemployment on wage growth are found, there is evidence that they weakened during the 1980s. An increase in government involvement in the labor market may have contributed to this, as may an increase in the duration of unemployment benefits for older workers (who represent a large and growing fraction of the unemployed). The finding is also consistent with the observed growing segmentation in job opportunities of the German labor market—between regions, ages, and skill levels—during the 1980s. There is also inertia in wage levels, which depend strongly on their own past values, implying slow adjustment to labor market disequilibria. Germany’s system of pattern bargaining, whereby the first wage contracts signed each year become the standard for subsequent ones, may contribute to this, as there is little opportunity for changed labor market conditions to be reflected after the first agreements have been concluded.

Rises in measured unemployment tend to be mitigated by declines in the labor force: there is evidence that the labor force responds negatively to unemployment, suggesting that discouraged workers leave the labor force. Thus, the detrimental effects of unemployment may be undetestimated if one looks only at the unemployment rate, without considering the loss of the productive potential of discouraged workers. Measures to keep the long-term unemployed in touch with the labor market may be especially important in this respect.

References

    Abraham, Katherine G., and SusanS. Houseman,1993, Does Employment Protection Inhibit Labor Market Flexibility? Lessons from Germany, France and Belgium, NBER Working Paper No. 4390 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    Franz, Wolfgang, and HeinzKönig,1986, “The Nature and Causes of Unemployment in the Federal Republic of Germany since the 1970s: An Empirical Investigation,”Economica (Supplement), pp. S219–S244.

    • Search Google Scholar
    • Export Citation

    Grubb,David, and WilliamWells,1993, “Employment Regulation and Patterns of Work in EC Countries,”OECD Economic Studies,Vol. 21, pp. 7–58.

    • Search Google Scholar
    • Export Citation

    Karanassou,Marika, and DennisJ. Snower,1994, “The Sources of Unemployment Persistence and Responsiveness” (unpublished; London: University of London, Birkbeck College).

    • Search Google Scholar
    • Export Citation

    Layard, R.,S.Nickell, and R.Jackman,1991, Unemployment: Macroeconomic Performance and the Labour Market (Oxford, England: Oxford University Press).

    • Search Google Scholar
    • Export Citation

    Mendoza,EnriqueG.,AssafRazin, and LindaTesar,1993, “A Comparative Analysis of the Structure of Tax Systems in Industrial Countries,”IMF Working Paper 93/14 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Soltwedel,Rüdiger,1988, “Employment Problems in West Germany: The Role of Institutions, Labor Law, and Government Intervention,”Carnegie-Rochester Conference Series on Public Policy,Vol. 28(Spring), pp. 153–220.

    • Search Google Scholar
    • Export Citation

    Streeck,Wolfgang,1988, “Industrial Relations in West Germany. 1980–87.”Labour,Vol. 2 (No. 3), pp. 3–44.

    van derWilligen,Tessa,1995, “Unemployment, Wages, and the Wage Structure,” in United Germany: The First Five Years—Performance and Policy Issues,ed. by R.Corker and others.IMF Occasional Paper No. 125 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
1

Insufficient data preclude the inclusion of east Germany, but the wholesale extension of west German labor market institutions to east Germany means the results should also be relevant to the new Länder.

2

For a detailed discussion of three key features of the German labor market—the wage bargaining system, income support for the unemployed. and employment protection—See van der Willigen(1995).

3

The Minister of Labor has the authority. under certain conditions, to extend the validity of collective agreements to employers who are not members of employers’ federations, through a “declaration of general validity.” However, these declarations involve a minority of contracts. usually relating to working conditions rather than to wages.

4

The only legislative change—much debated but, in the final analysis, rather minor—was the passage of Section 116 in 1986. Previously, unions had been able to conserve strike funds by striking only at selected firms and counting on other firms in their sector having to shut down for lack of supplies, but Section 116 made workers at the latter firms (who stand to benefit directly from the industrial agreement at stake) ineligible for unemployment benefit.

5

This is not to deny the benefits of centralized bargaining, which many authors have theorized includes greater overall wage restraint. See van der Willigen (1995) for a detailed discussion of the advantages and disadvantages of the German system of collective bargaining.

6

This is true whether the labor market is broken down by region, sector, or skill level. In the last case, it is the usually egalitarian objectives of the unions that make all wages tend to move together. See van der Willigen (1945) for a detailed discussion of wage differentiation in Germany.

7

An interesting example is provided by the 1995 wage round. The key metalworkers’ agreement was signed just before a sharp appreciation of the deutsche mark. Commentators generally agree that wage increases in the (export-oriented) metal working sector would probably have been lower had the appreciation occurred before the agreement was concluded, but these wage increases nevertheless set the standard for other sectoral agreements, concluded after the appreciation.

8

A government proposal to limit the duration of unemployment assistance to two years was recently rejected by the Parliament.

9

Social criteria that have been adduced include wealth, age, health, family responsibilities, and spouse’s income.

10

For a detailed description of employment protection in Germany, see van der Willigen (1995).

11

All equations were also estimated using an alternative strategy, which involved the estimation of long-run, cointegrating relationships and the subsequent estimation (both with the ordinary least squares (OLS) and with the three-stage least squares (3SLS) techniques) of error correction models (ECMs) based upon them. The ECM model is a restricted form of the ARDL model, with the restrictions reflecting a priori judgments about the form adjustment dynamics are likely to take. The results of the ECM equations ate not presented because they were not used in the final estimation of the measures of persistence and imperfect responsiveness.

12

While formal Dickey-Fuller and Augmented Dickey-Fuller unit root tests yield ambiguous results for some of the series, it seems safe to assume that all variables included in the regressions are integrated of order 1.

13

In all cases, there is tentative evidence that long-run relationships exist among the variables of interest. For the most part, the Dickey-Fuller and Augmented Dickey-Fuller statistics are not far from rejecting the null of no cointegration at the conventional levels. In all cases, the Johansen procedure yields a strictly positive set of cointegrating vectors, which typically do not significantly differ from those of the long-run relationships estimated using both approaches adopted in this study.

14

The data are drawn from German official sources and the IMF’s International Financial Statistics. All series refer to west Germany. All variables are in logarithms. Employment includes both dependent employment and self-employment.

15

The 3SLS estimation technique was not applied to the ARDL model, because the 3SLS results would have been practically identical to the OLS ones, because—in the final specification—none of the equations contained contemporaneous values of any endogenous variable.

16

The fact that eight lags are required when quarterly data are used is consistent with the finding that two lags are needed when annual data are used, as in Karanassou and Snower (1994).

17

The potential role of a number of other explanatory variables—including real interest rates, competitiveness indicators (the ratio of import prices to the GDP deflator), and (changes in) the real oil price—was explored, with no significant and robust relationship being identified, so that they are not included in the preferred specification.

18

Little evidence was found of endogeneity bias, related to the two-way causation between (demand for) output and labor inpur. In fact, in a cointegrating regression in levels, two-stage least squares estimation using trade-weighted foreign partner income—that is. an appropriately exogenous variable that affects the demand for, but not the supply of, domestic output (and/or oil prices) as an instrument for domestic GDP—yielded results very similar to the OLS estimates;

19

An analysis of the exact mechanisms through which aggregate product demand may affect aggregate labor demand is beyond the scope of this paper.

20

Because much of the impact of unemployment insurance in Germany is thought to come through the intermediary of duration of benefits, rather than their level, statutory replacement ratios were not included in the wage equation. No suitable data could be found for aggregate replacement ratios.

21

A model in which the tax wedge (defined as the ratio of nominal total labor cost to nominal take-home pay) and the price wedge (defined as the ratio of the CPI to the GDP deflator) were used as separate regressors showed that the coefficients on these two wedges were not significantly different. The preferred specification therefore aggregates the two wedges into a single “real tax wedge” (rtw).

22

While the apprenticeship system in Germany ensures that the majority of people have a qualification, the unemployment rate for the unskilled is still generally at least twice as high as for those with a qualification. In addition, those with low-level skills should be taken to include those whose specific skills have been devalued by structural change, helping to explain the concentration of unemployment in the heavily industrial northern regions of west Germany and among older people. See van der Willigen (1995) for a derailed discussion of labor market segmentation in Germany.

23

Since the labor demand equation contains output (demand) among its explanatory variables, the simulated system included a technical relationship between labor and output, using the coefficients obtained from prior estimation of a Cobb-Douglas production function for the German economy. In addition, two identities were part of the system: first, unemployment was defined as the log-difference between the labor force and employment, and second, productivity was defined as the log-difference between output and employment.

24

The measure of persistence is defined as the sum of deviations of unemployment from its baseline path in response to the temporary shock, and the measure of imperfect responsiveness as the sum of deviations of unemployment from the new long-run equilibrium resulting from the permanent shock.

25

Layard, Niekell. and Jackman (1991), however, also find more rigidity—in this case, real wage rigidity, interpreted as the extent to which real shocks translate into unemployment at constant inflation rates—in Germany than in France or Italy.

    Other Resources Citing This Publication