Germany: Selected Issues
Author:
International Monetary Fund
Search for other papers by International Monetary Fund in
Current site
Google Scholar
Close

This Selected Issues paper analyzes labor market asymmetries and macroeconomic adjustment in Germany. Empirical work reported shows that in Germany, negative demand shocks increase the unemployment rate by more than the decrease in the unemployment rate caused by a comparable-sized positive demand shock. The contribution of labor costs to explaining the high level of unemployment, particularly since unification, is studied. Empirical estimates are obtained for the wage gap—the deviation of actual labor costs from warranted labor costs based on estimated production functions assuming competitive factor markets and full employment.

Abstract

This Selected Issues paper analyzes labor market asymmetries and macroeconomic adjustment in Germany. Empirical work reported shows that in Germany, negative demand shocks increase the unemployment rate by more than the decrease in the unemployment rate caused by a comparable-sized positive demand shock. The contribution of labor costs to explaining the high level of unemployment, particularly since unification, is studied. Empirical estimates are obtained for the wage gap—the deviation of actual labor costs from warranted labor costs based on estimated production functions assuming competitive factor markets and full employment.

I. Labor Market Asymmetries and Macroeconomic Adjustment1

A. Introduction and Overview

8. Stubbornly high and rising unemployment remains Germany’s chief economic problem and reversing this trend poses a considerable challenge to policymakers. In the broader context of European unemployment, much attention has been focused on institutional and structural rigidities of labor markets that may have hindered their functioning. In particular, the structure of unemployment benefits along with aspects of the wage determination process have been studied repeatedly in search of explanations for Europe’s persistent unemployment problems. Also at issue is the extent to which unemployment reflects structural factors or stems from cyclical considerations and insufficient demand. In the context of traditional natural rate models of the labor market, unemployment is delineated into cyclical and structural components; in hysteresis models, these lines become increasingly blurred. In turn, different unemployment theories have produced a range of policy recommendations.2

9. This paper revisits many of these issues from the perspective of the wage-price mechanism—the linkage between wages, prices and economic activity. This reduced-form relationship can, in principle, encompass the implications of a variety of unemployment theories and allow for broad comparisons across countries with very different institutional features. This framework highlights the costs that economic rigidities in labor (and product) markets can have for their functioning and for overall macroeconomic adjustment. In particular, in the presence of various labor and product market rigidities, the wage-price mechanism may display important asymmetries in the unemployment-inflation process. If, for example, labor markets are characterized by significant downward rigidities in nominal and real wages, negative impulses would reduce employment to a greater extent than equivalent positive impulses would raise employment. Thus, the short-run Phillips curve would display a degree of convexity, reflecting this asymmetrical response.

10. Asymmetries in the unemployment-inflation relationship have several important economic implications. A central implication is that economic recoveries fail to generate sufficient employment gains over the cycle to offset the rise in unemployment suffered in economic downturns. As a result, economies tend to operate, on average, at a level of unemployment (economic activity) above (below) the level which would obtain in the absence of these asymmetries. Conceptually, unemployment in this framework can be classified into three broad components: (1) long-run structural unemployment-—representing the equilibrium which would obtain in the absence of any business cycles; (2) the gap between structural and average unemployment—resulting from asymmetric adjustment over the business cycle; and (3) the difference between observed and average unemployment, reflecting the prevailing stage of the business cycle. Average unemployment (i.e., averaged over the business cycle) associated with stable inflation over the longer run is often referred to as the “natural rate” or non-accelerating inflationary rate of unemployment (NAIRU). In traditional (symmetric) natural rate models with a linear Phillips curve, this long-run equilibrium rate is identical to the structural rate of unemployment, leaving two (rather than three) components of unemployment: cyclical and structural.

11. In the asymmetric model, the third component measured by the difference between the natural and structural rates of unemployment depends on the magnitude and frequency of shocks and the extent of the rigidities which underlie the fundamental asymmetries. Episodes of higher economic variability would tend to increase the natural rate of unemployment and this gap. Correspondingly, effective stabilization policies which lessen the impact of economic shocks can durably reduce (average) unemployment. By allowing business cycles to affect unemployment persistently, this framework combines aspects of both natural rate and hysteresis models. However, whereas the level of the economic activity over the business cycle usually affects longer-term unemployment in hysteresis models, its variability affects the natural rate of unemployment in the current context.

12. In the case of Germany, the empirical findings suggest that the unemployment-inflation trade off is characterized by asymmetries and significant upward drift in both the natural and structural rates of unemployment. In terms of cross-country comparisons (G-7 countries plus Denmark, the Netherlands, Ireland, Spain, and Sweden), Germany is grouped with countries (e.g., France, Spain, and Denmark) with the highest degree of asymmetry in the wage-price mechanism. It is also found to have had the largest increase in its structural rate of unemployment over the past twenty years. These findings are consistent with other studies that indicate that Germany has considerable structural rigidities in labor markets stemming from the wage bargaining process, labor market regulations, and social programs.

13. In terms of policy implications, the advent of European monetary union (EMU) presents several challenges for Germany. Under the Exchange Rate Mechanism (ERM), Germany occupies a unique role as the anchor currency and, with it, has the flexibility to conduct a monetary policy focused on domestic objectives. This policy flexibility has enabled the authorities to mitigate through their efforts the effects of relatively rigid and asymmetric labor markets, which provide strong incentives to pursue effective stabilization policies. With monetary union, this policy asymmetry vanishes under a common monetary policy, limiting the Bundesbank’s ability to address Germany-specific shocks and their effects. Consequently, in the absence of further structural reforms, the change in monetary regime could raise average unemployment rates in Germany relative to other countries that might participate in EMU, who have already largely relinquished independent monetary policies. Indeed, this regime change with a greater focus on area-wide conditions may improve average unemployment rates for other prospective EMU participants compared with the monetary policy regime under ERM.

14. Comparing the wage-price mechanism across countries, Germany also appears somewhat less flexible, in terms of labor market adjustment, than several other European countries (and even more so when compared with Canada, Japan, and the United States). Assuming that these existing differences in the Phillips curve continue, common macroeconomic shocks in the future could give rise to uneven outcomes among prospective EMU participants (or existing ERM participants). In particular, without the aid of greater flexibility in prices and wages, the high degree of asymmetrical adjustment to macroeconomic shocks in Germany could lead to relatively adverse unemployment consequences compared with some of its European partners. Moreover, larger structural increases in German unemployment than elsewhere (based on past trends) could further lead to difficult policy dilemmas under EMU (or existing ERM arrangements) in the absence of further structural reforms. Thus, the need for prompt action to increase labor market flexibility is highlighted.

15. This chapter is organized as follows: section B briefly reviews explanations for longer-term developments in German unemployment to motivate the issue of rigidities and asymmetry in the wage-price mechanism; section C outlines the underlying framework and its implications (an illustrative model is relegated to the appendix); section D provides the basic empirical results; finally, several policy implications of labor market asymmetries in Germany are further explored using MULTIMOD simulations in section E.

B. Unemployment Developments and Alternative Explanations

16. Before turning to the analytical framework, it is useful to review prominent developments in German unemployment as well as alternative explanations that have been advanced. Perhaps the most notable aspect of unemployment behavior in Germany has been the sustained or trend increase in the rate of unemployment over the past thirty years. The unemployment rate has risen from ½ percent in the mid-1960s to 9½ percent in 1996 (in west Germany).3 Although the unemployment rate has oscillated from year to year, since the 1960s each major economic downturn and subsequent recovery, or supply side shock—such as the oil. price hikes of the 1970s and German unification in the early 1990s, has left the unemployment rate at a higher level than previously (Chart I-1). Thus, the unemployment rate has been steadily ratcheted up.

Chart I-1.
Chart I-1.

West Germany: Unemployment Rate

(In percent)

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

Source: Deutsche Bundesbank, Monthly Report.1/As defined by Bundesbank.

17. In general, there are several alternative approaches to studying the “equilibrium” (noncyclical) unemployment rate, relying on natural rate, persistence, or hysteresis explanations. These approaches have certain common features. Noncyclical unemployment, for example, can be the result of job search, market failures, including regulations (e.g., minimum wage, restrictive labor laws) and unionization, and repeated shocks coupled with a slow adjustment speed to equilibrium. One major difference between traditional natural rate models and persistence/hysteresis models is that in the former, the equilibrium is independent of history but dependent on other economic and policy variables, while in the latter class of models, the unemployment equilibrium is also history or path dependent.4

18. Traditional natural rate models build on the framework of a linear Phillips curve and rely on real or structural factors to explain trend developments in the noncyclical unemployment rate. In the presence of adverse supply shocks and structural changes, the equilibrium rate of unemployment would increase over time and the adjustment period to shocks could lengthen. In SM/96/227, the staff estimated simultaneous price and wage equations to identify an expectations-augmented Phillips curve. An “equilibrium” unemployment or NAIRU for west Germany stood around 7¼ percent since unification compared with an annual average unemployment rate of 8¼ percent in 1995—the last year of the estimation period. This estimate—although not very precise with a standard error of nearly 1¼ percentage points—is not dissimilar from results obtained by other researchers.

19. The hysteresis approach can accommodate the persistent or ratcheting behavior of the unemployment rate more easily than natural rate models.5 Such models usually rely on insider-outsider effects or human capital depreciation to explain why the burgeoning unemployed are not easily reabsorbed into the workforce.6 The implication is that the NAIRU depends on the adjustment path and is thus affected by transitory disturbances in addition to the factors determining long-run equilibrium.7 An implication of standard hysteresis models is that the unemployment rates could just as easily ratchet downward as upward under appropriate conditions. For example, a prolonged economic expansion that steadily reduced unemployment would over time lower the “equilibrium” unemployment rate. However, while examples of countries getting stuck in high unemployment equilibria are not uncommon, examples of countries moving hysteretically into low unemployment equilibria are scarce.8

20. In this connection, the importance of asymmetrical responses in explaining developments in unemployment and inflation has recently had a resurgence of interest.9 Specifically, where excess supply impulses generate larger output and employment losses than the counterpart gains associated with excess demand impulses, the wage-price mechanism exhibits asymmetrical behavior. Correspondingly, the Phillips curve would be convex rather than linear. The implication of the asymmetric model is that economic expansions fail to generate sufficient employment gains over the cycle to offset the losses incurred during comparable downturns, leaving average unemployment higher. Hence, a combination of more economic noise (variability) and asymmetries in the wage-price mechanism can raise the average or natural unemployment rate even if the structural rate was unchanged.

21. Thus, the asymmetric model exhibits features of both natural rate and hysteresis models. Like hysteresis models, the (average) unemployment rate is time dependent, and the history of past shocks is reflected in the current unemployment rate. However, it is the variability—rather than the level—of economic activity as a result of those shocks that durably affects the longer-run unemployment rate. For a fixed distribution of shocks, the asymmetric model predicts a stable long-run natural rate at a fixed deviation from a given structural rate of unemployment. Of course, the structural unemployment rate can also change over time, representing a shift in the Phillips curve, under the influence of structural and institutional factors as suggested by traditional (symmetric) natural rate models.

22. The economic foundations for asymmetries in the wage-price mechanism derive from underlying rigidities in labor and product markets. In his original formulation, Phillips10 argued for a non-linear relationship between wage adjustment and unemployment, largely reflecting asymmetric wage demands and bargaining.11 Other explanations for asymmetries in the wage determination process have turned to segmented labor markets and efficiency wages which place an effective floor on real wage adjustments. At the microeconomic level, features of goods markets and the behavior of firms could generate asymmetries in price adjustment as well. For example, production lags and a floor on inventories suggest that firms with market power may adjust prices upward more quickly as economic conditions change.12

23. In the case of Germany, several prominent institutional features of labor markets are worth noting. Wage determination is largely based on a system of collective bargaining at the sector level, with the coverage of negotiated wage settlements extending far beyond the level of union membership.13 At the firm level, employers are free to set wages above but not below the tariff wages that are agreed upon centrally, under a provision known as the “favorability principle” (i.e., in favor of the worker). At the sectoral level, wage determination is characterized by pattern bargaining, where leading sectors (e.g., the metalworking industry) set the standard for wage increases in the economy;14 wage dispersion across industries is lower than in more decentralized systems (e.g., the United States). In terms of firing costs, extensive protection and stringent restrictions regarding dismissals in Germany are considered relatively high by European standards.15 Meanwhile, social assistance of unlimited duration and high replacement ratios characterize a generous social benefit system supportive of high reservation wages. And while the German authorities have recently embarked on an extensive program of structural reforms,16 the effects of various labor market rigidities are likely to be felt for some time to come.

C. Analytical Framework: The Wage-Price Mechanism

24. An approach to exploring the implications of rigidities in labor and product markets for unemployment and macroeconomic adjustment is through a closer examination of the aggregate wage-price mechanism. Derived from the interaction of underlying wage- and price-setting equations (see appendix), the wage-price mechanism is a reduced-form relationship which summarizes the linkage between wage and price adjustment and the level of economic activity (i.e., the unemployment or the output gap). This relationship can be broadly characterized in the form of an expectations-augmented Phillips curve:

π t π t e = f ( u t u t * ) , ( 1 )

where π and πe are inflation and inflationary expectations, defined to be a function of the unemployment gap—i.e., the difference between actual unemployment u and the level associated with non-accelerating prices u* (NAIRU). This latter measure is referred to—interchangeably in the linear case—as the structural or natural rate of unemployment; correspondingly, this approach has been described as the natural rate hypothesis. As written, equation (1) also implicitly embodies the Keynesian perspective on economic fluctuations. With predetermined or sticky prices, output in the short run is demand determined, with causality running from the level of economic activity to eventual price adjustment. In the traditional Keynesian approach, the wage-price mechanism would summarize the supply-side of the economy and the mechanism for (sluggish) price level adjustment.

25. In contrast to the Keynesian view, the new classical approach runs the causality in the opposite direction, specifying that changes in output and employment are a function of expectational errors;17 this revised relationship has been referred to as the Lucas supply function.18 However, treating equation (1) as a reduced-form (rather than structural) relationship, one could allow causality to run in both directions.19

26. Causation issues aside, both the Keynesian and new classical perspectives formulate the basic wage-price mechanism in the form of a linear relationship:

π t π t e = γ [ u t u t * ] , ( 2 )

where γ is interpreted as the (constant) short-run trade-off parameter between unemployment and inflation, related to the “sacrifice ratio” or output costs of disinflation.20 From the Keynesian perspective, this parameter reflects the interaction of nominal and real rigidities, where greater flexibility suggests a larger γ coefficient (steeper Phillips curve).21 While this specification does not allow for a lower bound unemployment rate (or fixed capacity constraints), the advantages of the linear specification is that it may provide a reasonable approximation to the short-run unemployment-inflation trade-off over the relevant range in which an economy typically operates.

27. Alternatively, a simple specification for the convex Phillips curve, which maintains a strictly positive unemployment rate, can be written as follows:

π t π t e = γ [ u t u t * u t φ t ] ( 3 )

where Φ represents the lower bound on the unemployment rate, below which the economy cannot extend (Φt < ut), regardless of the degree of excess demand.22 When the economy reaches this “wall” or capacity constraint, further excess demand pressures would only feed into greater price inflation without any further gain in output or employment. This lower bound could reflect frictional unemployment beyond which excess demand pressures would be manifested in a further rise in vacancies, rather than a further decline in unemployment, owing to technological and informational limitations to search and matching activities.23

28. As with the symmetric model, γ in equation (3) reflects the interaction of nominal and real rigidities. However, unlike the linear case, the degree of nominal and/or real rigidity would not be constant (or symmetric). In the asymmetric model, the behavior of these rigidities would further depend on economic conditions and the nature of shocks, which further underpins (beyond capacity constraints) the non-linear relation between inflation and unemployment (see appendix). For example, assuming an effective floor below which real wages would not fall, the absolute unemployment response to disinflationary and inflationary impulses could differ significantly, depending on the size of the shock and the prevailing level of excess demand or supply (i.e., whether the floor acts as a binding constraint).

29. If these asymmetries are significant, several important implications are worth mentioning. Once the tradeoff between unemployment and inflation is not constant but changes with economic conditions (i.e., the “unemployment gap”), the history of past disturbances and current conditions become important in determining the response of the system to new shocks. This consideration is essentially irrelevant in a symmetric model where the marginal responses are always identical. Moreover, in the presence of asymmetries, the average (or expected) level of unemployment is above the level which would prevail in the absence of these convexities or in the absence of stochastic disturbances. This is shown in Chart I-2.

Chart I-2.
Chart I-2.

Germany Convex Phillips Curve

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

30. In terms of equation (3), note that the “structural” rate of unemployment u* represents an underlying NAIRU only in deterministic (non-stochastic) equilibrium in the asymmetric model. In the presence of shocks, however, the average or natural rate of unemployment ū, representing the NAIRU in a stochastic setting, lies somewhat above this value depending on the magnitude and frequency of shocks and the degree of asymmetry in the wage-price mechanism. This asymmetry-based component of unemployment is represented by a—the gap between ū and u* in Chart I-2. Finally, the actual unemployment rate u could diverge from its average or natural rate ū, as a result of cyclical considerations stemming from the business cycle, which completes the model’s decomposition of observed unemployment behavior. It should be noted that this second component of unemployment reflected by the gap α—absent in traditional natural rate models—cannot be considered as purely structural or cyclical as it reflects the interaction of both factors.

31. By modifying this separation found in standard models between the natural rate and cyclical developments, the asymmetrical model is akin to hysteresis models where cyclical unemployment directly affects the path-dependent NAIRU.24 However, whereas changes in the level of economic activity can have direct permanent effects on unemployment in standard hysteresis models, only changes in the variability of economic fluctuations would permanently affect the natural rate of unemployment in the present context. Meanwhile, temporary episodes of higher economic variability would lead to a protracted rise in unemployment in the asymmetric model which would eventually reverse depending on the degree of inertia in the economy.

32. In this sense, the asymmetric model combines elements of both hysteresis and traditional natural rate models. Transitory disturbances can have persistent (variance) effects on average unemployment, but the economy eventually reverts back to a given long-run natural rate of unemployment. Of course, lasting changes in the unemployment rate in the asymmetric model could also arise, through changes in the “structural” rate of unemployment u*, representing shifts in the Phillips curve. These permanent or trend developments would largely reflect institutional or structural considerations, including supply-side shocks such as oil price increases, and, in the case of Germany, reunification.

33. In terms of policies, the model with asymmetries addresses a notable shortcoming of traditional (symmetric) natural rate models wherein macroeconomic policy was incapable of affecting the longer-term level of output and unemployment.25 In the current context, economies with a high degree of asymmetry will be hampered in their ability to respond to economic disturbances, leading to a larger decline in average activity and employment than would prevail if markets were more flexible; correspondingly, effective stabilization policies which mitigate the impact of these disturbances can durably reduce (average) unemployment. To the extent that the gap between ū and u* changes over time—because of changes in the policy regime, in the distribution of external shocks, or in structural factors affecting the degree of asymmetry—the asymmetric model can also explain sustained or persistent changes in the observed level of unemployment. Such explanations are absent from the symmetric model.

34. Over the business cycle, the asymmetric model also has broad implications for the short-run stance of financial policies. The convex Phillips curve, for example, specifies that the inflationary costs of overheating can be much higher than in the linear case, suggesting that policymakers may wish to make preemptive changes in the stance of monetary policy to forestall excess demand pressures. However, this caution against inflationary pressures needs to be balanced against the higher costs of excessive disinflation, in terms of larger output and employment losses (i.e., rising sacrifice ratio), associated with the region of excess supply. Thus, compared to the policy implications of the linear Phillips curve, the asymmetric model attaches a premium on timely and appropriate stabilization policies, as excess demand or supply pressures are more costly in terms of inflation and unemployment.

D. Estimation

35. To examine whether potential asymmetries in the wage-price mechanism are important in the case of Germany, the following empirical equation is considered:

π t = λ π t e + ( 1 λ ) π t 1 γ [ u t u t * u t φ t ] + ϵ . ( 4 )

This is essentially a stochastic version of equation (3) augmented for inflation persistence.26 To proxy for inflationary expectations which are unobservable, data on long-term interest rates are used to construct measures of expected inflation following Debelle and Laxton.27 Actual inflation is measured in terms of the GDP deflator.28

36. In equation (4), also note that u* is an unobserved component. Hence, to estimate the empirical equation and to allow for time variation in u*, Kalman Filter estimation of equation (4) is employed following Kuttner.29 Essentially, for a given set of parameters, incremental information in the difference between inflation and inflationary expectations is used to update time-varying estimates of u*.30 Recursive estimates of u* and the parameters are then revised based on information in the whole sample to maximize the likelihood function. In effect, the behavior of inflation and inflationary expectations and the nature of the wage-price mechanism itself are used to identify u*, in addition to the information contained in the times-series behavior of actual unemployment.

37. As mentioned above, while global asymmetry is imposed in this framework, which is a reasonable assumption, the model allows for more symmetric behavior locally—i.e., in the neighborhood of u*—depending on the parameters. Hence, this framework is sufficiently general, to allow the data to determine the degree of asymmetry in the output-inflation trade-off over the relevant range. Estimates of equation (4) and its linear variant for Germany are shown in the table below (for Φ=0).31

Linear vs. Convex Phillips Curves West Germany, 1962–95

Symmetric Model: πt=λπ¯t+(1λ)πt1γ[utut*]+ϵt;

Asymmetric Model: πt=λπ¯t+(1λ)πt1γ[utut*ut]+ϵt.

article image
Note: A *(**) indicates significance at the 5 (1) percent level; t-statistics in parentheses. σ^ = standard error of estimate; l = value of log likelihood function. Q-statistics (not reported) against low and high order serial correlation for both models were insignificant.

38. Based on the point estimates, plots of the Phillips curve under each specification at the end of the sample are shown in Chart I-3. In the symmetric model, estimates of the slope parameter γ suggest a fairly “flat” short-run Phillips curve as shown in Chart I-3. This finding is indicative of the presence of a high degree of nominal and real rigidities and a high “sacrifice ratio.”32 The estimate of γ in the asymmetric model is also low—as will become apparent in the cross-country comparisons later—also indicating significant rigidities and, in the current context, a fair degree of asymmetry in the wage-price mechanism (shown graphically in Chart I-3). Both models yield similar estimates of the degree of inflation persistence (1-λ).

Chart I-3.
Chart I-3.

Germany 1/ Linear versus Convex Model

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

Source: Staff estimates.1/ Estimates refer to West Germany only.

39. Using estimates for λ and the u* series obtained from both models, an auxiliary regression is shown in the table below, which compares each model’s ability to explain the difference between inflation and (fitted) inflationary expectations—adjusted for inflation inertia—πe=λ^π^+(1λ^)π1. The asymmetric model yields a somewhat better fit although not overwhelmingly so.33 However, where the two models do differ significantly is in terms of their implications for u*. Under symmetry, positive and negative unemployment gaps should be (in large samples) equally likely to occur. With the actual unemployment rate just as likely to be above or below the NAIRU, the level of u* with a linear Phillips curve should be closer to the level of observed unemployment. With the convex Phillips curve, the unemployment rate will generally lie asymmetrically above the structural unemployment rate u* as in Chart I-2. This level difference is seen in Chart I-3 by comparing where each model intersects the horizontal axis.

Linear vs. Convex Phillips Curves West Germany, 1962–95

π t π ^ t e = γ [ u t u ^ t * ] + v t or = γ [ u t u ^ t * u t ] + v t

(Symmetric model)

(Asymmetric model)

article image
Note: A *(**) indicates significance at the 5 (1) percent level; t-statistics in parentheses. σ^ = standard error of estimate; D.w. = Durbin-Watson statistic for serial correlation.

40. Another important difference between the two models pertains to the variability of u*.34 Because the symmetric wage-price mechanism has a constant inflation-unemployment tradeoff, it cannot fully explain both inflationary and disinflationary episodes without greater variability in u*. In other words, the Phillips curve must shift more often and to a greater extent under the linear specification in order to explain periods of high inflation or high unemployment. In the case of Germany, this restriction even leads (implausibly) to negative estimates of the NAIRU in the early part of the sample period, when the actual unemployment rate was quite low but disinflationary pressures were present.

41. Meanwhile, for Germany, the model with asymmetries yields sensible, and more stable estimates of u*, in accordance with the natural rate hypothesis. Chart I-4 depicts the time-series implications for German unemployment of the asymmetric model, showing actual, average (filtered), and structural unemployment rates.35 Notice again that the average unemployment rate ū lies uniformly above the structural rate of unemployment u*, consistent with convexity in the Phillips curve.36

Chart I-4.
Chart I-4.

Germany 1/ Unemployment Developments, 1960-95

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

Sources: OECD Economic Indicators and staff estimates.1/ Data refer to West Germany only.

42. Over time, most of the increase in the unemployment rate in west Germany between 1970 and 1995 is attributable to a rise in its underlying structural component. However, part of the sustained increase in unemployment also appears to stem from an expanded gap between the natural rate Φ (i.e., stochastic NAIRU) and the underlying structural rate u* (i.e., deterministic NAIRU). This wider gap stems from episodes of greater macroeconomic variability. In particular, this widening largely occurred during the mid-1970s and again around 1980. Both episodes correspond to major oil shocks, where downward real wage resistance may have exacerbated the effects of these supply shocks on unemployment.37

43. In 1995, observed unemployment in west Germany (8¼ percent) can be decomposed according to the asymmetric model as follows:38 structural unemployment u* is estimated around 6 percent, the natural rate of unemployment Φ is approximately 7-7½ percent (i.e., α between 1-1½ percent).39 These estimates suggest that three quarters of west German unemployment was structural, while the remaining one quarter was divided nearly equally between the asymmetrical (i.e., gap) component and the cyclical component of unemployment.

44. To view these results from a cross-country perspective, the basic model is also estimated for the remaining G-7 countries plus several of Germany’s European partner countries (Denmark, the Netherlands, Ireland, Spain, and Sweden). The individual country estimates of the wage-price mechanism are shown in Table I-1.40 Based on these estimates, plots of the Phillips curves for several of the major industrial countries, including Germany, are shown in Chart I-5. Among the G-7 countries, Germany and France have the lowest estimates for γ, suggesting a comparatively high degree of downward nominal or real wage rigidity. Real wage rigidity appears more important in the case of Germany based on the estimate of γ.41 In terms of the Chart I-5, this is manifested in a relatively “flat” Phillips curve in the region of excess supply (u > u*); given global convexity (i.e., the Phillips curve turns “steep” in the region of excess demand), this finding suggests a higher degree of local asymmetry (around u*) in the inflation-output process, in the sense that disinflationary shocks have proportionately larger (absolute) unemployment effects.

Table I-1.

(Convex) Phillips Curve Estimates, 1970–95

Model: πt=λπ¯t+(1λ)πt1γ[utut*utφt]+;φt=0.

article image
Note: A *(**) indicates significance at the 5 (10) percent level. Based on data obtained from OECD Analytical Database.

Data are for West Germany.

Data from 1972 to 1995.

Chart I-5.
Chart I-5.

Germany Phillips Curve Estimates, 1995

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

Source: Staff estimates.1/ Estimates refer to West Germany only.

45. As for overall inflation inertia, Germany has a comparatively moderate degree of nominal flexibility, somewhat higher than in the United States (low λ),42 but somewhat below that of Japan or Sweden (high λ). A non-exhaustive list of factors that could help explain cross-country differences in the degree of nominal rigidity would include the degree of indexation, length and staggering structure of nominal contracts, inflation levels and variability, policy credibility, and openness of the economy. As an open economy, with contracts typically lasting around one to two years, and with a strong, independent central bank, Germany fares well on most of these counts.

46. To account for significant upward trends in the unemployment rate in some instances, Table I-2 allows the lower bound for the unemployment rate to drift upward in the estimation.43 The corresponding Phillips curve estimates are shown in Chart I-6. If Φ were anchored at zero (as in Table I-1), high unemployment countries would by construction have fairly “linear” Phillips curve in the region of excess demand (less local asymmetry; larger γ). Allowing Φ to rise with the unemployment rate controls for the impact that location of the Phillips curve has on its curvature. Not surprisingly, the estimates change only for those countries with significant unemployment trends or very high unemployment levels (e.g., Spain and Ireland).

Table I-2.

(Convex) Phillips Curve Estimates, 1970–95

Model: πt=λπ¯t+(1λ)πt1γ[utut*utφt]+;φt=max[0;u¯t8].

article image
Note: A *(**) indicates significance at the 5 (10) percent level. Based on data obtained from OECD Analytical Database.

Data are for West Germany.

Data from 1972 to 1995.

Chart I-6.
Chart I-6.

Germany Phillips Curve Estimates, 1995

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

Source: Staff estimates.1/ Estimates refer to West Germany only.

47. It is interesting to note from Table I-2 that countries like Germany with low γ estimates (e.g., Spain, France, Denmark) are also countries which tend to have significant persistent increases in the level of unemployment over the sample period. In other words, countries which have prominent structural rigidities and asymmetries are also more likely to have experienced trend increases in the average unemployment rate over time. This result is perhaps not surprising considering that the institutional and structural factors that underlie labor market asymmetries are likely to be closely related to the factors underpinning the upward trend in structural unemployment. In the case of Germany, most of the increase in average unemployment appears to be the result of increases in structural unemployment; however, a comparatively larger share is attributable to a rise in the gap between average and structural unemployment for these other countries (see the table below). This finding suggests that increased variability in external shocks, increased asymmetries, and/or less effective stabilization policies over the sample period have featured more prominently in the unemployment increase for those European countries, while for Germany the problem has largely been structural.

Unemployment Developments 1973–95 1/

article image

Entries for u, u* are for end of sample (1995); Δ reflects change between 1973 and 1995.

Data for West Germany.

48. Comparisons among the G-7 also indicate that Germany has been quite successful in its overall economic stabilization, as measured by relatively low output and inflation (but not unemployment) variability and has the lowest mean inflation rate over the sample period.44 However, despite its otherwise solid macroeconomic performance, Germany (along with France) has the highest estimated gap between average and structural unemployment among the G-7, consistent with its higher degree of asymmetry.45 Overall, these findings highlight the importance of labor market flexibility and suggest that Germany’s stabilization record has been achieved more in spite of its labor market structure than because of it.

49. To summarize, estimation of the asymmetric model (convex Phillips curve) performs reasonably well across a fairly broad spectrum of countries. While individual country estimates show a good deal of variation from one another, countries are generally found to have some degree of asymmetry in the wage-price mechanism. In the case of Germany, low point estimates of γ indicate a comparatively high degree of rigidity and local asymmetry (around u*). From a time-series perspective, the rise in German unemployment over the past quarter century largely reflects a rise in the rate of structural unemployment; however, part of the rise can also be attributed to a further rise in average unemployment, reflecting the impact of underlying labor market asymmetries on macroeconomic adjustment in the wake of the oil shocks and German unification. From a policy perspective, the high degree of asymmetry provides ample incentive for the German authorities to pursue effective stabilization policies in order to contain the adverse effects of economic fluctuations on output and employment.

E. Dynamic Simulations and Policy Scenarios

50. To further illustrate the economic and policy implications for Germany of asymmetries in the wage-price mechanism, this section examines the effects of various shocks on output, inflation, and unemployment in the context of MULTIMOD simulations.46 To modify MULTIMOD accordingly, the new estimates for the wage-price mechanisms presented are introduced into the model.47 The revised labor market segment of the model is completed by estimation of a short-run adjustment equation for unemployment, characterizing the short-term behavior of cyclical unemployment.48

51. Based on the country model for Germany, the dynamic impacts of negative and positive demand shocks are shown in Table I-3 to examine the issue of asymmetry. Specifically, the response to a 10 percent increase and a 10 percent decline in the (target) money supply are simulated to illustrate the degree of asymmetry present at Germany’s current level of unemployment.49 In the previous (symmetric) version of MULTIMOD, the impact effects were identical in absolute value; as evident from the table, this is no longer the case. In particular, the negative effects of the contractionary shock on output and the unemployment rate are larger in absolute terms than their counterpart effects associated with the expansionary shock.

Table I-3.

Germany: Asymmetric Effects of a Change in Money Supply

(In percent deviation from baseline, unless noted otherwise)

article image
Source: IMF staff calculations

Percentage point deviations from baseline level.

Inverted scale where an increase denotes an appreciation for the home currency.

Average value for price level, money supply and exchange rate; otherwise cumulative value.

52. Looking at the cumulative deviations over the first five years, the cumulative inflationary effects under the expansionary shock raises the price level to greater extent than the decline associated with the contraction (10.9 versus -8.4 percent). Meanwhile, the cumulative gains in output and the rise in the unemployment rate in the first instance are 15 percent and 4½ percentage points respectively; while for the contractionary episode, the respective short-run losses are higher at 20 percent and 6½ percentage points. Hence, over the cycle the net loss would be 5 percent of GDP with a net increase of 2 percentage points in the unemployment rate. Note that these simulation results are sensitive to initial conditions. Here, they are based on prevailing economic conditions in 1996, which in the case of Germany reflects excess supply (i.e., cyclical unemployment). If the initial conditions were closer to full employment the asymmetric effects of the shocks would become more pronounced.

53. Changes in γ would also affect these calculations. If, for example, γ were increased from Germany’s point estimate of 1.5 to 2.5—close to the point estimates for Ireland and the Netherlands, the degree of asymmetry would be reduced appreciably.50 From the same starting point as before, the demand expansion would result in smaller cumulative gains in the unemployment rate (3½ percentage points), while the demand contraction would result in smaller cumulative losses (4¾ percentage points).51 The net difference in the unemployment rate (at 1¼ percentage points) would also be smaller than previously. In a stochastic environment, these results suggest that unemployment variability decreases with a higher γ (i.e., with a steeper and more symmetric Phillips curve), while the variance effect on average unemployment is smaller.52

54. Overall, these results highlight the fact that countries with significant asymmetries (low γ) in the wage-price mechanism will operate at a level of unemployment higher than in the absence of shocks or these asymmetric effects. A lasting increase in the variance of shocks would ratchet up the natural rate of unemployment rate, whereas episodes of high variability would tend to raise unemployment (on average) on a persistent rather than permanent basis. Conversely, countries with more symmetric adjustment will be less vulnerable to the variance effects of external disturbances on the average level of output and employment.

55. In terms of policy, structural reforms which address the underlying rigidities in labor and product markets will affect unemployment through two broad channels. There will be an indirect effect stemming from reduced asymmetry which would lower the gap between average and structural unemployment, reflecting the economy’s increased ability to respond flexibly to shocks; in addition, there will likely be a direct effect on the level of structural unemployment itself.53 The tabulation below illustrates the long-run effects of structural reforms in MULTIMOD when the Phillips curve is made more symmetric in Germany along the lines discussed above (i.e., raising γ to 2.5 and λ to 0.6).54 The long-run decline in the natural unemployment rate essentially reflects the decline in the gap (1 percentage point) attained through greater market flexibility, as well as the direct impact on structural unemployment (about 1 percentage point).55 56

Germany: Structural Reforms and Reduced Asymmetries

(In percent deviation from baseline, unless noted otherwise)

article image
Source: IMF staff calculations.

Percentage point deviation from baseline level.

F. Multi-Country Simulations

56. Applying the same basic framework to a multi-country context, one can examine the implications of asymmetries in the wage-price mechanism on macroeconomic adjustment under alternative policy regimes. In particular, the effects of various shocks on Germany are considered under ERM and a broad EMU, as well as the comparative country responses within a given policy regime.

57. Under the ERM regime, it is assumed that German monetary policy follows an inflation targeting rule geared toward domestic inflation (and output) objectives,57 while its ERM partner countries adjust domestic interest rates to maintain their bilateral exchange rates vis-à-vis the deutsche mark within a given exchange rate band. For Austria, Belgium, Italy, France and the Netherlands, the bands are “tight” (2¼ percent width); for Denmark, Finland, Ireland, Portugal and Spain, the bands are “loose” (15 percent).58 Under the EMU regime, which consists of all the ERM countries plus Sweden, the United Kingdom, and Greece,59 monetary policy is centralized, identical for all members, and follows a comparable inflation targeting rule but now geared toward EMU-wide indicators.

58. Table I-4 displays the effects of a country-specific demand shock on Germany on its major macroeconomic variables under each policy regime. Specifically, German investment and consumption levels are assumed to decline initially, reflecting an exogenous loss in private sector confidence; consumption and investment decline by about 2 percent and 6 percent (from baseline) in the first year, with some declining residual effect in subsequent years. The dynamic and long-run effects—expressed as deviations from baseline—of this “business-cycle” type shock are shown in the table.60

Table I-4.

Germany: Comparative Effects of a Demand Disturbance under Alternative Policy Regimes 1/

(In percent deviation from baseline, unless noted otherwise)

article image
Source: IMF staff calculations

Under inflation targeting based on either national or EU aggregates.

Percentage point deviations from baseline level.

Inverted scale where an increase denotes an appreciation for the home currency.

59. With monetary policy focused on EU-wide rather than national objectives,61 the adverse effects of the shock are more pronounced for Germany under EMU than under ERM; thus output would be lower (by about 0.3 percentage point of GDP) and the unemployment rate would be higher (by about 0.1 percentage point) in the first year. A potential increase in economic variability in Germany associated with country-specific shocks under an EU-wide (rather than German) monetary policy would expose the economy more to the effects of these asymmetries than in the past. This would tend to raise the NAIRU over time in the absence of further structural reforms. An increase in unemployment variability of one-third to two-thirds, for example, could over time raise the average unemployment rate in Germany (given its structural rate) by ½ to 1 percentage point, other things being equal.

60. An important caveat should be noted here: other things may not be equal across the two policy regimes. To the extent that a change in regimes affects the distribution of shocks, risk premia, etc., the level outcomes under each scenario may be somewhat different. The present analysis of shocks under EMU versus ERM neglects these potential differences in the baseline or control solution which might emerge and focuses instead on the deviations from each respective baseline. Hence, implications regarding the absolute welfare implications across the two regimes are difficult to draw. However, to the extent that the benefits of (say) monetary union accrue to all its member countries somewhat evenly, country differences across regimes concerning the response to shocks (as deviations from baseline) can be illustrative of the relative economic implications across countries.

61. In this regard, the differential effects of country-specific shocks under these alternative policy regimes may be less important for other countries. As ERM participants have to some extent already relinquished an independent monetary policy—particularly those operating under tight bands, monetary union with a monetary policy that seeks to stabilize area-wide economic conditions could represent a beneficial regime shift. For those countries, the effects of adverse country-specific shocks may be broadly similar, or even diminished, under EMU compared with ERM. Simulations of the same country-specific shocks in MULTIMOD (not reported) in the case of France, for example, support this proposition, showing quantitatively similar output and employment effects under the two policy regimes.62 Consequently, the removal of Germany’s asymmetric ability to conduct monetary policy according to national objectives may entail a rise in unemployment relative to many of its EMU partner countries when each are subject to country-specific shocks.

62. For a given policy regime, the current analysis also suggests that countries with different institutional and structural features and varying degrees of asymmetries may respond to common shocks differently. For example, a uniform increase in the volatility of external disturbances would have differential implications for unemployment across countries owing to substantial differences in their wage-price mechanisms. Although the comparative effects would reflect a variety of factors (e.g., openness, initial conditions, etc.) in addition to the shape of the Phillips curve, countries such as Germany, which are hampered by more severe rigidities and asymmetries, would tend to experience a larger rise in unemployment variability and, consequently, in the natural rate of unemployment. In the case of Germany, for example, an increase in the variability of shocks sufficient to raise average unemployment by 1 percentage point given present asymmetries (γ=1.5, λ=0.45) would lead to an increase of only ½ percentage point if its labor markets were more flexible (γ=2.5, A=0.6).63

63. These comparative differences in average unemployment effects are compounded when country differences in longer-term trends in structural unemployment are also considered. Extrapolating from past trends, unemployment levels in west Germany would further rise by l½ percentage points after five years, compared to smaller increases for Denmark (1 percentage point) and France (½ percentage point), unchanged longer-term unemployment in the Netherlands, and a trend decline for Ireland (½ percentage point).

64. From a policy standpoint, countries with rigid labor markets are confronted with challenging fiscal implications. With higher average unemployment through less flexible labor market adjustment or through continuing rises in structural unemployment, countries like Germany potentially face larger burdens on public finances through larger social transfers. Under the conditions of the Stability Pact, these higher fiscal burdens would leave countries with more rigid labor markets with less room to maneuver and could give rise to difficult policy dilemmas under EMU in the absence of further structural reforms. All in all, these simulations reinforce the proposition that entering a currency union places greater stress on reforming labor and product markets.

ANNEX: Illustrative Model

65. A very simple derivation of a non-linear Phillips curve is presented to motivate the empirics as well as to provide some further insight into the possible features of labor (and goods) markets which may underlie the implied nexus between wages, prices and employment. To proceed, we borrow from the framework discussed in Layard, Nickell, and Jackman (1991) and Clark and Laxton (1997). Specifically, we assume that price-setting and wage-setting behavior can be characterized respectively as follows:

p = w + δ o δ 1 u , ( A 1 )
w = p e + φ 0 φ 1 u . ( A 2 )

Equation (A1) specifies that (log) prices p are set as a (constant) mark-up over unit labor costs, expressed in terms of (log) wages w and the rate of unemployment u, which can also be related to output or capacity utilization via Okun’s law. Equation (A2) represents a target real wage expression in which (φ1 signifies the responsiveness of real wage demands to level of unemployment, and pe is the expected price level.

66. In the presence of nominal inertia, the observed price is assumed to only gradually adjust to the target price (as described in equation (A1); denoted now with a bar):

Δ p = λ 1 ( p ¯ p 1 ) + λ 2 Δ p 1 . ( A 3 )

Note that the first term on the right hand side represents an error-correction mechanism, while the second term introduces (higher-order) inertia in inflation (π=Δp) and not just the price level.64

67. Using these three equations, one can show that in expectational equilibrium (p=pe), with non-accelerating prices (Δπ=0), the equilibrium unemployment rate or NAIRU is given by:

u * = φ 0 + δ 0 φ 1 + δ 1 . ( A 4 )

Away from the NAIRU, a linear expectations-augmented Phillips curve summarizes the (reduced-form) relationship between inflation and unemployment:

π = λ 1 π e + ( 1 λ 1 ) π 1 λ 1 ( φ 1 + δ 1 ) ( u u * ) , ( A 5 )

where expected inflation is defined by: πe = pe − p-1. Note that in equilibrium, π-1 = π = πe in conjunction with an unemployment rate at its NAIRU level u* shown in (A4). In equation (A5), note that the tradeoff coefficient on the unemployment gap reflects the degree of nominal rigidity which depends on λ1, and—following the discussion in Layard, Nickell, and Jackman (1991)—the degree of real wage rigidity RWR = (φ1 + δ1)-1 implied by the price- and wage-setting equations. The interaction between these two considerations determines the slope of the Phillips curve or the constant short-run trade-off between inflation and unemployment. For example, greater nominal flexibility (largerλ1) implies a steeper linear Phillips curve.

68. However, if this coefficient is no longer fixed and independent of labor market conditions, non-linearities in the output-inflation process are introduced. For example, if we posit that the degree of real wage rigidity is not constant but is instead a function of the level of unemployment: RWR = h1>0, the Phillips curve would exhibit convexity. In the case where this functional form is linear: h(u) = >u − Ω, we can rewrite the Phillips curve as follows:

π = λ 1 π e + ( 1 λ 1 ) π 1 γ ( u u * u Ω / Λ ) ; γ λ 1 / Λ . A 6

69. Note that from the above equation that an increase in nominal flexibility (increase in λ1) or an increase in real wage flexibility (smaller Λ) raises the coefficient λ in the above expression. This equation is basically the specification in equation (3) in the case of inflation inertia.65 The motivation for the assumption on h(.) can be found in the implications of several labor market models. In many theories of unemployment, the degree of real wage resistance (and the wage gap) increases in the face of higher unemployment and market pressures for a wage decline.

70. For example, in incentive wage models, employers find it desirable to pay efficiency wages greater than the market-clearing wage in order to induce effort, sustain morale, reduce turnover, avoid adverse selection problems, etc., which places an effective floor (i.e., asymmetry) on adjustment in real wages regardless of the level of unemployment. The effect is that the degree of real wage rigidity and wage gap would increase with the unemployment rate. Asymmetric wage bargaining could also generate a similar implication. In effect, once market rigidities respond to the level of activity and employment, we can no longer expect the wage-price mechanism to be linear.

71. In the presence of convexity in the Phillips curve, u* no longer represents the natural rate of unemployment that would obtain in (stochastic) equilibrium. Instead, the economy would operate, on average, at a level of unemployment somewhat above the underlying NAIRU u* associated with a deterministic equilibrium. Around u*, disinflationary surprises lead to larger increases in unemployment than the decreases arising from symmetric inflationary shocks. Consequently, the extent to which the economy operates above the deterministic equilibrium level of unemployment depends on the variance of shocks, the effectiveness of stabilization policies, and the degree of convexity in the wage-price mechanism.

Table I-A1.

Summary Statistics 1970–95 1/

article image

Inflation π expressed in terms of the GDP deflator, y is log GDP; all variables expressed in percentage points.

Data for West Germany; output series from 1970 to 1990 (period prior to unification).

Table I-A2.

Alternative (Convex) Phillips Curve Estimates

Model: πt=λπ¯t+(1+λ)πt1γ[utut*utφt];0φ<u*

article image
Note: A *(**) indicates significance at the 5 (10) percent level. Model (1):Φt = 0; Model (2): Φt = u*t – 5 percent; Model (3):Φt = ut* – 2 percent; Model (4):Φt = min {ut}/2 For comparability, estimates using WEO and OECD data cover the sample period prior to German unification. The non-oil GNP price deflator is taken as the price measure in the WEO data; otherwise, inflation is in terms of the GDP deflator.
Chart I-A1.
Chart I-A1.

Germany 1/ Alternative Convex Models

Citation: IMF Staff Country Reports 1997, 101; 10.5089/9781451810318.002.A001

Source: Staff estimates.1/ Estimates refer to West Germany only.
1

Prepared by Hamid Faruqee.

2

See Dennis Snower, 1995, “Evaluating Unemployment Policies: What Do the Underlying Theories Tell Us?” Staff Studies for the World Economic Outlook, IMF for an evaluation of unemployment policies under alternative theories.

3

The unemployment rate for unified Germany stood at around 10V4 percent in 1996, owing to the very high unemployment rate in the new Länder. To maintain a comparable time series, only unemployment in west Germany is considered.

4

A more lengthy description of these models and associated empirical evidence can be found in SM/96/227.

5

The dynamic properties of labor markets as a source of the European unemployment problem is examined in S.G.B. Henry, and Dennis J. Snower, eds., 1996, Economic Policies and Unemployment Dynamics in Europe (Washington: International Monetary Fund).

6

The staff examined German labor markets from this perspective in 1995 (SM/95/202; 8/16/95).

7

In a pair of studies, Ansgar Belke, 1996, “Testing for Unit Roots in West German and U.S. Unemployment Rates: Do ‘Great Crashes’ Cause Trend Breaks?” Konjunkturpolitic, Vol. 42, and A. Belke and Münster Göcke, 1996 “Cointegration and Structural Breaks in German Unemployment: An Error-Correction Interpretation” Jahrbücher f. Nationalökonomie u. Statistik, Vol. 216, attribute persistence in the unemployment rate to structural breaks caused by oil shocks and unification, rather than a degenerative adjustment process (i.e., hysteresis). Their findings point toward a stable long-run unemployment trend or relationship once structural shifts are accounted for in the case of west Germany.

8

Some have argued that the United States in its current expansion represents a possible case where there has been downward drift in the natural rate of unemployment.

9

For a general description and associated literature see Peter Clark and Douglas Laxton, 1997, “Phillips Curves, Phillips Lines and the Unemployment Costs of Overheating,” IMF Working Paper, WP/97/17 (February).

10

A.W. Phillips, 1958, “The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957,” Economica, Vol. 25.

11

A traditional Keynesian explanation for downward nominal rigidity in wages focused on money illusion as an explanation for workers’ reluctance to accept nominal pay cuts. See Eldar Shafir, Peter Diamond, and Amos Tversky, 1997, “Money Illusion,” Quarterly Journal of Economics, Vol. 112 (May), for a recent paper on the prevalence of money illusion.

12

Another explanation for asymmetric price adjustment among imperfectly competitive firms is focal-point pricing; see Severin Borenstein, A. Colin Cameron, and Richard Gilbert, 1997, “Do Gasoline Prices Respond Asymmetrically to Crude Oil Price Changes?” The Quarterly Journal of Economics, Vol. 112 (February), for a discussion and evidence. Laurence Ball, and N. Gregory Mankiw, 1994, “Asymmetric Price Adjustment and Economic Fluctuations,” Economic Journal, Vol. 104 (March), motivate asymmetric price adjustment under imperfect competition in the presence of menu costs and trend inflation.

13

Overall union membership is around 40 percent; however, about 80 percent of employers in the manufacturing and financial sectors belong to federations whose members are obligated to pay their workers (who constitute 90 percent of the sectoral workforce) at or above the wage rates negotiated through collective bargaining. For a detailed description of labor market institutions and wage developments in Germany see Robert Corker, et al, 1995, “United Germany: The First Five Years—Performance and Policy Issues,” IMF Occasional Paper 125, and SM/96/227.

14

Lars Calmfors, and John Driffill, 1988, “Bargaining Structures, Corporatism, and Macroeconomic Performance,” Economic Policy, Vol 6, have argued that sector level bargaining may result in less efficient wage-employment outcomes, lacking the “coordination” benefits of more centralized systems and the “competitive” benefits of more decentralized systems. Wendy Carlin and David Soskice, 1997, “Shocks to the System: The German Political Economy Under Stress,” National Institute Economic Review, No. 159 (January), argue that a breakdown in consensus among the social partners following unification—and the sharing of its costs—underlies the lack of wage moderation in recent years and the subsequent labor retrenchment.

15

David Grubb, and William Wells, 1993, “Employment Regulations and Patterns of Work in EC Countries,” OECD Economic Studies, Vol. 21, rank Germany in terms of strictness of employment protection behind only Portugal, Spain, Italy, and Greece, while Kees Koedijk, and Jereon Kremers, 1996, “Market Opening, Regulation and Growth in Europe,” Economic Policy (October), find labor and product market regulation in Germany to be among the highest in Europe—only Italy and Greece are rated higher.

16

In January 1996, the Government adopted a 50-point “Action Program to Foster Investment and Employment.” Under the program, proposed measures cover a wide range of policy areas, including deregulation of goods and labor markets and reforms of work-related benefits.

17

The original work on the statistical relationship is attributed to Irving Fisher, 1926, Journal of Political Economy, Vol. 81, and, 1973, “A Statistical Relation Between Unemployment and Price Changes,” International Labor Review, Vol. 13., who also suggested that the causality ran from inflation to unemployment.

18

The new classical approach also gave equation (1)—appropriately inverted—a market-clearing interpretation under rational expectations and flexible prices. And while the Keynesian approach has largely adopted the rational expectations hypothesis, the presence of inflation inertia was used to retain a non-market clearing interpretation. See Alan Blinder, 1989, Macroeconomics Under Debate (Ann Arbor: University of Michigan Press), for a general review.

19

In the MULTIMOD simulations that follow, instead of considering shocks to either the unemployment gap or unanticipated inflation, “exogenous” disturbances (e.g., changes in target money supply) are considered and the endogenous responses of both the unemployment gap and inflationary expectations are examined.

20

The new classical framework also posited that the slope of the short-run Phillips curve was not stable and would react to changes in the macroeconomic and policy environment, leaving no systematic tradeoff even in the short run. See Robert Lucas, 1976, “Econometric Policy Evaluation: A Critique,” in The Phillips Curve and Labor Markets, edited by K. Brunner and A.H. Meltzer (Amsterdam: North Holland). In the Keynesian framework, to the extent that the nature of price and wage setting behavior changes (endogenous nominal rigidities), significant changes in macroeconomic environment (e.g., high or variable inflation) would also affect this tradeoff parameter. See Laurence Ball, N.Gregory Mankiw, and David Romer, 1988, “The New Keynesian Economics and the Output-Inflation Tradeoff,” Brookings Papers on Economic Activity.

21

As shown in the appendix, in the case of inflation persistence, equation (2) would be modified as follows: π = λπe + (1 − λ)π-1 − γ(u − u*), where the coefficient γ reflects both nominal and real rigidities.

22

Note that Φ must lie between zero and u* to ensure a positive, equilibrium rate of unemployment.

23

Apart from its level, another specification issue for Φ concerns time dependence. In countries where the unemployment rate has significantly drifted upward, Φ could increase as well. If, for example, the trend increase brought about a substantial increase in the numbers of long-term unemployed with depreciated human capital, it may prove difficult to entirely reabsorb these workers back into the workforce.

24

See for example Albert Jaeger, and Martin Parkinson, 1994, “Some Evidence on Hysteresis in Unemployment Rates,” European Economic Review, Vol. 38., for an empirical analysis.

25

This basic shortcoming led Lawrence Summers, 1988, “Should Keynesians Dispense with the Phillips Curve?” in Unemployment, Hysteresis, and the Natural Rate Hypothesis, ed. Rod Cross, Basil Blackwell, to question whether the traditional natural rate model was useful.

26

See appendix for an explicit derivation.

27

Guy Debelle, and Douglas Laxton, 1996, “Is the Phillips Curve Really a Curve? Evidence for Canada,” IMF Working Paper, WP/96/111.

28

Data are for West Germany from 1962 to 1995. Other country data cover the period from 1970 to 1995. Alternative price measures such as the consumer price index and absorption deflator were also used and do not qualitatively affect the results. Using the OECD standardized measure for unemployment also yields similar implications.

29

Ken Kuttner, 1994, “Estimating Potential Output as a Latent Variable,” Journal of Business and Statistics, Vol. 12. Note that Φ is statistically unidentified in the empirical model. However, theory places some structure on the feasible parameter space: Φt < ut and 0 ≤ Φt < u*t for all t. Consequently, various (static and time-varying) rules for (j) are employed to test for robustness of the estimates. See appendix Table I-A2.

30

equation (4) represents the measurement equation, and for a given Φ, the coefficient βt on (ut − Φt-1 is allowed to be time varying; hence, the transition equation is given by: βt = βt-1 + vt. The estimated coefficient γ on ut(ut − Φt)-1 in equation (4) then identifies ut* = β/γ. See Guy Debelle, and Douglas Laxton, 1996, “Is the Phillips Curves Really a Curve? Evidence for Canada,” IMF Working Paper, WP/96/111 (October), for further details of the estimation methodology.

31

Using Φ fixed at 0 in the case of Germany where u* is likely to rise over time tends to make the Phillips curve act more symmetric by construction in the region of excess demand (i.e., where u is above u*). Alternative models which better preserve the convexity are considered shortly.

32

Laurence Ball, N. Gregory Mankiw, and David Romer, 1988, “The New Keynesian Economics and the Output-Inflation Tradeoff,” Brookings Papers on Economic Activity, and Laurence Ball, 1994, “What Determines the Sacrifice Ratio,” in Monetary Policy, ed. by N. Gregory Mankiw (Chicago: Chicago University Press), finds similar results for Germany.

33

Allowing a strictly positive Φ based on (say) the minimum unemployment rate would further improve the fit of the asymmetric model. Estimates using alternative rules for Φ can be found in the appendix Table I-A2.

34

The implicit variability of innovations to u* in the transition equation is restricted a priori to be the same across the two models for comparability. Allowing greater variability in u* in the estimation procedure would allow either model to fit the inflation data better.

35

The estimates shown in this chart are based on a time-varying u* similar to the estimates presented in the appendix Table I-A2. This specification yields higher estimates for u* than with Φ=0 as shown in Chart I-3. The variance restriction in the transition equation is such that innovations in Δu* are allowed to be as variable a priori as actual Δu over the sample period; see Table I-A1.

36

Actual unemployment u also lies above u* except for 1969–1971. It is interesting to note that, during this period, vacancies rose markedly relative to unemployment, consistent with excess demand pressures; inflation also increased sharply. The symmetric model shows a similar result, albeit with a larger increase in u* (i.e., larger unemployment gap) necessary to explain the corresponding pick-up inflation.

37

Ansgar Belke, 1996, “Testing for Unit Roots in West German and U.S. Unemployment Rates: Do ‘Great Crashes’ Cause Trend Breaks?” Konjunkturpolitic, Vol. 42, and A. Belke and Münster Göcke, 1996 “Cointegration and Structural Breaks in German Unemployment: An Error-Correction Interpretation” Jahrbücher f. Nationalökonomie u. Statistik, Vol. 216, also identifies these important events in the German unemployment series, but using structural breaks rather than through asymmetries in the wage price mechanism (and α shifts).

38

The annual unemployment rate for unified Germany in 1995 at 10½ percent was higher than in west Germany, owing to much higher unemployment in the east which stood around 14 percent in 1995. This latter figure has been viewed as largely structural, suggesting a overall natural rate for unified Germany of 8 to 8¾ percent; OECD estimates place this latter figure at around 9 percent.

39

Using a second-order Taylor expansion of the convex function, one can obtain an approximate solution (conditional on u*) for the gap a between the natural and structural unemployment rates as a function of u*, Φ and the variability of u about u* (which depends on γ). Estimates of these parameters and the variance over the entire sample yield an approximate solution for α of around 1 percentage point or a ū estimate of 7 percent at the end of the sample; variability in the latter part of the sample was however somewhat higher; filtered unemployment estimates yield an estimate of 7½ percent at the end of the sample.

40

For (Φ=0. The case of a time-varying (Φ is presented in Table I-2 and Chart I-6. Guy Debelle, and Douglas Laxton, 1996, “Is the Phillips Curve Really a Curve? Evidence for Canada,” IMF Working Paper, WP/96/111 (October), report similar estimates to those in Table I-1 using quarterly data for the United States and United Kingdom.

41

An illustrative derivation of γ and its relation to real and nominal rigidities is shown in the appendix.

42

The finding of significant nominal rigidity or inflation persistence in the United States is well documented. See for example Robert Gordon, 1997, “The Time-Varying NAIRU and Its Implications for Economic Policy,” Journal of Economic Perspectives, Vol. 11.

43

The conservative rule which is used allows Φ (non-negative) to rise above zero as the average unemployment rate rises above 8 percent. Other time-varying rules are shown in appendix Table I-A2 for select countries.

44

Summary statistics for unemployment, inflation and output over the sample period are shown in Table I-A1.

45

For Germany and France, the estimates for u* in Chart I-6 are around 6 percent and 5½ percent, respectively, while average (filtered) unemployment stood at 8 percent and 10 percent respectively in 1995. The gap α between u* and ū for the United States, Japan, and Canada were found to be less than ½ percent.

46

A complete description of MULTIMOD can be found in Paul Masson, Steven Symansky, and Guy Meredith, 1990, “MULTIMOD Mark II: A Revised and Extended Model,” IMF Occasional Paper 71.

47

For Germany, the particular estimates of the wage-price mechanism which are used are based on model (2) in Table I-A2 which covers the longer sample period 1962–95; for other countries, the estimates are those taken from Table I-2 for the sample period 1970–95.

48

Following Leonardo Bartolini, and Steven Symansky, 1993, “Unemployment and Wage Dynamics in MULTIMOD,” Staff Studies for World Economic Outlook, the unemployment gap was expressed as a function of its own lag and the output gap (i.e., Okun’s law). Unemployment and output gaps were obtained through filtering. Coefficient estimates on the lagged unemployment gap were between 0.2 and 0.7, and point estimates on the output gap typically ranged from 0.25 to 0.4 in absolute terms; these estimates were generally found significant at the 1 percent level.

49

Here, the Bundesbank is assumed to follow a money targeting rule as described in Paul Masson, Steven Symansky, and Guy Meredith, 1990, “MULTIMOD Mark II: A Revised and Extended Model,” IMF Occasional Paper 71.

50

If λ were also increased from 0.45 to 0.6, back of the envelope calculations using the model in the appendix translate these parameter changes into a one-quarter decrease in the degree of both real wage and nominal rigidity.

51

A higher value for γ acts to “tilt” the Phillips curve, making it more vertical in both excess demand and supply regions; hence, smaller unemployment effects of the monetary shock obtain in both directions. In the limit, as γ → ∞, the Phillips curve becomes a vertical line at u*.

52

Stochastic simulations (i.e., repeated trials with random shocks) in MULTIMOD (not reported) were used to verify the effects of different degrees of asymmetry (γ) on unemployment variability and the size of the variance effects on average unemployment.

53

This latter channel is not modeled in this reduced-form framework, although many examples of such analysis have been conducted, usually through explicit consideration of wage and price setting equations. See Richard Layard, Stephen Nickell, and Richard Jackman, 1991, Unemployment: Macroeconomic Performance and the Labor Market, Oxford University Press, for an overview; see Leslie Lipschitz, and Donough McDonald, eds., 1990, German Unification: Economic Issues, IMF Occasional Paper No. 75 (Washington: International Monetary Fund), for analysis regarding Germany. See Olivier Blanchard and Lawrence Katz, 1997, “What We Know and Do Not Know About the Natural Rate of Unemployment,” Journal of Economic Perspectives, Vol. 11 for a critical review of this literature.

54

In above tabulation, Φ also declines. Changes in Φ alone largely affect the shape of the Phillips curve in the excess demand region (e.g., Chart I-A1). Changes in both 4Φ and γ can allow the adjusted Phillips curve to also lie uniformly below or above the unadjusted curve (tangent at u*) by changing the degree of curvature in both the excess demand and supply regions.

55

The indirect (variance) effect is approximated in the deterministic simulations analytically. Using a second-order Taylor expansion of the convex function, one can obtain—conditional on u*—an approximate solution for the gap α between the natural and structural rates of unemployment as a function of (Φ, u* and the conditional variability σ2 of u:

α E [ u u * ] = E [ ( u u * ) 2 ] ( u * φ ) or α = ( u * φ ) / 2 { ( u * φ ) / 2 } 2 σ 2 ;

where E[.] denotes a conditional expectation.

56

Using equation (A4) in the appendix, back of the envelope calculations show that an decline in real wage rigidity by 20 percent would lead to a proportional decline in u*.

57

The specification follows a forward-looking Taylor rule with conventional weights on inflation and economic activity. See for example Paul Masson, and Bart Turtelboom, “Characteristics of the Euro, the Demand for Reserves, and Policy Coordination Under EMU,” IMF mimeo. See Richard Clarida, Jordi Gali, and Mark Gertler, 1997, “Has the Bundesbank Followed A Taylor Rule?” Columbia and New York University, mimeo, for empirical evidence on the Bundesbank’s conduct of monetary policy in terms of the Taylor rule. Results are qualitatively similar under money targeting.

57

See Paul Masson, and Bart Turtelboom, 1997, “Characteristics of the Euro, the Demand for Reserves, and Policy Coordination Under EMU,” IMF Working Paper, WP/97/58, for details of the exact policy rules under both regimes a further description of the MULTEU version of MULTIMOD, and treatment of the risk premia under ERM and EMU.

59

All members of the European Union have been included in EMU.

60

The analysis of shocks under EMU versus ERM neglects the potential differences in the baseline or control solution which might emerge under the two policy regimes. Instead, the analysis focuses on the deviations from each respective baseline.

61

Richard Clarida, Jordi Gali, and Mark Gertler, 1997, “Has the Bundesbank Followed A Taylor Rule?” Columbia and New York University, mimeo, argue that German monetary policy has pursued both domestic inflation and output objectives over the past twenty years, largely through managing short-term interest rates (following a Taylor rule) in a manner broadly similar to that in the United States; comparative stabilization performances can be seen in Table I-A1.

62

Additional simulations also show that further structural reforms and more flexible labor markets in France would further improve these outcomes under either regime, highlighting the benefits of labor market reforms in countries other than Germany.

63

Analytically, the change in average unemployment as a function of the change in unemployment variability can be approximated as follows: dα = [(u* − Φ - 2α]-1dσ2, where the change in σ2 depends on γ. Stochastic simulations in MULTIMOD were used to quantify the effects of different degrees of asymmetry (γ) on unemployment variability.

64

We also impose that λ1 + λ2 = 1 in equation (A3) which translates it into an error-correction equation for inflation rather than the price level: Δπ=λ1(π¯π1); (this implies inflation persistence, given by 1 - λ1 and yields a non-zero (steady-state) equilibrium inflation rate where π1=π=π¯.

65

Note that in this very simple model that the lower bound on the unemployment rate Φ can be expressed solely as a function of the parameters Ω, Λ characterizing real wage rigidity, but perhaps unrealistically so, as a decrease in real wage rigidity raises this lower bound.

  • Collapse
  • Expand
Germany: Selected Issues
Author:
International Monetary Fund