Monetary and Exchange Rate Policies of the Euro Area: Selected Issues

This Selected Issues paper estimates the potential output and the associated nonaccelerating inflation rate of unemployment in the euro area. The study presents a conceptual framework for analyzing currency movements, and highlights the transmission of import price shocks on consumer prices. The paper compares different measures of trend money growth, and analyzes the monetary conditions. The study describes the stability and growth pact, outlines a simple framework for studying fiscal policy behavior, and estimates European Union countries' past cyclical fiscal policy responses to output growth fluctuations.


This Selected Issues paper estimates the potential output and the associated nonaccelerating inflation rate of unemployment in the euro area. The study presents a conceptual framework for analyzing currency movements, and highlights the transmission of import price shocks on consumer prices. The paper compares different measures of trend money growth, and analyzes the monetary conditions. The study describes the stability and growth pact, outlines a simple framework for studying fiscal policy behavior, and estimates European Union countries' past cyclical fiscal policy responses to output growth fluctuations.

I. Estimating Potential Output and the NAIRU for the Euro Area1

A. Introduction

1. The identification of the euro-area’s potential output and the associated non-accelerating inflation rate of unemployment (NAIRU) is an important empirical challenge. Knowledge of these unobservable variables allows policymakers to gauge the amount of existing inflationary pressure through the calculation of output and unemployment gaps. Modeling the behavior of the NAIRU in particular has taken on added significance as continued wage moderation in the face of rapidly falling unemployment and external terms-of-trade shocks has fueled speculation that recent labor market reforms and the start of EMU may have altered the wage formation process. This is an important issue for the ECB in its second pillar assessments of medium-term price pressures.

2. Modeling these concepts for the euro area, however, can be somewhat problematic. First, there is the question of data aggregation. To date, most work in this area has taken a bottom-up approach whereby potential output and NAIRU are estimated for individual countries that comprise the euro area and then added up to create an area-wide aggregate. A second issue is the correct econometric method and specification used to estimate and evaluate these unobservable variables. Recent attempts in an euro-area context have tended to focus on unobservable component-type (UC) models proposed by Apel and Jansson that consist of a system of equations organized around a standard Phillips curve framework. These models allow joint estimation of potential output and the NAIRU, while exploiting the mutual dependence inherent between output and unemployment. They also allow easy calculation of level and speed limit effects present in the economy. An additional difficulty has been the empirical specification of the unobservable trend components. A variety of authors have specified these components as random walks with stochastic trends (Mendez and Palenzuela (2001) and Fabiani and Mestre (2001)).2

3. The purpose of this chapter is to estimate the output gap and NAIRU in the euro area using area-wide aggregate data. After testing the integration properties of the data, we apply the basic Apel-Jansson UC systems model, which assumes potential output and the NAIRU follow random walks. The results suggest that although the implied output gap has closed considerably in recent years, by end 2000 a slight margin of slack still remained in the euro area. Moreover, the evolution of these estimates appears to be generally consistent with those found under a bottom-up approach used by many international organizations. Interestingly, the empirical estimates also indicate a substantial decline in the NAIRU since 1996, supporting the view that wage moderation has taken hold. In addition, the analysis highlights the importance of correctly specifying supply shocks in a Philips curve framework and demonstrates that output gaps from a UC model can be a useful tool in forecasting inflation.

4. Section B describes the specifics of the UC model, while Section C discusses the data and integration issues. Sections D presents the results of the UC model and Section E discusses the outcome of a simple inflation forecasting exercise. Finally, Section F presents some conclusions and policy implications.

B. The Unobservable Components Model

5. In this section we lay out the Apel-Jansson reduced form model used in this study, which includes the following equations:


where πt is the log difference of the CPI, ut the unemployment rate, utn the NAIRU, z, a vector of exogenous supply shock variables, yf the log of real GDP, and ytp the log of potential output. All innovations in the system (εtpc,εtol,εtn,εtp,εtc) are assumed to be independent and identically distributed, as well as mutually uncorrelated, with zero means and constant variances. Note that values of a provide estimates of the growth rate of potential output.

6. The Apel-Jansson approach3 allows joint estimation of the unobservable components—potential GDP and the NAIRU-as latent stochastic trends within a trivariate system of observables comprising information on unemployment, real GDP, and the change in inflation.4 Identification of the system is achieved through Phillips Curve and Okun’s Law relations (equations 1 and 2 above).

7. The introduction of the Phillips curve mechanism as an identifying restriction within the unobservable component system forces the estimated position of the NAIRU to depend on the actual inflation rate. Similarly, the Okun equation ensures that potential output and the NAIRU estimates will mirror each other. Thus the model takes into consideration the mutual dependency of output and unemployment through explicit co-variation restrictions on cyclical output and cyclical unemployment. In addition, the model includes as vector of supply side variables in the Phillips curve equation5. With the inclusion of supply shock variables the estimated NAIRU is that unemployment rate which is consistent with constant inflation in the absence of supply shocks.

8. The specification of the system’s (atheoretical) trend-cycle block (equations 3-5) follows established standards for decompositions in unobservable component models of this type.6 In the model, the NAIRU is assumed to follow a random walk, while potential output is assumed to follow a random walk with drift. Other trend specifications for the main unobservable components (equations 3 and 4), such as a random walk with stochastic drift would require slight modifications in the transition equations. Finally, to close the model, the evolution of cyclical unemployment is assumed to follow an autoregressive process.

9. For purposes of estimation, the model is rewritten in state space form and the unknown parameters of the model and the time series of the UCs are found through application of the Kalman filter and maximum likelihood estimation. Specifically, a sequence of optimal predictions of the observable variables for a given set of coefficients and a sequence of unobservable variables is found through the recursive Kalman filter algorithm. Forecast errors of the observables are calculated and inserted into the maximum likelihood routine to compute optimal parameters and corresponding estimates for the NAIRU and potential output. The maximization of the log likelihood function is achieved by minimizing the sum of these forecast errors.

10. Theory dictates that the sum of the coefficients on cyclical unemployment in both the Phillips curve and Okun’s law equations should be negative. In addition, the use of contemporaneous and lagged cyclical unemployment in the Phillips curve indicates that the unemployment gap may affect inflation through both level and change effects. This is revealed by rewriting the contribution of cyclical unemployment in the Phillips curve with one lag:




11. In the specification (6)‒(7), level effects are captured by the sum of the coefficients (η01) while “speed limit” effects are captured by the individual coefficient (η0). As with the level effects, we would expect negative coefficient values, i.e. an unemployment rate above the natural rate should reduce inflationary pressures.

C. The Data

12. Quarterly aggregate euro area data on real GDP, prices, unemployment, and four specific supply shock variables over the 1973.1 to 2000.4 period have been constructed.7 The four supply shocks are quarterly changes in relative real exchange rates, real oil prices, real import prices, and productivity.

13. Given the importance of the assumptions made about the non-stationary behavior of the unobservable components, the degree of integration in the unemployment rate and output series was tested. Real GDP is clearly an 1(1) series, so the random walk with drift assumption is valid. Perron8 unit root tests in the presence of structural breaks suggested that unemployment has a structural break—or shift in the slope of the trend around 1980—which causes near 1(2) behavior. Thus we have decided to estimate the data over the this shortened sample, which covers the last two decades.

D. Model Results

14. A variety of estimates were obtained using consumer prices, a GDP deflator, and a wage inflation series. Given that the estimates for the NAIRU and the output gap do not significantly depend on the inflation series used, we will focus our discussion on outcomes using the consumer price inflation series. In most cases, diagnostics suggested the use of four lags of past inflation changes to eliminate autocorrelation in the residuals, however, the final model residuals also tended to exhibit some minor amount of non-normality and heteroscedasticity in the unemployment and real GDP series. A general to specific modeling approach was used to determine lag lengths of the main supply shock variables.

15. In general, the results of the model are promising and in line with expectations.9 Table 1 provides a selected set of statistically significant (at least at the 5 percent level) parameter estimates from the UC model using consumer price inflation rates. Level and change effects are negative as expected, with the coefficients indicating that a one percentage point change in the unemployment gap results in a 0.4 percentage point change in the inflation rate in the current quarter. The level effect implies that an unemployment rate of 1 percent above the natural rate for one year would decrease the inflation rate by close to 0.5 of a percentage point. From the Okun’s Law relationship, the coefficients (φ0 φ1) indicated that each percentage point of the gap in unemployment can be associated with about a 1.7 percentage point of gap of output away from potential. Estimates of the trend growth rate of potential output a reveal an annual growth rate of around 2.4 to 2.5 percent—in line with current staff estimates. As expected, the sum of the four inflation lags implies a relatively high degree of price stickiness. Finally, the dummy variables on EMU points toward an upward shift in inflationary pressures under the regime change.10

Table 1.

Euro Area: Selected Parameter Results

article image

16. Figure 1 contains three panels that presents estimates of the evolution of the output gap and NAIRU in the euro area. The top panel compares the estimates of the output gap emanating from the UC systems approach against some well known statistical filters. The results are similar in terms of dating turning points but differ in their estimates of the magnitude of the gap. The output gap under the Ape-Wansson methodology, shows large negative gaps—reaching some 3½ of potential during the mid-1980s and about 3 percent through most of the 1990s—with a large positive output gap at the turn of the decade. The statistical filters, on the other hand, tend to place potential much closer to actual output, with relatively small negative gaps, and report various short periods of positive output gaps. The long period of negative output gaps under the model is not surprising given the declining trend of inflation over the last two decades.

Figure 1.
Figure 1.

Euro Area: Measures of Economic Slack

Citation: IMF Staff Country Reports 2001, 201; 10.5089/9781451812992.002.A001

Sources: European Commission, OECD, IFS and staff calculations.

17. As seen in the second panel, inclusion of the conditional supply shock variables can impart very different gap estimates. The two gap estimates indicate that the supply shocks provided beneficial deflationary effects, i.e., output gaps estimated without the supply shock variables tended to lie below those that included these conditioning variables during the expansion and contractionary phases of the late-1980s cycle. The evidence from the current expansion indicates supply shocks are greatly affecting inflation changes, with estimated gap differences of about 1¾ percent by end 2000. This highlights the importance of correctly conditioning on a set of supply shocks which allows a more refined attribution of inflation to demand pressures.

18. To check if the Apel-Jansson methodology suffers from an end-point bias problem, we also re-estimate the model using data ending in 1997:4. Although the result (not reported here) suggested an absence of an end-point bias, estimating over different sample periods can result in slightly different historical estimates of the output gap.11

19. The final panel contains the actual euro area unemployment rate and the evolution of the estimated NAIRU, both with and without supply shocks in the model. Throughout most of the 1980s the estimated NAIRU (using CPI data with supply shocks) gradually increased, reaching 11 percent by early 1994 before drifting downward to about 10 percent in late 1996. However, since the start of the 1997 expansion, the NAIRU has fallen more rapidly, reaching some 8¼ percent by end-2000.

20. How do these results compare to the available set of estimates from international organizations? The first two panels of Figure 2 compare the model based estimates of the output gap and NAIRU from those provided by the WEO, the OECD, and the European Commission’s databases, In addition, using the latest WEO assumptions we have forecast the output gap and NAIRU with the UC model. The OECD and the Commission estimate that the area-wide output gap has essentially closed by end-2000, with positive gaps opening up over the next two years. The IMF’s WEO assumes a small negative gap in 2000 with overall balance reached by the end of 2002. The model based estimate appears to be in line with the WEO estimate, indicating that the gap could persist over the next few years with a further decline in the NAIRU.12 Finally, the last panel of Figure 3 compares the estimated output gap with the annual rate of inflation in the euro area. As expected, the output gap from the model appears to be a good leading indicator of inflation, correctly anticipating four main turning points by about 3 to 4 quarters. It was also highly correlated—yielding a coefficient of about 82 percent—with other important activity indicators such as capacity utilization.

Figure 2.
Figure 2.

Euro Area: Comparisons of Slack Measures

Citation: IMF Staff Country Reports 2001, 201; 10.5089/9781451812992.002.A001

Sources: OECD, IFS, and staff calculations.1/ Annual data.

E. Inflation Forecasting Exercise

21. To provide a check on the informational content of the UC model output gap measure, a simple inflation forecasting exercise was undertaken using a standard Phillips curve:


with lagged values of the inflation, the output gap measure of interest and its first difference, and a relative oil price supply shock as a conditioning variable. 13 The change in the output gap has been inserted to account for speed limit effects, which may have a marked impact on inflation dynamics in a situation of rapidly accelerating activity. As usual, the level of the gap would indicate the overall state of inflationary demand pressures. In the exercise, output gaps from the UC model, a HP filter, and from a basic production function 14 were used for comparison purposes.

22. Using 1980.1 to 1997.4 as the estimation period, we construct 1-step ahead simulated forecasts of inflation over the next 12 quarters. One data point is added to the estimation sample, and the process re-run. Up to four lags of each of the regressors was considered in the specification of the inflation equation, thus resulting in 256 different specifications of this regression for each of the different output gaps. For each specification, the Theil U forecast statistic is calculated, namely the ratio of the root mean square error of the forecast under the model to the root mean square error for a “no change” or naive forecast. Values of this statistic greater than 1 indicate that the naïve forecast is superior to the regression specification. The results of the exercise reported in Table 2 confirm our prior that the output gap from the UC model does contain important information regarding future inflationary pressures.

Table 2.

Theil U Statistics from Inflation Forecasting Exercise 1/

article image
Source: Staff calculations based upon inflation regression equation (8) in the text.

F. Policy Implications and Conclusions

23. The results of our modeling exercise indicate that the NAIRU in the euro area has fallen substantially over the last five to six years. A falling NAIRU lends support to the notion that structural policy measures already implemented by euro-area governments have actually borne fruit. Encouragingly, movement toward peer review processes within policy making circles and comparable hard structural indicators have actually started to make a difference in area-wide labor market performance. Given that rigidity in labor markets usually goes hand in and with inefficient product markets, headway in reducing structural distortions in product markets—including cuts in state aids, implementation of the single market legislation, and liberalization of network sectors—may have also played a sizeable role in improving labor market outcomes.

24. The sizable degree of wage moderation in the face of sharp supply side shocks over the last few years suggests the possibility of a regime shift or structural change in the area’s wage inflation process. If a regime shift has actually taken place, it may indicate that the partial hysteresis process in effect in many euro area countries in the 70s and 80s—whereby wage compensation claims bore a diminished relationship to underlying tightness in labor markets—may have started to unwind.15 If the unwinding process has indeed begun, the monetary authorities’ need to preemptively tighten at the onset of inflationary external shocks has—to some extent—been reduced by the efforts of policy makers to make the euro area economy more flexible.

25. There are a number of indicators that support the hypothesis of a (wage) inflation process under EMU which could be different from the past. For example, the structure of the euro area economy has gradually shifted toward employment intensive service sectors, which typically have less militant unions and are less likely to push wage demands in excess of productivity gains. Also, wage setters have undoubtedly recognized that the EMU regime change has eliminated the national exchange rate adjustment lever, by which uncompetitive wage increases could be adjusted through currency realignments. Although labor markets are far from integrated, wage setters may have come to take into account that capital is more mobile than in the past, a change that increases the importance of wage competitiveness within the area. All these factors point toward moderating wage pressures now and in the near future.

26. Regarding fiscal policy, the expansion in potential output and decline in the NAIRU loosens, ceteris paribus, the usual constraints felt by fiscal policy makers. A falling NAIRU translates into an automatic improvement in budget balances which could be used to finance tax cuts without affecting underlying balances. This implies a virtuous trade-off where by the resultant surpluses from a falling NAIRU are used to “purchase” further declines. In this regard, the need to stimulate positive effective labor supply responses argues for stressing a fiscal policy agenda which focuses on reforming labor taxation and benefit systems in a comprehensive manner.


  • Apel, M., and P. Jansson, (1999a), “System Estimates of Potential Output and the NAIRU”, Empirical Economics, Vol. 24, pp. 37388.

    • Search Google Scholar
    • Export Citation
  • Apel, M., and P. Jansson, (1999b), “A Theory-Consistent System Approach for Estimating of Potential Output and the NAIRU”, Economic Letters, Vol. 64, pp. 27175

    • Search Google Scholar
    • Export Citation
  • Blanchard, O., and J. Wolfers, (1999): “The Role of Shocks and Institutions in the Rise of European Unemployment: the Aggregate Evidence”, (mimeo).

    • Search Google Scholar
    • Export Citation
  • European Commission, (2000), European Economy 2000 Review, No. 71.Prospects for Sustained Growth in the Euro Area”.

  • Fabiani, S., and R. Mestre, (2000), “Alternative Measures of the NAIRU in the Euro Area: Estimates and Assessment”, European Central Bank, Working Paper, No. 17.

    • Search Google Scholar
    • Export Citation
  • Fagan, G., Henry, J., and R. Mestre, (2001), “An Area-Wide Model (AWM) for the Euro Area,European Central Bank, Working Paper No. 42.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, (2001), World Economic Outlook, May 2001.

  • Laubach, T., (2001), “Measuring the NAIRU: Evidence from Seven Economies”, Review of Economics and Statistics, May, Vol. 83(2), pp. 218231.

    • Search Google Scholar
    • Export Citation
  • Lown, C. and R. Rich, (1997), “Is There an Inflation Puzzle?,Federal Reserve Bank of New York Economic Policy Review, December, pp. 5169.

    • Search Google Scholar
    • Export Citation
  • Mendez, G., and G. Palenzuela, (2001), “Assessment of Criteria for Output Gap Estimates”, Working Paper No. 54. European Central Bank.

    • Search Google Scholar
    • Export Citation
  • OECD, (2000), OECD Economic Outlook, No. 69.

  • Orphanides, A., and S. Van Norden, (1999), “The Reliability of Output Gap Estimates in Real Time,Board of Governors of the Federal Reserve System Working Paper, Washington.

    • Search Google Scholar
    • Export Citation
  • Perron, P., (1997), “Further Evidence on Breaking Trend Functions in Macroeconomic Variables,Journal of Econometrics, Vol. 80, pp. 355385

    • Search Google Scholar
    • Export Citation

Prepared by Kevin Ross, (


However, this assumes 1(2) behavior, implying that the change in unemployment or output is unbounded. Furthermore, as Laubach (2001) has pointed out, while many country’s NAIRU can be modeled as an 1(2) series, the resultant output gaps may not, in many instances, be sufficiently related to the inflation rate to be considered true NAIRU estimates.


See M. Apel and P. Jansson, (1999a), “System Estimates of Potential Output and the NAIRU”, Empirical Economics, 24:373-88 for details.


Although the standard Philips curve is formulated in terms of the level of inflation, Apel and Jansson (1999a) demonstrate that in the presence of a unit root, the Phillips curve can be restated in terms of the change in inflation. Given that unit root tests on inflation were inconclusive, the model was estimated in both level and change terms. The results were not materially different between either specification.


A dummy variable is also included in 1998:4, representing the startup of Stage III of EMU.


This atheoretical framework recognizes that information regarding the true structural determinants of potential output and the NAIRU may be limited. However, the use of supply shocks to identify cyclical components—and therefore the long run components-explicitly links potential output and the NAIRU to supply-side factors.


We use an unemployment series that does not contain the recent corrections made by Eurostat. End-2000 unemployment was 8.8 percent under this series as compared to 8.3 percent in the Eurostat series.


See P. Perron (1997), “Further Evidence on Breaking Trend Functions in Macroeconomic Variables, “Journal of Econometrics, 80:355-385.


An examination of the estimated inflation series suggested a well fitted model. In addition, the variances of the shocks to cyclical unemployment and the NAIRU were about ½ and ¾ of the shock to the variance of inflation, indicating that unemployment gaps explained a large part of the variance in inflation.


Given the small number of post-EMU observations, the likelihood ratio statistic may not correctly identify the direction and magnitude of the true regime change. Thus, this coefficient may be just picking up the changed direction of inflation.


This is an important point. As discussed in Orphanides and Van Norden (1999), for a realtime output gap measures to be useful for monetary policy, they should not change much from their original estimate when additional information becomes available over time. Mendez and Palenzuela (2001) found that output gaps from an Apel-Jansson type models were consistent when applied to euro area data, i.e., real time and historical output gaps did not differ significantly.


The somewhat larger implied differences in unemployment gaps versus the output gaps suggests that implicit estimates of the Okun coefficient may vary substantially.


The specification follows Lown and Rich (1997).


The production function estimates were calculated by following the methodology applied in the ECB’s area-wide model, (see Fagan, Henry, and Mestre, (2001)).


Blanchard and Wolfers, (2000), suggest that as the effects of negative shocks—e.g., rising real interest rates and declines in total factor productivity—fade and as European labor market institutions improve, employment growth should pick up.