Kingdom of the Netherlands—Netherlands Selected Issues

The Selected Issues paper of the Netherlands provides an overview of the Dutch pension system. It examines the strengths and weaknesses of the existing bargaining model, compares the Dutch wage bargaining in different periods, examines the past macroeconomic performance of the Netherlands, addresses the issues of wage dispersion and differentiation, and discusses future prospects and challenges for the bargaining model. It also estimates the potential growth in the Netherlands, using statistical detrending methods.

Abstract

The Selected Issues paper of the Netherlands provides an overview of the Dutch pension system. It examines the strengths and weaknesses of the existing bargaining model, compares the Dutch wage bargaining in different periods, examines the past macroeconomic performance of the Netherlands, addresses the issues of wage dispersion and differentiation, and discusses future prospects and challenges for the bargaining model. It also estimates the potential growth in the Netherlands, using statistical detrending methods.

III. Estimating Potential Growth and Output Gaps for the Netherlands21

A. Introduction

71. During the 1990s, the Dutch economy outperformed the euro area and was almost on a par with the United States (Figure III-1 and Table III-1). This robust performance was mainly a result of high trend employment growth. One key contributing factor was the expenditure-based fiscal consolidation program, which led to cuts in both the deficit and the tax burden. Labor market participation and employment were stimulated by cuts in taxes and social contributions, as well as by significant wage moderation and lower benefit replacement rates resulting from structural reforms.

Figure III-1.
Figure III-1.

Economic Performance: Netherlands, United States, and the Euro Area, 1991-2002

Citation: IMF Staff Country Reports 2003, 240; 10.5089/9781451829457.002.A003

Source: WEO.
Table III-1.

Sources of Growth, 1992-2000

(Annual percentage change)

article image
Source: IMF WEO database and Staff estimates.

Solow’s residual, using labor share of 0.6 and capital share of 0.4.

72. Since 2000, growth in the Netherlands has fallen sharply, in line with the euro area and the U.S. The slowdown has, however, been especially marked in the Netherlands: GDP growth was only 0.2 percent in 2002—one of the lowest in the euro area—and GDP could decline in 2003. This has raised concerns about the fixture course of the Dutch economy. Is the recent decline just a cyclical phenomenon, or the beginning of a more protracted slowdown? Looking ahead, as the population ages, sustaining the high employment growth of the past decades seems difficult without substantial structural reforms to further increase labor-force participation and employment rates.

73. This chapter presents estimates of potential growth and output gaps in the Netherlands, using a variety of approaches. These include two statistical detrending methods (HP filter and band-pass filter), a production function approach, and a system estimate of potential output and the associated non-accelerating inflation rate of unemployment (NAIRU). Based on this evidence, we now estimate annual Dutch potential growth at 2¼ percent, down from our previous estimate of 2½ percent. Estimating the output gap is important for assessing the stance of fiscal policies, but results are often sensitive to methods as well as sample selections used for estimation. Accordingly, we attempt to solve this problem by taking a different approach where the level of potential output is derived by using the estimated potential growth and the identified time point when the output gap closes based on empirical evidence. Specifically, the level of potential output is derived assuming potential growth of 2¼ percent over the period 1980–2004 and that the output gap closed near the end of 2002 (Figure III-2).

Figure III-2.
Figure III-2.

Potential Output, Growth, and Output Gaps

Citation: IMF Staff Country Reports 2003, 240; 10.5089/9781451829457.002.A003

Source: IMF Staff estimates.

74. The remainder of the chapter is organized as follows. Section B describes the empirical methods and data used for estimation, and discusses the benefits and weaknesses associated with each method. Section C presents the estimation results.

B. Estimation Methods

Statistical detrending methods

75. Hodrick-Prescott (HP) filter. Deriving potential output using the HP filter is a popular procedure because of its flexibility in smoothing fluctuations in output (Hodrick and Prescott, 1980). However, results from the HP filter are sensitive to the smoothing parameter λ, the choice of which is arbitrary, leading to at times excessive smoothing of structural breaks.22 Moreover, the HP filter is known for its high end-of-sample biases, because the weights in the HP filter change rapidly near the ends of the sample. This could result in substantial distortions of the trend at the end of the sample, just when an accurate estimate is most valuable.

76. Ideal band-pass filter (Ouliaris filter). Band-pass filtering selects components of the time series with periodic fluctuations between 6 and 32 quarters, while removing components at higher and lower frequencies (Baxter and King, 1999). Most recently, Ouliaris and Corbae (2002) propose an “ideal” band-pass filter based on Monte Carlo simulations. Compared to the Baxter-King filter and the HP filter, the Ouliaris band-pass filter is considered to be statistically consistent in the sense that the filtered series asymptotically converges to the true growth cycle.

Production function approach

77. Under this approach, potential output is estimated using a simple two-factor Cobb-Douglas production function (for example, see EC, 2002). In particular, our estimation of potential output growth is based on the trend growth of (i) the capital stock; (ii) the labor input based on an NAIRU estimate, a trend labor participation rate, and working-age population growth; and (iii) TFP estimates based on two alternative assumptions about the labor’s share (0.6 and 0.8).

78. The production function approach is very similar to detrending methods, in that the components of the production function are smoothed, in this case using an HP filter (with λ=100). It nonetheless provides useful information on the determinants of potential growth. Labor and capital inputs my be incorrectly measured due to labor-hoarding and changes in capacity utilization, as these are both cyclical in nature. However, using full-time equivalent employment and a net capital stock series adjusted for capacity utilization yield very similar results.23 Finally, since the capital stock data are only available for the business sector, we also estimate potential growth for this sector alone. This turns out to be only slightly higher than the estimated potential growth for the whole economy (Figure III-3 and Table III-2).

Figure III-3.
Figure III-3.

Total Factor Productivity (TFP) 1972-2004 1/

{Annual percentage change)

Citation: IMF Staff Country Reports 2003, 240; 10.5089/9781451829457.002.A003

1/ Solow’s residual, with labor share of 0.6.Source: IMF staff estimates.
Table III-2.

A Comparison of Output Gaps and Potential GDP Growth Using Various Approaches

article image

The pre-spccificd frequency band is [pi/16, pi/3], which implies that a cycle could be between 6 quarters and 8 years.

Based on Annual data 1969–2004.

Using the unobserved-components model of Apel and Jansson (1999), with sample 1970:1-2004:4.

Bureau for Economic and Policy Research, June 2003.

OECD database, as of June 23, 2003.

Multivariate model approach

79. The following unobservable component (UC) model was proposed by Apel and Jansson (Apel and Jansson, 1999) to jointly estimate the potential output and NAIRU:

Δπt=i=13ρiΔπti+j=01(utjutj)+k=04ωkztk+ε1t(1)
ytyt=l=01ϕl(ut1ut1)+ε2t(2)
ut=ut1+ε3t(3)
yt=α+yt1+ε4t(4)
utut=m=12δm(utmutm)+ε5t(5)

where πt is the CPI inflation, ut the registered unemployment rate, and yt the log of real output (all at a quarterly frequency). The NAIRU is assumed to follow a random walk, and the potential output is assumed to follow a random walk with drift. All error terms are assumed to be iid and mutually uncorrelated with constant variances.

80. The idea behind this method is to provide a economic structure by linking output gaps to unemployment gaps through Okun’s Law (equation 2), and both to changes in inflation through the Phillips curve (equation 1). For estimation, the Kalman filter and maximum likelihood method are used to obtain the estimates of the unknown parameters and of the unobserved variables, i.e. potential output and the NAIRU. Also included in the Philips curve equation are four variables to capture supply shocks: import prices, the oil price (both deflated by the CPI), the real exchange rate, and labor productivity. In addition, a dummy variable captures the impact of the structural reforms which took place in mid-1990s. Finally, in equation 2, a lagged employment gap is included to capture the effect of labor hoarding.

Data

81. Seasonally adjusted quarterly data are used for all variables for the period of 1970:1-2003:4, except for the production function approach, where annual data for 1970–2004 are used.

C. Estimation Results

82. Potential GDP growth rates derived from these methods prove very similar over the whole sample period of 1977–2003. They range from 2.1 to 2.4 percent, against the actual average growth of 2.3 percent during this period (Table III-2). However, the estimates for the short-run do vary considerably. Estimates based on the production function approach show that the higher potential growth during the 1990s, relative to 1980s, was a result of high trend employment growth (associated with a declining NAIRU) and trend capital accumulation, offsetting a decline in trend TFP (see Figure III-4). Assuming unchanged trend growth for TFP and capital stock accumulation, to sustained the output growth at 2¼ percent would require an employment growth of about 2 percent per year.

Figure III-4.
Figure III-4.

Sources of Potential Output Growth

(Annual percentage change)

Citation: IMF Staff Country Reports 2003, 240; 10.5089/9781451829457.002.A003

Source: IMF staff estimates.

83. All methods used yield similar estimates of the location of cyclical peaks and troughs (Figures III-5 and III-6). Cyclical developments (measured by the deviation between actual output and unemployment from potential output and the NAIRU, respectively) generated by the Apel-Jansson model are illustrated in Figure III-6. Overall, the Apel-Jansson approach is found to yield larger economic fluctuations (reflected in the larger output gaps), compared to other three methods. This result is consistent with previous studies.24

Figure III-5.
Figure III-5.

Potential Output, Growth, and Output Gaps

Citation: IMF Staff Country Reports 2003, 240; 10.5089/9781451829457.002.A003

Source: IMF Staff estimates.
Figure III-6.
Figure III-6.

Netherlands: Potential Output and NA1RU Based on Apel-Jansson Joint Estimation, 1980 Q1-2003 44

Citation: IMF Staff Country Reports 2003, 240; 10.5089/9781451829457.002.A003

Source: IMF staff estimates.

84. Overall, the evidence suggests that the large positive output gap, notably in 1999 and 2000, either closed in 2002 or will close in 2003. The various methods produce a range of results for output gaps as expected. Estimates based on the HP filter are sensitive to the choice of the smoothing parameter λ and the end point of the sample. Using the HP filter, the Ouliaris filter, and the production function approach, the Dutch economy fell to its potential in 2002, while Apel-Jansson estimation implies that the gap closes in 2003. Specifically, estimates of the output gap for 2002 range between 0.9 percent below potential (using Band-Pass filter) to 2.4 percent above potential (using the Apel-Jansson multivariate method), compared to the CPB estimate of zero gap and the OECD estimate of-1 percent.

References

  • Apel, Mikael, and Per Jansson, 1999, “A Theory-Consistent System Approach for Estimating Potential Output and the NAIRU,Economic Letters 64 pp. 271275.

    • Search Google Scholar
    • Export Citation
  • Baxter, Marianne, and Robert King, 1999, “Measuring Business Cycles: Approximate Band-Pass Filters For Economic Time Series,The Review of Economics and Statistics, November, 81 (4): pp. 575593.

    • Search Google Scholar
    • Export Citation
  • Cerra, V. and S.C. Saxena, 2000, “Alternative methods of Estimating Potential Output and the Output Gap: An Application to Sweden,IMF Working Paper, WP/00/59.

    • Search Google Scholar
    • Export Citation
  • European Commission, 2002, Public Finances in EMU 2002.

  • Netherlands: Selected Issues, July 2000.

  • Hodrick, R. and E.C. Prescott, 1980, “Post-War U.S. Business Cycles: An Empirical Investigation,Discussion Paper No. 451, Carnegie-Mellon University.

    • Search Google Scholar
    • Export Citation
  • Ouliaris, S. and P. Corbae, 2002, “Extracting Cycles from Non-Stationary Data,” in Econometric Theory and Practice, edited by P. Corbae, et al.

    • Search Google Scholar
    • Export Citation
21

Prepared by Jianping Zhou, (jzhoul@imf.org).

22

Consequently, there have been attempts to identify the “optimal λ,”, but in our case the identified “optimal λ (λ=374)” is associated with small output gaps through out the whole sample period (Table III-2).

23

Note that using the capacity adjusted capital stock did generate a much smoother TFP series.

Kingdom of the Netherlands—Netherlands Selected Issues
Author: International Monetary Fund