Back to Archives
Explaining Unemployment Trends in the Netherlands*
Robert A.J. Dur
In this paper, a small macroeconomic model of the Dutch labour market is estimated. The model is used to detect the causes of the rise in unemployment since the early seventies. In contrast to existing empirical work, we treat labour supply as an endogenous variable. This adjustment does not only yield a more informative model, but also appears to have serious consequences for the conclusions drawn. In particular, we show that the detrimental effect of the replacement rate on unemployment has been overestimated in earlier studies. Furthermore, we include contractual working time in the analysis. Our estimates imply that work sharing does reduce unemployment, but lowers total hours worked.
Keywords: Unemployment; Labour Market -The Netherlands; Wage Formation; Labour Supply; Working Time.
Persistently high unemployment has been one of the major problems in most OECD countries since the end of the seventies. Economists have put much effort on trying to find explanations for this particularly inefficient phenomenon (see Bean (1994) for a survey). One widely known explanation, suggested in the seminal papers by Layard and Nickell (1985, 1986), is that several factors have tended to generate pressure for 'too-high' real labour costs. These factors may include the level of unemployment benefits, taxes, and the gap between consumer and producer prices. The highly influential Layard-Nickell model, which is characterized by imperfect competition on both product markets and the labour market, has been applied to many OECD countries in order to account for the unemployment path (see for example Bean et al. (1986), Layard et al. (1991)). In this paper, we try to explain unemployment trends in the Netherlands (1970-1994) by using the Layard-Nickell approach. We estimate a small macroeconomic model of the Dutch labour market, consisting of an employment equation, a real wage equation, and an equation for the labour force. The model is used to detect the causes of the rise in unemployment since the early seventies in the Netherlands. Furthermore, the model sheds light on the usefulness of several policy options to combat unemployment.
Our analysis differs from the existing empirical literature in three ways. Firstly, we do not impose cross-equation restrictions on the model which ensure that capital formation, labour force growth and technical progress are neutral with respect to unemployment. Layard and Nickell assume that any change in these highly trended variables, induces a real wage adjustment which exactly neutralizes its impact on unemployment. Recently, Madsen (1994) criticizes this assumption by arguing that these restrictions are neither consistent with Layard and Nickell's assumption of imperfect competition, nor empirically valid for about half of the OECD countries (including the Netherlands). Layard and Nickell justify the cross-equation restrictions by stating that the model could otherwise imply a continuing rise or continuing decline in unemployment, a feature which is rejected by unemployment data over a (very) long time horizon (150 years). In our view, this empirical argument does not justify the imposed restrictions because it assumes a static institutional structure over a very long period. We do not want to exclude that capital formation, labour force growth and technical progress induced a persistent upward pressure on unemployment under the institutional structure present in the Netherlands over the period 1970-1994. Indeed, our empirical work presented in this paper indicates that labour force growth was not accompanied by sufficiently large wage adjustments, and thus pushed up unemployment. In contrast, productivity gains due to capital formation appear to be skimmed off by wage setters to a large extent. As a consequence, there has been a persistent upward pressure on Dutch unemployment. This result is consistent with the findings by Kuipers and Kuper (1991) for the Netherlands, but is in contrast with Bean et al. (1986) and Lever (1991) who restrict labour supply to be neutral with respect to unemployment.
As a second innovation, we treat labour supply as an endogenous variable. Most studies aimed at accounting for unemployment growth leave labour force participation unexplained (see however Pissarides (1991) and Madsen (1994)). There are theoretical arguments for two-way relationships between labour market outcomes and labour supply. Therefore, we might obtain a more informative model by treating labour supply as an endogenous variable. Furthermore, exogenous variables entering the employment and wage equation may also directly influence labour supply (e.g. taxes and the replacement rate). Inclusion of these effects appears to have serious consequences for the conclusions drawn. In particular, we show that the detrimental effect of the replacement rate on unemployment has been overestimated in earlier studies.
Finally, we allow the number of contractual working hours per labour year to affect unemployment. Earlier studies in the same spirit neglected possible effects of work sharing on unemployment (including all studies focusing on the Netherlands), or did not take productivity effects and wage responses into account (Madsen (1994)), or used average hours instead of contractual hours (Pissarides (1991)). Contractual working time has been reduced with some twenty percent from the beginning of the seventies until mid eighties in the Netherlands. Hence, even if the effect of reduced working hours on unemployment is small it might have had a significant effect on unemployment during the last decades. Furthermore, it is important to shed some light on its effect on unemployment in the light of the recent pleadings by unions for a shorter working week. Our empirical work indicates that a shortening of the work week does reduce unemployment, but at a high costs. Due to the rise in hourly wage costs in response to reduced working hours, the level of output is damaged quite strongly.
The paper is built up as follows. In section 2 we present the model. In the subsequent section, the estimation results are reported and commented on. In section 4 we use the estimated model to account for unemployment trends in the last decades. Section 5 concludes.2 The Model
In order to explain unemployment trends in the Netherlands, we use a small macroeconomic model which is characterized by imperfect competition on both product markets and the labour market. The employment and the wage equation are mainly based on the model of Layard and Nickell (1985, 1986). In addition to the Layard-Nickell model, we include an equation for the labour force. In the following, we will look more closely at the several parts of the model.
with is the share of labour in total income, and all variables are in natural logarithms. Equation (1) reveals that employment is negatively related to expected real hourly wages ( we - p), and positively to (the absolute value of) the price elasticity of demand the deviation of actual from expected aggregate demand technical progress, and the capital stock. A number of theoretical arguments support the notion that variations in aggregate demand lead to variations in the price elasticity firms met on their product markets (Rotemberg and Saloner (1986), Bils (1989)). To account for cyclical non-stability of the price elasticity of demand, we set:
with ytr is a trend value of yd. The constant term can be interpreted as the structural level of competitiveness on product markets and might be positively related to the severeness of antitrust policy and the openness of the economy. By substituting (2) into (1) it is easily seen that both aggregate demand fluctuations and the structural level of competitiveness relate positively to employment.
In the empirical analysis, we will add the number of contractual working hours per labour year as an explanatory variable. One reason for this is that worker effort might be affected (either positively or negatively) by reduced working hours, leading to a change in the marginal product of labour (see for example Booth and Ravallion (1993)). Furthermore, hourly labour costs may rise, due to fixed costs per employee (hiring costs, coordination costs etcetera; Calmfors and Hoel (1988)). Moreover, if the operating-time of the capital stock is positively related to working-time, employment may be negatively affected given the size of the capital stock (Calmfors and Hoel (1989)). Because the several arguments do not point to the same direction, we have no strong priors with respect to the sign of the coefficient.
* I am grateful to D.P. Broer, J.C. Siebrand and O.H. Swank for useful comments on earlier drafts. Furthermore, the analysis benefited from discussions with many colleagues. All views expressed and remaining errors are the author's own.
** A large part of this paper was written while the author was with Research Centre for Economic Policy (OCFEB),and Utrecht University, Economisch Instituut/CIA