Modelling and Empirical Evaluation of Labour Supply Behaviour

One of the major issues of policy makers in The Netherlands is to reduce the high unemployment rate. In 1988 economic growth was substantial in all OECD countries, which led to an increase in employment. The economic growth also induced extra labour supply, especially of married women, which altogether led to a smaller reduction in the unemployment rate than could have been expected in view of the economic growth (see Rapportage Arbeidsmarkt, 1989). The estimated official unemployment rate in 1988 is still 11.0% of the total labour force. Therefore, there is a strong interest in policies that seek to increase employment (the demand side of the labour market) as well as in understanding the factors that influence labour supply. In this thesis we try to further such understanding by constructing a detailed model of household labour supply.

In this chapter we point out how labour supply behaviour can be analysed within the framework of utility maximization subject to a linear budget constraint. The model we use is neoclassical in the sense that the individual is assumed to maximize utility subject to time and price constraints. Quite a few researchers have taken the neoclassical model as a starting point of their analysis of labour supply (cf. Pencavel (1986) and Killingsworth and Heckman (1986)).

Generally, the budget constraint will not have the linear form that was assumed in Chapter 2. Due to the tax and the social security and welfare system, it will be nonlinear and nonconvex. In Section 3.2 we will concentrate on modelling labour supply if the budget set is nonconvex, the nonconvexity being caused by the Dutch social security and welfare system.

It has already been stated in Chapter 2 that most likely utility functions will vary across households. For a good understanding of household labour supply it is important to identify systematic factors which cause this variation. We pay attention to two major causes of differences in utility functions across households: preference formation (in this chapter) and differences in household composition (in Chapter 2).

The major objective of this chapter is to develop a model of job choice, wages and labour supply, and to use this model to estimate the magnitude of the compensating wage differentials for qualitative differences in jobs (such as pleasantness, riskiness, dirtiness etc.). Like in the preceding chapters, we will maintain a utility maximizing framework as the starting point of the analysis. But we abstain from the idea that the household is a joint decision unit. Neither do we take into account preference formation. Both extensions would complicate the model too much.

In the preceding chapters various models of labour supply were presented. In this chapter we will in addition model demand side restrictions, although in a simple way. Our starting point is an individual (no household) utility function that leads to a linear labour supply equation. Neither preference formation nor job characteristics nor the social security and welfare system are taken into account.

In Chapters 2 through 6 several models have been developed and estimated. The estimation results are based on two different datasets: the OSA- and the SEP-sample. In this chapter we will compare the estimation results in several ways. The comparison will not only deal with the various models but also with the two different datasets. With respect to the data it should be noted that both samples are supposed to be drawn from the same subset of the Dutch population in 1985. namely those individuals who are between l8 and 65 years old and who are able to work. We start with a test of this hypothesis (concentrating on the hours variables only). To that end we use a Pearson x² test and two distribution free tests (Wilcoxon’s test and the Kolmogorov-Smirnov test). The last two tests are nonparametric in the sense that they do not depend on the functional form of distribution functions. That is we test whether two distributions are equal, without specifying these distributions. In Section 7.2 the three two-sample tests are discussed and the test results are presented. We will also present some sample statistics such as means and standard deviations. Keeping the conclusions from Section 7.2 in mind we will in Section 7.3 compare the labour supply wage elasticities generated by one specific model that was estimated on the two data sets.

The main point of emphasis in this thesis is the modelling of a neoclassical labour supply model for one-adult and two-adult households. The number of hours an individual prefers to work is assumed to follow from maximization of a utility function subject to a budget constraint. Throughout this thesis we have adopted the Hausman-Ruud specification of the utility function which leads to a labour supply equation quadratic in wages with the exception of Chapter 6 where we have opted for the linear Hausman specification. This was dictated by the complexity of the model in Chapter 6. The models developed in Chapters 2 through 6 have been estimated for four different groups of individuals: single males, single females, married males and married females. We distinguished between these four groups because we expected labour supply behaviour of these groups to be different. The estimation results confirm these expectations. Where possible, the models have been estimated on two different data sets, namely the SEP and the OSA Survey. This opened up the possibility of testing the stability of the parameter estimates across samples.