Transitions to retirement: a dynamic approach to the labour-market behaviour of older workers in the Netherlands

This paper aims at analysing labour market transitions of older workers (aged 50 and older) using the Dutch Socio-Economic Panel for the period of 1990 to 1998. The Netherlands are of special interest because they are among the countries with the lowest participation rates of older workers, both for men and women. In this paper, multinomial regression models are estimated for transitions from employment to various (competing) destination states. These are exit routes out of work into retirement, disability or unemployment. These transitions are modelled within a search-theoretical framework. The older worker is assumed to have some free choice, but at the same time is subject to constraints imposed by the rules and features of the alternative routes (e.g. firm’s retirement practices, retirement legislation, and the social security system). The dependent variable is the relative likelihood of exit from work to any of the various destination states. Among the covariates implemented in the model are individual variables, human capital variables, household variables, and labour market variables. First, a static multinomial logit model is estimated in order to analyse the influence of the covariates on the probability of being in non-employment (with a distinction between being in retirement, disability, unemployment or inactivity without receiving benefits) compared to being in employment. The results show that there is a significant positive effect of being a male, having a retired partner and individual non-wage income for most of the non-employment states, while significant negative effects are found of being the head of the household, having an employed partner, working in commercial services, job level, being self-employed, net wage income and total household income. No large differences were found between the several non-employment labour market states. For age only a significant positive effect was found for retirement, while for poor health only a significant positive effect was found for disability. Second, a dynamic multinomial logit model is estimated in order to analyse the effects of the covariates on the probability of making a transition to another state than employment. The results show a significant positive effect of age, being a male, having a poor health and non-wage individual income on the probability of making a transition into non-employment (regardless of what route is taken) while a significant negative effect is found for having an employed partner, working in the primary sector, being self-employed and net wage. Moreover, wanting to work more hours and tenure positively influence the relative probability of going into early retirement, while job level has a positive effect on the relative probability of going into inactivity without receiving benefits.