برنامه های بازنشستگی عمومی و دوران بازنشستگی زوج های متاهل در دانمارک

In this paper we study the economic determinants of the joint retirement process of married couples. We propose a tractable dynamic discrete choice model for retirement decisions which allows for non-trivial saving behaviour. We estimate the model on a 1% sample of Danish couples of potential retirement age drawn from a population-based administrative register. The introduction and subsequent reforms of a publicly financed early retirement programme provide us with variation in the data to insure identification of the parameters of interest: the elasticities of participation/retirement with respect to income flows. Our estimates imply a significant asymmetry in the sensitivity of retirement behaviour of men and women with respect to variation in their own, or their spouse's, income flows.

A growing fraction of people of potential retirement age are in two-earner married couples. In Europe and North America over recent decades the labour force participation rate of women of all ages has risen, and now an increasing proportion of women are reaching their 50's having had a long labour market career. Career married couples tend to retire close together in calendar time even though they are often of different ages. Given their numerical importance and correlated behaviour, couples are the natural unit for analysis of retirement decisions. Empirical studies have found support for different explanations of joint retirement: complementarity of leisure, correlated preferences, common shocks, shared finances and combinations of these factors. In most empirical work there is a balance to be struck between complexity of the behaviour being represented and the amount of information contained in the data. The amount of useful variation depends upon the sampling frame, the set of controls and the institutional setting. The latter is of overriding importance when observational data is used. A majority of published pension and retirement studies use the US Health and Retirement Study (HRS). This is a very rich set of variables following a modest number of seniors each second year since 1992. The US institutional setting poses several modelling challenges. Employer-provided health insurance, re-entry to the labour force after retirement from a career job, joint taxation and joint benefits all require careful attention be paid to several simultaneous processes.1 It is the aim of this study to address some of the difficulties that have hindered couples' retirement studies to date. We set up a dynamic structural model of joint retirement decisions. This is applied to a 1% sample of career married couples drawn from 25 years of population-based longitudinal data from Denmark. The Danish context is perhaps uniquely informative because of its combination of universal health insurance coverage, individual-based programme eligibility and rare labour market re-entry. This must be contrasted with the institutions and consequent modelling difficulties that researchers using US data in particular have to face. Furthermore, in Denmark a generous public pension programme (efterløn, hereafter PEW) was introduced at the end of the 1970s, with dramatic effects. Within two years, the participation rate for married men aged 61–65 fell from 70% to 50%, while the rate for those aged 54–58 remained almost unchanged. Eligibility to the programme was not exogenous, but was arguably pre-determined in the sense that it was a function of historical unemployment insurance contributions, which previously did not passport pension programme eligibility. These criteria were changed at different times during our sample period. Eligibility varied both within and between married couples, entitlement was all-or-nothing, and re-entry to the labour force after first receipt of PEW, meant permanent disqualification. An important contribution of this study is exploiting these discontinuities by way of carefully characterising income streams from potential future retirement ages to provide exogenous income variation that helps to identify the model. The population-based longitudinal administrative data we use here covers 1977–2001 and does not suffer from sample attrition.2 Furthermore it brings the relevant planning horizon within the observed time period for many couples. In comparison to the HRS for example, our Danish dataset can credibly sustain a good degree of model complexity. Three broad approaches have been followed in the literature on the economic determinants of retirement. First, structural modelling of the household as a decision unit where all the elements deemed important to describe current and future retirement options are characterised in detail and embedded in a structural discrete choice model.3 A second approach places more emphasis on the decision process within the household as the important determinant of association between retirement dates. Within this approach, two different routes (maintained behavioural hypotheses) have been followed: some studies assume that spouses are involved in bargaining over the outcomes of their labour force participation decisions (Nash or otherwise),4 while others posit that decisions within the household always achieve Pareto optimality.5 A third approach places less emphasis on structural modelling, or even on the details of decision making within household, and relies on reduced form dynamic (or even static) models of joint retirement ages as a response to variation of earnings in-work and income out-of-work.6,7 This has the advantage of allowing a complete statistical description of the data of interest, at the expense of less transparent economic interpretation. In particular, these models remain silent about the effects of out-of-sample policy reforms. Our stochastic dynamic model of married couples retirement ages is such that individuals are forward looking and we allow for non-trivial (although not general) consumption/saving behaviour. The household is the unit on which to base policy analysis because it is at this level that decisions are internalized. We do not set up a model of decision making within the household, and so implicitly take a unitary approach. However, our model structure allows us in principle to distinguish between alternative explanations of the joint timing of retirement within couples: correlated preferences due to leisure complementarity or substitutability, common shocks, and shared finances. In particular, we show the effect of changes in programme generosity and non-work income on the probability of each spouse retiring at each age. The rest of the paper is organised as follows. The next section describes the institutional specifics of the Danish social security system. This is followed by a description of the data. The fourth section presents our empirical model of couples' retirement. Section 5 presents and interprets the estimation results and provides some counterfactual policy analysis. Finally, we conclude.

We analyse the economic determinants of joint retirement decisions for career married couples. We set up a dynamic structural model and estimate its parameters on a 1% sample of the Danish population drawn from administrative registers. In contrast to the US, Denmark has several institutional features which simplify retirement modelling: universal health insurance coverage, individual-based programme eligibility and benefit entitlement and negligible partial retirement or labour market re-entry after retiring from a career job. Also, a generous public pension programme (PEW) was introduced within our sample period. Pre-determined eligibility criteria are used to break the simultaneity between work and potential non-work income flows and empirically identify the model. Estimated values of preference parameters are plausible with an average propensity to consume out of liquid wealth and current income of 0.5, a rate of time preference 0.85 and a coefficient of relative risk aversion of zero. Simulations show that the model fits the retirement effects of PEW introduction reasonably well. We decompose the cross-eligibility effects of spouses, and we show that male eligibility accounts for a much larger proportion of the programme-induced fall in female participation than female eligibility accounts for male. Female participation is more responsive to income changes than male participation. This is true for both own-income and cross-income elasticities. Spouse's leisures are found to be almost always complements. Furthermore, there is significant asymmetry in cross-income elasticities, with female participation responding much more to male income than male participation responds to female income.