The Parental Leave Benefit Reform in Germany: Costs and Labour Market Outcomes of Moving towards the Nordic Model

Abstract

Germany is known to have one of the lowest fertility rates among Western European countries and also relatively low employment rates of mothers with young children. Although these trends have been observed during the last decades, the German public has only recently begun discussing these issues. In order to reverse these trends, the German government recently passed a reform of the parental leave benefit system in line with the model practiced in Nordic countries. The core piece of the reform is the replacement of the existing means-tested parental leave benefit by a wage-dependent benefit for the period of one year. In this paper we simulate fiscal costs and expected labour market outcomes of this reform. Based on a micro-simulation model for Germany we calculate first-round effects, which assume no behavioural changes and second-round effects, where we take labour supply changes into account. Our results show that on average all income groups, couples and single households, benefit from the reform. The calculation of overall costs of the reform shows that the additional costs are moderate. As far as the labour market behaviour of parents is concerned, we find no significant changes of labour market outcomes in the first year after birth. However, in the second year, mothers increase their working hours and labour market participation significantly. Our results suggest that the reform will achieve one of its aims, namely the increase in the labour market participation of mothers with young children.

Appendix: Formal Description of the Labour Supply Model, Data and Estimation Procedure

The labour supply model used for the estimation of the behavioural effects resulting from the parental leave benefit reform is a structural household utility model. Household utility is assumed to be a function of net household income, the leisure of the mother and the leisure of the father. The estimation of the utility function is based on a discrete choice model. The underlying assumption is that households compare the utility U of J number of different choice categories, i.e. combinations of working hours and net household incomes, and choose the combination, which yields the highest utility.¹¹

We follow van Soest (1995) and specify the household utility function as a trans-log function, i.e.

$$ U_k (x_k ) = x^{\prime}_k Ax_k + \beta ^{\prime}x_k + \varepsilon _k $$

(1)

where U _k is the utility of the household in choice category k and the vector x contains the linear, quadratic and cross terms of the natural logs of net income, leisure of the mother and leisure of the father. Matrix A contains the coefficients of the cross and quadratic terms, vector β contains the coefficients of the linear terms and ε is a stochastic error term.

Assuming that households maximize utility, they choose category k if the utility in this choice category is greater than in all other categories. In probability terms, this decision rule can be stated as follows:

$$ P(U_k > U_l ) = P\left( {\left( {x_k 'Ax_k + \beta 'x_k } \right) - \left( {x_l 'Ax_l + \beta 'x_l } \right) > \varepsilon _l - \varepsilon _k } \right) $$

(2)

Assuming that the error terms ε are independently and identically distributed over all choice categories and follow an extreme value distribution, the probability to choose category k as stated in equation (2) can be estimated using the conditional logit model (McFadden, 1973):

$$ P\left( {U_k > U_l } \right) = \frac{{\exp \left( {x^{\prime}_k Ax_k + \beta ^{\prime}x_k } \right)}}{{\sum\limits_{j = 1}^J {\exp \left( {x^{\prime}_j Ax_j + \beta ^{\prime}x_j } \right)} }}{\text{, }}\forall l \ne k $$

(3)

In order to include individual or household specific characteristics, we also include interaction terms between these characteristics and leisure. These characteristics are age of the mother, age of the father, age of the youngest child, number of children and a dummy variable indicating whether the household is living in east Germany (see Table A.3).

For the estimation of the labour supply model, we include two-parent families (either married or cohabiting couples) who have at least one child less than 24 months. We exclude single parents because it seems plausible that the parameters of their utility function would differ from those of two-parent families. Unfortunately, we do not have enough observations to estimate a separate regression for single parents, who make up about 10% of our sample. Thus, our results refer to two-parent families only. We further exclude households in which one of the parents is still in education, self-employed or severely disabled. As we pool observations from three periods (2001–2003), some households enter the sample in two periods. We assume that the error terms are independently and identically distributed following an extreme value distribution, as specified above.

Table A.1 shows the distribution of households in our sample for the estimation of the labour supply equation across choice categories. The definition of the hours categories is motivated by both economic considerations and the actual distribution of hours in the sample. Although a relatively fine aggregation of hours into categories seems desirable in order to realistically approximate the household’s budget constraint, the actual distribution of hours in the sample severely restricts the number of possible categories. Furthermore, for couples, the feasible number of categories is not only restricted by the distribution of hours within one gender, but by the bivariate distribution of the two spouses’ working hours. For example, we observe few men in our sample who are either not working or are working part-time. Thus, for these two working categories of men, we define only two working hours categories for women, non-working and working.

Table A.1 Sample size and distribution of households across choice categories

Table A.2 presents further descriptive characteristics such as net household income, age of the parents and number of children in each hours category.

Table A.2 Descriptive statistics of the sample used for the labour supply estimation

The results of the labour supply estimations are presented in Table A.3.

Table A.3 Coefficients of the conditional logit model