The Labeling Effect of a Child Benefit System

Child benefit systems exist in many countries. While they show a lot of differences in terms of eligibility rules, amounts involved, and implementation (tax deduction, tax credit, or direct benefit), the common motivation for their existence is to increase children's welfare. In the Dutch child benefit system-the case I focus on in this paperone of the parents, usually the mother, is entitled to an untaxed child benefit amount which only depends on the child's age and the number of children in the household. Thus, child support does not depend on household income, marital status, or labor-market status. As a result, child benefit is exogenous to the household given the presence of children. This is in contrast with the situation in many other countries, including Germany, France, the United Kingdom, and the United States, where government-provided child support is (partly) means-tested; see Jonathan Bradshaw et al. (1993). In The Netherlands, child benefit is generally a nonnegligible addition to the household income; the median share of child benefit in the total net income of households with children is 8 percent. On the national level, expenditures on child benefits amount to 1.2 percent of GNP. The use of the child benefit is completely at the discretion of the parent. There are no legal requirements that a certain amount be spent on particular goods or services, nor does the government provide any guidelines regarding expenditure on children. A policy question that arises is to what extent children benefit from child benefits. Standard microeconomic demand theory allows no role for effects of the composition of income. Given the fungibility of income sources, it is only the sum of income components that is relevant in explaining expenditure patterns. Thus, within that framework, the answer would be that the marginal propensity to consume child goods from one guilder of child benefits is no different than from one guilder of any other income source. In the more general class of gametheoretic models of household behavior, however, the composition of household income will generally affect expenditure patterns; see, e.g., Shelly J. Lundberg and Robert A. Pollak (1993) and Martin Browning et al. (1994). If fathers and mothers have different preferences, and mothers have control over child benefits, the effect of child benefits on expenditures will differ from the effect of other income sources. Lundberg et al. (1997) provide empirical evidence showing that a transfer of control over child benefits from fathers to mothers in the United Kingdom due to a change in legislation resulted in a significant increase in expenditures on child clothing. Daniela Del Boca and Christopher J. Flinn (1994) have analyzed the effect of income composition on expenditure decisions of divorced mothers. They find that the coefficients associated with child support and alimony income differ from those for other income in Engel curves for expenditures on childspecific goods. Their results can be explained in terms of a noncooperative Nash model in which the child support transfer decisions of the noncustodial father and the expenditure decisions of the mother are determined simultaneously. In contrast to Del Boca and Flinn's analysis, in which child support is a decision variable, the present paper focuses on the possible differential effect of government-provided child support which is truly exogenous. In this paper I analyze the effects of child benefits on expenditures by running regressions in which child benefit enters as a separate explanatory variable. The enipirical analysis is based on a time series of 17 cross-section consumer expenditure surveys in The Netherlands, covering the period 1978 through 1994. It is important to note that in a single cross-section it * Department of Economics, University of Groningen, P.O.B. 800, 9700 AV Groningen, The Netherlands (e-mail: p.kooreman@eco.rug.nl). I am indebted to two anonymous referees for constructive comments. Helpful comments were also provided by Rob Alessie, Maarten Allers, Chris Bojke, Werner de Bondt, Bert Schoonbeek, and Tom Wansbeek. I thank Vincent Linderhof for organizing the data. Financial support was provided by the Dutch Organization for Scientific Research, NWO (HOMES-project grant).