Does Policy Affect Outcomes for Young Children? An Analysis with International Microdata - August 1999

1. Introduction

Previous

Contents

Next

This paper examines the role played by policy in shaping child outcomes. The central hypothesis is that child outcomes are influenced, not only by micro-level socioeconomic characteristics (e.g., family structure, age/gender of the child), but also by general macroeconomic conditions (e.g., the regional unemployment rate), 'social context' (e.g., percentage of the population who are immigrants; level of social cohesion), and, centrally for this paper, by social policy (e.g., social spending per capita).

Much of the excellent work which has already been done using the National Longitudinal Survey of Children and Youth (NLSCY) has focussed only upon the micro-level characteristics which are associated with child outcomes, for example, poverty status of the family, family structure (lone-parent versus two-parent family) or education level of the child's mother (see paper summaries on the HRDC web-site). Of course, both general macroeconomic policy and social policies (such as cash transfers for families with children), play a key role in determining, for example, family income, a micro-level characteristic. To the extent that this is true, studies which find a significant role for family income in determining child well-being, are also finding a significant role for policy, though this point is not always highlighted. Nor is this the central channel for policy effects studied here.¹

In this paper, account is still taken of micro-level factors, but the emphasis is on expanding the usual set of explanatory variables to include macroeconomic, contextual and general policy variables. The basic methodology is to exploit variation in policies and variation in child outcomes which exist across Canada, Norway and the United States, and across regions within Canada and the U.S. The goal is simply to test generally whether or not an econometric link can be found between policy and outcomes, and not to establish specific links between particular policies (e.g., child benefits or social assistance) and child outcomes. Further, the goal is to establish whether or not a link can be found, not to estimate the magnitude of effects.

Within a single cross-section of data for a single country, there is limited variation in contextual variables, but very little variation in macroeconomic or policy variables. A fairly long panel of data would be required in order to obtain significant variation in the structure of basic social programmes, in levels of social spending, in macroeconomic conditions, or in general social context. Yet, these factors may be very significant determinants of outcomes for children. Better quality schools or hospitals, more social cohesion or less economic insecurity, either because of more extensive transfers or because of less unemployment, for example, might all be expected to increase the well-being of children, yet such factors are not captured in models which focus only upon the micro-level determinants of children's outcomes. The strategy proposed in this research is to combine relatively similar cross-sections of microdata from 3 countries with very different social policies in order to obtain variation in policy which can be used to test for links with child well-being. Policy and outcomes variations for regions within countries are also exploited.

In a precursor study to this research (Phipps, 1998d), an effort was made to categorize, in particular, quantitative aspects of the mix of policies available for children in the 3 countries again studied in this paper. The focus of the earlier work was upon tax and transfer provisions available, but childcare, healthcare and education were also discussed as was the values context of the programmes available, the macroeconomic environment and the sociodemographic history of each country.²

To summarize this work, children in Norway receive much more in the way of cash transfers than do children in Canada, or particularly, the U.S. For example, all Norwegian children receive generous child benefits, by Canadian standards, many Canadian children receive a child benefit, but no children in the U.S. receive such a benefit.³ On the other hand, families with children in both Norway and the U.S. receive tax relief which is not available for families with children in Canada. Lone-mother families in Norway receive very generous transfers, including guaranteed child support payments and double the usual child benefit. Families with newborns benefit through paid maternity leaves in both Canada and Norway, though the Norwegian programme is much more extensive than the Canadian. No such benefit is available in the United States.

Canada and Norway both provide 'universal' public health insurance, while such a programme is not offered in the U.S. Within countries, there is also considerable variation in social policy. Within Canada, for example, many programmes which are of direct relevance to the well-being of children are matters of provincial jurisdiction (e.g., health, education and social services).

Section 2 elaborates upon the conceptual framework motivating this analysis. Section 3 of the paper provides a brief review of the literature on neighbourhood influences on children's well-being, the literature most closely linked conceptually to the analysis presented in this paper. Section 4 outlines the data sources used. Section 5 documents the extent of differences across countries and across regions for 6 child outcomes: incidence of asthma; experience of accidents/injuries; activity limitation; experience of anxiety/fear; restless/overly active behaviour; and disobedience at school. Section 6 presents national/regional differences in micro-level socioeconomic characteristics and asks whether national/regional differences remain after we control for key socioeconomic characteristics of the individual child/family in estimated probit equations for child outcomes. Section 7 introduces the macroeconomic, contextual and policy data and discusses national/regional differences in these variables. Probit regressions are re-run, adding this new information to the more standard micro-level controls, to provide a test of the hypothesis that policy matters for children's outcomes.

More specifically, we test a series of hypotheses. First, we add three very basic 'contextual' variables: the unemployment rate as an indicator of macroeconomic conditions; social spending per capita, as an indicator of over-all level of policy activity; percentage of heads of household who are immigrants as an indicator of 'social context' (i.e., homogeneity of population, in this case). Second, the social spending variable is replaced by 4 more disaggregated policy components likely to be important for the well-being of children: social transfers, student/teacher ratios, physicians per 100,000 population and percent of healthcare publicly funded. Third, an effort is made to control for the fact that social spending will be higher in regions experiencing economic hard times, even if the underlying structure of the transfer system, for example, is not especially generous. Finally, we also attempt to test whether the structure of the transfer system (e.g., more targeted versus more universal) plays a role in addition to the average level of benefits received. Section 8 concludes.

• ¹Arguably, the effects of policy on family income may be the most important channel of influence. But, since income is a well-understood determinant of child well-being, the effect of policy on family income and thus on child outcomes is not the focus of this study.
• ²See also Baker, 1995, Gauthier, 1996 or O'Hara, 1998. O'Hara, 1998 and Phipps, 1998d are complementary pieces of research conducted during the same period of time for CPRN and HRDC.
• ³However, children of 'working poor' parents can receive the 'earned income tax credit' in the U.S. (Kamerman and Kahn, 1997).