When Do We Really Know What We Think We Know? Determining Causality

1 Social scientists are often asked to determine whether or not one thing causes another. The answer to this question of causality may have important implications for public policy. However, it is generally difficult to establish that " A causes B " beyond a shadow of a doubt, and researchers often arrive at conflicting conclusions depending on their data sources and methods. This conundrum is of course, not confined to social science. Researchers in the " hard " sciences and medicine often come to conflicting conclusions regarding questions such as what killed the dinosaurs, whether there is global warming, and whether hormone replacement therapy is safe and effective for older women. This chapter considers some of the methods that social scientists working in the area of work and family use to get at the question of causality. It provides a general overview of some of the issues and problems, and then discusses these issues in the context of two specific examples: the effect of maternal employment on child well-being, and the effect of child care quality on children's outcomes. In both cases, studies have arrived at differing conclusions, but in the first case, a range of studies using different data sets and techniques is providing the basis for an emerging consensus, while in the second, the issue of selection has not yet been satisfactorily addressed. I conclude with the reminder that replication is at the heart of science, and findings must be reproducible before they can provide a reliable basis for policy. Moreover, given the ubiquity of the sample selection problem in social science, the issue must be addressed, preferably using a number of techniques. In social science, questions about causality are clouded by the problem of sample selection. For example, mothers who work tend to be healthier and better educated on average than those who do not. Hence, it would not be surprising to find that their children did better than those of non-working mothers on average, even if there were no causal relationship at all between maternal employment and child 2 outcomes. Similarly, children from more advantaged backgrounds are likely to be in better child care, making it difficult to distinguish between the effects of child care quality and the effects of family background.