The Causal Analysis of Change

This chapter discusses the casual analysis of change. In experimental research, spuriousness and direction of causality are evaluated by randomization and manipulation, respectively. The random assignment of subjects to a treatment or control group guarantees that there are no differences between the two groups on any variable apart from those because of statistical fluctuation, and probability theory can tell the likelihood of a fluctuation of given magnitude. The experimental design is frequently impractical, unethical, or too divorced from reality to shed light on the issues of concern. Under these circumstances, a statistical analysis of cross-sectional data based on the assumption that causal influences are at work in one direction only would run a serious risk of bias in estimating the causal influence assumed to be present. But the more sophisticated analyses that abandon the assumption of unidirectional causality cannot be carried out. Though not a complete substitute for the experimental design, panel analysis goes farther toward resolving the ambiguities in causal inference than other forms of analysis.