Problems of Model Specification and Measurement

Although the United States' existing welfare system no longer appears viable, there is no consensus over a program to replace it. Inability to choose among alternatives to the present system is partially, though significantly, caused by uncertainty over their potential effects on labor supply. One way to obtain estimates of these effects is through inferential studies based on cross-sectional data. So far, however, these studies have failed to sufficiently reduce the range of uncertainty about labor supply effects. Improving the usefulness of such studies depends on resolving problems in defining the sample population, measuring the principal variables — labor supply, income, and wage rates — and controlling for other factors. Based on a detailed review of methodologies used in previous studies, this report discusses the implications of making one methodological choice rather than another and points to areas where improved data or theories can make the greatest contribution to improving estimates of labor supply. Existing data are examined and tests for geographic differences in labor supply are discussed.