Jan 1, 2002
Nationally, the welfare caseload declined by more than 50% between 1994 and 2000. Considerable research has been devoted to understanding what caused this decline. Much of the literature examining these changes has modeled the total caseload (the stock) directly. Klerman and Haider (forthcoming) model the underlying flows and show analytically and empirically that previous methods are likely to be biased because they ignore important dynamics. However, due to their focus on the bias of the stock models, they present only limited results concerning the robustness of their findings and utilize only a single measure of economic conditions, the unemployment rate. This paper examines the robustness of the basic stock-flow model developed in Klerman and Haider (forthcoming), considering both richer dynamic specifications and richer measures of economic condition. We find that more complex dynamic specifications do not change the substantive conclusions, but richer measures of the economy do. While a model that only includes the unemployment rate attributes about half of the California caseload decline between 1995 and 1998 to the economy, models that incorporate richer measures of the economy attribute more than 90% of the decline to the economy.