Recently, welfare reform has received considerable attention because welfare programs (WPs) are costly, reduce work incentives, cause marital breakups, and result in the recipients' loss of selfesteem. Bassi [EI, 1990], Burtless [JEP, 1990], Chambers [JPE, 1989], Mueller [WSJ, 1993], and Gueron [JEP, 1990] evaluated the efficacy of WPs. De Borger [JPE, 1989] examined the welfare implications of in-kind programs. Garfinkel et al. [JHR, 1990] and Graham [CPI, 1990] investigated the relationship between child support and employment prospects of welfare recipients. This note examines the impact of monthly cash benefits, labor market conditions, family unity, and income on the number of AFDC recipients. It is hypothesized that the number of welfare recipients (WEL) is positively related to monthly cash benefits (BEN), tightness of labor markets, marital dissolution, and financial strength to support WPs. For estimation, tightness of the labor market is proxied by the unemployment rate (UNR), family dissolution is measured by the divorce rate (DIV), and society's financial strength is captured by real per capita income (INC). Rationale for the inclusion of these variables is as follows: higher BEN increases the reservation wage and thereby diminishes incentives to work. Increased UNR raises the probability to remain on welfare because of fewer job opportunities and it forces the newly unemployed to obtain welfare. A rise in DIV increases single parents who may become eligible for financial assistance. Higher INC means enhanced public ability to support WPs. However, higher INC is expected to lower poverty and thus decrease WEL. Therefore, INC's sign may be positive or negative. Data for the 1961-91 time period came from the Social Security Bulletin and the Statistical Abstract of the U.S. To reduce multicollinearity, the dependent variable is expressed as the ratio of AFDC families to total families in the U.S. BEN t and INC t are in 1982 dollars, with INC t being in thousands. The log-linear form is chosen because of its superior performance. Autocorrelation was tested and corrected, p is 0.860 and is significant at the 1 percent level. The results are reported below: