The dramatic decline in Temporary Assistance for Needy Families (TANF) caseloads in the 1990s has focused attention on the process of exit from, and, to a lesser extent, entry into, the welfare system. This paper charts the process of turnover in the TANF program in three major U.S. cities over an 18-month, post-PWORA period and documents its income, employment, other-program participation, and demographic correlates. The results show a high rate of turnover and significant increases in employment among all groups--stayers, leavers, and those never on welfare. The findings also indicate that the income gain from leaving welfare is modest, on the order of 11 to 18 percent, for the losses of benefits attendant upon leaving largely cancel out the earnings gains. Because of low benefit reduction rates, staying on welfare and working results in higher income than leaving for many families. Finally, the data suggest that caseload declines may have been as much the result of work and other requirements, and diversion policies, as the financial gain from leaving or entering welfare. A description of the Three-City Study is appended. (Author/SM)