Data from a recently-completed experimental program for out-of-work welfare recipients in Texas are used to examine the effects of a time-limited financial incentive coupled with post-employment services on recipients’ rates of entering and leaving employment. While there is strong evidence that such programs can increase overall employment, the crucial question of how these increases arise is not well-understood. This paper presents a rigorous analysis of employment entry and exit effects, using a fully-specified dynamic model of employment duration that accounts for non-random sorting into employment statuses through flexible specifications for duration dependence and unobserved heterogeneity. The results indicate that for the Corpus Christi site, short-term effects were due to both employment retention and employment entry but, over time (as the program ceased operation), the retention effects faded out but the employment entry effects persisted and grew. For the Fort Worth site, there were smaller effects overall and less evidence of impacts that lasted much beyond the program operation period.