In this paper, a model of job transitions based on job search theoretic arguments is specified and estimated. Because job search theory has implications for the optimal stopping rule followed, over time, by an unemployed individual, it also has implifications for the distribution of unemployment/job spells duration. Typically, stationary models predict that individual labor market histories are governed by a Markov process. The model analyzed here, allowing for nonstationarity and duration dependence in unemployment, predicts that the transition rate out of a current job may depend on the completed duration in the previous state (unemployment). The results show that the semi-Markov property is too restrictive and may therefore cause the loss of economic information that may help identify the source of nonstationarity.