In this paper, we have examined how individual components of the job search process influence the probability of reemployment. A reduced-form model of job search is estimated that takes account of the fact that users of a particular method of job search are not a random subset of all searchers. The empirical model produces estimates of the parameters governing the choice of search method, the contact rate per method, the process by which job offers are generated, and the acceptance or rejection of offers. Particular attention is paid to the role of observed characteristics in explaining differences in search behavior between the employed and unemployed. The results of this paper show that individuals who search for a new job while working are, on average, more successful at finding a job than otherwise similar unemployed searchers. As was indicated above, there are two possible explanations for this finding. One is that employed search is more effective than unemployed search, perhaps because of better search technology (e.g., access to internal career ladders and better contacts) or the stigma associated with unemployment. If this explanation is correct, then a typical unemployed job seeker would have a better chance of finding a desirable job by accepting the first offer received and continuing to search while employed. However, such behavior may not be optimal for the individual if there are important differences in search costs between the employed and unemployed. The other explanation is based on unobserved heterogeneity: employed searchers may simply search harder or may be better searchers than the unemployed in ways that are not captured by observed variables. In other words, employment status may be correlated with (or serves as a signal of) search ability or effort. In this case, there is no presumption that employed search is more effective than unemployed search for any particular individual; unobserved differences across individuals are respon... [ABSTRACT FROM AUTHOR]