This paper uses an unbalanced panel dataset to evaluate how repeated job search services (JSS) and other personal characteristics affect the quarterly earnings of the prime-age female welfare recipients in the State of Washington. We propose a joint dependent framework for the probability of employment and potential earnings or hours or hourly wage rates to facilitate the investigation of the issues of joint determination of employment and potential earnings and to allow for factors to have different impacts on employment status and on quarterly earnings. We have also suggested formulae to compute the dynamic impact of JSS on duration and earnings. Both the maximum likelihood (ML) and semi-parametric estimates are provided. We find that the results are sensitive to the choice of models and estimation methods. For a randomly assigned individual, the first, second and three or more JSS raised the short-run and long-run earnings by (5%, 0%, 0%) and (1%, 0%, 0%), respectively, based on the MLE, and by (56.6%, 36.2%, 36.9%) and (50.7%, 36.2%, 36.9%), respectively, based on the semi-parametric Tobit model. We have also conducted specification analysis. The results appear to favor semi-parametric Type II Tobit model estimates.
