This paper considers a model of earnings over time which incorporates individual effects and time effects without assuming that these effects are orthogonal to the variable of primary interest. The central coefficient is the effect of participation in a Manpower Development and Training Act training program on the earnings of trainees. Since the training status of (some) individuals in the sample changes during the period of the sample, both pre- and posttraining contrasts and trainee-nontrainee contrasts in earnings can be made. An estimate of the cross-section bias in a training coefficient can be made directly. The extent of the analogous bias in the education coefficient in regression studies is a point of current debate. The cross-section bias in the sample analyzed is large, and the estimated effect of training is small and positive.

This paper provides an empirically tractable version of a job-search model. The model is estimated using data on a sample of workers who were laid off when their plants closed. A generalization of the empirical model which allows for reservation wages to change over duration of unemployment is provided and estimated. Reservation wages are found to decline significantly with duration. Applications of the model and the estimates to explain diverse labor market phenomena are provided.

[The efficiency gain from observing the sample classification in a disequilibrium or switching model is analyzed. The problem is set up as one of comparing the precisions of estimates based on a joint density with those based on a marginal density. Asymptotic results are obtained for a simple model.]

This paper considers a model of earnings over time which incorporates individual effects and time effects without assuming that these effects are orthogonal to the variable of primary interest. The central coefficient is the effect of participation in a Manpower Development and Training Act training program on the earnings of trainees. Since the training status of (some) individuals in the sample changes during the period of the sample, both pre- and posttraining contrasts and trainee-nontrainee contrasts in earnings can be made. An estimate of the cross-section bias in a training coefficient can be made directly. The extent of the analogous bias in the education coefficient in regression studies is a point of current debate. The cross-section bias in the sample analyzed is large, and the estimated effect of training is small and positive.

This paper provides an empirically tractable version of a job-search model. The model is estimated using data on a sample of workers who were laid off when their plants closed. A generalization of the empirical model which allows for reservation wages to change over duration of unemployment is provided and estimated. Reservation wages are found to decline significantly with duration. Applications of the model and the estimates to explain diverse labor market phenomena are provided.