A general stochastic model of optimal job separation behavior is developed in the paper. The model nests both the job training and the job-matching hypotheses of the wage-tenure relationship as special cases. The purpose of the paper is to compare the implications of the two hypotheses for job turnover. That expected wealth-maximizing separation strategies are qualitatively identical under the two hypotheses is the principal theoretical result. Although the empirical implications of the two hypotheses for observations on the distribution of completed job-spell lengths are similar as a consequence, they are not quite identical.

This article investigates the consequences of labor market search for the theory of hedonic wages. We find that the introduction of search has surprising consequences for the theory of hedonic wages. In particular, we demonstrate that the equilibrium distribution of wage and nonwage amenity bundles generally bears little resemblance to workers' underlying preferences. A consequence of this analysis is that estimates of workers' marginal willingness to pay, derived from the conventional hedonic wage methodology, are biased. In addition, we demonstrate that search generates differences between firm‐level and employee‐level data that can cause substantial deviations in the estimates of hedonic wage equations.

The article structually estimates an on‐the‐job search model of job separations. Given each employer pays observably equivalent workers the same but wages are dispersed across employers, an employer's separation flow is the sum of an exogenous outflow unrelated to the wage and a job‐to‐job flow that decreases with the employer's wage. Using data from the Danish Integrated Database for Labour Market Research, the empirical results imply, as predicted by theory, that search effort declines with the wage. Furthermore, the estimates explain the employment effect, defined as the horizontal difference between the distribution of wages earned and the wage offer distribution.