In search theory, an important distinction can be drawn between models with directed search and with random /undirected search. In the present paper we first present a simple model of competitive on-the-job search, where firms' productivity differences emerge endogenously through the firms' investment choices. We believe this model gives some new insights in its own right. Then we discuss testable differences between our model and the BM model. In the empirical part of the paper, we use Norwegian data to discriminate between the theories. To overcome problems with noise in the data, the estimation is done in two steps, fitting a finite mixture model to classify data into two groups. Results are consistent with the competitive search model.
