LEARNING AND ENDOGENOUS SCRAPPING IN THE MACHINE REPLACEMENT MODEL

We take the standard machine replacement model and introduce uncertainty about the productivity differential of new vintages made available as time passes. In particular, agents are uncertain about the quality of their match with a new vintage until they adopt it. Uncertainty allows us to generate endogenous switching costs as agents will not necessarily replace their old vintages every period, even if there are zero (physical) scrapping costs. We analize the behavior of investment in the model and compare it to the case of fixed and proportional switching costs with no uncertainty. We also determine the value of information about vintage quality by comparing the path of investment with and without uncertainty in the baseline case (no scrapping costs).