Dynamic learning with hidden actions

Hidden characteristics and hidden actions are present in many observed phenomena involving dynamic processes. I study different dynamic environments where hidden information and unobservable actions play an important role on determining agents\u27 incentives and so behavior. Chapter 1 analyzes how quality-adjustment costs help firms to build reputation on producing high quality goods. The reason is that adjustment costs generate private information that is endogenously persistent in equilibrium. As a result, recent signals about the quality choices are informative about the quality of the goods that are being produced right now. So, firms are able to build trust on selling high-quality products. This provides an interpretation of reputation as a moral hazard phenomenon instead of an adverse selection problem. Also, I find that reputation dynamics are cyclical or diverging, depending on the signal structure. In Chapter 2 I analyze the effect of hidden actions in a dynamic signaling environment, where these actions are unobservable. Transactions of heterogeneous assets or high education are examples of situation where unobservable actions may be used to transmit information about characteristics of the asset or the student. I find that the signal is more informative when the market is pessimistic. Further- more, signaling is non-trivial even if the cost of signaling is arbitrarily high. Finally, Chapter 3, that is joint work with Fei Li, analyzes the role of college drop out as a signaling device. Dropping out is more attractive for low-productive students, since the cost savings of doing this are higher for them. We find that they dropout through time, making returns of education (weakly) decreasing. Also, the maximum length of education is found to be a decreasing function of the market average skill