Stock market volatility in regime shift models
The thesis studies how learning affects asset prices within rational expectations equilibrium models. All chapters have the following common framework: the economy is characterized by a discrete set of "regimes" which determine the payoffs of securities. Investors, who can choose between a risky asset and a risk-free asset, do not know the regime but receive signals on the current regime.
Chapter 1 and 2 study dynamic models in a continuous-time framework under the assumption that regimes--the drift rate of the dividend diffusion process--can shift randomly over time. In chapter 1, I assume that there are only two regimes, investors have CARA utility functions and they live in a "production economy" with a constant risk-free rate of return. I show that the price of the asset is increasing and convex in investors' posterior probability of the favorable regime. I also show that return volatility depends on the sensitivity of investors' posterior probability to news (the "uncertainty effect") and the sensitivity of the asset price to probability changes (the "risk aversion effect"). I finally show that equilibrium stock returns display GARCH and EGARCH behavior, excess volatility and predictability. Issues on the small sample bias are also studied within the equilibrium framework.
Chapter 2 studies a specification of the model that has been widely studied in the financial literature: investors have a CRRA utility and live in an "exchange" economy. I show that Chapter 2 studies a specification of the model that has been widely studied in the financial literature: investors have a CRRA utility and live in an "exchange" economy. I show that learning has the ability to generate changing volatility, but I also find that it makes it harder to explain high risk-premia and high volatility. This is in contrast to the existing literature which assumes that investors observe the current regime (no learning).
Chapter 3 studies the model in a static setting but now investors have the option to learn the regime by paying a fix cost. Contrary to the previous literature, here learning may be a strategic complement (the incentive to learn increases as more agents learn), the asset may be "Giffen" (its demand is increasing with price) and the price may have a negative information effect (higher price implies lower expected payoff).