| Chapter 1, Extreme Events and the Role of Learning in Financial Markets, investigates the role of parameter uncertainty in financial markets when the economy is subject to rare, extreme events of unknown probability and magnitude. In a dynamic portfolio optimization setting, Bayesian investors learn about the parameters of the economy in continuous time, which generates a hedging demand for the risky assets. This results in a significant reduction in the optimal stock allocation if investors are sufficiently risk-averse. The impact of learning on equilibrium asset prices is then examined in a model where economic fundamentals are subject to discontinuities of uncertain frequency and size. We find using an analytic expression for asset prices that learning plays a destabilizing role in equilibrium, magnifying the impact of economic shocks on asset prices.; Chapter 2, Behavioral Heterogeneity and the Income Effect, co-authored with Laurent Calvet, introduces HITS, a semiparametric model of consumer demand that allows for heterogeneous tastes. The strong variation of budget shares observed across income strata originates in individual income effects, and taste differences across households. We disentangle the two effects by developing a new microeconometric framework that parametrizes a class of Nearly Ideal Demand Systems with a unique taste parameter. Linear heterogeneity allows GMM estimation of the structural coefficients on aggregate time series, and the joint density of spending and tastes is recovered from cross-sections by a nonparametric deconvolution procedure. We report a strong correlation between income and tastes, which explains most of the observed variation of budget shares with income.; Chapter 3, Global Strategic Asset Allocation and Estimation Risk , co-authored with Lorenzo Isla, examines the role of predictability in international portfolio choice. Previous research indicates that emerging market returns exhibit a high degree of predictability; we argue that limited historical data and a high volatility make the estimation of the predictive relation of asset returns very uncertain. Accounting for this parameter uncertainty in a Bayesian framework, we find that estimation risk offsets the long-term benefits of returns predictability, and reduces the level and variability of the optimal allocation to emerging markets by long-lived US investors. |
