ESSAYS ON PORTFOLIO SELECTION AND FINANCIAL MODELING

Abstract

This dissertation contains two parts. In part I we consider a portfolio choice problem faced by an open-end fund manager who needs to deal with periodic money flows. We investigate the optimal portfolio and its implications on the costs of portfolio delegation and on the liquidity premia. We show that the money flows create implicit incentives for the fund manager and significantly increase the trading volume of the illiquid asset, which implies surprisingly high liquidity premia. We also show that the transaction costs could help to reduce the costs of portfolio delegation for conservative investor. In part II we propose interest rate models with random level shifts to capture the salient empirical features observed in the U.S. short-term treasury market. We fit affine yield models with one factor or two factors to the U.S. treasury bill yield data from 2002 to 2009 and test these fitted models in the out-of-sample period that extends from 2010 to 2011. We show that the affine yield models with random level shifts almost uniformly outperform their counterparts without random level shifts. In particular, we show that both the time series forecast power and the cross-sectional pricing power are almost uniformly improved by incorporating the random level shifts.