| The classic mean-variance model with a "plug-in" estimator often produces negative or extreme portfolio weights, and performs poorly in out-of-sample tests. Furthermore, empirical findings show that in practice, managers and households do not hold a constant share of risky assets, which runs counter to the conclusions of the mean-variance model. This dissertation presents two optimal portfolio choice models to improve the estimation of expected returns and variances. The relationship between actual portfolio choices and health status is also examined. I compare the effect of poor health on risky asset ownership and shares held across age groups.;The first model extends the Black-Litterman model with a minimum relative entropy principle. Historical information and subjective views are combined through an entropy approach. It is assumed that asset returns can be predicted in a four-factor model, which extends the Capital Asset Pricing Model (CAPM) used by Black and Litterman (1992). Subjective views on the factors are generated from a Bayesian Vector Autoregressive (VAR) model, although the entropy approach is flexible enough to allow for any form of subjective views. The empirical results compare the performance of the entropy approach to the Black-Litterman model and the "plug-in" method.;The second model proposes a recursive coefficient rule to estimate means and variances by maximizing expected utility. We apply the Mixed Data Sampling (MIDAS) method to assign different weights to the sample observations. Out-of-sample tests compare the performance of the recursive estimation to the classic "plug-in" approach and the equal weight rule. The Sharpe ratios and certainty equivalents suggest that the recursive coefficient rule performs better than the other two approaches.;Unlike earlier research that primarily focuses on portfolio choice after retirement, the effect of health in younger populations is also explored here. The results indicate that the decision of to purchase risky assets is negatively related to the deterioration of health. A negative health shock during an earlier stage of life might have a persistent effect on individuals' subsequent investment decisions. In addition, wealth and private health insurance significantly increase the probabilities of owning risky assets and shareholding. |
