Effect of estimation risk on efficient set : evidence from Thai stock market.

Abstract

This study explores effect of estimation risk on admissible efficient set and optimal portfolio basing on analysis under Bayesian framework assumed diffused prior density distribution function. In this study, we perform a test whether the historical average return can be used to estimate true parameter or not. We found that historical average return is an inappropriate estimator. We assume that future asset returns follow multivariate normal distribution and form three states of analyses. Assuming true values of parameters are known in the first state, true variance-covariance of asset return are known but true average return are not know in the second state, and both true parameters are not known in the third state. The empirical results based on sectorial index of Stock Exchange of Thailand (SET) indicate that when estimation risk is taken into account, the admissible efficient set is not changed. Two conclusions can be made. The first conclusion is true portfolio return can be represented by weighted average sample returns. The second conclusion is that when estimation risk is built into a decision, portfolio risk is affected by a scale factor. Therefore, Bayesian admissible efficient set will always lie to the right of the traditional admissible efficient set due to higher risk from estimation.