Mathematical Finance Seminar

April 5, 2001 , 5:30 PM to 7:00 PM

Jerome Detemple, Boston University School of Management

A Monte Carlo Method for Optimal Portfolios

This paper provides (i) simulation-based approaches for the computation of asset allocation rules, (ii) economic insights on the behavior of the hedging components and (iii) a comparison of numerical methods. For general utility functions with wealth-dependent risk aversion and diffusion state variable processes, hedging demands are abtained as conditional expectations of random variables depending on the parameters of the model, which can then be estimated using standard simulation methods. We propose a modified simulation approach which relies on a simple transformation of the underlying state variables and improves the performance of Monte-Carlo estimators of portfolio rules. Our approach is flexible and applies to (i) arbitrary utility functions, (ii) any finite number of state variables, (iii) general diffusion processes for state variables and (iv) any finite number of assets. The procedure is implemented for a class of multivariate nonlinear diffusions for the market price of risk (MPR), the interest rate (IR) and other factors (dividends). After calibrating the models to the data we document the portfolio behavior. We find that intertemporal hedging demands (i) significantly increase the demand for stocks and (ii) exhibit low volatility. In addition we show that (iii) non-linearities are important: significant biases in allocation rules are documented if one uses typical (linear-affine) processes calibrated to the data.