Portfolio Choice with Path-Dependent Preferences

Abstract

Many sophisticated investors rely on scenario analysis to select a portfolio. These investors define prospective economic scenarios, assign probabilities to them, translate the scenarios into expected asset class returns, and select the portfolio with the highest expected return or expected utility, given all these inputs. With this approach, the investor only considers single period outcomes. The authors propose a new approach to scenario analysis that enables investors to consider sequential outcomes. They define prospective scenarios, not as average values of economic variables, but as paths for these variables. And they measure the likelihood that these paths will prevail in the future based on their statistical similarity to the historical sequences of these variables. The authors also employ a novel forecasting technique called partial sample regression to map economic outcomes onto asset class returns. This process allows investors to evaluate portfolios based on the likelihood they will produce a certain pattern of returns over a specified investment horizon.

Keywords: Informativeness, Mahalanobis Distance, Partial Sample Regression, Relevance, Scenario Analysis, Statistical Similarity

JEL Classification: A10, C00, C01, C02, C10, G00, G01, G10, G11, G17