A Tractable Utility-Based Approach to Consistent Model Selection and Asset Allocation
Craig A. Friedman
Standard & Poor's - Quantitative Analytics
May 12, 2004
Investors, such as traders and portfolio managers, who face the practical problem of how to allocate their assets in financial markets, often use probabilistic models, in conjunction with some quantification of their preferences. In this paper, we explore a consistent, tractable framework, based on utility theory, for both selecting models and allocating assets. This framework leads us to a particular family of utility functions, the generalized logarithmic family, which is consistent with model selection via likelihood. We also discuss drawdown and optimality consequences of applying this framework as well as related calibration issues.
Number of Pages in PDF File: 18
Keywords: Likelihood Ratio, Drawdown Probability, Optimal Performance, Generalized Logarithmic Utility Function
Date posted: October 27, 2005