An Empirical Assessment of Characteristics and Optimal Portfolios

Abstract

We analyze characteristics' joint predictive information through the lens of out-of-sample power utility functions. Linking weights to characteristics to form optimal portfolios suffers from estimation error which we mitigate by maximizing an in-sample loss function that is more concave than the utility function. While no single characteristic can be used to enhance utility by all investors, conditioning on momentum, size, and residual volatility produces portfolios with significantly higher certainty equivalents than benchmarks for all investors. Characteristic complementarities produce the benefits, for example momentum mitigates overfitting inherent in other characteristics. Optimal portfolios' returns lie largely outside the span of traditional factors.