« Improving the Evaluation of Asset Pricing Models by Expanding the Set of Test Portfolios »
Recent studies show that the standard test portfolios do not contain sufficient information to discriminate between asset pricing models. To address this issue, we develop a novel approach that expands the cross-section to several thousand portfolios. Our large-scale approach is simple, widely applicable, and allows for formal comparison tests. Its ability to improve performance evaluation is confirmed by our empirical results—while the tested models are all misspecified, they exhibit striking performance differences. In particular, the human capital and conditional CAPMs produce lower pricing errors which suggests that labor income and recession risk are important concerns for investors.
« The Predominance of Real Estate in the Household Portfolio», with Sebastien Betermier
This paper investigates why household portfolios are heavily skewed toward real estate. Previous studies suggest that the large portfolio share of real estate primarily stems from non-investment-related motives as homeowners are often forced to invest heavily to buy the home they want to consume. In contrast, we show that homeowners would still invest the bulk of their wealth in real estate in a frictionless setting where they could own and consume separate amounts of housing. We provide empirical support to this argument and derive a dynamic portfolio model to study why real estate has such a strong investment appeal.
« The Economic Gains from Predicting Returns of Multiple Assets »
This paper determines the economic gains from predicting the returns of multiple assets. I set up a scenario where active investors receive timing or selectivity signals, and compute the economic gains they generate. The results reveal that investors with selectivity skills take more aggressive portfolio decisions, and, therefore, produce significantly larger gains. When predictive signals are persistent, the impact on short-term performance can be positive, in contrast to the single-asset case. In the long-run, however, this impact is always negative. Finally, moderate levels of estimation risk and trading costs–when considered jointly–greatly decrease performance. It implies that small levels of predictability do not necessarily produce large economic gains, as previously documented.