Web: home.uchicago.edu/~liorm Education

Thesis “Influential Observations in Cross-Sectional Tests of Linear Factor Models” I investigate the link between the variations in the cross-section of average returns and macroeconomic conditions. Instead of examining a specific factor, I ask which properties are needed for a factors to capture a large part of the cross-sectional variation of average return. I show that the return data of commonly used portfolios have few influential observations that are central in capturing this variation. The concentration of these observations around recession periods indicates a strong link between the cross-section of assets returns and the business cycle, providing support for conditional (C)CAPM models. However, the existence of influential observations and the small number of them also suggest that the power of the cross sectional tests is low, prohibiting any meaningful attempt to discriminate between different (macroeconomic) explanations. Any factor regression that loads on one or two of these observations is likely to exhibit large explanatory power. This concern is particularly important in the case of scaled factor regressions, commonly employed in the estimation of conditional models. As a principle solution, I advocate imposing restrictions on the test based on economic theory. In particular, instrumenting and hence modeling the first-stage betas as a function of firm characteristics mitigates the relation between the estimation error of the betas and average returns.