Revisiting Lettau and Ludvigson (2001): Does cay really matter for cross-sectional risk premia?

The authors revisit the asset pricing tests conducted in Lettau and Ludvigson (2001) (LL). Contrary to LL’s claim, their conditional models based on cay do not explain the dispersion in risk premia among the size/book-to-market portfolios. The pricing performance is either very poor (conditional CAPM/Consumption-CAPM) or quite modest (conditional Human-capital-CAPM), with all models largely underperforming the Fama–French model. Furthermore, the risk price estimates for the scaled factors are insignificant in several cases. When the zero-beta-rate is unrestricted, the authors obtain average pricing errors that are similar or above the raw average risk premia. Alternatively, LL’s extreme intercept estimates are inconsistent with the equity premium puzzle. Using alternative test portfolios and evaluating the identification of the risk prices substantially reinforces the negative outlook on their results. LL’s models also perform poorly out-of-sample. Overall, LL’s wrong conclusions stem from a combination of incorrect empirical choices and a misinterpretation of their results.