1. Lettau and Ludvigson find that ? is close to zero, but the slope on ? is around 0.06% or 0.07% quarterly depending on the specification. In principle, we could test whether this slope equals cov(? t , CAY t-1 ), but the statistics reported in the paper are insufficient to do so. Here we simply note that the estimates seem huge;Ludvigson Lettau;If we assume that consumption betas (? t ) are linear in CAY and the zerobeta rate is constant, their Table 3 shows returns regressed on ? and ? ? cov(? t,2001
2. ) Returns on size and B/M portfolios can be traced to three common factors (time-series R 2 s above 90%), and that betas on the factors explain most of the crosssectional variation in expected returns. In this setting, one can show that almost any multi-factor model will produce a high cross-sectional R 2 . (3) The papers don't report standard errors or confidence intervals for the R 2 , and simulations show that it is easy to find a high sample R 2 even when the population R 2 is zero. For example, we simulate the two-pass regressions of Lettau and Ludvigson;; However;Journal of Financial Economics,2001