note: just because a factor model happens to work well does not necessarily
mean that we are learning anything about the economic drivers of average
returns. To be fair, Fama and French (1996) themselves admit that their re-
sults can only have their full impact once it is explained what it is about in-
vestor preferences and the structure of the economy that leads people to
price assets according to their model.
One general feature of the rational approach is that it is loadings or
betas, and not firm characteristics, that determine average returns. For
example, a risk-based approach would argue that value stocks earn high
returns not because they have high B/M ratios, but because such stocks
happen to have a high loading on the B/M factor. Daniel and Titman (1997)
cast doubt on this specific prediction by performing double sorts of stocks
on both B/M ratios and loadings on B/M factors, and showing that stocks
with different loadings but the same B/M ratio do notdiffer in their average
returns. These results appear quite damaging to the rational approach, but
they have also proved controversial. Using a longer data set and a different
methodology, Fama, French, and Davis (2000) claim to reverse Daniel and
Titman’s findings.
More generally, rational approaches to the cross-sectional evidence face a
number of other obstacles. First, rational models typically measure risk as
the covariance of returns with marginal utility of consumption. Stocks are
risky if they fail to pay out at times of high marginal utility—in “bad”
times—and instead pay out when marginal utility is low—in “good” times.
The problem is that for many of the above findings, there is little evidence
that the portfolios with anomalously highaverage returns do poorly in bad
times, whatever plausible measure of bad times is used. For example,
Lakonishok, Shleifer, and Vishny (1994) show that in their 1968 to 1989
sample period, value stocks do well relative to growth stocks even when the
economy is in recession. Similarly, De Bondt and Thaler (1987) find that
their loser stocks have higher betas than winners in up markets and lower
betas in down markets—an attractive combination that no one would label
“risky.”
Second, some of the portfolios in the above studies—the decile of stocks
with the lowest B/M ratios for example—earn average returns below the
risk-free rate. It is not easy to explain why a rational investor would willingly
accept a lower return than the T-Bill rate on a volatile portfolio.
Third, Chopra, Lakonishok, and Ritter (1992) and La Porta et al. (1997)
show that a large fraction of the high (low) average returns to prior losers
(winners) documented by De Bondt and Thaler (1985), and of the high
(low) returns to value (growth) stocks, is earned over a very small number
of days around earnings announcements. It is hard to tell a rational story
for why the premia should be concentrated in this way, given that there is
no evidence of changes in systematicrisk around earnings announcements.
40 BARBERIS AND THALER