looks for novel predictions the theory makes. For example, Lee, Shleifer,
and Thaler (1991) test their model’s prediction that small firm returns will
be correlated with closed-end fund discounts, while Hong, Lim, and Stein
(2000) test the implication of the Hong and Stein (1999) model that mo-
mentum will be stronger among stocks with thinner analyst coverage.
Another sort of test is to look for evidence that agents actually behave
the way a model claims they do. The Odean (1998) and Genesove and
Mayer (2001) investigations of the disposition effect using actual market
behavior fall into this category. Bloomfield and Hales (2002) offers an ex-
perimental test of the behavior theorized by Barberis, Shleifer, and Vishny
(1998). Of course, such tests are never airtight, but we should be skeptical
of theories based on behavior that is undocumented empirically. Since be-
havioral theories claim to be grounded in realistic assumptions about be-
havior, we hope behavioral finance researchers will continue to give their
assumptions empirical scrutiny. We would urge the same upon authors of
rational theories.^38
We have two predictions about the outcome of direct tests of the assump-
tions of economic models. First, we will find that most of our current theo-
ries, both rational and behavioral, are wrong. Second, substantially better
theories will emerge.
A SURVEY OF BEHAVIORAL FINANCE 65(^38) Directly testing the validity of a model’s assumptions is not common practice in econom-
ics, perhaps because of Milton Friedman’s influential argument that one should evaluate theo-
ries based on the validity of their predictions rather than the validity of their assumptions.
Whether or not this is sound scientific practice, we note that much of the debate over the past
twenty years has occurred precisely because the evidence has not been consistent with the the-
ories, so it may be a good time to start worrying about the assumptions. If a theorist wants to
claim that fact Xcan be explained by behavior Y, it seems prudent to check whether people
actually do Y.