people invest heavily in those, and invest little or nothing at all in ambigu-
ous assets. Their portfolios therefore appear undiversified relative to the
predictions of standard models that ignore the investor’s degree of confi-
dence in the probability distribution of a gamble.
Not all evidence of home bias should be interpreted as a preference for the
familiar. Coval and Moskowitz (1999) show that U.S. mutual fund man-
agers tend to hold stocks whose company headquarters are located close to
their funds’ headquarters. However, Coval and Moskowitz’s (2001) finding
that these local holdings subsequently perform well suggests that an infor-
mation story is at work here, not a preference for the familiar. It is simply
less costly to research local firms and so fund managers do indeed focus on
those firms, picking out the stocks with higher expected returns. There is no
obvious information-based explanation for the results of French and Poterba
(1991), Huberman (2001), or Benartzi (2001), while Grinblatt and Keloharju
(2001) argue against such an interpretation of their findings.
7.2. Naive DiversificationBenartzi and Thaler (2001) find that when people dodiversify, they do so in
a naive fashion. In particular, they provide evidence that in 401(k) plans,
many people seem to use strategies as simple as allocating 1/nof their sav-
ings to each of the navailable investment options, whatever those options
are. Some evidence that people think in this way comes from the labora-
tory. Benartzi and Thaler ask subjects to make an allocation decision in
each of the following three conditions: first, between a stock fund and a
bond fund; next, between a stock fund and a balanced fund, which invests
50 percent in stocks and 50 percent in bonds; and finally, between a bond
fund and a balanced fund. They find that in all three cases, a 50:50 split
across the two funds is a popular choice, although of course this leads to very
different effective choices between stocks and bonds: the average allocation
to stocks in the three conditions was 54 percent, 73 percent, and 35 percent,
respectively.
The 1/ndiversification heuristic and other similar naive diversification
strategies predict that in 401(k) plans which offer predominantly stock
funds, investors will allocate more to stocks. Benartzi and Thaler test this in
a sample of 170 large retirement savings plans. They divide the plans into
three groups based on the fraction of funds—low, medium, or high—they
offer that are stock funds. The allocation to stocks increases across the
three groups, from 49 to 60 to 64 percent, confirming the initial prediction.
7.3. Excessive TradingOne of the clearest predictions of rational models of investing is that there
should be very little trading. In a world where rationality is common
52 BARBERIS AND THALER