subjects, the composition of the funds, and the pool of subjects^6 were all
identical to those used in the first experiment.
In addition to replicating the previous conditions we added a fourth con-
dition to this experiment. In this condition, subjects have to choose one
fund out of five different multiasset funds, labeled as Fund A, B, C, D, and
E. The proportion of stocks in the five funds varies from 0 to 100 percent
by 25 percent increments. Fund A, for example, invests all of its assets in
stocks whereas Fund E is invested completely in bonds. This means that a
subject can choose any asset allocation in 25 percent increments, and thus
is formally equivalent (up to rounding error) to the first condition where
the subject explicitly divides her assets between stocks and bonds. This con-
dition is motivated by the design of the privatized Chilean Social Security
system.^7 In that system participants must choose among an array of funds,
each of which is diversified across asset classes.^8 We are interested in
NAIVE DIVERSIFICATION STRATEGIES 577Table 16.1
Verbal Savings Questionnaire: Mean Allocation to Fund AMean Actual Mean Implied p-Value for
Allocation to Allocation to the Difference
Fund A Fund A in Means
Version N Fund A Fund B (Median) (Median) (Medians)
One 53 Stocks Bonds 54 percent 54 percent N/A
(50) (50) N/A
Two6 6 Stocks Half stocks 46 21 0.001
and half (50) (0) (0.001)
bonds
Three 61 Half stocks Bonds 69 87 0.001
and half (70) (100) (0.001)
bonds
Notes: Three groups of individuals were asked to allocate contributions between two funds,
labeled Fund A and Fund B, based on a verbal description of the composition of the funds.
The first group was asked to allocate contributions between stocks (Fund A) and bonds (Fund
B). The second group was asked to allocate contributions between stocks (Fund A) and a bal-
anced fund, which was half stocks, and half bonds (Fund B). The third group was asked to allo-
cate contributions between a balanced fund (Fund A) and bonds (Fund B). The table provides
the actual allocation to Fund A by group. The table also includes what should have been the
allocation to Fund A by the second and third groups to stay consistent with the choices of the
first group—i.e., the implied allocation.
(^6) Although the subjects in the various experiments were drawn from the same pool, no sub-
ject participated in more than one of the experiments we report in this work.
(^7) See Peter Diamond and Salvador Valdes-Prieto (1994) for details on the Chilean retire-
ment system.
(^8) We do not use actual data from Chile, because regulatory restrictions result in all the in-
vestment funds having similar asset allocations.