undermine their willingness and ability to lend to each other, but it does undermine the ability to
use the group as a source of credit, especially when the group runs in financial difficulties (see column
2). This may be due to the di∏culty of agreeing on how to allocate the common funds when these
are very limited. The next section will look more closely at the various forms of heterogeneity among
group members and at the relationship between heterogeneity and group organization.
4.3 Heterogeneity and group organization
One of the clear patterns that emerged from the previous analysis was that certain types of het-
erogeneity had a stronger impact than others. In particular, inequality in individual earnings from
the group significantly reduced the ability to borrow from the group as a whole, while inequality in
“wealth”, as measured by the index of durable goods, did not. It is useful to investigate why this is
the case.
In addition to asking about ‘objective’ wealth, the questionnaire asked each respondent the
following question: “Leaving aside the income from this group, do the members have the same
wealth?”; and prompted the following responses: “Mostly the same”, “Some differences”, “Very
different”. Figure 1 relates this ‘subjective’ notion of inequality to the objective ones used in the
regressions.
[Insert figure 1]
The vertical axis in each panel measures the fraction of members of each group saying that
members’ wealth is “very different”. The horizontal axis reports for each group the Gini coe∏cient
for individual wealth (top panel), the Gini for earnings from the group (middle panel), and the
ethnic fragmentation index (bottom panel). For example, a point such as (.23, .54) in the top panel
represents a group in which the Gini coe∏cient for wealth is .23 yet 54 percent of the members
think that differences in wealth among them are substantial. The size of the circles in the figure is
proportional to group membership: points identified by larger circles correspond to larger groups.
It is quite striking how in the top panel no relationship at all emerges between the actual
level of wealth inequality and that perceived by the members. On the contrary, the middle panel
15