From the studies reviewed in section 2, both ethnic and geographic fragmentation can be ex-
pected to play a significant role to the extent that they capture possibilities for information flows,
monitoring and enforcement. Education fragmentation instead is usually not employed in the liter-
ature, but becomes relevant in the present study because different education levels may represent
complementary production skills. As can be seen from the table, of the fragmentation indexes only
that related to education is significant: members of groups in which widespread education levels
are represented are more likely to borrow, possibly because of complementarities among members
themselves.
The last two regressors in column 1 are Gini coe¢cients of inequality either in wealth (durable
assets ownership) or in earnings from the group. It is noticeable that the latter has a negative and
highly significant coe¢cient, indicating that when members earn very different amounts from group
activities it is less likely that loans are advanced either by the group as a whole or by individual
members. Column 2 expands on this result by allowing the coe¢cient on group income inequality
to differ depending on whether the group has positive or negative profits. Interestingly, the negative
effect of inequality on credit is found in groups that have financial difficulties but not in those that
make positive profits. In other words, having members who earn very different amounts does not
harm their ability to borrow if the pie is large enough for a number of them, while it does undermine
lending by the group or among individuals when members have to struggle over resources.
[Insert table 6]
Borrowing from the group as a whole or borrowing from individual members could make a
difference, especially in the light of the above result. Table 6 reports the estimated coe∏cients from
two multinomial logit regressions in which the explanatory variables are the same as in table 5, but
the dependent variable is a categorical variable taking value 1 if in case of need the individual borrows
from the group as an ‘institution’, 2 if he or she borrows from a single member, and 0 if the lender
is outside the group. The latter is the omitted category in the table. Inequality in group earnings
decreases the likelihood of both types of loans, but is statistically significant only for ‘group’ loans.
This suggests that allowing for differences in remuneration of the members does not significantly
14