4.2 Access to credit
Aside from the cash flow provided by the group in the form of earnings, responses to the survey
suggest that many members value the possibility of resorting to the group for help in case of need.
This section will examine to what extent this occurs with respect to credit, i.e. what factors affect
members’ access to group loans.
[Insert table 4]
Table 4 estimates the probability of borrowing from the group or from individual members in
case of need as a function of the respondents’ characteristics plus group fixed effects. The dependent
variable takes value 1 if, when asked “If you need a loan, who do you borrow from?”, the respondent
answers “Group as a whole” or “Individual group member”. It takes value 0 if the answer is “Bank”,
“Relative”, “Friends not in the group”, or “Nobody”. The regressors are meant to control both
for demand and for supply factors, as will be clear from the discussion. Column 1 reports marginal
probit coe∏icients on demographic variables only. It is surprising that only the language dummies are
statistically significant. An interesting regressor is the dummy identifying respondents that speak
the same language as the chairperson when decisions are taken by the leaders. Ceteris paribus,
speaking the language of the chairperson increases the probability of borrowing from the group or
from members by 20 percentage points, which is a very sizeable effect. Column 2 adds two measures
of individual wealth, to control for the need to borrow. The first is a dummy equal to one if the
respondent or a relative owns a house, which for the reasons explained in section 3.1 is rather
uncommon in the informal settlements under investigation. The second is the index of durable goods
described in section 3.2. Neither index turns out to be significant in column 2. While home ownership
can be considered exogenous given the small size of group loans, one could argue that ownership of
durable assets is an endogenous variable or that it is measured with error. Column 4 of table 4
reports two stage least squares estimates of a linear probability model when wealth is instrumented
with three variables: income before joining the group (in logarithm), the number of groups to
which the respondent belongs (excluding the current group), and a dummy equal to one if group
work is the main source of income for the individual. Column 3 reports the uninstrumented linear
12