The coe¢cient on this variable is positive but not statistically significant.8
A second aspect to consider in order to assess the role of these groups for income generation is to
what extent they offer better earning opportunities compared to individual jobs in the informal sector.
When asked what was the most important reason for joining the group, 39% of the respondents said
that they did not have another job, 25% said that they had a job but the job the group was better,
19% indicated the access to side benefits such as health or training opportunities, and the rest gave
other reasons. At first sight, then, the answer to the above question is not obvious. A possible way
to rephrase the question is to ask to what extent people are dependent on these groups as their main
source of income, and estimate the determinants of this ‘dependency ratio’ through multivariate
analysis.
[Insert table 3]
The dependent variable in table 3 is the ratio of individual income earned in the group over total
individual income, expressed in percentage. The sample average for this variable is 75%, indicating
a fairly high dependence on group earnings. The controls in columns 1 to 3 are the same as those
introduced in table 2. The single most important correlate of group dependency is gender: other
things equal, the share of women’s income coming from the group is about 11 percentage points
higher than that of men. Though not statistically significant at conventional levels, the coe∏icients
on the age variables show that group dependency declines as members become older, suggesting
that these groups may serve a particularly valuable function for young people who do not have
access to alternative jobs. Column 4 attempts to control for alternative employment opportunities
by introducing among the regressors the income that the individual was earning before joining the
group. Interestingly, this variable is negative and highly significant: people who were earning more in
the past are less dependent on the group to earn a living. Notice also that from column 4 dependency
decreases with education up to Form 4 level, as one should expect.9
8 The variable ‘Chair’s language’ is an interaction between a dummy equal to 1 if the language of the respondent is
the same as that of the chariperson, and a dummy equal to 1 if the respondent declares that ‘leaders decide’ in the
group. Using only the former dummy also yielded an insignificant result.
9 The co∏cient on Advanced education cannot be estimated precisely due to the small incidence in the sample (only
2.6%) of respondents with a technical or college degree.
11