1 Introduction
Recent years have seen a growing interest on behalf of economists and other social scientists in the role
played by groups in the process of economic development. Drawing upon the conceptual framework
of Putnam (1993), some studies have looked at civic engagement in a variety of associations, including
recreational and socio-political ones, to argue that the mere participation in such groups can have
an economic impact by providing opportunities for members to share information, enforce informal
transactions, and coordinate on cooperative outcomes.1 Other studies have focused on the role
of groups in the informal credit market and have analyzed their incentive schemes and economic
performance.2 For some of the poorest individuals in developing countries, however, groups are more
than socio-political associations or saving devices: they are ‘employers’. People who do not have
access to the formal labor market and whose options in the informal market are relatively unattractive
can often benefit from pooling resources and working in groups. To what extent can informal groups
constitute a reliable source of income for the poor? What factors affect group performance, and in
particular, how does the ‘social’ composition of the group affect the organization of production and
the allocation of resources to members?
This paper attempts to address the above questions by employing a unique dataset on ‘self-help’
groups with income generating activities collected by the author in the informal settlements of Nairobi
in 1999. Information has been collected on each and every member of the surveyed groups, which
allows to construct exact measures of the composition of the group in terms of income, education, age
and ethnicity. This is particularly important when investigating the impact of heterogeneity on group
performance. The advantage of this methodology compared to the studies that infer within-group
heterogeneity from the heterogeneity of the population at large is that it accounts for the possibility
that people sort into groups that are more or less heterogeneous than the whole population. By having
a ‘census’ of the entire group, the matching between group composition and individual outcomes can
1In the context of developing countries, see for example the work of Narayan and Pritchett (1999), Grootaert (1999),
and Isham (2001).
2 See the surveys by Besley (1995) on ROSCAs and other nonmarket institutions, and by Ghatak and Guinnane
(1999) on group lending.