is also negatively related to ethnic fragmentation, while that of sanctioning absenteeism and other
irregular behavior is not. Finally, no pattern emerges between group composition and the criteria
for recruiting new members.
The remainder of the paper is organized as follows. Section 2 briefly reviews the related lit-
erature. Section 3 describes the setting, starting with a description of the informal sector and of
self-help groups in Kenya and then moving to the specific data used in this paper. Section 4 con-
tains econometric results on income generation, access to credit, and the organization of production.
Section 5 concludes.
2 Background literature
The issue of heterogeneity and group participation has received increasing attention in recent years.
Alesina and La Ferrara (2000) analyze the impact of racial, ethnic, and income heterogeneity on
individual propensity to join socio-political associations in the US and find that it has a negative
effect. La Ferrara (2001) models the role of wealth inequality in villages where people have the
choice of joining groups that provide different net benefits to the rich and the poor, and shows that
when access is unrestricted, higher village inequality translates into lower participation because the
relatively rich opt out of the groups. Both papers differ from the present work in two respects. First of
all, they address the question of how groups form (what determines the likelihood that an individual
will join), and not how they function. Secondly, they relate group formation to “society wide”
heterogeneity as opposed to within-group heterogeneity. In fact, the link tested by those papers is
from inequality (or racial fragmentation) in the “pool” from which members are drawn, to individual
choices regarding group participation. While this paper employs exact indexes of within group
heterogeneity, it has the limitation of taking membership as exogenous, without accounting for the
motivations that may induce people to join homogeneous versus heterogeneous groups. Developing
a comprehensive framework (and suitable data) to account for both group formation and internal
group composition and functioning seems an important task for future work.
A strand of the literature which is closer to the current paper is the one on heterogeneity and
group performance. This has been explored in the context of project maintenance by Khwaja (2000),