interpretations will have to be taken with caution, given the small number of observations on which
table 7 is based (there are only 18 different groups in the sample).
Starting from the division of labor, on average 33% of the groups assign the same task to
everybody, 45% allow for specialization, and 22% have different tasks in which the members alternate,
so that no one is systematically assigned to a given task. When we split the sample according to their
ethnic fragmentation (columns 2 and 3) the pattern is different. Highly fragmented groups tend to be
ones in which everyone does the same job, while in relatively homogeneous groups it is more common
to observe specialization in tasks. A possible explanation, which will be further advanced below, is
that when members are heterogeneous it is more di‰ult to reach consensus on something that would
generate disparities in economic treatment or in responsibilities, hence a “flat” rule becomes more
attractive. Notice that this does not hold for heterogeneity in wealth, which, as argued above, is
likely not to be common knowledge among the members. As for education levels (columns 6-7), more
heterogeneous groups display more specialization, while in homogeneous ones everyone tends to do
the same thing. This seems reasonable, given that people with different education are ‘naturally’
suited to different tasks.
Turning to remuneration schemes, groups tend to fall into two broad categories: those that
pay a fixed amount equal for everyone (55.6%) and those that pay in proportion to the number of
items produced (33.2%). Occasionally, they pay according to the number of hours worked (5.6%)
or to the contributions made by the members when the group started (5.6%). When the sample is
split according to ethnic fragmentation, more heterogeneous groups tend to pay a ‘flat rate’, while
more homogeneous ones pay in proportion to output. As above, this can be interpreted as resulting
from problems in reaching agreements that involve differential treatment of heterogeneous members.
Results for wealth inequality go along the same lines, while those for education heterogeneity again
follow the expected allocation of ‘competence’.
Previous studies (e.g., Gugerty and Miguel (2001)) indicate the sanctioning technology as the
key mechanisms through which ethnic fragmentation affects group functioning. The third panel of
table 7 reports data on the fraction of groups that employ a penalty (monetary fine or suspension)
in case of missed contributions and in the case of absenteeism and other irregular behavior. The
17