Crime as a Social Cost of Poverty
and Inequality: A Review Focusing
on Developing Countries
F Bourguignon
case of homicide than in the case of robberies, so that alternative
specifications may be estimated in that case on larger samples. The results
reported by FLL lead to the same general conclusions as above. In addition,
they confirm that crime deterrence variables, essentially police and
conviction rates, have a significant negative influence on homicides.
Convergent findings are reported by Londono and Cuerreo (1998) who
ran fixed effect regressions on homicides in a panel sample of 17 Latin
American countries between 1970 and 1995. The specification that they
chose to estimate is not as complete as FLL, so a detailed comparison is
not possible. But they find sizable effects of poverty and inequality on
homicide also. According to the figures they indicate, a 1 percentage point
increase in the poor population would produce on average an instantaneous
2.5 per cent increase in the number of homicides37. This does not seem
very different from the orders of magnitude seen above.
Having said this, the preceding estimates must be viewed with very much
care. We already have insisted on the natural limitations of pure cross-
sectional exercises. The introduction of fixed effects in samples where
observations of different countries at different points of time are pooled
together certainly should lead to more satisfactory conclusions. In the
present case, however, it must be kept in mind that both in the FLL study
and in that of Londono and Guerrero the corresponding samples of
observations are very limited so that the relevant effects may be estimated
on the basis of a few observations. All this definitely points to the need for
better and more consistent data being regularly collected on crime and
victimization, both across and within countries over time.
The Social Cost of Crime and Inequality:
Rough Estimates
Given the preceding evidence of a Iikclypositive association among crime,
poverty and inequality, we now seek to measure the social cost of crime
and then that part of the social cost of inequality which goes through
37 It would certainly be worthwhile to obtain estimates based on this sample comparable to those
given by FLl..
90