Crime as a Social Cost of Poverty
and Inequality: A Review Focusing
on Developing Countries
F. Bourguignon
These estimates, which are essentially based on the longitudinal dimension
of the data, confirm the results of the cross-sectional analysis and put
into evidence additional effects. The comparison is not totally valid since
the samples of countries used in each case differ due to distinct data
requirements. Nevertheless, such a coincidence between cross-sectional and
longitudinal estimates is somewhat remarkable and would suggest that the
phenomena put into evidence by all these regressions are rather robust.
This seems to be true, first of all, for the effect of income inequality upon
criminality. This effect is significant and substantial both for homicides
and robberies. In the short-run a 1 percentage point increase in the Gini
coefficient would produce on average in the countries included in the sample
a 3.6 per cent increase in the homicide rate and a 1.1 per cent increase in
the robbery rate. However, this effect is much stronger in the long run
because of the compounding effect of hysteresis in crime rates. The
coefficients of the lagged crime rate are such that the effect of inequality
would be multiplied by 3 for homicides and by 7 for robberies3'1. Ifone has
in mind major changes in inequality’ like those observed recently in the
1980s in the US and in the UK, but also in several Latin American countries,
increases of the Gini coefficients of 3 or 5 percentage points are not
unreasonable orders of magnitude for periods extending over 5 years or a
little more15. Other things being equal, this would correspond to increase
in crime rates from 40 to 60 per cent at a horizon of 10 to 15 years, a
rather frightening order of magnitude. This long-run effect may be
somewhat biased since the data sample does not include national time
series long enough for a satisfactory representation of the complete
dynamic processes governing crime rates.
A second effect discussed above in connection with the cross-sectional
model is that of the urbanization rate. It still fails to be sizable and
significant for homicides, but it is very much so for robberies. For the
latter, a once and for all 1 percentage point increase in the urbanization
rate by 1 per cent produces in the long run an increase in the nationwide
crime rate equal to 8 per cent. Although the full dynamics of this process
34 These mu Itiplicative factors are simply the inverse of 1 - I, where 1 is the coefficient of the lagged
crime rate in Table 1.
35 See Atkinson. Gottsehalk and Smeeding ( 1995). and Gottschalk and Smeeding (1998). for series
of Gini indices in OECD countries and Morlcy ( 1995) for Latin American countries.
88