Call Us: US - +1 845 478 5244 | UK - +44 20 7193 7850 | AUS - +61 2 8005 4826

Microeconomic analysis

Formally, the inequality ð1Þ is less likely to hold for any aggressor-victim
pair, and so—all other things being equal—the probability of attack, as
given by ð2Þ, must come down.
12 Recall that w is linear in average incomes and is therefore bounded above by a fraction
of Y if all incomes in society are smaller than Y. Moving Y down lowers w and must create
a crossover to the human protection technology at some positive level even if w* 5 0. This level
is sufficient for our needs ðit may be far from necessaryÞ. 13 By mixing across individuals who are indifferent between making this change, we can
always make sure that the graph of the protection function is continuous, so that an equilibrium exists.
732 journal of political economy
This content downloaded from on Tue, 9 Sep 2014 10:18:07 AM
All use subject to JSTOR Terms and Conditions
FIG. 3.
—A change in group fortunes: high income.
A, The defense response;
B, equilibrium. In panel
A, a victim’s optimal choice of defense shifts away
from the human mode. In panel
B, this leads to a sizable shift in both the attack and protection functions, so that the probability of attack changes in an
ambiguous fashion.
This content downloaded from on Tue, 9 Sep 2014 10:18:07 AM
All use subject to JSTOR Terms and Conditions
IV. Empirical Analysis
A. Data
Systematic statistical information on outbreaks of religious violence in
India is relatively hard to come by, and our choice of time period is
constrained by the available overlap of conflict data and economic information. On conflict, we use a data set compiled by Steven Wilkinson
and Ashutosh Varshney. ðSee, in particular, the recent use of this data
set in Wilkinson ½2004.Þ It summarizes reports from The Times of India,
a leading national newspaper, on Hindu-Muslim conflicts in India in
the second half of the twentieth century. This data set has information
on deaths, injuries, and arrests. It does not provide hard information
on which side initiated the violence, for in most cases that issue would
necessarily be mired in subjectivity. For every report of Hindu-Muslim
violence, the data set provides the date of incidence of the riot; the name
of the city/town/village; the district and state; its duration; the number of people killed, injured, and arrested; and the reported proximate
cause of the riot.