In this paper, we investigate the impact of security expenditures from military alliances involving third-party intervention on violent incidents from non-state actors. Our main learning is a rather surprising fact that at a lower level of security expenditure in the violence affected area, an increase in security expenditures leads to an increase in violent incidents (rather than a decrease); and only at higher level of security expenditure in the area, an increase in security expenditure leads to a decline in violent incidents. For the analysis, we use a novel dataset on naxalite violence obtained directly from the police head-quarters of the three most affected states in India. The data consists of 64 districts spanning over years 2001 till 2013 and includes information on the annual number of violent incidents and the size of the security forces allocated specifically to curb the naxalite violence. We use negative binomial regression model with the number of violent incidents as the dependent variable and lagged size of security forces as the independent variable, while controlling for other relevant variables. Further, to address issues of potential reverse causality, we use a propensity score matching technique to infer the causal nature of such an association.

We also argue that when the union government intervenes as a third party to support the state governments to fight the naxalites, the latter may be under-funding by free riding. However, despite such underfunding, if the overall contribution is positive and continued, eventually the concentration of security forces in an area will overwhelm the naxalites and reduce the incidents.

Lastly, we highlight that though the mainstream literature on civil wars has used per capita income as a proxy for security expenditures, it becomes inappropriate when a more direct measure of security expenditures is used.