Empirical Analysis of the Influence of Strategic Nuclear Weapons on Interstate Conflicts

This work involves an econometric analysis of the factors influencing the occurrence of militarized interstate disputes (conflicts). The empirical analysis was conducted based on the richest available dataset which was previously used by several researchers from leading US universities. There is the problem of perfect separation in the data for econometric estimation of the effects of the presence of nuclear weapons on the probability of interstate conflicts. This problem arises in the case of modeling of the probability of war between nuclear states. This makes impossible to estimate coefficients in discrete choice models such as binomial logit model. Technical aspect of this problem is solved by applying count outcomes regressions. Therefore, the modeling of the number of conflicts was done instead of the modeling of probability as in previous works on this topic. Another major advantage of this work is a comprehensive justification of the choice of methods as well as the comparison and explanation of the results for various panel data approaches. Estimation and comparison of coefficients was conducted for the fixed effects models and the population-averaged GEE models also known as between estimators. Additional specification for within-group correlation in the GEE model was applied in order to improve efficiency of estimates. Consistent estimation in the presence of unobservable individual effects correlated with explanatory variables appeared to be the main motivation for the usage of fixed effects models. These effects are also the cause of the inapplicability of random effects models for this task.