Inequality and Political Violence Revisited

In their 1987 article in this Review, Muller and Seligson used logged ordinary least-squares (LOLS) to estimate the effect of income inequality on cross-national levels of deaths by political violence. T. Y. Wang challenges the robustness of the main conclusion and argues for the application of a maximum likelihood approach—the exponential Poisson regression (EPR) model—rather than LOLS. He concludes that the widely used LOLS approach yields misleading conclusions when applied to event count data. Dixon, Muller, and Seligson replicate previous work using both LOLS and EPR approaches and conclude that in most—but not all—respects the two approaches yield similar results, supporting the effect of inequality when the specifications are identical. They also argue (in response to concerns expressed by Brockett 1992) that the inequality results are robust when account is taken systematically of the best information on underreporting of deaths.