New Methods for Conflict Data

Abstract

This chapter sketches out some new ways to look at conflict data sets. Since political scientists have more computer power available to them than at perhaps any point in the past, the paper emphasizes methods whose principle feature is that they purchase substantive realism at the cost of more compute cycles, not more advanced statistics. Existing statistical theory is sufficient to perform much more realistic analyses than are typically performed, but it is not necessarily found in the standard location. Most of the models and methods described here can be found in other guises in the field of machine learning. Political methodology is often accused of importing techniques wholesale from other disciplines, particularly econometrics, and by introducing machine learning as another field worth mining, this paper continues a long tradition.