Ecological Inference from Goodman to King

Abstract

Since it was introduced to historians nearly three decades ago, a statistical technique known as ecological regression has been widely used to analyze aggregate election returns and similar data in history, political science, and law, and methodologists have discussed problems with, extensions of, and alternatives to the technique. The literature has become so vast and complicated, and recent contributions to it are of such importance, that it is time for a comprehensive review. This article provides that review, starting at an elementary level, sorting through the major arguments and evidence, and explaining the nature of the most complicated as well as the simplest methods. Taking an intuitive, rather than a statistically rigorous approach, the article is aimed at historians and political scientists, particularly graduate students, who already have some statistical knowledge. Lawyers and expert witnesses in voting-rights cases may also find the article useful.