Ecological Regression and the Analysis of Past Politics

Abstract

Regression estimation of cell entries in contingency tables is among the most useful statistical techniques for political historians. Developed in the 1950s by statisticians who were attempting to circumvent the so-called "ecological fallacy," regression estimation has received a good deal of attention in other social scientific disciplines, but surprisingly little in history. In a recent article in this journal, Jones provided a short introduction to the technique pioneered by Leo A. Goodman. In addition, Jones tested its accuracy by comparing survey results with estimates of voting behavior from the 1960 presidential election, and urged the historical profession to utilize the Goodman procedure. Although helpful as far as it goes, Jones's paper does not treat the theory, mathematical background, and assumptions of the method in sufficient detail to enable historians to employ it creatively. Nor does he offer the researcher advice on how to deal with typical difficulties which arise in an actual analysis of past data-for example, the problems of nonlinearity and logically impossible estimates.