Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This paper evaluates the ability of common explanatory variables to predict who votes. Logit voting regressions are estimated with more than three dozen explanatory variables using survey and aggregate data for the 1979, 1980, 1984, and 1988 Canadian national elections. The authors find that the usual demographic variables such as age and education, and contextual variables such as campaign spending have significant effects on the probability of voting, but the models have low R-square's and cannot predict who votes more accurately than random guessing. They also estimate regressions using past voting behavior as a predictor of current behavior, and find that although the explanatory power rises it remains low. This suggests that the difficulty in explaining turnout arises primarily from omitted time-varying variables. In some sense, then, it appears that whether or not a person votes is to a large degree random. The evidence provides support for the rational voter theory, and is problematic for psycho/sociological approaches. Copyright 1999 by Kluwer Academic Publishers