Some Problems in Testing Two Rational Models of Participation

Two rational choice models of participation conflict over whether there is an expected benefit or a pure benefit component to participation (as modeled by the calculus of voting and a minimax regret model respectively). Data from the 1952-1972 elections are used to examine this difference. Turnout data for any given election are inconclusive on this point, our results being more indeterminate than others (who usually de-emphasize expected benefits) suggest. Differences between elections clearly support an expected benefit model. We also apply our procedures to other forms of campaign participation. Such data lead to the opposite conclusion, supporting the minimax regret model. The analysis then raises several problems associated with testing formal models, such as measurement error, validity and reliability of measures, etc. It is shown that these problems can materially affect conclusions.