Since Downs (1957) introduced the idea of the rational voter, there have been numerous empirical studies to test the construct. Typically, these studies have employed cross-section data to ascertain the predictive capacity of various demographic and election-specific variables on the probability of voter participation. These studies have usually failed to produce robust empirical support for the hypothesis. Indeed, students of rational choice theory have yet to find a satisfactory solution to the paradox of voting (Brazel and Silberberg 1973; Ashenfelter and Kelly 1975; Aldrich and Simon 1986; Cox and Munger 1989; Green and Shapiro 1994; Lapp 1999; Greene and Nikolaw 1999).
In perhaps one of the more comprehensive empirical investigations on the subject to date, Matsusaka and Palda (1999) employ nearly 40 different variables in a cross-section analysis of voting behavior in an effort to quantify the degree to which voting behavior can be explained. Despite the number of variables employed, they find the explanatory power of these variables to be very low. Traditional demographic variables can explain no more than 15 percent of voter turnout variation. Matsusaka and Palda conclude that "most of the inability to predict who votes appears to come from non-stationary factors" (1999: 442) and go on to suggest two possible paths for future research. One suggestion is to search for new nondemographic explanatory variables. This suggestion is echoed and even pursued in the more recent study using LOGIT techniques on micro data for even-numbered years from 1986-1996 by Copeland and Laband (2002). The second suggestion is to use aggregated voting data, which might allow the estimation of models with increased explanatory power. Both of these recommendations are followed in the present study.
In a related study, Greene and Nikolaw's (1999) empirical results do not support the redistributive theory of the state, a subset of the rational voter hypothesis of voting. Greene and Nikolaw state that there is "considerable doubt that the theory of the re-distributive state can help explain the pattern of voter participation rates across individuals" (1999: 224). They also identify a pattern of declining voter participation and further note that "cross-section results on counties do not control for time" (1999: 224).
In effect addressing the latter point, the present study provides a time-series analysis of voter participation rate determinants.
This investigation seeks to provide an additional dimension to the empirical study of voter participation rates. In particular, the purpose of this study is to empirically investigate determinants of aggregate voter participation rates over time. Unique to the present study is the inclusion of political action committee (PAC) contributions to congressional election campaigns, which arguably could reduce voter participation by reducing the expected benefits from voting. PAC contributions to congressional election campaigns grew very rapidly over the 1980s and 1990s. It is argued in this study that, as a result of this growth, prospective voters may have become dissuaded from voting because they expect that the influence resulting from these PAC contributions will strongly influence the actions of elected officials in favor of special interest groups, and that voters' political preferences will be subordinated accordingly.
Also arguably unique to this study are (1) the use of a dissatisfaction index, (2) the inclusion of two "money" factors, too slowly growing real GDP and excessive inflation, and (3) the use of aggregated (macro) time-series data. The dissatisfaction index is constructed as an equally weighted average of three normalized indices reflecting responses to the University of Michigan's Institute for Social Research (ISR) surveys concerning whether government officials as a whole (i.e., not only elected ones but also nonelected ones) can be trusted to do their jobs, whether they are dishonest, and whether government wastes tax dollars. …