The rational voter: an analysis of two Atlanta referenda on rapid transit

In this paper we have analyzed voting behavior in two rapid transit referenda with the objectives of testing the rational voter hypothesis and providing insights into voter preferences for public mass transit. A framework was formulated based upon previous voter behavior models but applied to referenda on public transit proposals. The framework focused on three different variables — quantity of transport, tax/price of transport and the quality of transportation services. From this fairly simple framework, several testable hypotheses were drawn. The analyses of the referenda were performed on voting-precinct level data obtained from two rapid transit referenda held in Atlanta in 1968 and 1971. The data were especially interesting since there were few, but well-defined, differences in the two proposals — in 1968, only rapid rail was emphasized and property taxes were to be used to finance the local portion of the costs, while in 1971 both rapid rail and a substantial increase in bus services coupled with a decrease in bus fares was proposed, along with the use of a local sales tax to finance the local portion of the costs. The principal hypotheses stemming from the voter behavior framework were substantiated in the weighted logit regressions used to analyze the voting pattern. Current use of transit (measured by relative importance of bus ridership in commuting to work) and the relative importance of the CBD as a work place (measured as the percent of those in a precinct working in the CBD) both positively affected the proportion of votes in favor of the issues. The distance from the train station relative to the distance to the CBD (a measure of the benefits of the rail system) was negatively related to the percent voting as hypothesized. The absolute distance to the CBD (a measure of the benefits of the bus and rail systems) had the hypothesized parabolic relationship. In addition, although the framework does not produce an unequivocal prediction of how income might affect the outcome of the referenda, the empirical results suggested that income was positively associated with affirmative votes on the issues. We also considered further variables which influenced changes in voting patterns. The outcome of the 1968 proposal within a precinct proved to be very important in explaining the 1971 vote. Bus ridership was also an extremely important positive factor in determining proportions voting in favor of the 1971 issue. This then lends credence to the suspicion that in planning the 1971 transportation package, MARTA helped to insure a positive vote by offering enough to everyone (i.e. current bus riders, potential bus riders as well as potential rapid transit riders or its beneficiaries) so that it more than offset the increased local costs and use of a sales tax. The regression results strongly support the hypothesis that individual voters act in their own economic self-interest. Not only are the variables that reflect the benefits and cost of proposed transit systems significant in explaining voter behavior in the two Atlanta elections, they also explain differences in results in the two referenda. The results also lead to rejection of the non-economic arguments regarding voter behavior including the public regardedness hypothesis. Finally, the analysis suggests that passage of public transit referenda in Atlanta was dependent upon constructing a package of features designed to attract the support of a broad cross section of the electorate. Copyright Martinus Nijhoff Publishers b.v 1978