Taft Foster and Dr. Eric Eide, Economics Department

Within the economic and political science literatures, a variety of theoretical models have been developed which attempt to describe the role of campaign contributions in the political process. These models can be separated into two groups, those which employ the assumption that campaign contributions can purchase legislative votes and those which do not. Proponents of this assumption, known as vote buying, argue that legislators reward special interest groups for their campaign contributions by voting on laws in a way that will benefit the contributor. Though widespread, the use of this assumption remains controversial despite many attempts by researchers to examine its validity. The first obstacle faced by these researchers is that, due to the nature of the problem, they are unable to test the hypothesis by performing a controlled experiment. As such, researchers must search for evidence within existing campaign finance and legislative voting data. Using a variety of statistical techniques, early research seemed to confirm the vote buying hypothesis by identifying a statistically significant correlation between campaign contributions and legislative voting behavior. However, these results have come under much criticism due to the possibility that campaign contributions may be what economists refer to as an endogenous variable.

A variable is endogenous if it is correlated with another variable that the researcher cannot observe, and this unobservable variable is also correlated with the phenomenon that the researcher is attempting to explain. If a variable is endogenous, the results associated with it cannot be trusted. In the case of vote buying, the researcher is trying to explain the voting behavior of legislators using campaign contributions. Thus, campaign contributions would be endogenous if they were correlated with something that researchers cannot observe that is also correlated with legislative voting behavior. It has been suggested that this unobservable variable could be the beliefs of legislators. In other words, if special interest groups contribute to politicians because they share similar beliefs, then campaign contributions would be correlated with politician’s beliefs, and those beliefs would also be correlated with the politician’s voting behavior. As a result of this possibility, the discovery of a statistically significant correlation between campaign contributions and voting behavior does not necessarily imply that vote buying is taking place.

In some cases, the endogenous variable problem can be overcome through the use of a method known as instrumental variables. However, this technique is unlikely to produce statistically significant results using a traditionally small campaign finance data set. Fortunately, my approach to examining the vote buying hypothesis is designed to circumvent the endogenous variable problem without the use of instrumental variables. Suppose, as the endogenous contributions argument goes, that the correlation between campaign contributions and legislative voting behavior exists simply because special interest groups contribute to candidates who hold views similar to their own. If this is the case, then, on average, we would expect legislators to vote consistently on similar laws regardless of changes in the amount of contributions that they have received from certain groups, assuming the politician’s beliefs remain constant. On the other hand, if changes in the amount of contributions received from certain special interests can be shown to be correlated with changes in legislative voting behavior, then this would weaken the claim that campaign contributions are endogenous. To my knowledge, my research represents the first attempt to identify this type of correlation.

Using data from the official websites of the U.S. Senate and the Federal Election Commission, I constructed variables indicating whether or not each senator’s most recent vote differed from their vote on the previous trade bill. This involved the creation of two variables, one that indicated if they had changed from a previous vote of “yea” to a vote of “nay”, and another indicating the converse. From previous research of mine, I had also learned that the effects of campaign contributions become more statistically significant when expressed as a percentage of total committee contributions received by the candidate. This implied that a simple difference from one year to the next would not be an adequate measurement of the change in contributions perceived by the senators. After considering several alternatives, I decided that a reasonable formula would be the actual change from the previous year divided by the sum of the total contributions from committees in both years. The other control variables in my data set included each senator’s political party, the voting behavior of the other senator in each state, the country involved in each trade agreement, and each senator’s years of tenure.

My final estimates of the effects of varying campaign contributions on legislative voting behavior were obtained using ordinary least squares (OLS) and probit estimators. Due to the fact that only 20 senators changed from a vote of “nay” to “yea” over the two year period in my sample, I was unable to obtain any significant results using that variable. However, using the 132 applicable observations of the “yea”-to-“nay” variable, my regressions returned some interesting results. The estimates for several special interest groups were statistically significant, and in all of these cases the signs of both the OLS and probit estimates were the same. To correct for heteroscedasticity, the t-statistics for my estimates were calculated using robust standard errors. Specifically, my estimates suggested that two special interest groups, the Communications Workers of America and the International Association of Machinists and Aerospace Workers, influenced the voting behavior of senators through their campaign contributions. In the case of the Communications Workers of America, my estimates revealed that, on average, the larger the increase in contributions from this group, the less likely senators were to change their vote from “yea” to “nay”. On the other hand, large changes in contributions from the International Association of Machinists and Aerospace Workers increased the probability that senators changed their vote. These estimates were statistically significant at the one and five percent levels, respectively. It should also be noted that since contribution changes are measured as a percentage of total contributions, these effects were greater for senators who experienced contribution changes that were large relative to the total amount of money that they had received from committees.

While these results are by no means definitive, if correct, they lend support to the vote buying assumption by casting doubt on the idea that campaign contributions are endogenous. Future empirical work might build on this study through the use of a more comprehensive data set that contains more control variables, such as information about the demographics of each senator’s state. However, the small number of observations in campaign finance data sets remains a problem.