During the 2012 election cycle, President Barack Obama sent an early fundraising email in which he told prospective donors, “Our campaign doesn't take money from Washington lobbyists or special-interest PACs. We're doing this the right way—with a whole lot of people like you taking the lead” (Bedard Reference Bedard2011). Such a statement is typical in American politics, as candidates stress their popular appeal and/or insinuate that their opponents are beholden to “special interests” if they accept large sums from political action committees (PACs). It is unclear, however, whether touting a funding portfolio marked by more donations from individuals than interest groups is beneficial to a politician. In this article, we therefore engage the following question: Do voters consider the source of campaign funds as a relevant piece of information when evaluating candidates?
We present the results of two survey experiments in which we randomly assigned a funding portfolio to a hypothetical candidate. In general, we find that people do consider the sources of funding when evaluating candidates, but that the effect of this information is often moderated by whether the candidate and voter are of the same political party. Our results have three broad implications. First, in terms of campaign strategy, campaigns may be wise to trumpet their individual contributions and make the interest group donations to their opponents known. Second, in terms of public policy, several states (e.g., Alaska and Massachusetts) have instituted reforms such as real-time electronic disclosure of donors. Our results suggest that making donor information more readily available may inform voters’ evaluations of candidates at the margins. Finally, our findings inform academic debates concerning the effect of money on election outcomes.
DOES IT MATTER WHERE CAMPAIGN MONEY COMES FROM?
Money is a powerful force in American elections. All campaigns must have it, and challengers in particular realize vote returns when they spend it (Abramowitz Reference Abramowitz1991; Gerber Reference Gerber2004; Jacobson Reference Jacobson1990; Reference Jacobson2006; Krasno and Green Reference Krasno and Green1988). Luckily for candidates, donors seem all too happy to put money into the political system, as evidenced by the consistent growth in campaign spending since the 1970s (Jacobson Reference Jacobson2009, ch. 3; Gross and Goidel Reference Gross and Goidel2003; Lott Reference Lott2000). Conventional wisdom holds that candidates use their political skills, favorable conditions, or both to raise money from donors (Maestas and Rugeley Reference Maestas and Rugeley2008). Once raised, this money is spent with the aim of delivering returns in the form of higher vote totals.
Within this framework, the percentage of a candidate's overall funding that is comprised of any single source (for federal candidates: individual contributions, PAC donations, party support, or self-financing) should not be correlated with that candidate's vote percentage. In other words, if money is a fungible commodity in terms of its ability to persuade voters, then it should deliver the same benefit to the campaign regardless of the source from which it is obtained. Yet, controlling for a variety of other factors, prior observational work finds positive associations between PAC (interest group) donations and candidate vote share (Alexander Reference Alexander2005; Brown Reference Brown2013). The dominant explanation for this finding is a selection effect, with stronger candidates being more likely to receive PAC funding in the first place (Brown Reference Brown2013).
In short, this paradigm holds that candidates who do well on Election Day received PAC contributions precisely because they were expected to do well. The strategic decision-making of PACs has received substantial attention in both federal and state elections, where previous research has consistently found that incumbents, leadership, and powerful members of the rank and file benefit disproportionately from PAC donations (Cassie and Thompson Reference Cassie, Thompson, Thompson and Moncrief1998; Herrnson Reference Herrnson1992; Herrnson and Curtis Reference Herrnson, Curtis, Magleby and Corrado2011; Snyder Reference Snyder1990; Stratmann Reference Stratmann1992; Thielemann and Dixon Reference Thielemann and Dixon1994; Thompson et al. Reference Thompson, Cassie and Jewell1994). Consistent with this narrative, Brown (Reference Brown2013) argues that external funding correlates with electoral success because good candidates attract donors and voters alike.
An alternative—or perhaps complementary—explanation is that voters use interest group funding as a cue that indicates a high-quality candidate worthy of their support. For instance, voters might respond directly to information about campaigns’ sources of funding, recognizing a PAC-backed candidate as a “stronger” one. Voters might take cues from other sources of funding as well. If so, there is no shortage of these hints in the political environment. Information about candidate funding sources—or at least, intimations about those sources—is often reported by media and reform-minded interest organizations who disseminate summary financial information for mass consumption, particularly as part of “horse-race” coverage to describe which candidate has raised more money at each quarterly deadline.Footnote 1 Indeed, political campaigns themselves might attempt to use their funding profile to their advantage, as evidenced by the email from President Barack Obama to prospective donors referenced above.
Prior work suggests that campaign finance information can be informative to voters. Lupia's (Reference Lupia1994) seminal study, for instance, found that individuals who knew the insurance industry's position on California's 1988 ballot initiatives on insurance policy—presumably as a result of the industry's advertising campaign—were able to use that information as a cue and vote in the same way as someone who had “encyclopedic” knowledge of the ballot initiatives. More recently, Sances (Reference Sances2013) randomly assigned survey respondents to receive information about political contributions and found that information about the source of those contributions (business or labor) allowed respondents to more accurately place candidates on an ideological spectrum. Other recent experimental work finds that giving voters information about the sponsors of negative ads, such as the identity of major donors to sponsoring organizations, results in the ads having a smaller persuasive effect on voters than if that information is not available (Dowling and Wichowsky Reference Dowling and Wichowsky2013; Ridout et al. Reference Ridout, Franz and Fowler2015). In this article, we address a more general question: Does knowledge about the major sources of funding for candidates—and not the specific ideological source of those funds—matter for how candidates are evaluated?
RESEARCH DESIGN
To assess whether voters respond to information about candidates’ funding sources to evaluate candidates, we conducted two survey experiments that provided a short biography of a fictitious candidate for an open congressional seat.Footnote 2 The first experiment was included on the 2010 Cooperative Congressional Election Study (CCES), an opt-in Internet survey administered by YouGov/Polimetrix, which uses a combination of sampling and matching techniques in an effort to approximate a random digit dialing (RDD) sample (Vavreck and Rivers Reference Vavreck and Rivers2008; see the Supplementary Material for more information). The second experiment used a convenience sample of US residents recruited using Amazon.com's Mechanical Turk (MTurk) interface (see Supplementary Material for more information). Table 1 describes the experimental design and provides question wording for each experiment.
Experimental Designs and Question Wording
The key manipulation of interest in each study is the “funding source treatment.” In the CCES experiment (panel A of Table 1), there were four funding source treatments: the candidate financed “his campaign primarily with. . .” (1) “money he made in the private sector,” (2) “money he inherited,” (3) “contributions from individual citizens and interest groups,” or (4) “contributions from individual citizens.” In the MTurk experiment (panel B of Table 1), the funding source treatment conditions were either that the candidate raised $700,000 from one of three possible sources (his own funds, interest groups, or individual contributions) out of a total of $1.3 million, or that his “campaign is funded with a mixture of individual contributions, contributions from interest groups, and his own money.”Footnote 3 Providing participants with dollar figures (as opposed to informing them directly about the “primary” source of the candidate's funding as in the CCES experiment) requires participants to make their own judgments about the extent to which various funding sources were important in the candidate's funding portfolio.
We also randomized the party of the candidate for two reasons. First, partisanship provides a point of comparison for the size of any funding source treatment effects. Second, party agreement (between the respondent and the candidate) may moderate the effect of the funding source treatments. For example, co-partisans may see many contributions from individuals as a positive indicator, whereas out-partisans may not, as they are likely from individuals with whom the voter does not agree politically. In the CCES experiment, we randomly assigned the party of the candidate to be Republican, Democratic, or not included.Footnote 4 Because there were no significant differences across funding source treatment conditions in the CCES experiment when no party was included (see Figure 1 below), in the MTurk experiment, we simply randomly assigned the candidate's party to be Republican or Democratic.
Effect of Funding Sources on Vote Intent (CCES Experiment).
The design of the MTurk experiment differed from the CCES experiment in two other ways. First, we included an “extra information” condition; participants randomly assigned to this condition read additional biographical material about the candidate (see Supplementary Material). Supplying this information at the end of the vignette more closely mimics the sort of background provided in a news story that a voter might encounter during a real-world election. Second, we included a control condition in which participants saw no information about the candidate's funding sources. This control condition likely reflects the lack of knowledge many voters might have of a typical candidate's funding portfolio, thereby providing a logical point of reference in the experimental design.
ANALYSIS AND RESULTS
We asked several candidate evaluation questions after each vignette. Question wording is included in Table 1 and coding details are in the Supplementary Material. We restrict analysis of both experiments to the respondents who provided responses to all of the questions used in the analysis presented below (756 for the CCES; 1,463 for the MTurk study).Footnote 5 Summary statistics for the outcome measures and demographic and political characteristics for the CCES sample by funding source treatment condition are presented in Table S2, and Table S3 provides the same for the MTurk experiment. Tables S4 and S5 in the Supplementary Material present differences of means on relevant outcomes by treatment condition, and also by whether the candidate shared a partisan affiliation with the respondent (or whether no party information was provided in the vignette).Footnote 6 In the text, we present figures for the vote intent outcome measure only. We present and discuss figures for the other outcomes of interest in the Supplementary Material, although we also highlight some of those key findings in the text.
CCES Experiment
Figure 1 presents the results from the CCES experiment, depicting mean levels of our Vote Intent item (0 = not very likely; 100 = very likely) with 95% confidence intervals. One aspect of Figure 1 that immediately stands out is the fact that individuals are much more likely—on the order of 30 to 40 points—to vote for a co-partisan candidate (middle pane) than one from the other party (right pane), no matter what funding source treatment condition they saw. Moreover, the effect of partisanship is much greater than the effect of the funding source conditions.
Of most interest to us is the difference between a candidate primarily relying on interest group funding and one primarily relying on funding from individuals. In each pane of Figure 1, the interest groups condition rates lower than the “individual contributions” condition. However, none of the differences between the interest groups and individual contributions conditions reach conventional levels of statistical significance (left-to-right across the three panes, p = 0.151, 0.135, 0.130, two-tailedFootnote 7 ). Examining across funding source conditions within the “same party” and “different party” panels does reveal some significant differences with respect to the self-financing conditions, though.Footnote 8 When told that a co-partisan candidate had primarily spent money he had inherited, respondents were about 14 points less likely to vote for him compared to one who was described as receiving funds primarily from individuals (p < 0.01; 14 points represents a shift of almost one half of a standard deviation on the outcome measure). These two conditions are not statistically significantly different when the candidate is of the opposite party from the respondent, however (p = 0.705). Instead, primarily using private sector money resulted in a candidate being less likely to be voted for when he was of a different party from the respondent, by about 10 points compared to the “individual citizens” condition and 12 points compared to the inherited money condition (p < 0.05 for both comparisons). These findings suggest that partisan agreement moderates the relationship between funding sources and vote intent, at least with respect to the different self-financing conditions; co-partisan (but not out-party) voters downgrade candidates who spent money they did not earn.Footnote 9
MTurk Experiment
Figure 2 depicts the results of the MTurk experiment's five (funding source: control, individual contributors, interest groups, self-financing, or mixture) × two (party: same party of respondent or different party) × two (extra information or no extra information) design for the Vote Intent item (−3 = not very likely; 3 = very likely). The two top panes of Figure 2 represent means for when the respondent is of the same party as the candidate, while the bottom two panes show responses for opposite party respondents. Within each pair of panes, the chart on the left is for those respondents who received no extra information and the chart on the right is for those respondents who did receive the extra biographical information.
Effect of Funding Sources on Vote Intent (MTurk Experiment).
As the top-left pane of Figure 2 shows, respondents were significantly more likely to vote for same-party candidates who raised a majority of their funds from individuals, relative to the control group, the interest group condition, the self-financing condition, and the mixture condition (p < 0.05 for all) when no extra information was given. Candidates in the mixture condition also did well when no extra information was given, at least compared to those in the interest group condition (p < 0.05). However, all those significant differences for same-party candidates disappear, and the substantive differences are muted, when respondents were shown extra information, displayed in the top-right pane.Footnote 10
For opposite-party candidates, when no extra information was provided (bottom-left pane), candidates in the interest group condition were significantly less likely to be voted for than those in the control group, the individual contributor group, and the self-financed group (p < 0.05 for all). Opposite-party candidates who received a majority of their contributions from individuals were also more likely to be voted for than those in the mixture condition (p < 0.05). Unlike with same-party candidates, however, many of these differences endured when extra information was provided for opposite-party candidates. Specifically, interest-backed opposite party candidates in the extra information condition (bottom-right pane) were significantly less likely to be voted for than the control group, the individual contributor group, and the self-financing group (p = <0.05 for all). Self-financers also fared better than candidates in the individual contributor group and the mixture group (p < 0.05 for both).
DISCUSSION
Mandated disclosure of funding sources has long been a cornerstone of American campaign finance law, and has gained renewed attention in the wake of Supreme Court decisions in recent campaign finance cases that have resulted in an environment in which independent groups can spend unlimited sums during American elections, sometimes without disclosing their donors (Dowling and Miller Reference Dowling and Miller2014; Kang Reference Kang2010). To combat this new environment, non-partisan reform groups such as the Sunlight Foundation have begun to petition Congress for reforms such as real-time electronic disclosure of contributions and/or mandated donor disclosure for “social welfare” or 501(c)3 organizations; as mentioned above, similar movements have also gained traction in a number of states. The assumption underlying such reforms, consistent with the Court's position since Buckley v. Valeo 424 U.S. 1 (1976), is that knowledge of candidates’ financiers might be an important element in forming preferences about them.
The results of our experiments suggest that voters may incorporate information about candidates’ funding sources in their judgments when such information is provided, and that the candidate's partisan affiliation—relative to that of the voter—often moderates the manner in which this information is utilized. In the MTurk experiment, for example, when the candidate is a member of the voter's preferred party, our findings are not as robust to the inclusion of the extra biographical information. When the candidate is not a member of the voter's preferred party, however, out-party voters are less likely to vote for the candidate who takes a large amount of interest group contributions, even with the provision of the extra biographical information.
The negative association between interest group funding and vote intent for opposite-party candidates stands in contrast to the positive correlation between external (PAC) funding and vote share observed in previous observational work (Alexander Reference Alexander2005; Brown Reference Brown2013). However, we believe that the negative (or, in some cases, null) effects of interest group supportFootnote 11 that we report are likely outweighed by much stronger, positive ones resulting from candidate traits that are difficult to observe—namely, that interest groups contribute to candidates with a good chance of winning.Footnote 12 In other words, we believe that the resolution to the dissonance between our experimental findings and those of observational studies in the area of interest financing is that both are probably correct. There is some risk of a backlash from taking PAC money, but the very fact that a candidate attracts large PAC contributions can be viewed as a proxy for inherent strength that generally makes such a risk worth bearing.
It is important to note that as is the case with all research designs, our survey experiments have their limitations. Our study examines immediate responses to information about candidate funding sources and, in so doing, informs debates about the effect of money on politics. Although these initial reactions are important in advancing our understanding of voters’ preference formation, we cannot say how they are affected in the temporal context of a long political campaign. Nonetheless, we believe that our findings are valid with regard to assessing voter reaction to candidate funding sources. In particular, by providing raw dollar amounts and additional biographical information, the MTurk experiment delivers treatment in a setting designed to closely simulate the conditions in which a voter would encounter information about a candidate's financial backers. Granted, the effects we observe in this lab setting could get “washed out” over the course of an actual campaign. However, they could also be larger, if candidates trumpet their funding sources—or those of their opponents—to a greater degree than our more subtle one-sentence treatments do. Future work could consider the consequences of repeat exposure to campaign finance information, and also include other sources of funding, such as public financing, or vary the amount of funds from the various sources to see if that matters for how individuals evaluate candidates.
SUPPLEMENTARY MATERIALS
For supplementary material for this article, please visit Cambridge Journals Online: http://dx.doi.org/10.1017/XPS.2016.5.
