Voter Information and Campaign Finance: How News Coverage Affects Contributions

Transcription

1 Voter Information and Campaign Finance: How News Coverage Affects Contributions J. Baxter Oliphant Bryn Rosenfeld Lizette M. Taguchi July 16, 2014 Abstract Using a plausibly exogenous measure of press coverage, we provide the first rigorous empirical test of a classic result in formal models of electoral competition: that campaign contributions increase with the share of uninformed voters. We show based on an analysis of U.S. campaign finance data from that House members who represent low-information districts raise more money from political action committees (PACs) than incumbents who serve higher-information districts. However, we also demonstrate that voter information affects PACs contributions differently, depending on the nature of their agenda. Groups with narrower (corporate) agendas give more when voter information is low, while donations by more broadly representative groups (e.g. organized labor) are unrelated to a district s information environment. We wish to thank Kosuke Imai, Tali Mendelberg, Michael Donnelly, and participants of the 2011 Society for Political Methodology Annual Meeting and the 2012 Midwest Political Science Association Annual Meeting for valuable feedback on previous versions of this paper as well as David Strömberg and Jim Snyder for providing their data. We are indebted to the Center for Responsive Politics for their campaign finance data. Ph.D. Student, Department of Politics, Princeton University, Princeton NJ oliphant@princeton.edu Ph.D. Student, Department of Politics, Princeton University, Princeton NJ brosenfe@princeton.edu Ph.D. Student, Woodrow Wilson School, Princeton University, Princeton NJ ltaguchi@princeton.edu 1

2 Introduction Though two decades have passed since Baron (1994) observed that substantial campaign expenditures and considerable voter ignorance are among the most salient characteristics of U.S. elections, his observation still rings true. Campaign expenditures have continued to grow, while the majority of voters remain uninformed about and uninterested in congressional politics. This paper investigates the empirical relationship between interest groups campaign contributions and the electorate as mediated by press coverage of Congress. Our central question is the following: What is the effect of voter information 1 on interest group donations? The link between voter information and campaign finance matters for normative debates about accountability, transparency, and representation in congressional elections. Does less information about congressional politics mean more special interest money in elections? If interest group donations are used to inform the electorate, do some types of interest groups donate more than others to fill informational gaps? Does an inverse relationship between voter information and campaign donations by corporate interests help to explain poorer quality representation? While an exhaustive answer to all of these questions is beyond the scope of this paper, we shed light on many of these connections using a novel measure of voter information and comprehensive data on campaign contributions from Despite its important normative implications, there has been little empirical research to date on the relationship between voter information and campaign contributions. A major stumbling block is the inherent endogeneity of voter information to campaign contributions. Consider how voters gain information about congressional politics. Most voters learn about their House member both from objective news sources and campaign advertisements. 1 Throughout, we use the terms press coverage, information environment, and voter information interchangeably to refer to our independent variable of interest. Evidence supporting the interchangeability of these terms for the measure we use can be found in Snyder & Strömberg (2010). The operational measure is described below. 2

3 Clearly, campaign advertisements are related to campaign contributions, and, hence, campaign contributions are likely to be correlated with voter information. Survey-measured, individual-level data on voters knowledge of congressional politics suffer from similar endogeneity problems. Thus, to obtain an unbiased estimate of the effect of voter information on campaign contributions it is necessary to separate the reciprocal effect in which campaign contributions are used to inform and persuade voters. Building on the work of Snyder & Strömberg (2010) and others, we employ a measure of voter information that is plausibly exogenous to campaign contributions and captures the degree to which media markets and congressional districts are overlapping, with press coverage of Congress increasing in the extent of overlap. We describe this measure together with our strategy for identifying the causal effect of voter information on campaign contributions at length below. Relevant Literature Much of the existing campaign finance literature focuses on the link between campaign contributions and candidate success (Chappell 1982, Ashford 1986, Currinder 2009) or on the relationship between donations and the legislative behavior of House members once in office (Wright 1990, Ansolabehere, de Figueiredo & Snyder 2003). In addition, a number of scholars address the pros and cons of campaign finance reform (Goidel, Gross & Shields 1999, Coleman & Manna 2000) or analyze the voter welfare effects of various campaign finance proposals (Coate 2001, Coate 2004, Ashworth 2006). More closely connected to our own research, the political economy literature offers two related ways of thinking about the relationship between interest-group contributions and voter information. First, and most fundamentally, politicians solicit interest group contributions in order to influence voters (Baron 1994, Morton & Myerson 1992). Contributions are used to persuade 3

4 uninformed and less informed voters, whose votes can be purchased, in essence, through political advertising. 2 It follows that the larger the share of uninformed voters in the total voter population, the more productive are an interest group s funds in buying votes for favored candidates, and, therefore, the larger are its campaign donations all other things equal (Baron 1994, Grossman & Helpman 1996). In this tradition, influential models of electoral competition, such as Baron s (1994), imply an inverse relationship between voter information and campaign contributions. Again, the negative relationship between voter information and campaign contributions arises in Baron s model, because candidates in districts with many uninformed voters have an incentive to attract contributions from interest groups and use these donations to inform and persuade. Although the basic prediction of a negative relationship between voter information and campaign contributions has long been posited in the theoretical literature, empirical tests are lacking. Our first contribution is thus to test whether this relationship holds empirically, using more than a decade and a half of data on contributions to members of the U.S. House of Representatives and a plausibly exogenous measure of voter information. Second, another line of existing research views interest group contributions as signals of a candidate s competency. In models of electoral competition such as Prat s (2002), interest groups possess insider information about candidates quality. Based on this information, they contribute to candidates in exchange for favorable policy positions. The level of a candidate s campaign spending (and not the content of political advertising, per se), thus, serves as a costly signal to uninformed voters about a candidate s competency. One shortcoming of this literature, we argue, is its failure to theorize differences in the strategic interaction between candidates and political action committees (PACs) which rep- 2 Zaller (1992) makes a distinction between inattentive, uninformed public who are essentially unpersuadable and the less informed, slightly attentive public who pay just enough attention to possibly be persuaded. The persuadable uninformed voters in these theories, empirically, are Zaller s moderately informed, moderately attentive citizens. 4

5 resent more and less encompassing interests. By encompassing, we mean groups, such as labor PACs and some types of membership PACs, whose interests are more closely aligned with those of the median voter. We define narrow interest groups as those whose interests are less closely aligned with those of the median voter examples of these are corporate and trade PACs. Most scholars seem to assume that all PACs fall into the latter category of groups whose objectives differ from the median voter s objective. 3 We agree that campaign contributions come from groups whose preferences are often at odds with the preferences of the majority of voters, but we acknowledge that in this regard interest groups can be arrayed along a continuum. We begin with the assumption, shared by others in the literature such as Ashworth (2006), that contributions are obtained in exchange for promises to pursue policies favorable to the contributors, if the candidate is elected, and that voters understand this and form expectations about the extent to which a candidate is beholden to special interests. Is it realistic to assume that voters discriminate in this manner? As Ashworth points out, The fact that candidates who do not accept money from PACs make a point of emphasizing that fact in their campaigns suggests that at least some voters will look more favorably on them if they have not made deals with interest groups (57). Given this, we expect strategic divergence in campaign contributions by groups which represent more and less encompassing interests. Our second contribution, then, is to demonstrate that while corporate PACs (with less encompassing interests) contribute more in districts with a larger fraction of uninformed voters, there is no evidence that labor PACs (with more encompassing interests) share this tendency. 3 For an explicit formulation of this idea, see Prat (2002, p. 999). 5

6 Accountability and Representation Why should we care about the combination of voter ignorance and substantial campaign contributions? Is there evidence that accountability and representation are diminished in those districts where these factors coincide? Evidence suggests that PACs aim to influence the legislative process, with industries that stand to gain greater potential benefits from government assistance contributing systematically more than others (Grier, Munger & Roberts 1994, p. 911). Moreover, PACs purposively invest in representatives who are more likely to offer a return (Gordon, Hafer & Landa 2007, p. 1057). While there is a great deal of debate in the empirical literature about whether contributions affect congressional roll-call votes (see Ansolabehere, de Figueiedo, and Snyder 2003 for a recent review), contributions may nevertheless elicit favors in other kinds of legislative activities. Hall & Wayman (1990), for example, find that contributions buy members time, rather than their votes. According to these authors, the effects of moneyed interests appear not on the floor but in committee and are evinced by members patterns of legislative involvement. 4 Interest groups goal, Hall and Wayman write, is not simply to purchase support but to provide incentives for supporters to act as agents (802). Representatives whose voters have less information about their behavior are a more enticing target and have a freer hand to act as agents for favored interests. Previous research suggests that local media outlets are more likely than national outlets to provide information about members of Congress and that local coverage plays an important role in shaping voters knowledge about their member of Congress and in influencing how congressional members vote (e.g. Hutchings 2003, Arnold 2006). It follows 4 Like Hall & Wayman (1990), Esterling (2007) argues that contributions change how elected officials approach committee work. By examining the characteristics and campaign finances of representatives participating in hearings on Medicare, Esterling finds that representatives who are more analytical and technical raise more hard money from PACs. Conversely, representatives who focus on the symbolic aspects of policies, aspects that are of interest to constituents, tend to receive fewer contributions (106). 6

7 that representatives from low-information districts will see the costs of acting as agents for interest groups as lower than members from high-information districts. Their activities on behalf of constituents gain them less because they are less likely to be covered by the local press. As such, representatives from low-information districts will be less motivated to engage in constituent service activities and allocate more time to advancing interest groups agendas, which in turn furnish them with greater campaign contributions. Indeed, Snyder & Strömberg (2010) find that voter information affects politicians actions in Congress as well as their policy choices. In particular, representatives in low-information districts provide fewer constituent services; they deliver to their districts less federal money, make fewer appearances before constituent-oriented committees, and are less likely to vote against the party line than representatives in high-information districts. Our analysis shows that these same underperforming representatives receive higher levels of interest group donations, especially from corporate and trade PACs. Theory We contend that interest groups campaign contributions do not always serve as a positive signal to voters about a candidate s quality. Instead, campaign contributions signal what types of policies the candidate is likely to pursue in office. Put differently, campaign contributions serve as an indicator of the extent to which interest groups are likely to exert influence on a candidate s legislative activities. These signals can be perceived positively or negatively by voters. Whether a contribution is perceived positively or negatively by an individual voter depends on the degree to which her preferences are consistent with those of the interest group. Contributions by popular membership organizations are thus likely to be perceived as positive signals. Of course, collective action problems make it relatively more difficult for large groups to 7

8 organize, contributing to a situation in which the vast majority of interest groups do not represent the preferences of a majority of voters in a district. In particular, the largest share of interest-group contributions come from corporations and trade organizations, for which a central priority is often to gain particularistic benefits, subsidies, or influence on regulatory policy. These preferences tend to be at odds with those of a majority of voters (or are, at least, viewed suspiciously by them). Hence campaign contributions from such interest groups are more likely to be perceived as a negative signal. Voters ability to hold politicians accountable depends on the level of information that voters possess. In particular, we consider campaign contributions as a signal to voters about divergence from their preferred legislative priorities. The extent of political news coverage within a given district determines the probability that voters receive such signals regarding interest-group contributions. Therefore, more news coverage implies smaller equilibrium campaign contributions by narrow interest groups. In sum, candidates face a trade-off between campaign contributions which can be used to sway persuadable voters and electoral support from informed voters. For groups whose interests diverge from those of voters, the marginal benefit of contributions increases as the link between candidates and voters becomes more tenuous. These points lead us to two substantive hypotheses: 1) Total interest group campaign contributions will be higher in low-information districts and 2) The effect of press coverage depends on the type of interest group: whereas less encompassing interests, such as corporate and trade PACs, will contribute more in low-information districts (where a candidate s tradeoffs in terms of public perceptions are less acute), this relationship will not hold for groups representing more encompassing interests, such as labor PACs and some membership organizations. 8

9 Measures The fundamental empirical problem of estimating the effects of the media on important political outcomes like campaign contributions and quality of representation is that press coverage is endogenous to these variables. Snyder & Strömberg (2010) offer the strategy to address this problem by constructing a plausibly exogenous measure of a district s newspaper press coverage. 5 Congruence, the main independent variable in our analysis, is defined for a given congressional district as the market-weighted reader share of the newspapers in that district: N Congruence dt = MarketShare ndt ReaderShare ndt n=1 where n denotes newspaper, d denotes district, and t denotes time. MarketShare 6 captures newspaper n s share of total newspaper sales in district d, while ReaderShare reflects the share of newspaper n s readers living in district d. 7 This measure captures the degree to which newspapers and congressional districts are overlapping. A low level of Congruence implies that a newspaper draws readers from multiple congressional districts, and is therefore less likely to cover any one representative. The key assumption is that the process which determines the structure of newspaper markets is exogenous to the process that determines the boundaries of House districts. 8 Note that 5 The choice to focus exclusively on newspaper coverage arises from the absence of equivalent data about other media sources such as radio and television. Nonetheless, we believe this choice is justified. Mondak (1995), for example, finds that local newspapers are a key source of information about local politics, including House elections. With their only major local newspaper on strike, Mondak found that the knowledge voters in Pittsburgh held about local House elections declined. In short, while other media outlets may provide coverage of national and international politics, local newspapers are often the only source to adequately cover local Congressional campaigns. 6 Constructing Congruence in terms of market shares avoids a definition that is dependent on the total newspaper readership in a given area, as total newspaper readership is endogenous to political knowledge as well as to demographic characteristics like education and income. 7 For an extended discussion of the definition and measurement of Congruence, see (Snyder & Strömberg 2010). 8 Substantively, this assumption implies that redistricting is not shaped by newspaper markets. If, for 9

10 instead of adopting the arbitrary simplification of voters into the binary categories informed and uninformed, as have many formal models of political economy, we use Congruence as a plausibly exogenous and, more realistically, continuous measure of voter information in a given congressional district. Data To analyze how Congruence affects the strategic environment of campaigns and campaign giving, we use data from The Center for Responsive Politics (CRP) on campaign contributions over the period and merge this data with Snyder & Strömberg s (2010) original dataset on press coverage and political accountability. 9 The analysis presented in this paper applies to incumbent candidates 10 running for reelection in the same district as in the previous election cycle. 11 All contributions have been converted to constant year 2000 dollars using CPI figures collected from the Bureau of Labor Statistics. The CRP data come from Federal Election Commission (FEC) campaign summary files that aggregate campaign contributions for each candidate in all House races (primary and general elections) for each election cycle. We focus exclusively on campaign donations from interest-group PACs, and do not analyze donations by individuals or party committees. The histogram in Figure 1 describes the distribution of total interest group donations in our instance, politicians intentionally gerrymandered districts to avoid congruence with newspaper markets in order to facilitate taking larger donations from interest groups, this assumption would be violated. However, we believe this scenario unlikely. 9 We use CRP s data instead of the FEC s because it disaggregates trade and membership PACs, which we hypothesize to have agenda that are less and more encompassing, respectively. The data can be found on the Center for Responsive Politics s website: 10 Though we focus only on incumbents, note that donations made to incumbent candidates make up the vast majority of campaign contributions; challengers raise comparatively little. For instance, over the period considered, incumbent s raised, on average, approximately $435,000 from PACs, while challengers raised, on average, approximately $90, The data did not allow us to include incumbent candidates who ran for office in another district after redistricting. These accounted for very few cases and are not problematic with respect to our hypotheses, because we expect political accountability to be stronger in cases where voters, donors, and representatives have time to learn from their information environment. 10

11 dataset. Incumbents received a mean of about $435,000, and the median amount raised from PACs is about $358,500. Number of Representatives Density 0.0e e e e Total Contributions) Total Contributions Figure 1: Total Interest Group Contributions Note: These figures show the empirical distribution of total interest group contributions for all races in which incumbents ran for reelection between 1990 and The dashed vertical line marks the mean contribution of approximately $435,000. Data on interest group contributions can be disaggregated into several categories of political action committees (PACs), including corporate, trade, labor, and membership. 12 Examples of corporate PACs include those associated with Pfizer, UPS and Wal-Mart, while examples of trade PACs include those of the American Medical Association and the National 12 Gopoian (1984, p. 259) notes that generic treatments of labor and corporate PACs mask the diverse orientations of individual PACs. Though we acknowledge that such categorical treatments have drawbacks, we believe that these categories nonetheless capture essential distinctions among the agendas of various types of PACs 11

12 Table 1: Amount Raised by Incumbents from Political Action Committees, Total Corporate Labor Trade Membership Mean $434,841 $160,497 $88,260 $130,435 $17,057 St. Dev. $338,027 $154,929 $ $103,399 $17,215 N 2,914 2,896 2,699 2,896 2,804 Association of Homebuilders. Emily s List and NARAL Pro-choice America support PACs which fall into the membership category. The UAW and SEIU support labor PACs. Table 1 shows that contributions by corporate and trade groups together make up roughly two-thirds of total interest group donations. 13 Model Specifications Following Snyder & Strömberg (2010), we exploit the panel features of our data to estimate four different fixed-effects specifications of each of our models. These specifications allow us to control for time-invariant but unobservable characteristics, better identifying the causal effect of congruence. The models take the following general form: D dt = δcongruence dt + X dt β + α st + ɛ dt where d denotes district, t denotes time, s denotes state, X dt is a matrix of control variables, α st denotes state-by-year fixed effects, and D dt denotes campaign donations. The dependent variable in the first analysis is total interest group contributions. In the subsequent analysis, we examine interest group contributions by type and sector. State-by-year fixed effects control for time-invariant but unobserved characteristics of a particular state in a given year. Similarly, our second specification employs district and year fixed effects, thus focusing on variation in congruence within a particular district over time 13 The FEC breaks PAC contributions into additional categories that are not reported here: Non-stock, Cooperative, and Non-connected. These are omitted because they are not a central part of the analysis. 12

13 and within a given year across all districts. A third specification controls for time-invariant but unobservable characteristics of states and congressional election cycles (state and year fixed effects), while a fourth does the same for districts and congressional election cycles in a given state (district and state-by-year fixed effects). Of course, since our dependent variable is measured for a given congressional race at the district level, we cannot use district-by-year fixed effects, as there would be no source of variation. In sum, we use these several fixed effects specifications to minimize threats to inference which might arise from unobserved variables that are correlated both with Congruence and campaign donations, thus violating the standard exogeneity assumption of OLS regression and biasing estimates of the causal effect of Congruence. Control Variables All specifications include a comprehensive set of controls for the congressional race, representative, and district. These controls are described in detail and descriptive statistics are given in Table A.1 in the appendix. In brief, the controls fall into three categories: congressional race controls, district controls, and candidate controls. The race controls include the closeness of the election and a dummy for whether the incumbent is running unopposed. The representative controls are age; whether the incumbent is a freshman; tenure in office; whether the incumbent is a member of the majority party in congress; if the incumbent is a party leader; if the incumbent serves on a powerful committee; and if the representative is a chair or ranking member. At the district level we control for median income, urbanity, population density, whether a city contains the majority of a district s population, and the total number of congressional districts in any city containing a majority of the population, as well as the percent of the population that is elderly, employed by the military, employed in agriculture, employed in blue-collar occupations, and foreign-born. 13

14 Results and Discussion H1: Total interest group contributions will be higher in low-information districts Table 2 reports the estimated coefficients and standard errors from linear regressions of total interest group contributions on Congruence. The dependent variable in the table aggregates contributions across all categories of interest groups. Thus, the first row of the table tests the hypothesis that the strategic interaction between donors and candidates varies with the amount of information available to voters, and, specifically, that interest group contributions are likely to be higher in districts where voters are less well-informed. Table 2: Predicted Effect of Congruence on Total Interest Group Contributions Model 1 Model 2 Model 3 Model 4 All PAC Contributions Congruence -274,428*** -214,214* -252,733*** -299,707** (72,674) (116,774) (67,622) (123,529) Observations 2,616 2,616 2,616 2,616 R-squared Fixed Effects State Year Year, District Year, State District, State Year Controls Cluster robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: This table reports OLS regression coefficients predicting the effect of Congruence on PAC contributions to the campaigns of incumbents running for reelection. The coefficients reported capture the effect of varying Congruence from zero to one, 14 all other variables being equal. The dependent variable is total interest group contributions. The coefficients for the control variables are not reported. Standard errors are clustered by representative. Column (1) gives the baseline specification including all controls and state-by-year fixed effects. Column (2) includes district and state fixed effects, while column (3) includes state and year fixed effects. Column (4) introduces both district and state-by-year fixed effects and thus reflects within-district variation in the level of interest-group contributions over 14

15 time. Across all specifications the coefficient on Congruence has the expected negative sign, and, in all but the second model, it is significant at the 0.05 level. In Model 2, the coefficient for Congruence is marginally insignificant at the 0.1 level. All standard errors are clustered by representative. The use of robust standard errors allows observations for a given representative to be correlated in some unknown way, inducing correlation in the error terms across observations for a given member across several election cycles. These results offer clear support for the hypothesis that interest group contributions are likely to be higher in districts where voters are less well-informed, and suggests, in accordance with our theoretical framework, that the strategic interaction between donors and candidates varies with the amount of information available to voters. Figure 2 gives a more intuitive picture of the substantive significance of these results. Each point in the figure represents the estimated effect of a one standard deviation shift 15 in Congruence on total contributions by interest groups. The estimates and both their 90 and 95 percent confidence intervals are reported in $10,000s. The effect of Congruence on interest-group giving is substantial. A one standard deviation increase in Congruence decreases the level of contributions from PACs in a given race by more than $60,000. This is more than 13% of the average total interest-group contributions to incumbent candidates over the period we study. H2: The effect of voter information on campaign donations will depend on interest group type In this section, we report results from a second set of tests which distinguish between types of interest groups. If Baron s (1994) logic holds and all interest groups are rational actors seeking to maximize the benefit of their campaign contributions by spending in low-information 15 Obviously, this is a much more realistic increase in Congruence than the shift from 0 to 1 that is represented by the OLS coefficient. 15

16 Model 1 (State x Year) Model 2 (Year, District) Model 3 (Year, State) Model 4 (District, State x Year) Contributions in Constant Dollars ($10,000s) Figure 2: Effect of Congruence on Total Interest Group Contributions Note: This figure shows the estimated effect of a one standard deviation increase in Congruence (0.24) on the level of total PAC contributions using four different fixed effects specifications. The bars represent 95% confidence intervals, while the hash marks denote 90% confidence intervals. districts where the marginal cost of persuasion is lower, then we should see no differences in strategy across types of interest groups. All interest groups should recognize the greater efficiency of contributing in low-information districts and ceteris paribus donate more in those races. We hypothesize, however, that the strategic behavior of interest groups and candidates depends on whether an interest group is encompassing or narrow. By encompassing, we again mean interest groups, such as labor PACs, whose interests are more closely aligned with those of the median voter. By narrow, we mean interest groups, such as corporate PACs, whose interests are often at odds with those of the median voter. Hence, we expect the strategic behavior of PACs and candidates to differ with regard to these broad classes of interest groups. As voter information increases, we expect contributions by narrow interest 16

17 groups to decrease while contributions by encompassing interest groups remain unaffected. In sum, we propose that PACs decisions to give and politicians choices to accept campaign donations depend on how easily voters are able to hold them accountable that is, on the transparency of the campaign environment and the level of information available to voters. In support of our hypothesis, as Table 3 shows, contributions by corporate and labor PACs do not follow the same pattern. Representatives receive less money from corporate PACs in high-information districts, whereas contributions by labor PACs are unaffected. Further, consistent with our argument, the effect of Congruence on contributions by trade PACs is similar to the effect of Congruence on corporate contributions. For membership PACs, the coefficient on Congruence is small and unstable. Two of the four models are statistically insignificant and the coefficient is never larger than $12,000, a fraction of the estimated effect of Congruence on corporate contributions, for example. Thus, the effect of Congruence on donations by membership PACs more closely resembles its effect on labor PACs than its effect on corporate or trade PACs. Figure 3 clearly makes the point that not all interest groups behave as previous theoretical models have predicted. In fact, they do not all act strategically such as to maximize the efficiency gains of donating to candidates in low-information districts. Though corporate contributions decrease by roughly $25,000 for each one standard deviation increase in Congruence, the effect of Congruence on labor contributions is negligible. On average, labor PACs donate no less in high-congruence districts than in low ones. In light of existing theory on voter information and campaign finance, this result is surprising. Under the framework we propose, it is not. As Snyder & Strömberg (2010) show, when candidates stand for reelection in high information districts, they face stricter standards of accountability and do more for their constituents. Both candidates and interest groups face a different strategic environment in which donations from narrow interest groups may be less desirable. 17

18 Corporate Labor Model 1 (State x Year) Model 2 (Year, District) Model 3 (Year, State) Model 4 (District, State x Year) Contributions in Constant Dollars ($10,000s) Contributions in Constant Dollars ($10,000s) Figure 3: Contrasting Effects of Congruence on Contributions by Corporate and Labor PACs Note: These figures show the estimated effect of a one standard deviation increase in Congruence (0.24) on the level of corporate and labor PAC contributions using four different fixed effects specifications. The bars represent 95% confidence intervals, while the hash marks denote 90% confidence intervals. Finally, Figure 4 summarizes our results by plotting the estimated marginal effect of voter information on campaign contributions. It shows how total contributions from PACs are predicted to change across the entire range of Congruence, while holding all controls at their observed values. As voter information increases from low to high, the total raised from PACs (top, solid line) falls by about $272,000. More modest changes in voter information also have significant effects. A shift from one standard deviation above the mean to one below it decreases predicted PAC contributions by $130,

19 $600 All Corporate Labor $500 Total PAC Contributions (hundreds of thousands of dollars) $400 $300 $200 $100 $ Congruence Figure 4: Predicted Contributions from Political Action Committees by Congruence Note: The figure shows predicted total contributions across the range of Congruence by all PACs (top, solid line), corporate PACs (middle, dashed line) and labor PACs (bottom, dotted line), based on the first column in Table 2 and Table 3, respectively. All other variables are held at their observed values. Predicted contribution levels are shown with their 95% confidence intervals. The dashed lines in Figure 4 show predicted contributions by corporate (dashed line in the middle) and labor (dotted line on the bottom) PACs across the range of Congruence. While giving by labor PACs is flat as voter information rises, donations by corporate PACs have a significant negative slope. Around a Congruence of 0.75 differences in the contribution strategies of corporate and labor PACs cease to be statistically meaningful. A district s information environment clearly has a different effect on giving by narrow interests compared to more encompassing groups. Our results suggest that corporate money flows predominately into lower information districts, where corporate PACs outspend more broadly representative groups, such as organized labor, by a wide margin. Fewer donations by large corporate donors offers one plausible mechanism to explain why politicians do more 19

20 for their constituents in high-information districts (Snyder & Strömberg 2010). Conclusion Based on an analysis of Center for Responsive Politics data over the period , we find that incumbent members of the House of Representatives who represent low-information districts raise more money from PACs than incumbents who serve higher-information districts. The data also suggest that PACs with narrower agendas are responsible for most of this difference. Previous research by Snyder & Strömberg (2010) finds that representatives with less informed constituencies have weaker relationships with their constituents across a variety of dimensions. Our results suggest that ceteris paribus these same representatives accept more money from interest groups particularly corporate and trade PACs. As such, our findings lend support to the idea that corporate contributions may be one possible mechanism to explain these districts democratic deficit. These results add an important new dimension to our understanding of how elected officials represent their constituents: representatives from low-information districts tend to have stronger financial relationships with certain interest groups compared to their peers. Furthermore, these financial ties may have real implications for members legislative behavior. Our evidence suggests that not all interest groups share the same contribution strategies. Corporate and trade PACs, in particular, appear to give more to candidates from low-information districts than they would if the representatives constituencies were better informed. These findings suggest that the donations from PACs that are most likely to be viewed skeptically by constituents are more common when there is less information available to voters. These results also support the theory that PACs give more to representatives who are able to dedicate more time and energy to their agenda. That said, further analysis is needed to determine whether these representatives are 20

21 actually doing more for interest groups that give more (Hall & Wayman 1990) or are targeted because interest groups hope they will serve their interests (Fox & Rothenberg 2011). 16 We suspect it is a mix of both, but more research is necessary to shed light on these post-election outcomes. At this point, there is evidence that the financial relationship is different, which suggests that the legislative relationship may be different. Future studies would benefit from a more precise categorization and identification of PACs and their agendas. CRP provides data that divides PAC contributions by economic sector, but these divisions do not always mirror the types of agendas we wish to study. Notwithstanding these limitations, preliminary evidence suggests that sectors with stronger regulatory interests (finance, real estate, health care) do tend to give more in low-information districts. 17 Developing a better taxonomy of interest groups and further exploring the agendas that lead to more giving in low-information districts is a promising avenue for future research. The analysis in this paper deals primarily with the era before the Bipartisan Campaign Reform Act (McCain-Feingold) 18, the creation of Super PACs 19, and the 2014 Supreme Court ruling which struck down overall limits on individuals donations. 20 While both the McCain-Feingold legislation and the Citizen s United decision left in place the restrictions on PACs direct contributions to candidates, interest groups gained the ability to spend 16 Fox & Rothenberg (2011) build a formal model that outlines possible complications from unlimited independent expenditures from Super PACs. Their non-contractual model highlights how PAC contributions do not result in quid pro quo relationships between representatives and PACs, but suggests, nonetheless, that incumbents do adjust their behavior to attract direct and indirect contributions from ideologicallyallied interest groups because interest groups match their donations to ideologically sympathetic elected officials. 17 See Figure A.1 in the appendix for these results. 18 BCRA was passed in 2002 and became effective in 2004, which is the beginning of the final election cycle in our dataset 19 The Supreme Court s 2010 decision in Citizens United v. the Federal Election Commission created Super PACs that can raise and spend unlimited amounts of money as long as they do not coordinate with candidates and their campaigns. 20 The 2014 decision in McCutcheon v. Federal Election Commission struck down overall limits for how much individuals can give to all candidates and party committees during an election cycle. 21

22 unlimited funds independently. Super PACs thus represent a new way of giving that our data cannot address. How press coverage will affect spending by Super PACs is an important topic for future investigation as the data cumulate. Local information and news environments affect voters political knowledge, the quality of congressional representation, the party loyalty of representatives, and, as we have shown, the financial relationships of House members with interest groups. Recent research has illustrated how information environments affect voters (see e.g. Prior 2007), and the findings here compliment that work and expand our understanding of how representatives are affected. For the foreseeable future, both money and uninformed voters are likely to remain a fixture of congressional elections. 22

23 References Ansolabehere, Stephen, John M. de Figueiredo & James M. Snyder Why is There so Little Money in U.S. Politics? The Journal of Economic Perspectives 17(1): Arnold, R. Douglas Congress, the Press, and Political Accountability. Princeton University Press. Ashford, Kathryn L The Role of Corporations in the 1980 U.S. Congressional Elections. Sociological Inquiry 56(4): Ashworth, Scott Campaign Finance and Voter Welfare with Entrenched Incumbents. The American Political Science Review 100(1): Baron, David P Electoral Competition with Informed and Uniformed Voters. The American Political Science Review 88(1):pp Chappell, Henry W., Jr Campaign Contributions and Congressional Voting: A Simultaneous Probit-Tobit Model. The Review of Economics and Statistics 64(1):pp Coate, Stephen Political Competition with Campaign Contributions and Informative Advertising. National Bureau of Economic Research Working Paper See Coate, Stephen Pareto-Improving Campaign Finance Policy. The American Economic Review 94(3): Coleman, John J. & Paul F. Manna Congressional Campaign Spending and the Quality of Democracy. The Journal of Politics 62(3):

24 Currinder, Marian Money in the House: campaign funds and congressional party politics. Boulder, CO: Westview Press. Esterling, Kevin M Buying Expertise: Campaign Contributions and Attention to Policy Analysis in Congressional Committees. The American Political Science Review 101(1): Fox, Justin & Lawrence Rothenberg Influence without Bribes: A Noncontracting Model of Campaign Giving and Policymaking. Political Analysis 19(3): Goidel, Robert K., Donald A. Gross & Todd G. Shields Money Matters: Consequences of Campaign Finance Reform in House Elections. Lanham, MD: Rowman & Littlefield Publishers. Gopoian, J. David What Makes PACs Tick? An Analysis of the Allocation Patterns of Economic Interest Groups. American Journal of Political Science 28(2): Gordon, Sanford C., Catherine Hafer & Dimitri Landa Consumption or Investment? On Motivations for Political Giving. Journal of Politics 69(4): Grier, Kevin B., Michael C. Munger & Brian E. Roberts The Determinants of Industry Political Activity, The American Political Science Review 88(4): Grossman, Gene M. & Elhanan Helpman Electoral Competition and Special Interest Politics. The Review of Economic Studies 63(2): Hall, Richard L. & Frank W. Wayman Buying Time: Moneyed Interests and the Mobilization of Bias in Congressional Committees. The American Political Science Review 84(3):

25 Hutchings, Vincent L Public Opinion and Democratic Accountability: How Citizens Learn about Politics. Princeton University Press. Mondak, Jeffery J Newspapers and Political Awareness. American Journal of Political Science 39(2): Morton, Rebecca B. & Roger B. Myerson Campaign Spending with Impressionable Voters. Discussion Papers 1023, Northwestern University, Center for Mathematical Studies in Economics and Management Science. Prat, Andrea Campaign Advertising and Voter Welfare. The Review of Economic Studies 69(4): Prior, Markus Post-Broadcast Democracy: How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. New York: Cambridge University Press. Snyder, James M. & David Strömberg Press Coverage and Political Accountability. Journal of Political Economy 118(2): Wright, John R Contributions, Lobbying, and Committee Voting in the U.S. House of Representatives. The American Political Science Review 84(2):pp Zaller, John The Nature and Origins of Mass Opinion. New York: Cambridge University Press. 25

26 Appendix Finance, Insurance, & Real Estate Health Care Model 1 (State x Year) Model 2 (Year, District) Model 3 (Year, State) Model 4 (District, State x Year) Contributions in Constant Dollars ($10,000s) Contributions in Constant Dollars ($10,000s) Figure A.1: Effect of Congruence on Finance, Insurance, Real Estate, and Health Care PACs Note: A sector-level analysis shows that the inverse relationship between interest group contributions and information is particularly strong for finance, insurance, real estate, and health PACs. The left panel of this figure shows the estimated effect of a one standard deviation increase in Congruence (0.24) on the level of contributions by finance, insurance, and real estate PACs, while the right panel shows the same for health care PACs. Four different fixed effects specifications are shown with similar results. The bars represent 95% confidence intervals, while the hash marks denote 90% confidence intervals. 26

27 Table 3: Predicted Effect of Congruence on Interest Group Contributions by Type Model 1 Model 2 Model 3 Model 4 Corporate Congruence -124,930*** -54, ,739*** -101,494* (36,184) (54,316) (33,630) (56,846) Observations 2,598 2,598 2,598 2,598 R-squared Trade Congruence -81,716*** -69,910* -72,359*** -98,010** (25,155) (35,999) (23,235) (38,887) Observations 2,598 2,598 2,598 2,598 R-squared Labor Congruence -2,667-15,309-5,604 3,887 (14,574) (25,204) (13,832) (25,566) Observations 2,424 2,424 2,424 2,424 R-squared Membership Congruence -11,187*** -4,567-10,802*** -4,123 (3,665) (6,949) (3,509) (6,927) Observations 2,509 2,509 2,509 2,509 R-squared Fixed Effects State Year Year, District Year, State District, State Year Controls Cluster robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: This table reports OLS regression coefficients predicting the effect of Congruence on PAC contributions to the campaigns of incumbents running for reelection. The coefficients reported capture the effect of varying Congruence from zero to one, all other variables being equal. The dependent variable is contributions by a particular category of PAC. The coefficients for the control variables are not reported. The standard errors are clustered by representative. 27

28 Table A.1: Control Variables Standard Variable Name Description Mean Deviation close The closeness of an election, measured by the negative absolute difference between the Democratic share of the vote and open seat Indicator for an open seat uncontested Indicator for an uncontested race age Age, measured on a continuous scale freshman Indicator for freshman status tenure Tenure in office, measured on a continuous scale majparty Indicator for member of the majority party in Congress party leader Indicator for whether the representative is a party leader powercom Indicator for whether the representative serves on a powerful committee (Appropriations, Budget, Rules, and Ways and Means) chair Indicator for whether the representative is a chair rankmem Indicator for whether the representative is a ranking member dwnom1 DW nominate score measuring a representative s ideology pct urban Percent urban in the district urb2 Indicator for the second quintile of urbanity urb3 Indicator for the third quintile of urbanity urb4 Indicator for the fourth quintile of urbanity urb5 Indicator for the fifth quintile of urbanity ldensity Log of district s population density ld2 Indicator for the second quintile of logged population density ld3 Indicator for the third quintile of logged population density ld4 Indicator for the fourth quintile of logged population density ld5 Indicator for the fifth quintile of logged population density cd per ci 14 dummy variables for the total number of congressional districts in any city containing a majority of the population CPI lmed income Log of the district s median income in year 2000 constant dollars pct old Percent of the district s population aged 65 or older pct military Percent of the district s population employed by the military pct farmer Percent of the district s population employed in agriculture pct foreign Percent of the district s population that are foreign-born pct bluecol Percent of the district s population employed in blue-collar occupations pct veterans Percent of the district s population that are veterans