PARTIES AS NETWORKS

Though the founders of the United States wasted no breath nor ink in decrying their ill-effects, political parties have remained a vitally important part of American politics since their emergence during the dawn of the republic. Political parties perform a myriad of functions crucial to the maintenance of government such that American democracy as presently conceived would be “unthinkable” in their absence (Schattschneider Reference Schattschneider1942). From electing candidates to office to guiding their votes once seated, parties are intimately connected to almost every facet of American government.

Despite their ubiquity, however, parties are notoriously hard to define. Scholars from the classic era such as Key (Reference Key1964) offer complementary visions of what makes a party as well as what functions a party performs. Aldrich (Reference Aldrich2011, 5) goes further in describing parties as vehicles for election—“the creature of the politicians, the partisan activist, and the ambitious office seeker and office holder.” Crucially, Aldrich (Reference Aldrich2011) locates these office-seeking actors at the center of party network. This is the crucial distinction highlighted by Bawn et al. (Reference Bawn, Cohen, Karol, Masket, Noel and Zaller2012) and what later came to be known as the “UCLA School”—that it was interest groups rather than office seekers who were the true center of gravity that other satellites in the party constellation orbited around (e.g., Cohen et al. Reference Cohen, Karol, Noel and Zaller2009; Grossmann and Dominguez Reference Grossmann and Dominguez2009; Noel Reference Noel2014). Most recently, Masket (Reference Masket2016, 114) developed a useful definition of this theory. “Parties are not rigid entities, limited to their appearances in legal definitions or business filings,” he writes. “They are, rather, networks of intense and creative policy demanders…working both inside and outside the government to determine the sort of people who get elected to office and thus change public policy.”

Contemporaneous to the emergence of the “EPN” theory were major changes in campaign finance laws which altered the way the EPN deployed resources during elections. Skinner, Masket, and Dulio (Reference Skinner, Masket and Dulio2012) trace the position of 527 groups within party networks by tracing common employees. The authors find that shifts in campaign finance law in the early 2000s led to party organizations turning toward extra-party electioneering organizations to skirt prohibitions on fundraising and spending in federal elections. They also found that the networks of 527’s mirror the hierarchical structures of the parties which they represent.

This interest-group centric view of parties holds that these coalitions work in concert to control nominations—effectively deciding contests before voters even get a chance to decide. Yet, as McCarty and Schickler (Reference McCarty and Schickler2018) point out, this theory has a central deficiency—namely how the incentives of elected officials and formal party leaders line up with those of interest groups such that the agency relationship theorized by the UCLA school can materialize. As works such as Barber (Reference Barber2016) and La Raja and Schaffner (Reference La Raja and Schaffner2015) point out, the goals of formal party organizations and interest groups are often at odds with one another. Others, however, reconcile this by placing formal party organizations at the center of the EPN (e.g., Karol Reference Karol2009; Kolodny and Dwyre Reference Kolodny and Dwyre2018)—coordinating the larger orbit of interest groups.

The U.S. Supreme Court’s decision in 2010’s Citizens United v. FEC and the subsequent overhaul of campaign finance law in its aftermath have led some to argue that empowered interest groups have hijacked American democracy and are fueling the rise of extremists by overwhelming the traditional party apparatus. Notably, La Raja and Schaffner (Reference La Raja and Schaffner2015) found evidence of this phenomenon in the states by observing variation in their campaign finance laws prior to the decision. However, Dwyre and Kolodny (Reference Dwyre and Kolodny2014) test the idea with federal data and find that formal party organizations have remained the central actors that coordinate the efforts of others. Others argue that the independent expenditure organizations borne by the decision—notably super PACs (Political Action Committees)—are mere party creatures in wearing different names but fulfilling the same functions are their centralized fore-bearers did prior to 2010 (see Mann and Corrado Reference Mann and Corrado2014; Drutman Reference Drutman2015).

Following the 2010 election cycle largely due to the Supreme Court’s decision in Citizens United v. FEC, so-called “outside spending” or advertising purchased and produced by organizations not officially affiliated with a candidate, exploded (Hasen Reference Hasen2016; Jacobson and Carson Reference Jacobson and Carson2015). While the majority of this spending came from so-called “super PACs,” or PACs unencumbered by traditional spending and fundraising limits as they were in theory “independent” from a candidate’s campaign (see Dowling and Miller Reference Dowling and Miller2014), a sizable portion of outside spending came from so-called “dark money” organizations (see Figure 1). These organizations are so named because their method of incorporation—as IRS designated 501(c) nonprofits—carries no legal requirement to disclose their donors. Dark money spending accounted for an average of about 17% of all independent expenditures in each election cycle between 2012 and 2018, according to data collected by the Center for Responsive Politics, a nonprofit that monitors trends in campaign finance.

FIGURE 1.

FEC Reported Spending by Dark Money Groups and All Outside Spending (Inflation Adjusted to 2018 Dollars)

Source: Center for Responsive Politics.

Over this same time period, the amount of money ideological interest groups (see Barber Reference Barber2016) spent in American elections increased. While parties dominated outside spending in the years prior to Citizens United (see Figure 2), the three election cycles following that decision saw ideological group spending dwarf that of all other categories. In those three cycles, ideological interest groups made more independent expenditures than all three of those other categories combined.

FIGURE 2.

Outside Spending by Type

Source: Center for Responsive Politics.

As important as how much money these outside groups were spending in elections is also the degree of coordination between them. Are these organizations just conduits for the same donors as before? Moreover, are these groups funneling money through the same channels in the era before the explosion of dark money in the 2010s?

One can answer these questions by looking to the degree of interaction between these groups and also to the extent that they interact with other interest groups. If their activity remains within relatively closed networks, it implies a separate fundraising ecosystem and a departure from established networks of the past. Further, the presence of these groups may alter the relative influence of other previously important nodes and thus shift the scholarly perception as to the true gatekeepers of money in politics.

Until recently, finding the connections in the dark money networks was virtually impossible. Though federal tax law dictates that all nonprofits fill out a yearly disclosure form detailing various facets of their financial operations, prior to June 2016,Footnote ¹ these forms were housed in a non-machine readable format as image files. Aside from being tremendously cost prohibitive (prior to the 2016 release, the only method of obtaining a complete universe of these disclosure forms was by spending $2,100Footnote ² for a DVD (Digital Video Disc) of just 1 year’s worth of filings), anyone interested in harvesting data from multiple nonprofits would have to perform the arduous task of manually inputting data from the several hundred fields on each disclosure form. Given that political nonprofits are often also recalcitrant to make their disclosure forms available despite being under legal obligation to do so (see Stevenson Reference Stevenson2013), the only way to begin collecting data on the finances of political nonprofits would be to collect Form 990s, as they are known under their IRS designation, directly from the IRS itself.

For this analysis, I collected data from over 2.3 million nonprofit financial disclosure forms filed with the IRS as of September 2019. This provided me with data on about 540,000 unique nonprofits who had filed a disclosure form for activities between the years 2007 and 2016 (typical filing deadlines mean that these disclosures are often filed up to 12 months after the calendar year’s end). From these forms, I have harvested data on cash transfers between one nonprofit group to another to create network linkages. Next, I connect the IRS data to data from the FEC using politically active nonprofits identified by the Center for Responsive Politics as bridges. With these combined data, I am able to document a vastly expanded and more complete picture of the financial network of money in American federal elections (see the “Data Collection” section of the Supplementary Material for a more detailed description of the data gathering process).

Understanding why these new financial data are important for the study of party networks is best demonstrated by looking at a congressional race where dark money played a large role. In the 2012 Virginia Senate race between former Virginia Governors George Allen and Tim Kaine, dark money groups spent over $19 million, $16 million of which was spent opposing the election of the eventual winner Kaine, according to the Center for Responsive Politics.Footnote ³ The Virginia Senate race was one of six congressional races in that cycle where over $5 million was spent by dark money groups and one of four races where non-disclosed outside spending totaled over $10 million.

Relying only on FEC disclosures provides a somewhat limited view of the relationships between dark money groups spending in that race. As Figure 3 indicates, a total of 18 outside groups made independent expenditures either in favor or opposing Kaine. Examining only FEC data, these groups appear autonomous and disconnected from one another. Without including IRS data, all appear as outpourings of “spontaneous speech” to use Justice Anthony Kennedy’s verbiage in his opinion to 2003’s McConnell v. FEC which Citizens United would overrule 7 years later. Compare Figure 3 to the diagram of underlying financial connections depicted in Figure 4. Figure 4 depicts all the organizations that spent to oppose Kaine in that race and all the groups that had given money to them (i.e., a second-degree ego graph). Pictured as well are the “communities” that each group belongs to. These communities were determined by a “random-walk” bisection of the network which takes random paths through the various linkages of the network and groups each node based on how densely they cluster with other nodes (for further explanation, see Newman Reference Newman2010; Pons and Latapy Reference Pons, Latapy, Yolum, Güngör, Gürgen and Özturan2005). With the IRS data included, the deeper layer of coordination behind these previously autonomous-looking groups is more clear. These independent groups are actually anything but, as they draw from common pools of funding.

FIGURE 3.

Dark Money Spending in 2012 Virginia Senate Race Depicted using Only FEC Data

Note: Unlike in Figure 4, deeper layers of coordination are invisible. Solid lines = negative independent expenditures. Dashed lines = positive independent expenditures. Dotted lines = direct contributions.

FIGURE 4.

Dark Money Spending Opposing Democratic Candidate Tim Kaine in 2012 VA Senate Race Depicted with Linked FEC and IRS Data

Note: Unlike in Figure 3, deeper layers of financial coordination are apparent when including IRS data. Letters indicate differing communities detected using random-walk technique. Solid lines = negative independent expenditures. Dotted lines = grants made between nonprofit organizations. Nonprofits making grants to more than two other nonprofits in the graph and nonprofits active in other 2012 races are named.

Eight distinct communities emerge from this method, each revolving around a nonprofit that spent against Kaine. Community D centers around “Americans for Tax Reform,” a long-established right-wing advocacy group founded by conservative activist Grover Norquist. Yet while those browsing FEC disclosure forms would only see the combined approximate $580,000 Americans for Tax Reform spent on negative independent expenditures in the 2012 VA Senate race, the contributions of several well-known industry groups to Americans for Tax Reform such as the Motion Picture Association of America and the National Cable and Telecommunications Association would remain unseen. In the three years prior to the 2012 election, the groups depicted in the above community plot gave a combined $5,649,000 to Americans for Tax Reform, according to the IRS disclosure forms. Similar communities exist around Crossroads GPS (Community C), the largest dark money organization of the 2012,Footnote ⁴ and the U.S. Chamber of Commerce (Community E). Crossroads GPS, which was founded by political strategist Karl Rove, spent a combined $71 million in the 2012 election, nearly twice what the second highest spending dark money group spent.

Also interesting is the constellation of nonprofits that form around the “American Future Fund” and “American Commitment” (Community B). With innocuous names typical of dark money groups, these two groups spent over a half-million dollars opposing Kaine’s election. However, the new IRS data show these groups firmly ensconced within the orbit of more recognizable nonprofits—Americans for Prosperity, the Freedom Partners Chamber of Commerce, and American Encore. What is available instantly to researchers with the development of this new dataset was revealed by the Washington Post (see Gold Reference Gold2014) 2 years after the 2012 election through laborious tracking and analysis of paper tax returns—these groups are part of a maze of nonprofits developed by the Koch brothers.

The aforementioned Koch network is emblematic of the labyrinthine structure that these nonprofit networks that spend money to influence elections assume. Looking at the funding structure for Americans for Prosperity in the 2012 cycle (Figure 5), one sees that similar patterns emerge. Figure 5 shows a first-degree ego graph for that organization—that is, the organizations that made grants to Americans for Prosperity and organizations that received grants from that organization. One can see that Americans for Prosperity shares connections with many other political nonprofits. Most importantly, all of these nonprofits receive funding from two common sources—Freedom Partners Chamber of Commerce and American Encore (formerly known as the Center to Protect Patient’s Rights)—whose giving dwarfs any of the other grants in the network. Similar to the party network graphs, money flows from more central nodes into a disparate network of satellite organizations before making its way to political candidates.

FIGURE 5.

Nonprofits Making Grants to Americans for Prosperity (2011–12)

Note: Utilizing just FEC data, these other organizations funding Americans for Prosperity are invisible, especially Freedom Partners Chamber of Commerce, American Encore, and the TC4 Trust, which act as financial backers for several other dark money groups (see Figure 7). Other nonprofits that spend in candidate races are marked as triangles.

To illustrate the validity of the community detection strategy, compare the membership of one network created via this algorithmic process with a similar map (Figure 6), created by the Center for Responsive Politics and featured in the Washington Post in 2014 (Gold Reference Gold2014). In that network, one can see the flow of money outward from Freedom Partners, the TC4 Trust and the Center to Protect Patient Rights to a myriad of political nonprofits. The data for this analysis were painstakingly compiled by hand with IRS disclosure forms housed in disparate locations. (While the 990 forms must be made available to the public, the law does not mandate they be held in a central location. This has led many organizations either making interested parties arrive in person to obtain them or simply stonewalling requests for these data.)Footnote ⁵

FIGURE 6.

Diagram of Koch Brothers Funding Network, 2012

Source: Center for Responsive Politics (January 7, 2014).

Compare Figure 6 to the graph of a community created by a five-step random-walk algorithm from the donation network of nonprofits in the 2012 election (Figure 7). The network map created via the community detection algorithm produces an almost identical graph to the one built by the Center for Responsive Politics (the Center to Protect Patient Rights was renamed “American Encore” following the 2012 election and carries that name in Figure 7). Thus, this agnostic method of partitioning the graph produces results equivalent to what experts well-versed in the subject matter would produce. Because nonprofits operate in tight orbits around one another, this graph-partitioning strategy gives us useful communities of interest for further analysis.

FIGURE 7.

The Koch Network of 2012 as Drawn via a Community Detection Algorithm using the Combined FEC and IRS Data

Note: Americans for Prosperity, shown in Figure 5, is but one constituent portion of this wider movement of electorally focused organizations. Names of organizations depicted in Figure 6 are bolded and in italics.

CONNECTING THE DARK MONEY NETWORK

Linking nonprofits via grants made to one another, I now apply the same algorithm used to draw the communities in the previous section to all races for Congress during the 2012, 2014, and 2016 cycles (though Citizens United was decided in January 2010, growth in dark money spending did not begin in earnest until the 2012 election given the time necessary to incorporate these groups with the IRS). This method is ideal for detecting communities within the nonprofits grant network as it accounts for the directionality of the linkages as well as being computationally efficient enough to run on a network of this size. It also provides a semi-agnostic method of partitioning the dark money network that can be compared to known linkages between organizations and communities of nonprofits as described by media outlets. In running the walktrap algorithm, I weight the strength of the network ties by the size of the grant given such that larger donations create “stronger” ties than smaller ones.

I begin with network graphs of 2012–2016 congressional elections, of which there were 183,907, 223,724, and 244,343 nodes, respectively. I then subset these graphs to the second-order neighborhoods (i.e., the nonprofits they received money from and the nonprofits they gave money to) for each of the nonprofits that purchased advertising in a given election cycle. In addition, I excised the vertices representing politicians, so advertising in support or opposition of a common candidate did not create a network linkage. This methodology produced 636 unique communities for 2012, 613 for 2014, and 901 for 2016. Further, I expand the previous network using FEC data to include super PACs receiving money from a dark money network. For instance, in the 2017 special election in Georgia’s 6th Congressional District, out-of-state dark money groups contributed about $3.75 million to super PACs involved in the race (Davis Reference Davis2017). Funneling money into an already existing super PAC is a common practice for these networks. Using FEC records, I link these networks of super PACs and dark money groups to the ultimate destinations of their money—the congressional candidates their ads either support or attack.

Some communities in this network were relatively dense, with a maximum of 93 organizations in one community, whereas most others were relatively sparse with 2,040 communities containing only two or fewer organizations. Overall, the average number of nonprofits per community was 1.47 with a median value of 1. Given the relatively few steps the walktrap algorithm took in determining communities and given that not all independent expenditure-making organizations received support from nonprofit entities, such a large number of singletons is to be expected. However, the conservative approach taken in delineating communities ensures that one can be quite confident that each community drawn by the algorithm represents an actual network of nonprofits and independent expenditure-making organizations.Footnote ⁶

Adding these additional financial data from the IRS disclosures changes the structure of the overall network. As shown in Figure 8, about 10% of the dyads that were in the same community when utilizing only FEC data were not in the same community once the IRS data were added. In 2016, a full 15% of these dyads switched from both nodes belonging to the same community to each node being grouped separately.

FIGURE 8.

Percentage of Dyads Remaining in the Same Community when Graphing the Network with FEC Data and Combined FEC and IRS Data

Neglecting these communities gives one a skewed view of just how dominant the spending of these groups are. The prevailing wisdom, as typified by works such as Mann and Corrado (Reference Mann and Corrado2014) or Drutman (Reference Drutman2015), is that despite the Citizens United ruling, the top independent expenditure-making organizations are dominated by party affiliated groups. That hold true when considering just FEC data. As depicted in Table 1a, in the list of the top 10 independent expenditure makers in the 2012 race, one can see a list dominated by party affiliated groups with the exception of American Crossroads, Crossroads GPS, and Americans for Prosperity. However, when accounting for these interwoven communities, the picture shifts dramatically. Now the interest groups in the orbit of both Crossroads GPS and Americans for Prosperity dominate all other groups in terms of independent expenditures made in that election cycle except for Mitt Romney’s super PAC. Americans for Prosperity’s network spent more than President Obama’s super PAC Priorities USA Action and more than both the Republican and Democratic Congressional Hill Committees (see Table 1b). Given the high degree of financial coordination between these groups, looking at their total spending gives a more realistic view of their influence in American elections.

Table 1.

When Observing the Spending Levels of Dark Money Organizations as a Whole, Their Spending Levels Can Often Rival the Most Prominent Party-Affiliated Organizations

Note: Dark money networks are denoted with a “***.”

DOES DARK MONEY RESHAPE THE EXTENDED PARTY NETWORK?

The emergence of dark money as a potent spending force in American elections also allows us to reassess our conceptualizations of the EPN. Previous work utilizing FEC data to identify the EPN (e.g., Desmarais, La Raja, and Kowal Reference Desmarais, La Raja and Kowal2015; Grossmann and Dominguez Reference Grossmann and Dominguez2009; Herrnson and Kirkland Reference Herrnson, Kirkland, Victor, Montgomery and Lubell2013; Kolodny and Dwyre Reference Kolodny and Dwyre2018; Manento Reference Manento2019) excludes possible deeper levels of financial coordination that take place outside of the jurisdiction of federal election law. Accounting for these unseen financial linkages could reorder understanding of the EPN and force previously peripheral groups to the center. Conversely, these dark money networks may not be integrated into the EPN—operating peripherally and outside the orbit of central coordinating entities. Both scenarios would lead scholars to develop an updated understanding of the relative influence of interest groups in the party network framing.

To measure this, I borrow an approach used by Koger, Masket, and Noel (Reference Koger, Masket and Noel2009) and examine network centrality measures for all organizations spending in the 2012, 2014, and 2016 election cycles. I create two separate directed networks, one using only data from the FEC and a second that incorporates both FEC data and my IRS dataset of financial transfers. For the second graph, I combine the FEC data with a one-degree ego graph accounting for funds received and grants made for all dark money groups active in each of the two cycles. This approach gives a conservative estimate to financial coordination between groups while limiting spurious linkages that may result from extending the ego graph backwards a second or third degree. With both graphs, I weight linkages between each organization by the dollar amount of the financial transfer. Finally, I prune the network to exclude all connections formed by a transfer of $5,000 or less. This reduces the size and complexity of each network to a level where it is computationally feasible to analyze it as well as removing spurious connections.

Like Koger, Masket, and Noel (Reference Koger, Masket and Noel2009), I examine the degree centrality of each node of the network, but I extend their approach to include a weighted in-degree and out-degree measure to account for the amount of money passing through each connection—the intuition behind this approach being that both collecting and disseminating more money from fewer sources is a better proxy for overall node importance than just the number of connections. I also include a betweenness centrality measure (see Newman Reference Newman2010) as a measure of how important each node is to the flow of political money in the network. Finally, I utilize the PageRank centrality of each node (Brin and Page Reference Brin and Page1998), a measure of network centrality ideal for this purpose in that both graphs are directed. This centrality measure helps identify the most important nodes of the network while obviating the issue of other centrality measures (see Borgatti Reference Borgatti2005) assigning high values to nodes linked to other nodes with a high out-degree—as would be the case when dealing with campaign finance data. Furthermore, PageRank centrality is optimized to handle directed networks better than eigenvector centrality, an older measure of centrality from which PageRank centrality was derived.

As shown in Table 2, the graphs of both networks produce similar statistics—though the addition of IRS data produces a much denser graph for obvious reasons. The average path length in both networks is roughly similar—as is the largest component of each. In addition, running a five-step walktrap community detection algorithm on both networks produces similar modularity scores for each.

TABLE 2.

Network Statistics

I next look to the most central nodes of each network, beginning with the network created using only FEC data (Table 3). Looking at the top 15 most central nodes as determined by the PageRank Centrality algorithm, I find a familiar listing of the most important organizations of the 2012 and 2014 elections. At the top of the list in 2012 were fundraising committees for both presidential candidates (Romney Victory Inc and Obama Victory Fund 2012), as well as Act Blue, a small-donor giving platform for Democratic candidates. Other top entries were the four Hill Committees as well as the Republican and Democratic National Committees. In 2014, we see a similar group as well as the fundraising organizations for both former Speaker of the House John Boehner and Republican Senator Tom Cotton (AR)—both prolific congressional fundraisers in an important midterm election year. The only issue-oriented groups appearing in both years were NORPAC, a political action committee promoting U.S.–Israeli relationships, EMILY’s List, which sponsors Democratic women’s congressional campaigns, and Club for Growth, a conservative organization. Ranking these organization by betweenness centrality, which prioritizes a node’s importance as a connector of other nodes, yields a similar list—though importantly one that elevates the ranking of several state-level political parties.

TABLE 3.

Centrality Measures, FEC Data Only, Sorted on PageRank Centrality

Interestingly, this list does not shift appreciably when adding in the IRS data from political nonprofits. As Table 4 illustrates, the most central nodes of this expanded network are quite similar to those from the original. In this expanded dataset, party affiliated committees are still the most central as measured by their PageRank centrality score. When looking at the betweenness centrality measure, a similar list emerges—again with more state party organizations rising to the top.

TABLE 4.

Centrality Measures, FEC + IRS Data, Sorted on PageRank Centrality

This would seem to indicate that despite the Citizens United ruling and the emergence of dark money, party-affiliated organizations still retain their central place in the EPN. Party-affiliated organizations still acted as the main coordinators for campaign money into candidate races in the two cycles following that 2010 decision. It would appear then that the party-centric theory of the EPN has garnered new evidence and that fears of parties losing their control over the EPN are somewhat inflated.

However, the inclusion of the IRS data to account for money flowing outside the jurisdiction of the FEC complicates this understanding of the role parties play as conduits for political money. Looking at the two centrality measures depicted in Tables 3 and 4, I perform a difference-of-means test between both the PageRank and betweenness centrality measures for party and dark money organizations. Party organizations are those designated as such by the Center for Responsive Politics in their taxonomy of interest groups, a taxonomy used by works such as Barber (Reference Barber2016) and others. Dark money organizations are those categorized as such by the Center for Responsive Politics.Footnote ⁷ Expressed mathematically, the centrality measure for organization i is modeled as a function of whether or not it is a party organization, according to the following models:

(1) $$ \begin{array}{rl}PageRank\hskip0.3em Centralit{y}_i=\alpha +{\beta}_1{(Party\hskip0.3em Organization)}_i+\epsilon & \end{array} $$

and

(2) $$ Betweenness\hskip0.3em Centralit{y}_i={\displaystyle \begin{array}{l}\alpha +\hskip2px {\beta}_1{\left(Party\hskip0.3em Organization\right)}_i+\epsilon .\end{array}} $$

When measuring PageRank centrality, as shown in Figure 9 and detailed in Table 5, party organizations remain more central than nonparty organizations both with and without including the IRS data. Given the nature of the measure, one might expect this. In studying Twitter conversations involving Spanish politics, Borondo et al. (Reference Borondo, Morales, Benito and Losada2014) find that politicians’ accounts were the most central (as measured by PageRank centrality) to these online discussions. Looking at the most central nodes as measured by PageRank in Table 4, one finds not just party organizations, but the party organizations most heavily involved in electoral politics such as Hill committees and presidential campaign committees. Given a weighted directed network, it is logical that such nodes would be not just the ultimate destinations of political money, but a source of funds for other campaigns. PageRank centrality measures the degree to which a given node is an influential reference to other nodes.

TABLE 5.

Measured with PageRank Centrality, Party Organization Remain More Central to Fundraising Regardless of Whether IRS Data Is Included

Note: $ {}^{***}p<0.001 $ ; $ {}^{**}p<0.01 $ ; $ {}^{*}p<0.05 $ .

FIGURE 9.

Party Organizations Have a Higher PageRank Centrality than Dark Money Organizations in Both Datasets

Note: Error bars show a 95% confidence intervals.

However, the interesting wrinkle when adding the IRS data is the shift in betweenness centrality. This measure of centrality indicates the degree to which a specific node is an influential gatekeeper in the network, not an influential referrer. Nodes with a high betweenness centrality act as conduits to other more disconnected nodes in the network—forming crucial bridges between them. Song and Yeo (Reference Song and Yeo2017) provide an intuitive example of this metric in looking at the betweenness centrality of world airports to measure their relative importance in the world air-travel network. While airports in the United States served a higher number of flights, airports in Asia had far higher betweenness centrality, which means that a disruption at one of these airports would have a greater influence on world air traffic.

Applying this measure to the financial networks of U.S. congressional elections, party organizations have a higher betweenness centrality than dark money organizations when examining just FEC data, a distinction that is statistically significant at the p < 0.05 level. When one adds the IRS data, however, this distinction disappears and suddenly, parties lose their primacy in connecting other nodes (see Figure 10 and Table 6). Thus, parties may not be the crucial gatekeepers that can open or close the spigot of political cash that scholars examining only FEC data once thought they were.Footnote ⁸ The ability of these networked interest groups to affect arenas such as the nomination process via primary elections or the formation of intraparty organizations (e.g., Rubin Reference Rubin2017) following an election is a subject ripe for future inquiry.

TABLE 6.

Measured with Betweenness Centrality, Party Organization Are No Longer More Central to Fundraising Regardless of Whether IRS Data Is Included

Note: $ {}^{***}p<0.001 $ ; $ {}^{**}p<0.01 $ ; $ {}^{*}p<0.05 $ .

FIGURE 10.

Party Organizations Have a Higher Betweenness Centrality when Measuring Only with FEC Data

Note: The distinction disappears once IRS data is added. Error bars show a 95% confidence intervals.

Similarly, if one uses an alternate specification, another picture begins to emerge than the one painted by past scholars of the EPN. Looking toward degree centrality—a basic measure of how many organizations either gave money to an organization or how many organizations a specific group gave money to—one sees a different image of the EPN. Several scholars have used this rather crude metric to measure the influence of groups in the EPN previously (e.g., Grossmann and Dominguez Reference Grossmann and Dominguez2009; Koger, Masket, and Noel Reference Koger, Masket and Noel2009; Kolodny and Dwyre Reference Kolodny and Dwyre2018; Skinner, Masket, and Dulio Reference Skinner, Masket and Dulio2012). I modify this tactic with an improved metric—the weighted degree centrality (Barrat et al. Reference Barrat, Barthelemy, Pastor-Satorras and Vespignani2004)—to weight this tally by the amount of money that formed each connection. As shown in Tables 7 and 8, when looking at the amount of money flowing through various nodes in the network, political nonprofits begin to emerge in dominant positions.Footnote ⁹

TABLE 7.

Centrality Measures, FEC + IRS Data, Sorted on Weighted In-Degree Centrality

Note: Bold = dark money organization.

TABLE 8.

Centrality Measures, FEC + IRS Data, Sorted on Weighted Out-Degree Centrality

Note: Bold = dark money organization.

Examining weighted in-degree centrality (Table 7), American Encore, the central funder for the Koch network depicted in Figure 5, occupies a spot just under both 2012 presidential campaigns and just above the Democratic National Committee. The Sierra Club and American Crossroads, both 501(c)(4) nonprofits, occupy the 9th and 10th spots, respectively, above both the Republican and Democratic Congressional Hill Committees. A similar ranking occurred in 2014, with the Sierra Club, NextGen Climate Action Committee, and the National Rifle Association all making the top 15 list. The presence of some nonprofits such as the Schwab Charitable Fund and the East Bay Community Foundation represents a deficiency with the IRS data—namely that while these organizations’ support for politically active groups can be known, the exact amount each spends on political activity is not a matter of public record.

Measuring out-degree centrality (Table 8), one finds three Koch network organizations in the top 15 during the 2012 cycle—the three financing organizations depicted in Figure 5. In addition, both the Sierra Club Foundation and Planned Parenthood also spent several million dollars on grants to groups that were later active in the 2012 cycle. In 2014, several nonprofit groups sit prominently with party Hill Committees—though again, this is not to say that the entire sum depicted eventually wound up in a candidate race. Interestingly, Tom Steyer, founder of the NextGen Climate Action Committee and a mega-donor to environmental causes and Democratic candidates, was himself one of the principal financiers of organizations active in the 2014 election.

In both Tables 7 and 8, one should also note the PageRank centrality measure of all these organizations. Because of the insularity of the dark money communities described in the previous sections, the PageRank centrality of these organizations was at or near zero—indicating that they did not link to other nodes with many in-bound linkages. By this measure, these interest groups occupy a peripheral part of the EPN—in contrast to the central role theorized by Bawn et al. (Reference Bawn, Cohen, Karol, Masket, Noel and Zaller2012) and more in line with the theory advanced by works like Hassell (Reference Hassell2017) and earlier works like Aldrich (Reference Aldrich2011).