Detecting and Understanding Donor Strategies in Midterm Elections

Political Contributions as a Form of Consumption

To date, most of the research on individual political contributions has focused on explaining why people choose to donate to campaigns. While a variety of factors influences this decision, three play a central role: income, political extremism, and political interest. Unsurprisingly, income is positively associated with political giving—people with higher incomes have more discretionary funds to allocate to political contributions (James 2009; Rosenstone and Hansen 1993; Verba et al. 1995). Beyond this monetary factor, political extremism strongly influences individuals’ decisions to contribute to campaigns (Barber 2016; Barber and McCarty 2015; Bonica 2014, 2016; Francia et al. 2005). Finally, people who are very interested in politics and care deeply about electoral outcomes are more likely to donate to political candidates or organizations (Grant and Rudolph 2002; Souraf 1992).

Making note of these patterns in giving, Ansolabehere, de Figueiredo, and Snyder (2003) argue that individuals contribute to campaigns primarily out of enjoyment rather than an expectation of material return. In other words, campaign contributions are a way of “consuming” the experience of politics, rather than “investing” in candidates or policies. By focusing on the consumption value of contributing, Ansolabehere, de Figueiredo, and Snyder account for why some people contribute and others do not: in the end, it comes down to intensity of preferences and interest, as well as ability to pay. This approach also sheds light on the otherwise puzzling fact that most contributions are of trivial dollar amounts. If donors sought access—or the purchase of votes—they would presumably have to make large contributions to have reasonable expectations of returns on their investment. However, if contributions are simply means for participating in politics, this experience can be purchased at low cost. While some donors—particularly very wealthy individuals with strong partisan attachments—undoubtedly attempt to purchase access or influence candidate positions with contributions (Francia et al. 2003; Gimpel, Lee, and Kaminski 2006; Gimpel, Lee, and Pearson-Merkowitz 2008; Gordon, Hafer, and Landa 2007; Schlozman and Brady 2012), the fact that donations are overwhelmingly of relatively modest amounts suggests that consumption motivations predominate.

Yet this approach does not illuminate another prominent feature of the contribution landscape: the dramatic variation in both the targets and spread of campaign contributions. Some donors contribute simply to congressional races within their district and state, while others make contributions to out-of-jurisdiction House and Senate races, and still others target contributions to political parties, PACs, or interest groups. In addition, some donors concentrate their contribution resources on one or a few targets, while others spread their contributions out over a much wider range of recipients. This variation is not readily explainable within the “contributions as consumption” framework.

Contribution Recipients as a Source of Information about Contributor Types

This variation is worth explaining—in no small part because different types of contribution recipients vary substantially and in politically relevant ways. House candidates are typically both ideologically extreme and focused inordinately on district concerns (Bafumi and Herron 2010; Thomsen 2014), presidential candidates are usually national figures who adopt fairly moderate positions on a wide range of issues (Sigelman and Buell 2004), political parties are large organizations with diverse interests that are focused on winning elections (La Raja and Schaffner 2015), and ideological interest groups are frequently extreme organizations that specialize in the development of policies in particular issue areas (Bawn et al. 2012; Herrnson 2012). In addition, potential recipients vary substantially in their visibility and familiarity. For example, presidential candidates and political parties are widely known, in-state/district congressional representatives are somewhat less familiar, while most out-of-jurisdiction House and Senate candidates, PACs, and ideological interest groups may be obscure to all but the most attentive observers.

We argue that because contributors must make conscious choices both about where and how broadly to spread their contributions, these decisions convey important information about contributors’ underlying preferences and strategies. Indeed, donors who make a point of contributing to out-of-jurisdiction House and Senate candidates are likely quite different from those who focus their political contributions on political parties, while donors who expend their donations on congressional candidates within their districts are probably distinct from those who contribute primarily to PACs or ideological interest groups.

Indeed, it is likely that these choices convey information about donors’ characteristics and strategies. For example, deliberate selection of more “obscure” targets (out-of-jurisdiction congressional candidates, PACs, or interest groups) over more “obvious” ones (political parties) may imply a higher degree of interest and motivation, greater financial means, or both. If a donor has greater resources, he or she is better positioned to contribute to a wider array of potential recipients, all things being equal. But selection of less “obvious” recipients also implies a heightened degree of motivation and intentionality on the part of the contributor. Just as more partisan, ideological, and interested individuals are more likely to donate in the first place (Barber 2016; Barber and McCarty 2015; Ensley 2009; Francia et al. 2005; Grant and Rudolph 2002; McCarty, Poole, and Rosenthal 2016; Souraf 1992), so too may they be more motivated to engage in more discerning patterns of contributing.

We know of one major study that categorizes types of major congressional donors, based primarily on surveys of donors (Francia et al. 2003). That study identifies three categories of donors, including investors seeking material gains, “ideologues” seeking to promote public goods, and “intimates” engaged for social reasons. Subsequent work emphasizes the dominance of “sincere” giving, which is assumed to be ideologically motivated, over several models of “strategic” behavior, including partisan-motivated giving to win races, and an investor model that posits donor giving for the purpose of purchasing legslative services (Bonica 2014). Our work expands on these studies by looking at small, medium, and large donors who give to a broad range of political committees (PACs, parties, and ideological organizations, as well as congressional candidates). Specifically, this analysis demonstrates that while donors tend to behave sincerely, they vary on other factors that predict their contribution strategies. Through our use of LCA (described in detail below), we are able to inductively identify distinctive donor groups based on donors’ contribution behavior rather than their self-described preferences and beliefs. We then identify the shared and unique characteristics of donors in each of these communities. This approach builds on Francia et al. (2003) by applying a new, inductive approach for identifying donor groups, enabling (as we explain in detail below) the identification of donor classes that were not categorized in this previous work.

Given that we employ LCA, our study is not conducive to classic hypothesis testing about the existence of donor categories. In fact, our method of donor categorization, which is intentionally inductive in approach, differs significantly from previous research that uses either deductive strategies or self-reporting in surveys to articulate donor motives (Francia et al. 2005). Our model uses information about donors’ choices of contribution targets to probabilistically classify donors into groups based on common contribution patterns. We then make inferences about motivations based on identification of individual characteristics that significantly influence the prior probability of assignment into one or another donor class. Nonetheless, we can still offer a broad set of expectations about donor behavior based on research about factors that motivate individuals to donate money. We focus here on income, political extremism, and political interest.

First, we expect rising income to give donors more options to express their preferences. For this reason, we expect the wealthy to give to a broader range of institutions and do so at the national level as well as at the state and local levels. Second, we expect political extremists to be particularly interested in giving to candidates who conform to their passionately held preferences, as well as groups that promote such candidates. Consequently, these individuals should be more likely to give to out-of-jurisdiction candidates, as well as to PACs and issue organizations. They should also be less likely to give to parties because these organizations may often support candidates with whom they disagree.

Finally, political interest will affect the degree of engagement and types of institutions supported. The most interested should be focused in their giving, deliberately picking candidates and organizations that best reflect their preferences. Consequently, we may expect those with greater interest to be more likely to donate to out-of-jurisdiction candidates, specific within-jurisdiction candidates, PACs, and political organizations. In contrast, those with middling political interest may have patterns of political giving that focus on the most obvious targets, such as the political parties, because they do not take the time to learn about particular candidates. We analogize to casual investors in the stock market who do not spend time researching individual firms to pick stocks but instead purchase an index fund. Similarly, the party heuristic provides the diversified portfolio of candidates without much information cost. Finally, the least interested donors may adopt idiosyncratic donation patterns that are difficult to categorize in systematic ways.

LCA as a Means of Categorizing Campaign Contributors

In this paper, we use LCA to discover underlying categories of donors from information about these contributors’ choices of recipients. The latent class model attempts to stratify a number of observed, or manifest, variables by an unobserved latent variable, and thereby probabilistically groups each observation into a latent class. This grouping process in turn produces expectations about how each observation responds on each manifest variable. Because the unobserved latent variable is nominal, the model effectively “sorts” observations into different types. Observations with similar sets of responses on the observed variables will tend to cluster within the same latent classes.

A nice feature of LCA is that, with a simple extension, the model permits the inclusion of covariates to estimate prior probabilities of latent class membership (Linzer and Lewis 2011). While in the basic model (without covariates) each observation has the same probability of belonging to each latent class prior to observing responses on the observed variables, with this extension the prior probabilities are allowed to vary by individual as a function of included covariates. Inclusion of covariates thus enables a more accurate representation of the process by which individuals are sorted into classes.

In what follows, we employ LCA to identify donor types based on the targets and spread of their political contributions. Based on an assessment of the relationship between donor classes and responses on the manifest variables, we establish interpretations of the various donor classes. We then investigate the relationship between selected covariates and the probability of membership in these different donor types.

Data and Method

The data for this project come from the 2010 and 2014 CCES. The CCES is a national stratified sample survey of more than 50,000 respondents administered biannually by YouGov. We analyzed data from two midterm elections to determine whether our observations about donor classes are generalizable beyond a single election. Our focus on midterm elections is deliberate. In presidential election years, presidential campaigns dominate donor attention: indeed, 80 percent of donors in 2012 reported giving to a presidential campaign, leaving fewer resources for disbursal among nonpresidential recipients. In contrast, when presidential candidates are not on the ballot, donors are more inclined to consider allocating resources among a wider variety of potential recipients. Consequently, midterm elections are arguably more useful for understanding the (nonpresidential election) factors influencing contributors’ donation decisions.

Identifying Donors

In this paper, we focus on individuals who said that they had donated money to “a candidate, campaign, or political organization in the past year.” Thanks to the large sample size of the CCES and the inclusive nature of the question itself, we are left with a large number of reported donors in each election year—13,699 donors in 2010 and 8,689 in 2014. Given these large sample sizes, it is important to address the extent to which this self-report question can be used as a valid way of identifying donors.

The first issue in understanding these self-reports involves reconciling the data we use with what is available from other studies that attempt to identify the percentage of Americans who contribute to politics. Importantly, validated records of federal contributions can only capture individuals who give enough to actually be forced to report. For example, Open Secrets estimates that just 0.3 percent of American adults made a federal contribution in excess of US$200—the reporting threshold for federal donations—during the 2014 cycle. However, the question we draw on in our study is much broader than that, allowing people to indicate whether they contributed to a candidate at any level of government (state, local, or federal) or any organization that they might view as political (many of which are not organizations that report to the Federal Election Commission (FEC)). Our question also includes donors who gave any amount; indeed, 31 percent of those who identified as donors in response to our question reported giving a total of US$200 or less to all candidates and organizations.

Thus, to make a valid comparison to the Open Secrets estimate, we would need to focus only on those individuals who reported in the CCES that they gave to federal candidates, political parties, or PACs and who reported giving over US$200 to these targets. When we do this, we find that 5 percent of adults in the CCES sample gave to federal targets and gave a total of over US$200. Of course, this figure is still not directly comparable because donors only have to report to the FEC when they give more than US$200 to a single candidate, party, or PAC. Thus, many of the donors we identify who give more than US$200 in total may still have not given more than US$200 to any given source and thus would not show up in the Open Secrets records. Indeed, Open Secrets finds that donors who had to report because they made at least one donation of more than US$200 actually gave an average total of US$1,665 to all federal candidates and committees during the cycle. By comparison, 1.5 percent of the 2014 CCES sample of American adults reported giving to federal candidates and committees and reported giving at least US$1,665 in total. Thus, while the initial large number of self-reported donors in the CCES seems quite high compared with figures on those who give enough to report their donations, the discrepancy is much smaller once we make an effort to identify only those who would likely show up in FEC records.

In the online appendix, we provide an additional validation exercise showing that self-reported federal donors in the CCES survey have a similar demographic and political profile as federal donors identified in FEC records. In a separate study, Hill and Huber attempted to match respondents from the 2012 CCES to the Database on Ideology, Money in Politics, and Elections (DIME) (Bonica 2013). Of the CCES respondents who reported making a contribution, 25 percent were successfully matched to the DIME database, but the match rate was 54 percent for those who reported giving US$300 or more. This is a fairly high match rate given that the DIME database is limited to donations that exceed disclosure thresholds. Indeed, it is likely the case that overreporting is much less of an issue when it comes to making political donations than it is for turnout because the social pressure to donate is much less than it is to vote. ¹

While the validation we conducted focused solely on those giving more than US$200, in the analysis that follows we include all self-reported donors. We do this to create a profile of the donor base at-large, rather than just those giving large enough contributions to show up in FEC disclosure reports. Nevertheless, we take an additional step to attempt to reduce the rate of false positives in our data. Specifically, we limit our analysis to self-reported donors who are also validated as registered voters. We do this because it seems unlikely that individuals who are not registered to vote would be active contributors to political organizations. That is, it would be the rare individual who would donate her money to political candidates, but not register to vote. ² After applying this additional filtering to the data, we are left with 13,020 donors in 2010 and 7,102 in 2014.

To identify donor classes based on the identity and spread of contribution recipients, we used responses to a follow-up question querying whether or not respondents donated money to a variety of potential recipients. This question simply asks donors to identify each type of candidate and political organization that they contributed money to during the past year, including a candidate for Senate in the respondent’s state, a candidate for Senate in another state, a candidate for the House in the respondent’s state, a candidate for the House in another state, a candidate for state office, a political party committee, a PAC, or a political organization.

Table 1 shows the percentage of donors indicating that they had given to each source during each of the three years of our study. The table reveals that there is considerable variation in the extent to which each potential recipient is a target of donor contributions. While political parties receive contributions from 47 percent of donors, in-jurisdiction House and Senate candidates obtain contributions from just 22 to 27 percent of donors, while other potential recipients receive support from even smaller proportions of donors. ³ The substantial variation in the degree of donor support for each of these recipient types points to the possibility that there are a variety of different donor classes, each with distinctive preferences about where to give.

OPEN IN VIEWER

Table 1.

Percentage of Donors Giving to Each Target by Year.

Targets of Donations	2010	2014
Own House member	27%	22%
Another House member	17%	11%
Own Senator	31%	25%
Another Senator	24%	20%
State office	23%	26%
Political party	47%	47%
Political group	23%	15%
PAC	14%	22%
n	13,020	7,102

PAC = Political Action Committee.

The recipient groups presented in Table 1 are the manifest variables from which we derive our latent classes. In the two midterm election years we examine in this paper, there are eight such indicators.

Determining the Correct Number of Latent Classes

The first step in conducting LCA involves determining the correct number of classes to be estimated by the model. With a large number of different combinations of giving patterns across our donation target variables, it is possible to estimate models with a large number of latent classes. However, as the number of classes increases, the gain in model fit is often minimal, and empty or very small classes with little analytic utility may be produced.

To determine the optimal number of classes to estimate, we conducted a search for the number of classes that provided the best model fit according to the Bayesian information criterion (BIC). The BIC is most appropriate for this task because it calculates the fit of a model accounting for the number of parameters being estimated. Thus, for two models that fit the data relatively similarly, the BIC will advantage the model estimating the smaller number of classes. For each year of our study, we estimated models searching for anywhere between two and seven latent classes and then we plotted the BIC for each of these models. Figure 1 presents this plot.

OPEN IN VIEWER

Figure 1.

BIC values for latent class models estimating two through seven classes.

The BIC plots in Figure 1 reveal a consistent pattern across the two years. (Note that a smaller BIC value indicates a better fitting model.) The figure shows that adding additional latent classes to the model improves model fit up through the fourth latent class; however, including further latent classes results in significantly higher BIC values. The models with four latent classes also appear to meet face validity criteria, as each class is sufficiently large and meaningfully distinctive to justify inclusion. Thus, in the analyses that follow, we estimate four class models for each election cycle.

Covariates for the LCA

LCA allows us to simultaneously estimate the latent classes as well as a set of covariates expected to determine selection into each class. As LCA is computationally intensive, these models generally include a limited number of covariates. In our latent class regression model, we account for four covariates—a donor’s total self-reported family income, ideological extremism, political engagement, and partisanship. We include these measures because they most directly capture the two factors that are most likely to impact donor strategies—motivation and capacity. In terms of motivation, we expect that donors who hold issue positions that are more extreme, who are more politically engaged, and who are more partisan will be more motivated to pursue highly strategic donation strategies that exhibit indications of discernment in the choice of recipients (in particular, selection of out-of-jurisdiction House and Senate candidates, PACs, and issue organizations); while more moderate and less engaged donors will adopt strategies that require less deliberation and thus result in the selection of more obvious recipients (especially political parties) or the adoption of idiosyncratic giving patterns.

For our measure of ideological extremism, we eschew traditional ideology self-report questions and instead create an ideological scale based on responses to a large set of issue questions included on the CCES. Scaling issues provides us with a more direct measure of individuals’ ideologies and also provides much more granularity on this measure. We use a two-parameter logistic item response model to create the issue-based ideology scale. The latent variable extracted from this model is standard normal (with a mean of 0 and a standard deviation of 1). We create our variable for ideological extremism simply by taking the absolute value of this variable. Thus, a value of 0 on the extremism variable would indicate that the individual’s ideology is at the mean for all respondents in the CCES sample, and a value of 2 would indicate that he or she is two standard deviations more extreme than the average individual. Figure 2 shows the distribution of the extremism variable for both of the years included in our study.

OPEN IN VIEWER

Figure 2.

Distribution of ideological extremism among donors.

We also use a two-parameter logistic item response model to create our measure of political engagement. In this model, we include five component terms: whether the respondent indicated that they (1) are very interested in news about public affairs, (2) worked for a campaign, (3) attended a local political meeting, (4) put up a political sign during the past year, and (5) had ever run for elected office. After scaling these measures, we have a measure of political activity that is represented by the plot in Figure 3. Again, this is a standardized variable; thus, a value of 0 identifies a respondent with an average level of political engagement, while a value of 1 would indicate a respondent who was one standard deviation above the mean in terms of his or her engagement.

OPEN IN VIEWER

Figure 3.

Distribution of political engagement among donors.

In addition, we account for the partisanship of donors by including two dummy variables—one indicating whether the respondent identified as a Democrat and the other for whether the respondent was a Republican. ⁴ Across the two election cycles that we examine, 45 percent of donors in the CCES were Democrats and 42 percent identified as Republicans. The rest were independents or affiliated with a minor party.

Finally, a donor’s ability to act on strategic imperatives will be constrained by her capacity to give more money. Accordingly, we expect that income also influences donors’ latent class memberships. Because they can contribute larger sums, wealthier individuals are better positioned to spread their donation dollars over a wider variety of recipients. In contrast, less wealthy individuals may be constrained to focus their more modest contribution dollars on a narrower range of recipients. We thus expect family income to be positively associated with the probability of membership in donor classes characterized by greater diversity in the types of recipients. We use the CCES’s question about family income and code each individual as having an income equal to the lower bound of the income category that he chose. Thus, for an individual who said they earned between US$100,000 and US$120,000, we coded that individual as having an income of US$100,000. In 2010, the top-income category in the CCES was US$150,000 and above. However, in subsequent surveys, the CCES added larger categories to the question, including US$150,000 to US$199,999, US$200,000 to US$249,999, US$250,000 to US$349,999, US$350,000 to US$499,999, and US$500,000 and above. Thus, our analysis of the 2014 donor patterns allows for a more fine-grained investigation of the role of very large incomes in affecting donor types.

Identifying the Four Donor Types

The first step in understanding the results from the LCA is to examine the relationship between the classes themselves and the manifest variables used to partition class membership. In Figure 7, we present information characterizing the profiles of each of the four donor classes identified by the latent class model for 2010 and 2014, respectively. The charts plot the predicted probability that contributors in each class contributed to each of the potential recipient types (Figure 4). Figure 5 shows the proportion of donors who fall into each of the four types as well as the proportion of all funds that CCES donors reported giving in each year that came from each of the four donor types. ⁵

OPEN IN VIEWER

Figure 4.

Patterns of targeting donations for three classes of donors.

OPEN IN VIEWER

Figure 5.

Estimated share of total donors and total campaign funds accounted for by each class of donors.

As the figure indicates, each of the four donor types is characterized by a distinctive contribution profile. The first class of donors are those we refer to as Idiosyncratic Donors. These individuals have a relatively small probability of giving to any of the targets. Accordingly, their profile does not suggest any kind of overarching pattern or strategy. Idiosyncratic Donors comprised 28 percent of all donors in 2010 and 30 percent in 2014. They accounted for an even smaller share of the total funds given in both years (12% and 17%, respectively).

A second class, which we dub Local-Oriented Donors, is distinctive for the tendency of its members to donate to candidates from their own states, though members in this class also are inclined to donate to the political parties. This class of donors has a .4 probability or higher of giving to candidates for state office, political parties, and candidates for a Senate or House seat in their own state. But note that Local-Oriented Donors almost entirely eschew contributing to congressional candidates in other states. Local-Oriented Donors comprise about one-third of all donors in both election years, and they account for roughly the same share of total dollars given.

The latent class model identifies a third class of donors we label Nationalized Donors. This contributor class—while the smallest of the three classes in terms of the number of donors—is distinctive for the extent to which its members contribute to all eight potential recipient types. In addition, what really differentiates Nationalized Donors is their singular tendency to contribute to congressional candidates outside their state. Whereas Idiosyncratic Donors, Local-Oriented Donors, and Party-Oriented Donors (described below) generally ignore these potential recipients, Nationalized Donors have very high predicted probabilities of contributing to them. Because they concentrate contribution dollars on less obvious targets located outside their home turf, Nationalized Donors display a high level of interest in making their influence felt broadly.

While Nationalized Donors only account for about one-in-six of all donors in the CCES data, our analysis reveals that they are disproportionately comprised of medium and large donors. Thus, the relatively small share of this class as a proportion of all donors belies the significance of their contributions in the financing of contemporary campaigns. In 2010, despite accounting for just 16 percent of all donors, Nationalized Donors comprised approximately 45 percent of the total funds contributed. In 2014, Nationalized Donors accounted for about one-third of all money given by individuals, despite making up just 15 percent of all donors.

A final class, which we label Party-Oriented Donors, is characterized by its focus on donating overwhelmingly to the political parties. Members of this class have very high predicted probabilities of donating to political parties, but low predicted probabilities of contributing to any other potential recipients. Party-Oriented Donors make up 23 percent of all donors in 2010 and 27 percent in 2014, but they account for a much smaller share of total money given (just 8% and 16%, respectively).

Explaining Donor Types

We now turn to examining the factors that are associated with donor types. First, Figure 6 shows the predicted probabilities of membership into each donor type based on an individual’s party identification. Several patterns in this figure are of note. First, donors who do not identify with either political party are overwhelmingly classified as Idiosyncratic Donors. Second, while Democrats and Republicans have a similar probability of being Nationalized Donors and Local-Oriented Donors in 2010, there were some distinctions when it came to the other two classifications in that election cycle. In particular, Democrats were more likely than Republicans to be classified as Party-Oriented Donors. This pattern likely reflects the fact that the Republican donor class largely sees the party establishment as too moderate and chooses to give directly to preferred candidates instead, a dynamic that does not exist on the Democratic side of the spectrum (see La Raja and Schaffner 2015, chap. 2).

OPEN IN VIEWER

Figure 6.

Effect of partisanship on donor class membership.

While the partisan patterns are noteworthy, our primary interest is in examining whether membership in one or another latent donor class is closely related both to the impetus to act strategically and to the financial capacity to do so. We expect that ideological extremists and those donors who are more politically engaged will prefer to extend their influence to recipient types who most closely share their preferences, like out-of-jurisdiction House and Senate candidates, PACs, and political organizations; while less engaged moderates will be inclined to focus their contributions on within-jurisdiction recipients and “moderate” organizations like the political parties. All things being equal, therefore, we anticipate that ideological extremism and political engagement should be associated with membership in the Nationalized Donor class.

All things are not equal, however: after accounting for ideology and engagement, donors vary considerably in the amount they are able to contribute to campaigns and political groups. Because the ability to donate large amounts facilitates the spread of contributions across a diverse array of recipient types, we anticipate that individuals who have the capacity to contribute more should be more likely to be members of the Nationalized Donor class.

As estimating the effects of these variables across four classes in two separate years produces a large number of coefficients, we present the full estimates from our models in the online appendix to this paper. In this section, we show the practical implications of these coefficients in terms of predicted probabilities of class membership. Figure 7 shows the influence of each of these three variables—extremism, engagement, and income—on the probability of belonging to the four different donor types.

OPEN IN VIEWER

Figure 7.

Effect of covariates on predicted probabilities of class membership.

The far left column of plots in Figure 7 shows the relationship between ideological extremism and the predicted probability of being in each class of donors. The lack of noteworthy effects for ideological extremism is consistent across the two years we examine. There is an increase in the likelihood that one becomes a Nationalized Donor as she becomes more ideologically extreme, but this effect is modest.

Much more important in structuring donor class membership is our measure of political engagement—a measure of how interested and involved in politics each donor is. Increasing political engagement has a dramatic effect on one’s probability of being in each donor class, and the pattern is consistent across the two election cycles. As we anticipated, donors who are generally more interested and active in politics are much more likely to be Nationalized Donors. In a modest departure from our expectations, donors who are more interested and active in politics are also more likely to be Local-Oriented Donors, as well. This finding suggests that, like their Nationalized Donor colleagues, Local-Oriented Donors are also very selective in making contributions, only donating to within-jurisdiction congressional and state candidates who share their views.

Consistent with our overarching expectations, donors with lower levels of political engagement tend to be either Idiosyncratic or Party-Oriented in their contribution strategies. That is, these less engaged donors either demonstrate no clear strategy in their giving behavior or choose to simply focus on giving to what is arguably the most “obvious” target of donations—their preferred political party.

While political engagement clearly structures what type of donor an individual is, capacity to donate is also an important influence. This can be most clearly seen in the plot for 2014. Recall that in 2010 the top-income category the CCES coded was US$150,000 or above. Thus, the level of granularity at very high incomes is lacking. This is clear becauise we have put both plots on the same x-axis scale. In 2014, the top category was US$500,000 or above. By comparing these two plots, we can see that using the higher income categories is important for understanding the role of financial capacity in affecting contribution strategies. Indeed, what we find in 2014 is that as a donor’s income grows quite large (e.g., above US$250,000 per year), they become more likely to be Local-Oriented or Nationalized in their donation strategies. ⁶ Donors who are at the lower end of the earnings scale tend to be much more likely to exhibit either Party-Oriented or Idiosyncratic donation behavior.

Thus, the Nationalized and Local-Oriented Donor groups appear to represent classes of donors who lie at the intersection of wealth and high engagement/motivation. When donors are more engaged and have the capacity to donate more money, they are more likely to choose to either give to within-jurisdiction congressional candidates and candidates for state offices, or attempt to spread their influence more widely by contributing to out-of-jurisdiction congressional candidates and ideologically congruent political organizations and PACs. In contrast, donors who lack the capacity to give large amounts of money or who are less politically engaged tend to focus much more on giving to the political parties or exhibit idiosyncratic donation strategies.

Discussion

Individual donors play a central role in funding American political campaigns and organizations, yet to this day their motivations are not well understood (Barber 2016). The dominant paradigm conceives of political donors as “consumers” of politics, who contribute primarily as a means of self-expression and entertainment (Ansolabehere, de Figueiredo, and Snyder 2003). Studies of congressional donors, however, suggest that there is more variation in motives among individual donors to congressional campaigns than simply consumption (Francia et al. 2003). Our study sheds additional light on key aspects of contributor behavior, particularly the factors underlying the considerable variation in donors’ choices of targets (including parties, PACs, and political groups, as well as congressional candidates) for contributions.

Our approach, which leverages LCA to identify donor types inductively based on their choices of contribution targets, yields a nuanced narrative about money in politics by showing how differently situated individual donors exhibit different contribution patterns. Rather than being a mass of political “consumers” (or “investors”), the donorate is comprised of several distinctive classes, each with its own characteristic contribution pattern. Of particular note, our analysis provides evidence of a class of engaged and wealthy political donors who spread their dollars widely, especially focusing on giving to out-of-jurisdiction candidates. These donors reflect a small fraction of all donors, yet they have an immensely large footprint in financing the nation’s politics.

These different classes, in turn, differ in their relationships to politically important variables such as political extremism, political engagement, and fiscal capacity. Somewhat contrary to our expectations, political extremism was not strongly associated with assignment to donor classes (Local-Oriented Donors and Nationally Oriented Donors) that most clearly exhibit careful selection of contribution targets. That finding suggests that the donor population, regardless of contribution strategies, is commonly motivated by ideology (Bonica 2014). However, as we expected, another indicator of motivation (political interest) and our measure of capacity (income) both strongly affected the probability of assignment to these more discerning donor classes. Consistent with our overarching expectations, donors with lower political interest and more limited funds were more likely either to donate in an idiosyncratic manner, or to concentrate their contributions on a relatively obvious target, the political parties.

Equally, if not more important, our analysis provides new insights on contemporary debates about the role of campaign contributors in the United States’ highly unequal political system. A popular political motif today is that big donors with extreme views are dominating the American political system and using contributions to make elected officials less responsive to the preferences of moderates (Bonica et al. 2013; Hacker and Pierson 2005). An implication of this view is that empowering smaller donors would help encourage elected officials to become more responsive to the moderate majority (Green et al. 2015; Malbin, Brusoe, and Glavin 2012). The reality is somewhat more complex. As previous work has shown, both small and large donors are both polarized, suggesting that empowering smaller donors would not facilitate moderation in American politics (La Raja and Schaffner 2015). Rather, and as this paper reveals, what really differentiates big and small donors is their respective capacity to act on their views. Politically engaged wealthy donors are positioned to behave as Nationalized Donors and thus make contributions to a wide range of recipient types. In contrast, politically engaged people with more modest resources are constrained from acting on their desire to spread their contributions more widely and as such must content themselves with contributing to their party or in idiosyncratic ways. In other words, what really differentiates big and small donors is the capacity of the wealthiest individuals to pursue “surrogate representation” from candidates and groups outside their congressional district and state.

Based on our findings, some reformers might understandably argue that one remedy for constraining the influence of wealthy, Nationalized Donors is to restore aggregate limits on the total amount of campaign contributions from individuals to political committees. These limits were eliminated by the Supreme Court decision in McCutcheon v. FEC. The Court is unlikely to alter its position anytime soon, but more critically, the most ideologically committed wealthy donors such as the Koch brothers have been able to circumvent contributions limits by supporting independent groups that face no constraints on fund-raising and spending. Given this history, our findings suggest an alternative strategy of constraining the power of Nationalized Donors by encouraging more money to flow through political parties. Such a strategy would encourage more small donors who typically give to parties, while limiting the ability of Nationalized Donors to personally pick-and-choose their favored candidates across the nation.