The Party Reacts: The Strategic Nature of Endorsements of Donald Trump

Assessing the Substantive Content of Evaluations

Again, the primary task in this study is to move beyond simple endorsement counts by identifying and grouping candidate supporters based on their stated reasons for endorsing a candidate. To accomplish this task, coders read the article(s) associated with each endorser/endorsement dyad and searched these texts for stated justifications. ² Only reasoning that is clearly attributable to the endorser—rather than, for instance, interpretation by the article’s author—was coded.

Our coding scheme categorizes reasons for assessments into four nonexclusive categories, meaning that any single actor can have multiple reasons for their evaluations of Trump. First, pragmatic reasons represent instances when electoral or policy victory is explicitly prioritized over other concerns. Policy reasons refer to endorsements based on explicit policies supported or opposed by Trump, which can range from narrow (e.g., opposition to the Trans-Pacific Partnership) to entirely abstract (e.g., support for Trump’s policies generally). Ideological reasons, unlike policy justifications, are based on broader normative considerations of how society “ought to be” and the proper role of government in accomplishing this normative vision. Finally, personal reasons for assessments draw on Trump’s individual characteristics, including his personality, background, experience, rhetoric, or behavior. To validate this content analysis, we conducted intercoder reliability tests on a subset of 54 articles that were shared by two coders. The Krippendorff’s alpha coefficients for the four categories were as follows: pragmatic = .709, policy = .782, ideology = .757, personal = .768. We manually adjudicated any disagreements in coding and spot-checked all remaining coding that did not overlap between research assistants.

For each unique actor/endorsement dyad, these four justifications are binary variables coded as 1 if the actor offered such a reason and 0 otherwise. The reason(s) provided by each actor represents the main variable used to connect supporters and group them into clusters based on similar motivations for their endorsement. To gain further insight into Trump’s coalition of supporters, coders also recorded open-ended text to describe, in greater detail, the specific reason provided for each assessment. ³ After data collection was completed, we standardized these text reasons into broad categories inductively identified by the authors, detailed in Table SI.1 in the supplementary information. This standardization attempts to simplify the analysis of reasons for Trump support while retaining valuable information that distinguishes different types of supporters.

Constructing an Endorsement Network

Because the extended party is thought to operate as a coalition—with actors coalescing around “electable” candidates with preferable policy positions—we leverage our original data set to construct endorsement networks. Other scholars have conducted social network analysis on campaign finance records and endorsements as a means of identifying partisan coalitions in Congressional elections and policymaking (e.g., Desmarais et al., 2015; Grossmann & Dominguez, 2009), with the understanding that revealed connections between outside actors, candidates, and politicians from a particular party are representative of the extended party network. In many of these studies—as in ours—ties in the network do not represent explicit connections between actors. Rather, similar revealed behavior is used to map out the Republican Party network in relation to Trump. As such, the “communities” we identify are more akin to subsets of supporters than real-world groups. Our assumption is that by grouping actors according to their reasons for supporting Trump, we can operationalize the subgroups of actors who supported him, and analyze the internal dynamics of the Republican Party network. Our data cannot provide evidence that these actors were directly coordinating as a support coalition, but their reasoned support for Trump’s nomination sheds significant light on coalition building and intraparty conflict. This methodology is an important improvement over raw counts of candidate endorsements, allowing for the identification and clustering of distinct segments of the party and their various incentives and goals.

We construct bipartite endorsement networks, where actors are connected to their endorsement types through the reasons they offer. ⁴ We construct two separate networks. The first network includes all evaluations of Trump with actors linked to the broad reasons for their assessment. The second network includes all actors who made a positive endorsement at some point during the campaign, with actors tied to the standardized textual reasons they provided for their support and, if applicable, their other evaluations of Trump. In the first network, we are able to operationalize the overall response to Trump’s candidacy and assess early support for, or opposition to, his primary run. In the second network, we are able to identify distinct clusters of Trump supporters and use their stated reasons for support to better understand the ideas, demands, and trends that defined the Republican Party’s response to Trump.

The analysis proceeds as follows. First, we introduce the full endorsement network and provide an overview of assessments of Trump. We also present preliminary evidence that personal considerations, rather than policy concerns, were the stated justifications behind many endorsements and most opposition. Next, we measure the extent of party support for Trump during the “invisible primary,” when party insiders are said to screen candidates for their electability and acceptability and converge behind the eventual nominee. Our analysis shows that Trump broke sharply from this model. Finally, we use community detection methods to group supporters with similar reasons for their support, thereby identifying distinct factions within the Republican-extended party coalition. This analysis reveals a stark divide between true Trump supporters and individuals who reluctantly supported the nominee for more practical reasons.

Endorsements During the Invisible Primary

The fact that policy rationales played a relatively minor role in evaluations of Trump runs counter to the “Party Decides” (2008) framework of presidential candidate emergence. In short, Cohen et al. (2008) predict that early party insider support for a candidate and his or her policy stances manifests in endorsements and predicts whether a candidate will emerge as the party nominee well before party caucuses and primaries. This preprimary period is often referred to as the “invisible primary,” when candidates attempt to display their viability by accumulating endorsements, developing campaign infrastructure, and building up financial war chests (e.g., Aldrich, 2009; Cohen et al., 2008; Steger, 2000). By most popular accounts, Trump did not operate in this postreform model.

For this reason, in this section, we quantitatively examine the extent of early support for Trump and compare the policy demands of early supporters with theoretical expectations. To begin, Figure 1 tracks the “net affect” toward Trump among formal and informal party members. Our definition of formal party members (roughly 61% of the sample) is straightforward, including any current or former elected officials (e.g., governors or members of Congress), candidates for office, and partisan employees (e.g., staffers). All other actors in the sample were coded as informal party members, including public figures (e.g., actors or athletes), nonparty organizations, media outlets, and business elites. Net affect is calculated by assigning numeric values to each endorsement type and tracking the sum of these values across every week of the campaign. Specifically, endorsements are valued at +1, opposition or withdrawn endorsements at −1, and criticisms at −0.25. Thus, higher positive values in Figure 1 indicate greater overall support for Trump, at least among the actors in this sample. ⁶ An additional line tracks the total number of endorsements (minus withdrawn endorsements) over time. The figure also marks five key events during the 2016 cycle.

OPEN IN VIEWER

Figure 1.

Net affect toward Trump.

The most striking finding from this figure is the fact that Trump’s net approval was close to zero leading up to and following the earliest Republican caucuses and primaries. Although we do not compare this early support with preprimary support for other Republican candidates, the figure, nevertheless, reveals that Trump did not emerge from the invisible primary with strong, diverse party support. This is true for both formal and informal party members, though the net affect among informal party members was somewhat higher than formal party members throughout the time period. ⁷ Indeed, for both types of actors, Trump did not build substantial support until he had amassed a significant number of caucus and primary victories. Roughly half of his endorsements came after he secured the delegates needed to win, and his highest net favorability similarly occurred between the time when he had secured the nomination and his victory at the Republican convention. After this point, net affect toward Trump plummeted following several political stumbles during and after the convention, with growing numbers opposing his candidacy. The exposure of his lewd comments in the Access Hollywood tape accelerated this trend, and by Election Day 2016 Trump was, on average, opposed by the party establishment.

Who, then, supported Trump’s early candidacy? Although very few actors offered assessments of any kind prior to primary voting—suggesting that the usual suspects were largely inactive during the invisible primary—it is possible that a limited number of influential party insiders aligned behind Trump. Further qualitative analysis, however, suggests that this was not the case. Only 12 actors had positively endorsed Trump prior to the Iowa caucus, and this group—which includes Sarah Palin, Carl Icahn, and Dennis Rodman—is far from the notion of party insiders referenced by Cohen et al. (2008). ⁸ In fact, in the Party Decides analysis (1980-2004), the average eventual nominee received 328 endorsements prior to the Iowa caucus. ⁹ Even relative outsider Ronald Reagan received 117 endorsements in his successful 1980 bid, or nearly 20 times as many preprimary endorsements as Trump.

The members of the early Trump support coalition offered a variety of reasons for their endorsements (see Table 2), with several citing his rhetoric and temperament, his business experience, and his policy positions on national security. A related strand of reasons drew on his populist message, citing the fact that Trump, as an antiestablishment populist, would “Make America Great Again.” It is fair to say, then, that Trump drew his preprimary support not from the typical party establishment insiders but rather from like-minded political outsiders and nationalists. Importantly, none of these early supporters named Trump’s electoral prospects as a reason for his or her endorsement, and the few policy reasons offered are far from the wide ranging list we would expect if Trump had assembled a “big tent” coalition of various segments of the Republican Party.

OPEN IN VIEWER

Table 2.

Specific Reasoning Offered for Endorsements During the Invisible Primary (Prior to the Iowa Caucus).

Reasons for endorsements in the invisible primary
Ideological	Personal	Policy	Pragmatic
“MAGA” (3)	Temperament (4)	National security (4)	Will aid CA secession (1)
Antiestablishment (1)	Business experience (3)	Islamic policy (1)
Populism (1)	Rhetoric (2)
Will change country (1)	Leadership (1)

Note. The number of unique actors who mentioned each reason is in parentheses.

Only 10 other actors made assessments of any kind regarding Trump’s early candidacy. Four of these actors were indifferent toward Trump and, like later actors who offered indifferent assessments, these people were often supportive of Trump’s business experience and the fact that he broke from politics as usual. Still, these actors were apprehensive about Trump’s electoral prospects, particularly due to rhetoric that they perceived as racist and sexist. Surprisingly, only five actors in our sample opposed Trump before the first caucus, and two of these critics (Bobby Jindal and Lindsay Graham) were Trump’s direct primary opponents. Once again, we see that party insiders simply abstained from making assessments of Trump early in the campaign. His later success, then, cannot be attributed to early insider or establishment support. Rather, Trump operated outside of the traditional avenues of political success, perhaps by exploiting free media to stand out among a crowded primary field and appeal to dissatisfied voters (Francia, 2018; MacWilliams, 2016; Wells et al., 2016). Regardless of the cause, Trump’s victory seems to upend theoretical expectations. We do not believe this fact invalidates coalitional theories of candidate emergence, but it does raise an important question that the theory and data in existing studies do not address: In the face of an undesirable, yet victorious, party nominee, how does the party react?

The Party in the Postnomination Period

Despite Trump’s lack of early party support, he ultimately ended his campaign with positive endorsements from 117 actors. Did this convergence behind the Trump candidacy represent true support—perhaps due to Trump policy concessions that appealed to a new segment of the Republican Party—or simply reluctant acceptance of the nominee? In this section, we address this question by focusing on the positive Trump endorsement network, including all actors who endorsed the candidate at some point during the election. To understand the shifting nature of this coalition, we also include any other types of assessments that these endorsers made.

This positive endorsement network connects actors to the standardized textual reasons they stated for their assessment. Thus, this network provides far more information on potential variation in the justifications and policy demands of actors in the support coalition. These more nuanced ties also allow for the use of community detection methods to uncover meaningful subgroups in the network. Specifically, we use the Spin-Glass community detection algorithm (Reichardt & Bornholdt, 2006) to identify clusters of actors with dense internal connections and sparse ties across subgroups. ¹⁰ Again, these groups are not communities in the traditional sense, but rather, clusters of actors with similar reasons for their support of Trump. This analysis resulted in 12 unique clusters; the various types of endorsements and reasons provided by actors in each community can be found in Table SI.2 in the supplementary information.

Broadly, these support clusters can be classified into three distinct types. First, several can be described as policy-demanding coalitions (26% of actors), consisting of supporters who mainly drew on specific policy reasons to justify their endorsements. The most homogeneous policy demanding subgroup, which we call the Islamic policy wonk cluster, is relatively small but consists entirely of actors who supported Trump’s approach to Islamic extremism. A second group (immigration policy wonks) drew on a more heterogeneous list of typically conservative policy demands—including guns, health care, national security, and taxes—though a large majority of members (72%) share an emphasis on immigration policy. Similarly, actors in the third policy-based community (conservative policy wonks) referenced a mix of hot-button conservative issues, including the economy, energy policy, gun control, and health care and immigration reform.

Importantly, the median dates of endorsements in these policy-based communities are relatively late in the cycle. Only the immigration-focused cluster had a median endorsement date before Trump secured the delegates needed to win the nomination, making it unlikely that the support of these policy-demanding groups and individuals was responsible for Trump’s eventual success. Still, Trump could have conceded to other policy demands from these actors as the campaign progressed. On immigration and Islamic extremism policy, this seems unlikely, as Trump’s public statements on these topics began well before these endorsers had announced their support. It is possible that Trump adjusted his positions on gun policy, health care, or the economy to accommodate the demands of his elite supporters, though these issues are only mentioned by a handful of policy-focused endorsers (between 7% and 22%). Regardless, these policy emphases—limited as they are—are representative of the priorities of both the Trump campaign and the early Trump administration and, thus, have implications for our understanding of the Republican Party’s response to Trump and his governing priorities and capacity as President.

Besides this limited number of policy-based endorsements, the majority of the network can be broadly classified into two other camps: candidate-based endorsers (45% of actors) and reluctant supporters (29%). Candidate-based supporters tended to draw on Trump’s personal characteristics and style as well as the general antiestablishment, protectionist rhetoric of his campaign. For example, members in one cluster frequently referenced Trump’s business and leadership skills, with several suggesting that these traits provided Trump with relevant experience to improve the economy and manage the executive branch. Importantly, some of these endorsers, like Representative Chris Collins (NY-27), disagreed with Trump on key policy issues—especially immigration and terrorism—but, nevertheless, expressed the belief that Trump’s background made him the best candidate. Once again, this support seems to be based not on existing policy commitments or the expectation of future concessions, as coalitional theories might suggest.

Other candidate-based support came from antiestablishment populists and nationalists who sought a “new direction” for the country. Some simply implored Trump to “please make America great again.” Similarly, most actors in these communities sought “change” in the abstract, whether that meant a different culture in the capital, a move away from establishment “politics as usual,” or simply working “to defeat anyone who promises to finish what President Obama started.” Importantly, these endorsers viewed Trump as the best catalyst for such change, not because of the policies he advocated (though some did reference the odd policy in their endorsements), but rather, because of Trump’s style, temperament, and message. Indeed, many of these actors endorsed Trump despite their criticism of his sexist remarks or particular policies they deemed ineffective or even impossible. In these candidate-centric endorsement communities, then, we find some evidence of Polsby’s (1983) hypothesis that the reformed presidential nominating system would lead to the nomination of candidates with strong personal followings but who lacked consensual support from a majority of the party. By campaigning on his past business success and as the champion of populist, nationalist causes, Trump was able to garner support from actors who otherwise might have opposed his candidacy in a contest based around policy positions and “electability.”

Not all endorsers, however, were willing to ignore the more politically toxic aspects of the Trump candidacy. Rather, the Republican establishment was decidedly opposed to Trump in early months. The fact that many of these actors eventually supported the nominee suggests that party insiders—faced with an undesirable nominee—did not strongly assert themselves but rather reluctantly supported the candidate. What, then, led these reluctant supporters to fall in line behind the eventual Republican nominee? The answer, in short, is political pragmatism and partisan electoral goals. The reasoning of these reluctant supporters reveals that, in some cases, a party establishment that fails to proactively select its nominee is willing to adapt in the name of party unity.

Importantly, the vast majority of these reluctant supporters (approximately 75%) can be considered “political insiders” (i.e., current or former politicians and party leadership or staff), including prominent names like Reince Priebus, Paul Ryan, and Mitch McConnell. It is unsurprising, then, that many of the actors found in the reluctant supporter communities were initially indifferent about Trump’s candidacy. The reasons for this indifference were varied, including Trump’s rhetoric, temperament, and dishonesty; his sexist remarks; and his atypical ideological and policy stances. Generally, then, party insiders viewed Trump as unelectable and unconventional in his policy platform. At the same time, some indifferent endorsers paired these criticisms with vows “to support the Republican nominee” or claims that “Clinton is not the answer,” however unpalatable Trump might be. Combined, these statements reveal the profoundly difficult choice faced by many party insiders in 2016: to oppose Trump and risk alienating his populist base and losing the election, or else, support a candidate who most believed to be a danger to the Republican brand.

Our sample of endorsements provides insight into actors who chose the latter option. Among reluctant endorsers, the two most cited reasons for their eventual support were the facts that Trump was “not Clinton” and that they would ultimately support any Republican, highlighting a strategic pursuit of electoral goals but not necessarily a set of policy demands. For these supporters, then, party unity played a crucial role in their decision. Trump’s character and campaign, however, did not: Examining the reasons offered by these communities, most notably what we call the post hoc rationalizers community, reveals that many insiders were searching for some legitimate reason other than pure party politics to justify their endorsements. These reasons varied, with some even pointing to vice presidential candidate Mike Pence as their only reason for support, but in almost every case, the errant policy or ideological justification was paired with a pragmatic statement about the need to win and the need to prevent a Clinton presidency. A significant faction of the Trump endorsement network, then, consisted of unwilling party insiders who pragmatically supported the nominee, often after vociferously criticizing him earlier in the cycle.

A small number of these reluctant supporters, however, proved to be fair-weather friends. All five of the actors who eventually withdrew their endorsements—each one in response to Trump’s comments in the Access Hollywood tape—are classified as reluctant supporters, suggesting that there are limits to the willingness of party actors to overlook major transgressions. Similarly, the vast majority of postendorsement criticism came from insiders who were hesitant to endorse Trump in the first place. Taken together, these findings suggest that the Republican Party apparatus was largely a passive player in Trump’s victory, at least as we have operationalized influence. To the end, many party elites were lukewarm at best toward Trump and his chances of victory, whereas those who fully supported the candidate were often “outsiders” in traditional Republican politics.