The dynamics of Supreme Court confirmation hearings

At their core, Supreme Court confirmation hearings should be a routine exercise of Article II’s power of the Senate to offer “advice and consent” to the executive’s appointees. Yet, the speed and success of this process are contingent on the Senate Judiciary Committee, whose decisions to favorably discharge nominations ultimately determines whether the entire chamber will be given the opportunity to pass judgment in a final floor vote. As one of the original standing committees established in 1816, its broad legislative mandate provides it with oversight responsibility and the duty to manage the confirmations of all executive appointments to the judiciary. However, “what began as a largely nondescript committee with only limited input into the fate of Supreme Court nominees developed into a highly visible institution that plays a starring role in the confirmation process” (Collins and Ringhand Reference Collins and Ringhand2016, 143). With this comes a unique political opportunity for committee members to hold hearings and question potential jurists for the nation’s most important legal institution. It is here that we find a platform for senators to engage in strategic rhetorical behaviors through position-taking, posturing, and, more prevalently, grandstanding.

Recent observers note how the starring role of committee members has become increasingly reflective of the contentious nature of contemporary politics. Although several studies have drawn on a collection of theories to frame this development, I contend that two of the most frequently discussed hypotheses serve in conjunction as the primary mediators. First, observers often point directly to the importance of seminal events in the recent history of Supreme Court nominations (e.g., Totenberg Reference Totenberg2011; Elving Reference Elving2018; Wheeler Reference Wheeler2018). These notably include the advent of televised hearings (1981), the contentious rejection of Robert Bork (1987), the hostility framing the Thomas (1991) and Kavanaugh (2018) hearings, and perhaps even the outright refusal to grant Merrick Garland a hearing (2016). Second, more recent literature has increasingly framed these developments as a reflection of political opportunities and interbranch power dynamics. Yet, much of this literature focuses on observable behaviors such as confirmation votes on the Senate floor (Epstein et al. Reference Epstein, Lindstädt, Segal and Westerland2006) or explores rhetoric more holistically through topical analyses or explorations of question and answering behaviors (Collins and Ringhand Reference Collins and Ringhand2013, Reference Collins and Ringhand2016; Farganis and Wedeking Reference Farganis and Wedeking2011; Schoenherr et al. Reference Schoenherr, Lane and Armaly2020). Nonetheless, it could justifiably be expected that these theoretical motivations serve concurrently to explain the variation in rhetorical behaviors. I address these motivations below and develop a theoretical framework that considers these notable events and the conditions underlying competing party goals and the balance of interbranch political power.

A core set of narratives have become commonplace to frame the burgeoning hostility framing modern confirmation hearings. Among them, scholars often analyze these behaviors as a reflection of grievances and other resentments directly or indirectly emanating from key events, most notably the failed 1987 nomination of Robert Bork (Guliuzza et al. Reference Guliuzza, Reagan and Barrett1994; Ogundele and Keith Reference Ogundele and Keith1999; Epstein et al. Reference Epstein, Lindstädt, Segal and Westerland2006). Before Bork, historical accounts would often reinforce that even opposing partisans appeared to recognize that the executive’s power toward judicial appointments was not only solidified in the Constitution, but the Senate’s advice and consent powers were largely marginalized to confirming that the nominee was qualified (Collins and Ringhand Reference Collins and Ringhand2016). However, in the wake of Bork, the Senate’s political dynamic supposedly altered toward a structure that placed a greater burden on the nominee’s ideology and prospective voting behaviors. An extensive collection of literature draws several causal assumptions to frame this development, much of which centers on the proliferation of ideological extremism in the Senate toward all presidential nominees (Silverstein Reference Silverstein1994; Krutz, Fleisher, and Bond Reference Krutz, Fleisher and Bond1998), the growth of interest group involvement in the confirmation process (Caldeira Reference Caldeira1988; Vining Reference Vining2011), or perhaps even the politicization of the Court following the Warren period.Footnote ¹ In effect, beginning with Bork, critiquing a nominee’s ideological or political views was no longer out of bounds. Instead, modern nominees can expect a sort of “democratic reckoning,” whereby they “…are expected to provide their perspectives on a wide range of significant issues and affirm the existing constitutional consensus before taking their seats on the high Court” (Collins and Ringhand Reference Collins and Ringhand2016, 141). Even considering alternative theories, this research collectively depicts an evolving set of behaviors that frames the Bork nomination as a “high-water mark” (Caldeira Reference Caldeira1988, 538).

Yet, it is entirely plausible that Bork is not the only major inflection point, let alone the most important. More recent events, including the tumultuous Thomas hearing (1992), the Senate Republican majority’s refusal to grant Merrick Garland a hearing (2016), and most recently the hostility plaguing the Kavanaugh hearing (2018), lend support to an evolving trend of interparty conflict in Supreme Court confirmation hearings. Approaching this framework through the lens of sentiment-driven rhetorical behaviors relayed by committee members in the hearings, it would be reasonable to expect tangible shifts toward greater hostility in the post-Bork era. Specifically, I expect that the rhetoric employed by committee members will exhibit greater negativity in the post-Bork era that will iteratively become more pronounced over time. That is, notwithstanding other unique conditions framing each hearing, the supposed effect of these key moments should correspond with greater prospects for negative sentiments to be observed among committee members’ statements and remarks.

However, it is important to recognize that while longstanding resentments emerging from events such as the Bork hearing are routinely a point of contention, it should not be lost that committee members operate in response to political opportunities. Mayhew (Reference Mayhew1974) posited that members of Congress are in constant pursuit of reinforcing their electability through advertising their public image, credit-claiming on policies of importance, and using their offices to stake out popular positions. It is reasonable to expect these indicators of opportunity to be exercised when these hearings attract a captivated audience. Members should be motivated to boost their personal and party prestige by advertising their politics through sentiment-structured rhetoric. Indeed, recent developments have highlighted their growing dependence on politicking and grandstanding during hearings (Farganis and Wedeking Reference Farganis and Wedeking2011; Schoenherr, et al. Reference Schoenherr, Lane and Armaly2020). Collins and Ringhand (Reference Collins and Ringhand2016) went so far as to attribute much of these behavioral dynamics directly to the motivations outlined by Mayhew (Reference Mayhew1974). However, while the Senate maintains the power to refuse nominees – and has done so before – confirmation hearings are virtually akin to a formality. Supreme Court confirmation hearings nearly always result in a nominee receiving a floor vote and a successful appointment to the Court (Krutz, et al. Reference Krutz, Fleisher and Bond1998; Moraski and Shipan Reference Moraski and Shipan1999; Nemacheck Reference Nemacheck2008). This expectation was historically reinforced by the fact that parliamentary tactics and norms such as rejecting nominees through blue slips are seen as beyond the accepted practices of decorum for Supreme Court nominees (Denning Reference Denning2001; Black, Madonna, and Owens Reference Black, Madonna and Owens2014). Yet, the idea of a formality in this sense has experienced an evolving definition in recent decades. Even when committee members might not feel empowered to change the outcome, they have become adept at using the hearings to their advantage. Rather than viewing their role as facilitating a neutral fact-finding exercise, contemporary hearings facilitate a dynamic that tends to become more about a charged referendum on the contemporary political climate.

Although these expectations could potentially be replicated for all executive appointments, the Supreme Court produces an exceptional set of circumstances. Presidential administrations recognize that they are poised to nominate an individual to serve indefinitely in a prestigious position. However, their choice of a nominee is bound by interbranch constraints and institutional support (Nemacheck Reference Nemacheck2008). That is, presidents would have less flexibility to appoint a strong partisan when facing an opposition Senate during divided government, and perhaps even more so when the administration lacks strong public support.Footnote ² In the most preferable political environment, an administration would be able to freely select a nominee, construct a supporting narrative that champions their merits, and view confirmation proceedings as a formality. Even if a minority of out-party senators aimed to obstruct, the president could rest assured that a successful appointment was inevitable. However, in situations of divided government, committee members recognize that the change in interbranch party alignments alters the strategic calculus (Schoenherr et al. Reference Schoenherr, Lane and Armaly2020). Under these circumstances, a majority out-party not only has the motivation to engage in obstructionist behaviors, but now they also possess the capacity to impress real consequences that could ultimately derail the nomination. This variation in the balance of interbranch power could serve as a primary mediator of rhetorical behaviors because sophisticated in-party and out-party members recognize that these changing conditions necessitate different approaches to the hearings. As a result, I expect that in-party members will be more likely to employ hostile (i.e., negative) rhetoric as a response to mutually hostile and obstructionist behaviors by out-party members during periods of divided government.

I draw this assumption from findings observed by Schoenherr et al. (Reference Schoenherr, Lane and Armaly2020, 339) and provide a reproduction of their theoretical expectations in Table 1. Regardless of the interbranch power balance, it would be reasonable to expect out-party members to use combative rhetoric to instigate contentious hearings. They specifically note how “out-party senators seek to mitigate the damage of a confirmed nominee” (338), and a primary mechanism to achieve that would be to organize a narrative that frames the nominee as negatively as possible. However, the balance of interbranch political power ultimately determines whether negative framing by a contentious out-party bears the weight of real consequences. If it does not, then a majority in-party presumably has little motivation to reciprocate in a way that deters from the president’s framing of the nominee’s merits and acumen. Alternatively, if a minority in-party is suddenly thrust into a political battle, they would indeed be motivated to fight for the nominee. As a result, I expect “members of the president’s party [to] combat the out-party onslaught” with a similar usage of negative rhetoric (339). However, this expected behavior does not entirely indicate that minority in-party members are channeling their frustration through the nominee, but rather toward obstructing out-party members.

Table 1. Senatorial Motivations and Expectations for Grandstanding

Note: Reprinted from Schoenherr, J.A., Lane, E.A., & Armaly, M.T. 2020. “The Purpose of Senatorial Grandstanding During Supreme Court Confirmation Hearings.” Journal of Law and Courts 8(2): 333–358.

As noted previously, in-party members in the majority effectively pursue a coronation for their president’s nominee. To that end, maintaining the status quo position that their nominee would be an effective jurist remains their primary goal. Yet, this pursuit should not be entirely abandoned when the in-party is weakened to minority status. Instead, while they might be forced to relinquish their ability to pursue a coronation, they must attempt to maintain control of the narrative that their nominee is qualified while ensuring that their grievances are targeting the majority out-party, rather than the nominee directly. This is not to say that majority in-party members will otherwise allow the out-party to lambast their nominee unchecked, but rather that behaviors of in-party members will be dictated by the balance of interbranch political power. While in the majority, the preponderance of in-party rhetoric is likely to preference positivity in pursuit of their coronation and only deviate in occurrences directly targeting out-party members, lest they risk losing their control of the narrative altogether. Alternatively, with their power reduced to minority status, their usage of negative rhetoric will become more pronounced as they are forced to reconcile with a majority out-party whose hostility bears the weight of real obstructionist consequences.

Constructing a text-based algorithm

To measure the behaviors of Senate Judiciary Committee members during Supreme Court confirmation hearings, I expressly focus on their rhetoric between the joint hearings of William Rehnquist and Lewis Powell in 1971 and the Amy Coney Barrett hearing in 2020.Footnote ^{3 ,} Footnote ⁴ Transcripts for hearings between 1971 and 2018 were retrieved from the R-Street Institute (Weissman and Marcum Reference Weissman and Marcum2019). Alternatively, statements and other remarks from the Barrett (2020) hearing were manually coded from C-SPAN’s video transcriptions. The unit of analysis is measured as an individual statement or remark offered by a committee member,Footnote ⁵ rather than as the continuous dyad of interaction shared between them and a nominee. As such, the number of observations is equal to the total number of statements or remarks offered by members of the Judiciary Committee, which collectively totals to (n=) 20,984.

I observe behaviors by measuring the sentiments expressed in committee member rhetoric, which considers the underlying positive, negative, or neutral feelings they are relaying. The emergence of sentiment classification models represents a promising collection of methodological strategies in the social sciences, though they require careful specification to derive meaningful and reliable inferences. My analysis incorporates a supervised learning approach for Naïve Bayes classification of committee members’ rhetoric. The algorithm applies Bayes’ Rule to observe the probability that a statement or remark is in a categorical classification given an approximate mapping of distinct features from a set of training data (Grimmer, Roberts, and Stewart Reference Grimmer, Roberts and Stewart2022). Doing so provides a corresponding score to represent the sentiment of the rhetoric, which can subsequently be converted to a categorical or binary scale to represent its underlying positive, negative, or neutral features. The result is a process that is “simple, fast, and shown to be surprisingly robust to classification tasks with sparse and noisy training data” (Clark et al. Reference Clark, Staton, Wang and Agichtein2018, 107).Footnote ⁶ It further provides relief for concerns that might arise from alternative classification procedures. For example, among the most popular alternatives is the classification of text strings based on adjacent matches with a list of positive and negative words (e.g., Hu and Liu Reference Hu and Liu2004). However, these are routinely prone to validity errors because they “assume the measures created from a dictionary are correct and then apply them to the problem” (Grimmer et al. Reference Grimmer, Roberts and Stewart2022, 182). In an environment like confirmation hearings, where frequently used words and phrases can present alternative meanings based on their usage,Footnote ⁷ universally defined sentiment associations will almost surely produce misclassifications. Instead, a Naïve Bayes approach corrects these concerns by allowing a sample set of data to guide the classification process.

I began by manually coding a sample of 1,177 statements and remarks from the full set of observations, which was subsequently used as training data to compute a term-document matrix of frequent features. These features represent the most common words found across the training sample and are used to estimate their relative predictive weights. In essence, the classifier can estimate rhetorical sentiments using Bayes’ theorem guided by the relative frequency of unique terms and other features observed in the training data.Footnote ⁸ I provide the accuracy report of the Naïve Bayes classifier in the appendix (Table A1). Even with a training set that represents approximately five percent of the total observations, the classifier was able to internally validate negative and positive remarks from a sample testing set with precisions of 82 percent and 76 percent, respectively. Notwithstanding any knowledge of the contextual substance of each remark or statement, the classifier’s ability to correctly estimate classifications at a consistent rate underscores its reliability and validity.Footnote ^{9 ,} Footnote ¹⁰ I offer the distribution of classified rhetoric on a categorical scale in Table 2. Footnote ¹¹

Table 2. Distribution of Committee Member Rhetoric Sentiment Classifications by Hearing

Total observations = 20,984.

Note: Categorical labels were determined by assessing the frequency distribution of the rhetorical classifications, where Negative = [min, –0.77], Neutral = [–0.76, 0.77], and Positive = [0.78, max].

Research design

Using the sentiment classification procedure, I observed the rhetoric offered by members of the Senate Judiciary committee across 18 Supreme Court confirmation hearings between 1971 and 2020. However, the underlying forces that steer each member’s approach to the hearings are likely multidimensional. To evaluate this more directly, I constructed a multinomial logit design with standard errors clustered by each hearing and committee member (dyad) that considers the probability of positive or negative rhetoric. Furthermore, owing to the theoretical framework that rhetorical behaviors vary as a reflection of key events in the contemporary history of confirmation hearings, I conducted two sets of estimations controlling for additional key events beyond the Bork hearing and the advent of televised hearings.

While the mean sentiment score was measured as –0.551, I specifically define positive or negative rhetoric as those whose estimated score was classified as greater than 0.78 or less than –0.77, respectively. The purpose of this was to mitigate concerns of misrepresenting procedural or otherwise neutral rhetoric as relaying positive or negative sentiments. This is a real concern given a situation where the classifier observed an insignificant yet common feature found in the training data and subsequently estimated its sentiment as being marginally greater than or less than absolute zero. Creating these thresholds, which represent approximately 80 percent of the total observations, helps to ensure that rhetoric is indeed presenting sufficient features to qualify as definitively positive or negative.

Measuring positive or negative sentiments as the dependent variable rather than incorporating an additional dependent variable of neutral speech was a conscientious choice to best discern what underlying factors motivate members of the committee to incorporate their partisanships or other rhetoric with strong sentiments, rather than simply approaching the hearings as neutral arbiters. Further, using a multinomial, as opposed to a continuous dependent variable, helps to reduce concerns of overinflating the coefficients. Estimated scores derived from the Naïve Bayes classifier vary considerably from approximately –246 to +1117, though approximately 98 percent of the observations are captured between –25 and +24. Normalizing the scale would still conserve the outliers and risk biasing the coefficients. Even then, sizable scores are almost surely the result of a statement containing voluminous text, rather than being definitively more negative or positive. For example, Senator John McClellan’s opening statement from the joint Rehnquist and Powell hearing in 1971 contained over 1,200 words and was measured as a –17.55 by the classifier. Offered only a few minutes later, another comment from Senator McClellan was measured as a –7.67 but contained only 51 words. Rather than measure the variance in sentiment between the two as a difference of 10 negative units, it would be more appropriate to recognize that the scale is likely reflective of there simply being more negatively associated words in the long opening statement, as opposed to it being more negative outright. The independent variables are grouped into two primary categories that consider the factors mediating committee members’ rhetorical behaviors as a result of key inflection points in the recent history of confirmation hearings, as well as competing interests of party control and the balance of interbranch political power. Apart from those directly concerning the theoretical framework and corresponding hypotheses, I further include a set of additional terms focusing primarily on demographic considerations that have been identified in prior literature as other factors of importance.

The first group considers pivotal moments in the recent history of Supreme Court confirmation hearings that have been identified as key inflection points. The expectation is that evolving behaviors observed during the hearings are the result of long-standing grievances emanating from pivotal moments, as well as potentially in response to the hearings providing a platform to pursue political goals before a captivated audience. I measure this using a series of temporal fixed effects that include the Bork nomination in 1987, as well as the periods following the Thomas (1991) and Kavanaugh (2018) hearings. Apart from contentious hearings, I also incorporate periods following advent of televised hearings beginning with the O’Connor hearing in 1981, as well as the period following the Senate’s refusal to grant consideration to Merrick Garland in 2016. However, to diminish concerns of inadvertently introducing multicollinearity by not disentangling contemporary hearings such as Kavanaugh from registering as satisfying multiple key events variables, I frame the Bork, Thomas, Garland, and Kavanaugh variables as eras to reflect the existence of a hearing during periods when they would be purported to have the greatest effect.Footnote ¹²

The Bork term is specifically employed to test the substantive effects of the hearing as it has been assessed in the prior literature (e.g., Guliuzza et al. Reference Guliuzza, Reagan and Barrett1994; Ogundele and Keith Reference Ogundele and Keith1999; Epstein, et al. Reference Epstein, Lindstädt, Segal and Westerland2006; Collins and Ringhand Reference Collins and Ringhand2013), most notably being the belief that earlier hearings generally displayed diminished expectations of contentious behaviors than those following and including Bork. The second variable draws on theories established in prior research (Farganis and Wedeking Reference Farganis and Wedeking2011; Schoenherr, et al. Reference Schoenherr, Lane and Armaly2020) that the capacity for committee members to pursue personal and party goals was elevated as a response to the hearings being televised, which began with the O’Connor hearing in 1981.

The second group considers the dynamic linking party goals and the balance of interbranch political power. At the core of this dynamic is the importance of party alignments and the balance of interbranch political power identified by Schoenherr et al. (Reference Schoenherr, Lane and Armaly2020). I frame these conditions in the form of interaction terms that indicate whether a committee member is aligned with the in-party, as well as whether the hearing occurred during a period of divided government. While out-party members have an incentive to frame their rhetoric as a hostile interrogation of the nominee, in-party members face competing goals depending on whether they are in the majority. I expect that minority in-party members will respond in kind to hostile rhetoric when the majority out-party’s onslaught has the potential to derail the nomination.

However, it is also necessary to consider what other factors might embolden or stymie rhetorical behaviors motivated by this dynamic. Chief among them is the recognition that rhetoric offered by committee members is not exclusively reserved for nominees, nor would we expect in-party members in either a majority or minority position to direct negative sentiments at the nominee. As such, to consider the effect of whom their rhetoric is targeting, I include two dichotomous terms indicating whether a member’s rhetoric is directed at the nominee or a fellow committee member. The purpose of two dichotomous terms, rather than one indicating whether their rhetoric was followed by a nominee’s response, is to compensate for the (albeit limited) breadth of interactions with non-committee members. Rather than risk conflating the coefficients of a single dichotomous term, I was sure to observe whether rhetoric directed at committee members was not directed at a witness or other speakers such as Anita Hill (Thomas) or Dr. Christine Blasey Ford (Kavanaugh).

Furthermore, I incorporate additional considerations as a reflection of the contemporary political climate and the ideological congruity shared between committee members and nominees. I begin by including the public approval of the nominee’s appointing president as observed by Gallup (2021) in the period nearest the confirmation hearing. The purpose of this is to consider how the president’s institutional support potentially mediates their nominating power and the behaviors of both in-party and out-party members (e.g., Nemacheck Reference Nemacheck2008). While both coalitions are expected to pursue party goals, it would be reasonable to expect diminished rates of negative rhetoric if the president were especially popular. Even if an out-party would prefer to frame the hearings using negative rhetoric, they might view a contentious interrogation of a nominee as counterintuitive to their long-term electoral success if doing so forces them to reconcile with popular support for the president’s agenda. Apart from presidential approval, I include two variables that represent the influence of the political environments framing the hearings. The first is a continuous variable observing the time (in months) between a hearing and the next presidential election, and the second is a dichotomous term indicating whether the hearing occurred during a midterm or general election year. These variables consider factors that mediate the committee members’ pursuit of Mayhewian goals. Namely, it addresses whether the proximity to an upcoming election cycle motivates different behaviors among committee members who might choose to frame the hearings in ways that benefit their personal and party goals.

Finally, recent literature has shown how ideological extremism and congruity shared between senators and nominees have become increasingly important components of confirmation hearings for executive nominees, both including and beyond the Supreme Court (Martinek et al. Reference Martinek, Kemper and Van Winkle2002; Epstein et al. Reference Epstein, Lindstädt, Segal and Westerland2006; Epstein, Segal, and Westerland Reference Epstein, Segal and Westerland2007; Boyd et al. Reference Boyd, Collins and Ringhand2018). While the collective behaviors of in-party and out-party members might be best understood through the lens of competing party goals (Schoenherr et al. Reference Schoenherr, Lane and Armaly2020), it should not be lost that committee members themselves are political actors with distinct beliefs. Even as every member of an out-party might be expected to approach the hearings with greater hostility than those who share the appointing president’s party, the magnitude of that hostility could vary reflective on how much ideological congruence they share with the nominee. As such, I included a variable measuring the comparative ideological distance separating a committee member from the nominee in a NOMINATE-JCS common space (Epstein, Martin, Segal, and Westerland Reference Epstein, Martin, Segal and Westerland2007; Lewis et al. Reference Lewis, Poole, Rosenthal, Boche, Rudkin and Sonnet2022),Footnote ¹³ where an absolute value greater than zero corresponds with diminishing ideological congruity.Footnote ¹⁴ I further consider the ideological extremity of the individual committee member. Prior literature measuring variation in senatorial behaviors toward executive appointees has found that intensifying polarization of legislators often corresponds with diminished confirmation rates and greater expectations of contention among the parties (Krutz et al. Reference Krutz, Fleisher and Bond1998). I frame this condition by measuring a committee member’s extremity as the absolute value of their respective NOMINATE score, where scores greater than zero denote greater ideological extremity.

Apart from the core theoretical frameworks, I include a final group of controls that measures a collection of demographic and other occupational conditions. The first is a set of dichotomous terms indicating a committee member’s legal education and prior public legal experience. Schoenherr et al. (Reference Schoenherr, Lane and Armaly2020) drew from prior research that observed markedly different senatorial behaviors among those with prior experience as lawyers (Miller Reference Miller1995). They specifically observed how their skillset would reasonably lead to performing better at “engaging with nominees” (345) because they have more experience with the hearings’ cross-examination format. The second variable is another dichotomous term indicating whether the committee member was female. While only seven percent of the committee members between 1971 and 2020 were female, behaviors such as those of hostility displayed by then-Senator Kamala Harris during the Kavanaugh hearing (2018) raise questions about whether female members approach their rhetorical behaviors dissimilarly from their male colleagues (Arnholz Reference Arnholz2020). The final variable draws from research that merges judicial confirmations and intersectionality studies by indicating whether the nominee was female. Between 1971 and 2020, five hearings included female nominees.Footnote ¹⁵ While this does not constitute a majority of the nominee population, observers have noted that the experiences of female nominees are pointedly different from their male counterparts. For example, Ringhand and Collins (Reference Ringhand and Collins2010) observed significant variation in committee members’ question-asking behaviors toward female nominees (629–631),Footnote ¹⁶ most notably as it relates to female nominees expecting more significant volumes of interrogatory questions related to their judicial philosophy. I expect that variation in a nominee’s gender will serve as an additional mediator of rhetorical behaviors, at least insofar as interrogatory questions challenging their judicial acumen and philosophy would sensibly be perceived as relaying negative rhetorical behaviors.Footnote ¹⁷