Ruth Bader Ginsburg died, as a sitting Supreme Court Justice, on September 18, 2021. In the weeks prior, the Court was quietly on recess while attention focused on Donald Trump, COVID-19, and other issues. In the days after, despite having not issued a full opinion for months, the Court was caught up in intense partisan contestation over an open seat. Ginsburg’s death in the closing weeks of the Trump presidency made a shift in the Court’s ideological makeup likely. It also roused the politics of the failed nomination of Merrick Garland, and the disputed confirmations of Neil Gorsuch and Brett Kavanaugh. Donations to Democratic Senate candidates surged (Goldmacher Reference Goldmacher2020), and discussions of judicial reform ideas such as term limits edged into the mainstream. For all of these reasons, the fall of 2020 provides a rare and important opportunity to further our understandings of institutional support for the Court in the face of political and partisan shocks.
In this short paper, I ask: how did Justice Ginsburg’s death, and the intense politics around the unexpectedly open seat, affect views of the Supreme Court? I do so using data from a survey that was in the field at the time. I compare those who responded in the days just before her death to those that responded just after it. My focus is whether this period changed how much legitimacy people granted the Court, and whether it sparked new interest in institutional reform. I also assess whether it changed the extent to which people saw the Court as political or ideological. I find that Democrats who responded just after Justice Ginsburg’s death saw the Court as less legitimate, by about 1/3 of a standard deviation. They also expressed more interest in learning about reform proposals. However, Democrats’ other views about the Court were unaffected, as were Republicans’ views in general.
Legitimacy and Legitimacy Change
‘Legitimacy’ and ‘diffuse support’ broadly refer to people’s general loyalty to the Supreme Court as institution. This loyalty is seen as a key source of the Court’s influence, a primary reason people accept the authority of its controversial decisions on critical matters, and a bulwark against calls for fundamental institutional change. It is also seen as a deep and durable ‘reservoir of goodwill’ that the Court can generally tap into (e.g. Caldeira and Gibson Reference Caldeira and Gibson1992; Gibson, Caldeira, and Baird Reference Gibson, Caldeira and Baird1998).
I test the broad hypothesis that the events surrounding Justice Ginsburg’s death reduced diffuse support for the Court – at least among Democrats. Extant findings and theories offer a mix of reasons to expect, and to not expect, diffuse support to respond in this case (e.g. Sen Reference Sen2017; Gibson and Nelson Reference Gibson and Nelson2017; Armaly Reference Armaly2018; Rogowski and Stone Reference Rogowski and Stone2019; Badas Reference Badas2019b; Nelson and Tucker Reference Nelson and TuckerForthcoming). The two most similar recent studies to this one offer no consensus. Using a conjoint experiment, Rogowski and Stone (Reference Rogowski and Stone2019) find that partisan contestation over judicial nominees polarizes views about legitimacy. In contrast, Armaly (Reference Armaly2018) finds that views about the Court were unaffected around the recent contested seat opening following Justice Scalia’s death.Footnote 1
While the importance of legitimacy as a concept is widely accepted, there is less consensus about the mechanisms through which it can change. The literature offers at least three theories that can support the prediction that this case would affect legitimacy (see e.g. Clark and Kastellec Reference Clark and Kastellec2015; Armaly Reference Armaly2018; Rogowski and Stone Reference Rogowski and Stone2019, for more thorough elaborations). The first is changed perceptions of the way the Court makes decisions. According to perhaps the most influential theory of diffuse support, legitimacy is high and stable unless people come to believe the Court is ‘not different’ and that it makes unprincipled decisions like other political actors (e.g. Baird and Gangl Reference Baird and Gangl2006; Christenson and Glick Reference Christenson and Glick2015; Gibson and Nelson Reference Gibson and Nelson2017, Reference Gibson and Nelson2015). While there is good reason to discount this mechanism in this instance since the Court itself did nothing, it is plausible that the salience and rhetoric changed minds about how the Court operates. The second is expected changes to the Court’s ideology and decisions. Some recent work argues that people focus on the Court’s outputs and see it as less legitimate when they think it is out of sync with their own views (e.g. Bartels and Johnston Reference Bartels and Johnston2013; Christenson and Glick Reference Christenson and Glick2015). This perspective would predict that a seat switching from a Justice appointed by a Democrat to one appointed by a Republican would lead Democrats to see the Court as less legitimate. The third potential mechanism, partisan source cues, is rooted in the application of broader political behavior ideas to questions of support for the Court (Clark and Kastellec Reference Clark and Kastellec2015; Sen Reference Sen2017; Rogowski and Stone Reference Rogowski and Stone2019). Simply, people have limited knowledge about the Court and thus rely on what co-partisan elites are saying such that Democrats’ support declines when their elites are questioning the Court’s legitimacy and discussing reform (Clark and Kastellec, Reference Clark and Kastellec2015; Rogowski and Stone Reference Rogowski and Stone2019).
This study also connects to additional issues in the literature concerning the Court and legitimacy. First, while much of the literature focuses on the effects of the Court’s own actions (e.g. Gibson, Caldeira, and Spence Reference Gibson, Caldeira and Spence2003; Christenson and Glick Reference Christenson and Glick2015), some recent work asks whether legitimacy is at risk even if the Court itself does nothing? Armaly (Reference Armaly2018) includes a thorough review to frame the question about outside influence, while not finding much of it. However, experimental work has found that others’ rhetoric can shape views of the Court (e.g. Clark and Kastellec Reference Clark and Kastellec2015; Rogowski and Stone Reference Rogowski and Stone2019; Nelson and Gibson Reference Nelson and Gibson2019; Armaly Reference Armaly2020). This study also contributes to the broader literature focused on the politics of appointments and confirmations (e.g. Caldeira Reference Gibson and Caldeira2009; Gibson and Sen Reference Sen2017; Chen and Bryan Reference Chen and Bryan2018).
While a secondary focus, I also engage with work focused on measuring institutional support. Most studies of Supreme Court legitimacy use a common battery of survey questions to measure the latent concept (e.g. Gibson, Caldeira, and Spence Reference Gibson, Caldeira and Spence2003). However, recent work proposes alternatives (Badas Reference Badas2019a,Reference Badasb; Bartels and Kramon Reference Bartels and Kramon2020). Most pertinent to my analysis, Badas (Reference Badas2019a,Reference Badasb) argues for asking direct questions about support for reform proposals. Clark and Kastellec (Reference Clark and Kastellec2015) also use items such as support for term limits as key outcomes. Since legitimacy means loyalty to the institution and resistance to structural change, interest in reform should correspond to less legitimacy. I have both a conventional legitimacy battery and a question about interest in reform. Thus, I can evaluate whether the two types of measures move together.
Data and Design
My data come from a broad ranging survey I had in the field in the late summer and early fall of 2020. Justice Ginsburg died on September 18th. I have roughly 1,000 ‘pre’ responses collected between August 27th and September 16th, and more than 500 ‘post’ ones from between September 19th and 30th. The latter period begins just after Justice Ginsburg’s death and ends days after Justice Barrett’s nomination was announced. Qualtrics recruited participants from its panel of potential respondents and quota-sampled on race, gender, education, and age to match census demographics. This resulted in a sample that was also representative on political dimensions. The party identification breakdown was 44–39 Democrats at a time when Pew’s comparable estimate was 49–45. President Trump’s approve/disapprove was 41/51 in the sample and 43/53 in the 538 average.Footnote 2 Below I show comparability between pre and post respondents in Table 1.
Table 1. Comparison of respondents who participated just before Justice Ginsburg’s death to those who did so just after
1 Partisanship is measured with a 7 point branching question with 7 = “Strong Democrat,” Political Info is a four question battery.
2 College, Female, White, Homeowner and “Identify John Roberts (as Chief)” are proportions of dichotomous variables.
I fielded the survey to collect data for a few disparate projects. Respondents spent roughly 15 minutes answering questions across a variety of topics. Modules other than the Supreme Court one on which I focus covered issues such as housing preferences, COVID-19 rhetoric, and state politics.Footnote 3 Key to this paper, the survey included an experiment about the Court’s recent decision regarding President Trump’s tax returns. Respondents were randomly assigned to read different plausible presentations of the same case outcome. For example, some read a version that was framed in terms of the ‘winners’ and ‘losers’ in a horserace analysis. Others read a version that emphasized the Court’s internal coalition building and ability to strategically foreclose any real action until after the election. After reading their assigned presentation of the tax case, respondents answered a battery of standard questions about legitimacy, politicization, and legal realism.
In this paper, I do not analyze this framing experiment. Instead, I leverage the fact that it happened to be in the field on both sides of Justice Ginsberg’s death. I only use the 373 respondents that were in the control arm. Even those in the control were presented with basic information about the tax returns case. Specifically, they read that the Court recently decided cases about subpoenas for President Trump’s tax returns and financial records, that ‘seven out of the nine Justices rejected the President’s argument that presidents are immune from subpoenas while in office,’ and that ‘the Court allowed the investigations to continue subject to other disputes and arguments about evidence in the lower courts.’ The full text is in the Supporting Information. The control did not include the stronger political frames that the other conditions did. Of course, the ideal design would include a very large pool of respondents who only answered questions and did not read anything. However, the tradeoff in this case is external validity and timeliness. Importantly, any potential contamination from the control condition text should work against my findings about Democrats. It reminded them that the Court did not simply act like part of the Trump team as it rejected his immunity claims.
I analyze three measures that are common in the literature using a three item standardized index for each (e.g. Gibson and Nelson Reference Gibson and Nelson2017). Question order was randomized and full wording is in the Supporting Information. The primary variable of interest is institutional legitimacy or diffuse support. I include three standard questions about whether: 1) the Court ‘gets too mixed up in politics,’ 2) its power ‘to decide certain types of controversial issues should be reduced,’ and it ‘should be made less independent.’ I also include a common measure of legal realism – perceptions of how much sincere legal ideologies influence judicial decisions. Finally, I measure perceptions of court politicization with questions about, for example, whether judges are ‘politicians in robes.’
In addition to these conventional indices, the survey also included a more direct measure of respondents’ interest in knowing more about the Court, politics, and institutional change. At the end of the survey, respondents were asked to select up to two of the survey’s topics that they would most like to ‘learn more about.’ Two of the response options were ‘ideas for restructuring the courts’ and ‘ideas about control over filling judicial vacancies as a campaign issue.’ The other three options pertained to unrelated issues such as housing policy. The court reform item nods to the applied legitimacy index in Badas (Reference Badas2019a). Given that legitimacy ostensibly captures institutional loyalty, interest in institutional reform should be negatively related to it.
Before turning to the results, I briefly address a couple of important questions about the data. The first is how well the respondents balance across the pre and post groups. Table 1 summarizes respondent traits across a variety of attributes. It shows no significant differences in means or proportions. Balance on political knowledge is especially important since the results of a real world intervention in a relatively small sample could be quite sensitive to awareness. Since one of the knowledge items was about the position that John Roberts holds, I also report this separately as an indicator of Court knowledge.
The second basic question about the data concerns base rates and variation. On the questions about wanting more information: five percent of respondents choose both the item about reform ideas and the item about appointments as a campaign issue. Seventeen percent chose the reform option but not the appointments one. Fourteen percent chose the appointments item only, and 64 percent choose neither. Legitimacy scores were negatively related to curiosity about changing the Court. The mean standardized legitimacy score for the 17 percent who wanted information about reform proposals but not appointments was −0.25. It was 0.12 for those who did not want more information about either.
Results
Figure 1 shows the effect of completing the survey after Ginsburg’s death on five outcomes. It reports results for Democrats, Republicans, and all respondents. For simplicity, I plot estimates from linear models for all five variables with the demographic controls documented in the figure footnote. Estimates to the right of the zero line indicate an increase after Ginsburg’s death. I also report all coefficients from the models using the legitimacy and reform dependent variables in Table 2. For robustness, I replicate the substantive results for the two dichotomous outcomes using logit models in the Supporting Information where I also report full results tables for all models. The Supporting Information also includes a bivariate plot similar to Figure 1 that shows the same main results without controls.
Table 2. Change in Legitimacy and Reform Interest
Note: Standard errors in parentheses. Models: 1-3 are OLS on standardized DVs, 4-6 linear estimates with dichotomous DVs.
+p<0.1, *p <0.05, **p<0.01.
Figure 1. Effects of “open seat” condition (being a post Sept 18th respondent) on key DVs by party. The Legitimacy, Politicized, and Realism effects are standardized continuous estimates from OLS. The Court Reform and Court Appointments results are estimates from linear probability model (logit equivalent in Supporting Information). All are relative to the “pre” respondents. Full table output with controls in the Supporting Information and output for the legitimacy and reform variables are also in Table 2.
The effects on legitimacy and resistance to institutional change are displayed in the two left-hand panels of Figure 1. In the upper left of the plot, legitimacy scores among Democrats who responded after Ginsburg’s death were roughly one third of a standard deviation lower than they were among those who responded before it. Legitimacy among Republicans held steady. For context, a 1/3 standard deviation reduction is comparable to the change that Nelson and Gibson (Reference Nelson and Gibson2019) find among those with ‘complete confidence’ in former President Trump after they are exposed to him attacking the Court.Footnote 4 These results are also generally consistent with the within subject changes reported in a working paper that uses MTurk panel data over a similar time frame (Armaly and Lane Reference Armaly and Lane2022), and to recent experimental findings (Rogowski and Stone Reference Rogowski and Stone2019). The lower-left of the figure shows that interest in learning more about Supreme Court reform increased by about 18 points among Democrats.Footnote 5 P-values from tests comparing the ‘post’ effect from the models using only Democrats to the equivalent effects among Republicans are 0.09 for legitimacy and 0.01 for reform interest.Footnote 6
There is suggestive evidence, that the effects may have been more pronounced for moderate Democrats than for strong ones. In Figure 2, I plot mean legitimacy and reform interest in the pre and post periods by type of Democrat. The plot shows that, relative to self identified ‘strong Democrats,’ legitimacy started higher but then declined considerably for ‘other Democrats’ (including leaners). Interest in reform did the opposite. In both cases, all types of Democrats’ views came to look quite similar after Ginsburg’s death. However, all of these estimates reflect small subgroups (pre/post), of a subgroup (partisan strength), in a modest sample and thus have wide confidence intervals. The p-value on the interaction between strength of partisanship (among Democrats) and being a ‘post’ respondent (with the same controls as in the other models) is 0.07 for legitimacy and 0.36 for reform interest.
Figure 2. Descriptive Data: Mean legitimacy scores and reform interest by partisan strength among Democrats. Self identified “Strong Democrats” are those who place themselves in the most extreme category on the seven point branching measure. Ns: Strong Democrats (65 pre, 34 post), Other Democrats (43 pre, 24 post).
Finally, in contrast to the diffuse support results, there is no indication that perceptions of legal realism or politicization on the Court were affected. These null results hold with and without controls. Interest in learning more about judicial appointments as a campaign issue was remarkably similar across specifications and subgroups.
Additional Results: Measuring Diffuse Support and Relationships Among Variables
I now briefly turn to two additional analyses. The first concerns measurement. One would reach the same basic result above whether one used the conventional legitimacy battery or the question about interest in reform ideas. While this conclusion, rather than its measurement, is my focus, having both outcomes connects to recent methodological work that advocates asking direct questions about support for institutional change (e.g. Badas Reference Badas2019a). Having both measures around the same real world event helps validate both and helps to connect some substantive magnitude to changes on the legitimacy scale. In this case, a one third standard deviation drop in legitimacy corresponded with a nearly 20 point increase in curiosity about structural reform. To further investigate the connections between the two measures I estimated 1) models in which legitimacy is the dependent variable and interest in reform is an independent variable along with response timing (pre/post) and all of my demographic controls, and 2) the same models but with reform interest as the dependent variable and legitimacy as an independent variable. All else equal, curiosity about institutional reform is associated with .27 standard deviations less legitimacy (p<0.05) when pooling all respondents. Likewise, less legitimacy is significantly associated with more interest in reform (p<0.05). (Full models and a plot are in the Supporting Information).
My second additional analysis addresses the question: do previously documented relationships between key views about the Court hold up under stress? As Figure 1 above shows, Ginsburg’s death lead to aggregate changes in Democrats’ legitimacy views without corresponding changes to their perceptions of legal realism or politicization. However, these aggregate results cannot tell us whether the associations among these variables at the respondent level remained consistent. This question is important because these well established relationships are central to the argument that people differentiate a Court influenced by sincere ideological inclinations from one influenced by politics (e.g. Gibson and Nelson Reference Gibson and Nelson2017). Prior work on these relationships has been conducted during ‘normal’ times. Here, I test them when we know that a) Court politics were very salient, and b) some views about diffuse support were in flux.
To assess how well these relationships hold up, I estimate OLS models in which standardized legitimacy is the dependent variable and standardized realism and politicization are the independent variables of interest. I include the same controls as above which makes the models similar to the baseline model in Gibson and Nelson (Reference Gibson and Nelson2017). I split the sample by pre and post respondents. I plot the key results in Figure 3, and include a full table in the Supporting Information. The first finding is that legitimacy and realism are positively and significantly associated in both the pre and post periods. Legitimacy and politicization are negatively and significantly associated in both. The magnitudes of these relationships also appear stable. The second finding is that while these relationships are notably persistent under the stress of this period, the effect of partisanship changes considerably. In the pre period, there was no relationship between partisanship and legitimacy with realism, politicization, and other variables in the model. However, in the post period, there was a significant negative relationship between being a Democrat and legitimacy above and beyond all of the other variables. The p-value comparing the effect of the ‘Democrat’ variable in the pre and post models is 0.03. Finally, these relationships are where the legitimacy measure and the interest in Court reform measure most differ. They do not manifest in the same way using the reform interest outcome variable.
Figure 3. Linear estimates of effects on standardized legitimacy index on the left and the dichotomous reform interest measure on the right. Both panels separately estimate for “pre” and “post” responses. Both include controls for age, college education, white, female, homeowner.
Discussion and Conclusion
There were good reasons to expect that this case would yield externally valid evidence, consistent with recent experiments, that legitimacy is responsive to external events outside of the Court’s control. There were also good reasons to expect that this study could bolster, in a challenging setting, arguments that legitimacy is resistant to even short term effects. My primary finding is consistent with the former. Democrats’ legitimacy scores shifted by one third of a standard deviation without the Court itself doing anything. A question capturing curiosity about Court reform picked up on the same dynamic. Notably, this movement happened even before a salient and contested confirmation hearing or vote.
In the scope of legitimacy change, even acknowledging that these data cannot speak to effect durability, this is substantial short term movement. However, it is simultaneously possible that this change represents a ceiling and thus also bolsters arguments that legitimacy is fundamentally robust. The promise of a very unpopular president replacing the other side’s judicial icon, in the weeks before an election, replete with mainstream discussions about ‘court packing,’ moved legitimacy by a fraction of a standard deviation among one party’s identifiers. Because the legitimacy change was partisan, aggregate support for the Court did not change significantly. Moreover, this period did not undermine the fundamental relationships between legal realism, perceptions of politicization, and legitimacy.
Unfortunately, these descriptive results cannot parse the potential mechanisms I introduced at the outset. However, highlighting some of the context around this period connects back to some of the theoretical possibilities.Footnote 7 The likelihood of a seat flip and the identify of the aggrieved side were very clear. Republicans controlled their own destiny so to speak. All that was required was for the Republican President to make a nomination and for Republican senators to accelerate a confirmation. This near certainty of the seat changing from Justice Ginsburg to a Trump nominee is consistent with the notion that legitimacy is connected to one’s perception of the Court’s ideology. By this interpretation, it was not the process or rhetoric that mattered. Rather, it was that Republicans would quickly fill the seat. On the other hand, Ginsburg’s death occurred in the midst of a polarizing Presidential election that would likely maximize the intensity and salience of political contestation and signals. It also occurred after the Garland blockade, the Gorsuch appointment, and the contested Kavanaugh confirmation. This context likely made it more difficult for Democrats to overlook partisan cues in favor of more abstract values about the Court as an institution. Notably, neither of these mechanisms is the same as the Court losing legitimacy through being seen as more political in its own behavior. This is not surprising since the Court itself did nothing. Of course, it is also possible that some other mechanism is driving the results. For example, there were certainly complaints about process and norm violations. Perhaps Democrats reacted to what they perceived as a basic procedural unfairness or hypocrisy, or perhaps they felt a sense of ownership over a seat that they saw as taken from their side.
Finally, as in many studies in this literature, my data cannot speak to the longevity of the observed changes or whether the respondents’ diffuse support rebounds with time. These durability questions may be especially interesting and fertile in the context of a contested appointment battle that led one party to call the process and the nominee illegitimate (e.g. Everett and Levine Reference Everett and Levine2021). It could be that the legitimacy changes arose and then faded in response to a specific set of salient events in which elites briefly argued about the Court in very partisan ways. Perhaps diffuse support is so robust that it rebounds even when the Court comprises members whose very presence was initially seen as illegitimate. However, the alternative is also possible and intriguing. Maybe these delegitimizing origin stories stick to new Justices or the Court such that threats to diffuse support that emerge from the process of seating Justices are unusually durable. Even if these views fade into the background, perhaps they lie dormant and ready to return when the ‘tainted’ Court makes controversial decisions, or when elites want to invoke them.
Acknowledgments
Thank you to Miles Armaly, Alex Badas, Michael Nelson, Jessica Blankshain, and three anonymous reviewers for very helpful comments.
Data Availability Statement
Replication materials available in the JLC’s Dataverse.
Supplementary Materials
To view supplementary material for this article, please visit http://doi.org/10.1017/jlc.2022.6.
Open access
