Is the U.S. Supreme Court a Reliable Backstop for an Overreaching U.S. President? Maybe, but is an Overreaching (Partisan) Court Worse?

The Roberts Court has been called the most “___” Court in history, with many different adjectives being offered. Surprisingly, our study of voting data from Supreme Court Terms 1937-2021 shows that the Roberts Court is the most “anti-President” Court in that period: it has ruled against the President at a greater rate than any other Court. Should we take this to mean that the Court will be there to protect democracy if an overreaching President tries to trample constitutional limits? Not necessarily. Additional analysis and a deep dive into the cases and reasoning reveal a more complicated picture, of a Court exhibiting historic levels of partisan and loyalty bias as well as a strong penchant for judicial supremacy.

steel mills (Youngstown 1952). But not always. Nixon's assertion of executive privilege came in a crisis of his own making (United States v. Nixon 1974), as did Trump's claim of immunity from state prosecution (Trump v. Vance 2020).
When presidents overreach, do the courts, especially the U.S. Supreme Court, go along or push back? Our systematic analysis of voting data and doctrine since the 1937 term suggests that the Court has often served as a reliable backstop. Although it tends to hold for the president across all cases (a 65 percent win rate), in highly important disputes implicating executive authority, the justices are far less deferential. In these weighty cases, the Roberts Court (2005-2021 terms) was especially tough on the president, ruling in his favor in only 35 percent of them.
These results might be understood to suggest that democracy is in safe hands. The Court under Chief Justice Roberts, in particular, seems more than willing to act to constrain power-grabbing presidents. But "seems" is the operative word. Two findings must temper any conclusion that the Roberts justices are likely to be a reliable check on an overreaching president. First, the data show that the Roberts justices are especially and uniquely willing to put the brakes on a president who does not share their partisan affiliation, while far less likely to check a president of the same party, especially if the president appointed the justice (the "loyalty bias"). These results, we suggest, sit comfortably with intensifying affective polarization-the tendency to dislike and distrust those from the opposing party" among Americans (Druckman, et al., 2021, 28)-the justices not excepted (Devins and Baum 2017;Das, Epstein, and Gulati 2023).
Second, looking more granularly at the way that courts have acted either to invalidate or to uphold presidential actions over time, we find that the Roberts Court has consistently done so in a way that injects itself more directly in matters of policy across the spectrum of legal issues than prior courts. It has provided indications that it views its role as imposing substantive agendas through its interpretations of the powers of the other branches. Again, this finding is compatible with the larger political environment in which the Court operates: What with increasing polarization and the resulting gridlock, the elected branches may lack the wherewithal to retaliate against even extreme judicial overreaching-likely enhancing the justices' selfconfidence (Hasen 2013;Epstein and Posner 2018). Seen in this way, characterizing the Court as "imperial," as one scholar has, seems apt (Lemley 2022; see also Liptak 2022).
Putting together the existing commentary and the results of our quantified data and doctrinal analyses leads us to ask whether the branch that was thought to have "no influence over either the sword or the purse" (Hamilton,Federalist No. 78) is, in fact, transforming itself into the branch with the greatest ultimate control over matters of national policy, all without benefit of democratic selection or accountability.
Whether this is too high a price to pay for checks on a power-grabbing president, we leave to others to debate. Our goal here is rather to provide groundwork for that debate by documenting trends in the Court's support for presidential power. To that end, we begin with a discussion of our data sources and then turn to several descriptive analyses focused on the president's win rate in the Court. These analyses show that the Roberts Court has been far less willing to hold for the president than its predecessors were. Next, using multivariate models, we explore explanations for the justices' votes in cases implicating president power. Yet again, the results point to significant differences between the Roberts justices and all others-especially the role of partisanship in their decisions. Finally, acknowledging that a focus on votes and outcomes cannot capture all the nuances in the Court's decisions, we report the results of our deep dive into the doctrine on presidential power. That exploration, too, reveals a Roberts Court that stands out as both aggressive and political.

The Data
To explore Court-president interactions, we developed a dataset of all orally argued U.S. Supreme Court decisions, issued between 1937-38 term (the start of the modern Court; Leuchtenburg 1995), and the 2021-22 term, that implicated presidential power. These are decisions in which the United States, an executive department (or department head), an independent agency, or the president himself was a party. 1 At first blush, including all these decisions may seem inapt in a study of presidential power. Take agency cases. Scholars have long debated the level of control 1 To develop the dataset, we used the same procedures as Epstein and Posner (2016). Specifically, we began with the U.S. Supreme Court Database (Spaeth, et al. 2022), organized by citation, with decisionType=1 or decisionType=7 (that is, we included only orally argued cases excluding per curiams). To initially identify the cases, we used the petitioner and respondent variables in the Database. But because the Database identifies only the lead parties to the case, cases otherwise meeting our criteria were improperly excluded. To address this omission, we combined Collins's (2008) and Epstein and Posner's (2018) data and hand-collected data.
the President has over the bureaucracy (e.g., Howell and Lewis, 2002;Wood and Waterman 1991), and so the question naturally arises: Is it sensible to include cases in which, say, the Federal Communications Commission (FCC) is a party if the President lacks control over rules issued by the Commission? To address this concern, it is worth pointing out that virtually all cases implicating the federal governmentincluding agency cases-are brought to the Supreme Court not by the agency but by the Office of the U.S. Solicitor General (OSG). For example, when the FCC wanted to appeal a lower court decision invalidating ownership rules (FCC 2021), it was OSG lawyers who asked the Court to hear the appeal, filed the brief, and argued the case; in short, the OSG controls the executive branch's "access to the Supreme Court, which gives [it] the power to coordinate US appellate strategy, centralize it, and make it speak with one voice" (Black and Owens 2012, 21). And, importantly, because the OSG is headed by the U.S. Solicitor General (SG), who works under the Attorney General, a presidential appointee, it is highly unlikely that the SG would litigate against the president's interests and policies. (For those readers who remain skeptical: We later examine subsets of case, including core constitutional decisions directly implicating presidential power.) With these notes, let's turn to the resulting "presidential power" dataset. It consists of 3,660 cases (31,401 votes cast by 48 justices)-or 42 percent of all orally argued cases decided between the 1937 and 2021 terms. That percentage was even higher for the earliest terms in our dataset, as Figure 1 shows, with the executive branch a party in a majority of the cases. During more contemporary eras, the percentage dropped but still lingers at about a third-meaning that disputes involving the president continue to occupy a substantial portion of the merits docket. Figure 1. Percentage of orally argued cases implicating presidential power, by Chief Justice era. Terms are in parentheses. The total number of all orally argued cases for each era is: Hughes= 580, Stone=707, Vinson= 728, Warren= 1,538, Burger= 2,235, Rehnquist= 1,802, Roberts= 1,108.
We use these data to conduct two related analyses. The first supplies a statistical portrait of voting patterns in cases connected to presidential power. We are interested in identifying the extent to which the justices uphold the president (presidential "win rates") and in exploring various explanations for why they rule in the president's favor or not. We conducted this analysis on all cases, as well as on various subsets of the data.  Stone (1941-45) Hughes  Term, implicating presidential authority that appear in constitutional law casebooks. 2 Because there are only 50 of these "core constitutional decisions" (all of which are in our dataset), this inquiry takes the form of a close reading of the Court's opinions to identify any patterns or trends.

The President's Win Rate
Our analysis begins simply enough, with the percentage of cases in which the president prevailed (the president's "win rate" or, if you prefer, the Court's "deference rate"). 3 Overall, across the 85 terms in our dataset, the rate is 65 percent. 4 But, as Figure 2 shows, that percentage masks some term-by-term variation-from a high of 84 percent during the 1981 term (Ronald Reagan's first year in office) to a low of 38 percent in 2017 (Donald Trump's first year 5 ).
2 We consulted 16 constitutional law casebooks: Barnett and Blackman (2021), Bowie (2022), Calabresi and Lawson (2020), Chemerinsky (2020), Choper et al. (2019), Epstein, McGuire, and Walker (2021), Farber, et al. (2019, Feldman and Sullivan (2019), Friedman and Mortenson (2021), Maggs and Smith (2020), Murphy, et al. (2019), Paulsen, et al. (2021, Shanor (2016), Stone, et al. (2017, Varat, Amar, andCaminker (2021), andYoung (2017). 3 Identifying whether the president won or lost is mostly a straightforward task. If the Office of the Solicitor General represented the United States, a federal agency, or various executive actors, we noted whether the president was the petitioner or the respondent. We then used the variable partyWinning in the Supreme Court Database to determine whether the president won or not. In those few cases in which the OSG did not represent the president (e.g., Clinton v. Jones), we followed the rules in Epstein and Posner (2016).
We should also note that Posner (2016, 2018) undertook a similar analysis of presidential win rates (through the Obama administration) but they aggregate by the president in office at the time of decision. Breaking the data down by term and Court era unearths interesting patterns undetected in analysis by president, as we note in the text. 4 Studies of the Office of the Solicitor General report a similar win rate (e.g., Black and Owens 2012). 5 Technically, Reagan became president during the Court's 1980 term (and Trump, the 2016 term) but 1981 was his first full term in the Supreme Court (as was 2017 for Trump). Even more relevant for our purposes is that the variation appears to be patterned-with deference to the president declining markedly in recent years.
During the 1937-1940 terms, under the leadership of Chief Justice Hughes, the Court ruled for the president about two-thirds of the time, as Figure 3 shows. Except for the highly deferential Burger Court, 6 that percentage remained fairly stable until the current era, when the percentage barely broke 50. In statistical terms, the Roberts Court's 52 percent support for the president is significantly lower than in all other periods shown in Figure 3. The suggestion here is that the contemporary Court, which 6 The Burger Court was significantly more deferential to the president than all other Court eras (except Hughes); and the Roberts Court was significantly less deferential than all other Court eras. Note: Here and throughout, we use the term "significant" or "significantly" only when p < 0.05.  1940 1950 1960 1970 1980 1990 2000 2010 2020 has spanned the terms of four different presidents (Bush, Obama, Trump, and Biden), is uniquely willing to check executive authority.

Administrative State Cases
The decline in support for the president has not gone unremarked (e.g., Liptak 2017). The commentary, though, tends to center on presidents rather than on Court eras-noting, for example, that since the Reagan administration, "each succeeding president did worse than the last" (Liptak 2017; see also Epstein and Posner 2018).
Depicting support by the Chief Justice, as we do in Figure 3, suggests another possibility: the decline in the win rate traces to the Roberts Court-not to any trend toward less deference.  Stone (1941-45) Hughes  Either way, some scholars attribute the decline in presidential win rates to the contemporary Court's disdain for the administrative state rather than to its mistrust of presidential power (Chemerinsky 2019, Herz 2015. Metzger (2020, 2) explains it this way: "[C]onservative groups have put sustained efforts into fostering academic attacks on core features of administrative government… And … the Trump administration [selected] judges who are receptive to these conservative attacks." Those judges include Trump appointees Gorsuch, Kavanaugh, and likely Barrett, who join Thomas and Alito as "strident administrative skeptics" (Metzger 2020, 3).
This skepticism has manifested itself in several important doctrinal shifts related to the administrative state. 7 Most relevant here is the widespread belief that the deference accorded to executive branch interpretations of statutes by Chevron v. NRDC (1984) has reached zombie status, dead but still not interred (Howe 2022, Pierce 2022, Herz 2015. In the 2021 term, for example, the Court decided two cases in which Chevron theoretically should have been applied to resolve the issue, but the Court ignored it (American Hospital, Becerra v. Empire Health). Despite Chevron's having established that courts must give deference to reasonable agency interpretations of an ambiguous statute, in both cases, the Court assessed the agency's regulation based on its own interpretation of the statute, not considering or citing to Chevron, either when upholding the agency rule or when overturning it.
Considering the Chevron retreat, it seems possible that the Roberts Court's less deferential posture reflects its "anti-administrativism" and not disdain for presidential power more generally (Metzger 2020). If so, this might temper any conclusions about the Court's willingness to check an overreaching president, as opposed to overreaching agencies.
To assess this possibility, we created three subsets of government cases: 8 (1) independent regulatory agency cases (e.g., the Federal Trade Commission, the Security and Exchange Commission); (2) executive department or agency cases, including those in which the president is a named party; 9 and (3) cases in which the United States was the named party, many of which are criminal cases. Table 1 reports the president's win rate in each, as well as the overall win rate in col. 4.

All Cases in the Dataset
(4) Hughes (1937)(1938)(1939)(1940) Table 1. Percentage of U.S. Supreme Court decisions favoring the president, by the type of government party. The numbers in parentheses are the total number of government cases in each category. The percentages in the "All Cases" column are also shown in Figure 3.
In line with existing commentary, the Roberts Court was far more hostile toward independent regulatory agencies (IRAs) than any other chief justice era: its 39 percent support rate is 30 percentage points below the average of 69 percent. Still, the current Court's less-deferential posture extends to all types of disputes, not just to those involving IRAs. Note that its support for the president hovers around 50 percent compared to averages closer to the two-thirds mark.

High-Stakes Disputes
However revealing the data in Table 1  To identify cases that may be especially weighty (at least in the eyes of journalists, the public and so likely the president too), we used a measure common in the political science literature: whether the decision was reported on the front page of the New York Times on the day after it was issued (Epstein and Segal 2000). 10 We supplemented this list with the 50 cases mentioned in constitutional law casebooks implicating executive power. All in all, 13 percent of the cases in our dataset fall into one or both categories (467/3,660).
A comparison of the president's lower win rates in these weightier cases to overall president win rates seems to reflect favorably on the Court as a check on a powergrabbing president. 11 Over the 85 terms in our dataset, the Court was significantly less deferential to the president in these high-stakes cases: 56 percent versus 67 10 This measure is hardly perfect. It comes after the Court's decision, meaning it is post-treatment in models of outcomes and votes and so cannot be used as the basis for causal claims (see Howell 2003). Then again, based on news coverage of decisions granting review, oral argument attendance, and predecision coverage, the justices can probably form a reasonable prediction about the likelihood of frontpage coverage.
Another potential problem is that Times sometimes features decisions that are splashy but not especially weighty. Worth noting, though, is that all 50 decisions appearing in the constitutional law books in our sample were covered on the front page of the Times on the day Court issued them. 11 Another possibility is that when the president loses, the Times is more likely to highlight the case. Even if true, the main focus of our analysis-the drop off from previous eras to the Roberts Court-is quite noticeable (see Figure 4). percent in the lower-stakes cases. 12 In fact, during the same terms, the president received less deference from the justices than unions (60 percent) or civil rights plaintiffs (56 percent).
Looking at the prior Court eras, as Figure 4 does, the trend is similar for all but one Court. Only the Hughes Court in the 1930's held for the president more often in the weightier cases than in other cases. But the Roberts Court had the lowest presidential win rate of all: In just 35 percent of the high-stakes cases did the president prevail. We are not the first to document the Court's tendency to rule against the president in salient or weighty cases (see note 12). The question we confront is why the Roberts Court is so much less deferential in these high-stakes disputes than its predecessors.
One possibility follows from Howell's (2003) analysis of presidential win rates. He demonstrates that the president is far more likely to prevail when he enjoys "broadbased support" among key constituencies. Our data lend some support to this explanation. Using presidential job approval as a proxy for support, 13 we find that the president had the lowest approval rate, on average, in high-stakes disputes during the Roberts years, as Figure 5 shows. Then again, approval during the Vinson Court, which more often than not held for the president in important cases (see Figure   4), was statistically the same as in the Roberts years. Moreover, in the various models to come of the justices' voting, public support for the president has only a modest effect on votes.

Figure 5.
Mean percentage public approval of the president in high-stakes decisions, by Chief Justice era. The dark vertical line is the overall mean (55%). "High-Stakes Decisions" are those that received front-page New York Times coverage on the day after they were issued or that appeared in constitutional law case books. The percentage is the percentage of respondents approving "of the way [President's name] is handling his job as President" (see note 13).
Another possibility may lie less with the president than with the larger political environment, specifically with political polarization. When Howell was writing in the early 2000s, partisan polarization was on the upswing but today it is at unprecedented levels (see, e.g., Lewis 2022  Stone (1941-45) Hughes  activism (measured by federal laws invalidated) ever since the Clinton administration; 14 and Narechania's (2023) text analysis provides evidence of "the Roberts Court's historically unparalleled interest in using its docket discretion to select cases for the purpose of revisiting, and perhaps overruling, precedent." Finally, the data we offer in the pages to come are also consistent with justices seemingly more concerned with their own political agenda than with reactions from the president (and Congress).

Explaining President Win Rates
So far the data depict a Supreme Court that has grown less deferential to presidential power especially in weighty disputes. Even if it is difficult to nail down why this trend emerges, it would seem to be welcome news for scholars and other commentators questioning whether the justices would put the brakes on a runaway president.
But not so fast. Identifying the individual justices' motivation for invalidating presidential action is critically important to our inquiry. Think about a scenario in which a Court filled with appointees of one party is eager to rein in a power-grabbing president of the opposite party, but supportive of an equally encroaching president of their own party. This would amount to neither a deferential nor a backstop Court, but rather an aggressively political group of justices acting in service of its own partisan agenda with little concern for preserving democracy.
This scenario cannot be ruled out in light of the time-honored literature on the effect of the judges' partisanship on their decisions (for a review, see Peretti 2020).
Within this literature virtually all studies find that Democratic appointees, relative to Republican appointees, are significantly more likely to favor criminal defendants, plaintiffs in discrimination cases, and labor unions; and Republican appointees similarly are more likely to rule for their favored interests, such as business (Epstein, Landes, and Posner 2013). But the studies are more mixed when it comes to executive power cases. Although some research finds that political compatibility between the president and the justice "matters a great deal" to how the justice will vote in these cases (Robinson 2012, 915), 15 the more conventional view is that political preferences play less of a role in executive power disputes than in culturally controversial issues such as, say, abortion or religion cases (Black and Owens 2012;Segal and Spaeth 2002). This view may rest on a sense that, while the political parties differ as to certain preferred rights, they coalesce around the importance of structural protections in the Constitution. After all, the trope reminds us, six Democratic appointees ruled against Democratic President Truman in the 1952 Youngstown case and a unanimous Court (which included three Democratic and five Republican appointees) ruled against the Republican President Nixon in 1974. This is a 15 Many studies (mostly of circuit court panels) have considered whether partisanship comes into play in Chevron deference cases (e.g., Cross and Tiller 1998;Miles and Sunstein 2006;Czarnezki 2008;Barnett, Boyd, and Walker 2018). On balance, the studies show that judges are more likely to affirm an agency's interpretation when the judges' partisanship and the agency rule are aligned.
comforting story of justices putting country over party, but perhaps of questionable reliability in current times when bipartisan commitments to democratic norms have been called into serious question across the government landscape.
To get to the bottom of the matter, we consider whether the justices' partisanship influences their votes for or against the president. For the analysis, a justice and president are partisan-compatible 16 if, at the time of oral argument and decision, 17 they share the same party affiliation. 18 Because partisanship is not the only factor likely to influence the justices' votes, the statistical models also account for several other well-known determinants. One is ideology, such that a liberal (conservative) justice is more likely to decide in favor of the liberal (conservative) side in any given case (Segal and Spaeth 2002). To assess Ideological Compatibility between the president and the justice, we began with the U.S. Supreme Court Database, which enables identification of whether the government took a liberal position (e.g., defending government action protecting civil rights) or a conservative position (e.g., defending a search and seizure) in each case. 19 Then we determined whether the justices in our dataset were ideologically predisposed to vote in the government's favor. 20 In addition to Ideological Compatibility, the analysis considers whether the president was the petitioner (appellant) or respondent (appellee). The government's chances of winning should be greater when it is the petitioner (appellant) because the Court more often than not reverses the lower court. We also incorporate the president's job approval rating, 21 reflecting research we have already mentioned, which suggests the justices are more willing to rule against an unpopular president (Howell 2003;Yates 2002)-with Youngstown an often-cited example. Finally, the analysis considers whether the case was weighty or not (using the same measures as in Figure 4). (Appendix 1 provides summary statistics for all variables in the analysis.)

Basic Results
Table 2 provides the results of the analysis across the 85 terms in our dataset.
Looking first at the "All Votes" model (1), Partisan Compatibility does not seem to play much of a role (the coefficient is positive but not significant), whereas ideology does. When the justice and the president's position are ideologically compatible (i.e., 20 Ideology is the justices' term-by-term Martin-Quinn score (Martin and Quinn 2002). To determine ideological compatibility in each case, we rely on a measure that takes on higher values as the justice is more ideologically inclined to vote for the president, constructed using the procedures detailed in Johnson, Wahlbeck and Spriggs (2006) and Nelson and Epstein (2022), among others. Specifically, if government's position was liberal, the variable is the negative value of the justice's term-by-term Martin-Quinn (ideology) score; if the government was on the conservative side, this variable takes on the value of the Martin-Quinn score (Martin and Quinn, 2002).
Because this variable, Ideological Compatibility, is akin to an interaction between the side represented by the government and the justice's Martin-Quinn score, we also include each as a separate covariate in the analysis. 21 For the analysis, we aggregated the data to the term. the justice is ideologically predisposed to side with the president's action being challenged), the predicted probability of a vote in support of the president increases by nearly 0.15, 22 relative to a justice less ideologically inclined to vote the president's way (all else equal). 23 These basic results hold even if we drop Partisan Compatibility from the model; and likewise, Partisan Compatibility remains positive but not significant if we exclude Ideological Compatibility (see Appendix 2).  Table 2. Logistic regression of votes in favor of the president, 1937-2021 terms. Robust standard errors (clustered on justice) are in parentheses. *p < 0.05. All models include four covariates representing the components of two variables (Ideological Compatibility and Partisan Compatibility). See notes 18 and 20. The models exclude cases in which the party of the president changed between oral arguments and decision.
Other variables are related to the vote, notably presidential job approval and case importance. As the literature anticipates (e.g., Howell 2003), the higher the approval rating, the higher the likelihood of a justice voting for the president. But the effect, though statistically significant, is substantively small. Moving from a president with 45 percent approval rating (25 th percentile) to a 64 percent rating (75 th percentile) raises the predicted probability of vote in the president's favor by only 2 percentage points. 24 High-stakes cases, in contrast, exert a significant and larger effect on the justices' votes. All else equal the justices are nearly 7 percentage points less likely to rule for the president in weighty disputes. 25

Partisan Compatibility
Because this result, yet again, suggests that the justices will act as a backstop in the cases that are most important to the president, models (2) (lower-stakes cases) and (3) (higher-stakes cases) in Table 2 above explore it in more detail. Note that model (1) and model (2) produce analogous results-no surprise as the vast majority of cases in the dataset do not register as weighty on our measure. Not so of votes in the higher-stakes cases (model 3). While Ideological Compatibility remains significant, Partisan Compatibility now also plays a non-trivial role in these important disputes. All else equal, a justice is about 8 percentage points more likely to vote for the president when the justice and the president are of the same political party. 26 (Once again, these results remain if we exclude Ideological Compatibility from model 3. See Appendix 2.) The suggestion here is that in high-stakes disputes-disputes that may truly matter for democratic backsliding-the Court is not especially deferential to the president; but neither is it above partisan politics. A boundary-pushing president may well succeed if for no other reason than his party controls the Court.
It turns out, though, that partisan compatibility has not always figured into the justices' votes. Actually, looking at the four most recent Chief Justice eras, 27 as Table   3

does, for only the Roberts justices have partisan considerations significantly infiltrated their decisions. This result, it is worth noting, holds even if we exclude
Ideological Compatibility from the model (see Appendix 3); and holds in all cases, whether weighty or not. The Roberts justices, it seems, are uniquely partisan-and not just ideological-in their voting.  Table 3. Logistic regression of votes in favor of the president in high-stakes cases. Robust standard errors (clustered on justice) are in parentheses. *p < 0.05. All models include four covariates representing the components of two variables (Ideological Compatibility and Partisan Compatibility). See notes 18 and 20. The models exclude cases in which the party of the president changed between oral arguments and decision. This is an important distinction between the Roberts justices and all others. For one thing, the effect size is non-trivial. The predicted odds of a Roberts justice voting in favor of the president in these highly salient disputes decline by over 25 percentage points when the president is of a different political party (all else equal). Second and even more relevant to our concerns is this: Because a Democratic president was in office for over 50% of the (Republican-dominated) Roberts Court's high-stakes cases, it seems entirely possible that an aggressively political group of justices is acting in service of its own partisan agenda-and not necessarily protecting democracy from a reckless president.

Votes for the President in High-Stakes Cases
But why? Why does partisanship play such an outsize role on the Roberts Court relative to its predecessors? A likely explanation again traces back to the partisan polarization of our times. As partisan sorting-a growing alignment between partisanship and ideological leanings-took hold in the U.S. Senate, it trickled down to the Supreme Court, such that by 2010 the Court was perfectly sorted: all the Democrats were on the Court's left side and all the Republicans on the right. Never before had such "clear ideological blocs…coincided with party lines" (Devins and Baum 2017, 301). But there's more to the story, for partisan sorting is not the only manifestation of societal divisions; affective polarization has also intensified. Indeed, "fear and loathing across party lines" (Iyengar and Westood 2015) is so extreme that when confronted with two policies-say, on healthcare-that are otherwise identical except for the party endorsing them, Americans rate their own party's policy far more favorably (e.g., Munro, et al. 2013). Partisan loyalty, in other words, beats out policy considerations.
Of course, under idealized versions of apolitical and neutral judicial decision making, the justices would be above the "us-against-them" mentality associated with affective polarization. But theoretical accounts and empirical analyses-including our own-suggest they are not. 28 As Justice Benjamin Cardozo put it (1921, 168), "We [judges] like to figure to ourselves the processes of justice as coldly objective and impersonal." But "[t]he great tides and currents which engulf the rest of men, do not turn aside in their course, and pass the judges by."

The Loyalty Bias
The partisanship of the Roberts Court is not the only point of differentiation between it and its predecessors. More than any other modern Court era, the Roberts justices exhibit unusual loyalty to their appointing president in high-stakes cases (see Table 4). Although three of the four eras show a significant difference in voting between justices who were appointed by the president litigating the case and those who were not, 29 the gap is widest for the Roberts justices. Even more striking, Burger and Rehnquist Court justices not appointed by the sitting president still voted in the president's favor in the majority of high-stakes cases. Not so of the Roberts justices: they cast only 36 percent of their votes cast in favor of the non-appointing president. Table 5 probes these findings in more detail by accounting for the other factors related to voting. The loyalty effect remains for the Burger, Rehnquist, and Roberts 29 Although our study may be the first to show increased loyalty during the Robert Court years, we are not the first to uncover a loyalty effect among U.S. Supreme Court justices (see, e.g., Epstein and Posner 2016); and other studies have shown that the effect is not limited to the United States (see, e.g., Dressel, Inoue, Bonoan 2023). justices: In the high-stakes cases, they are significantly more likely to favor their appointing president than all others.  The effect size, though, is far greater for the Roberts justices, as Figure 6 shows.

Vote for the President in High-Stakes Cases
All else equal, the predicted probability of a justice voting for the president increases by over 20 percentage points when the president is the appointing president (versus 10 for the Burger justices and 10 for the Rehnquist justices). Figure 6 also shows that predicted probability of voting for the president-regardless of whether he was the appointing president-never falls below 0.50 for the Burger and Rehnquist justices (with all other variables at their mean). Not so for the Roberts justices, with a 0.33 predicted probability of voting for the non-appointing president.

Beyond Win Rates
The data on outcomes and votes paint a picture of a Roberts Court that is quite distinct from its predecessors. In cases that may matter most to the president, the current justices seem uniquely willing to assert judicial authority over the executive branch, especially when the president is an opposing partisan.
But, of course, a focus on outcomes and votes cannot convey all the important variables in play in the cases that may lead the justices to guard (or not) against presidential overreach. For this reason, we carefully reviewed "core constitutional cases"-those 50 making appearances in constitutional law books. That review only reinforced the data findings: the Roberts Court stands out as both political and aggressive.
One indicator is that in a comparison of the Roberts Court to all other eras, the tenor of the justices' opinions has changed radically, even when outcomes may not Notice that in both Pink and Zivotofsky the Court ultimately affirmed the presidential power at issue, and both would be reflected as wins in our dataset.
However, in the earlier case, the Court nodded to the primacy of the decision-making power of the political branch, while in the later case, the Court claimed for itself the right to determine the contours of the powers of the branches and exactly how they are permitted to interact. Doctrine did not change much on the surface, but the Court's sense of self had notably changed.
A second indicator of the activist posture of the Roberts Court is its tendency to regard issues before it as opportunities for aggressive judicial determination, with less and less regard being shown for the principle of deference, whether to other branches or to agencies. We have already mentioned the retreat from Chevron. But two other developments are equally notable for their aggressive posture toward the president (and Congress).
One is a move toward a stringent non-delegation doctrine. In Gundy v. United States (2019), four justices-with Gorsuch leading the charge-expressed the view that Congress's decisions to delegate legislative authority to the executive branch with the kind of broad "intelligible principle" that it has been using for the past 90 years should no longer be respected. 31 Rather, it is the Court that should decide whether Congress has limited agency discretion sufficiently to satisfy the Court. If that view gains one more vote, which seems likely with the additions of Kavanaugh (who did not participate in Gundy) and Barrett (appointed since Gundy), it will represent a major shift of power from the elected branches, on the one hand, to the Court on the other.
Gorsuch's plea for a new vigor in judicial enforcement of delegation insists that "[judicial] abdication is not part of the constitutional design" (Gundy 2019). To avoid "abdication," he would require a judicial inquiry into the legislative choice to delegate power, and an evaluation by the Court of whether policy choices have remained sufficiently anchored in the Congress, with the power given to the agency limited sharply to the ability to "fill in the details." This is in contrast to the more traditional view, described by Justice Scalia, that "we have almost never felt qualified to secondguess Congress regarding the permissible degree of policy judgment that can be left to those executing or applying the law" (Mistretta 1989, 416). Thus the Court appears to be reimagining its role as evolving from Congress's servant to its master.
31 Alito, Gorsuch, Roberts, and Thomas. The intelligible principle comes from J. W. Hampton, Jr. & Co. v. United States (1928), holding that "If Congress shall lay down by legislative act an intelligible principle to which the person or body … is directed to conform, such legislative action is not a forbidden delegation of legislative power." A related move by the Roberts Court, with perhaps even more implications for executive power, has been the articulation of a new principle of statutory interpretation declaring Congress will not be understood to have delegated power to decide so-called "major questions" unless the language used in the statute clearly specifies such power. This is related to the non-delegation principle because it appears that at least Gorsuch, if not a majority of the Court, believes that delegating such major-questions authority would likely be an unconstitutional delegation in any event. 32 The major questions doctrine, for that camp, enforces the underlying separation-of-powers rule. But the "major questions doctrine" itself presents as a statutory rule, not a constitutional one, leading to judicial invalidation of executive agency action on the ground that Congress must speak clearly if it intends to allow an agency to "exercise powers of vast economic and political significance" (National The Court acknowledged that demands for presidential documents have not historically ended up in court; rather, they tend to be worked out in the "hurly-burly, the give-and-take of the political process between the legislative and the executive." But rather than affirm the lower courts' deferential finding that the requisite legislative purpose was present and the subpoenas therefore satisfied the minimum constitutional standard for legitimacy, the Court adopted a more intrusive analysis giving judges the ultimate authority to decide the reach and bona fides of Congress's subpoena power in any given case. Under this approach, "courts would carefully assess whether the asserted legislative purpose warrants the significant step of involving the President and his papers." In addition, courts would scrutinize the nature of the evidence offered by Congress in support of its subpoena and assess for itself the degree of burdens imposed on the President by the subpoena. In so writing, the Court rejected both the President's plea for a formally higher legal standard of justification for a subpoena and the House's claim that it should receive deference in determining what is a valid legislative purpose. Instead, the Court appropriated to the judiciary the formerly political role of balancing the needs of the political branches and choosing a winner and a loser in a given dispute about access to documents. 33 These two examples, not to mention trends in other legal areas, suggest that even data showing a lower rate of deference to presidents from the Roberts Court can mask what appears to be a crescendo of Court involvement in matters involving the president that will allow the Court to make final rulings in separation-of-powers 33 A related, more comprehensive analysis suggests that the judicial supremacy exhibited by the Court in Vance and Mazars was for an even more nefarious, partisan, purpose of protecting the president from electoral consequences of the material sought by these subpoenas (Chafetz 2021,139-144).
disputes the justices may choose to resolve. The justices' preference for allowing such issues to be resolved on a retail basis rather than prescribing constitutional rules going forward weakens any inference they are motivated by a desire to provide a structural backstop against presidential overreach.

The Takeaways
The title of our paper poses two questions. The answer to the first-Is the U.S. Which takes us to the second question posed in our title: Is an overreaching (partisan) Court worse than an overreaching president? Although we are not sure of the answer, we think asking the question is important if only because the Court is so often thought to be the hero in narratives of overreaching presidents: think United States v. Nixon. This decision is widely viewed as an act of judicial supremacy (and, fittingly, includes the famous "emphatically" language from Marbury), but in that situation widespread, bipartisan agreement existed that a national crisis needed to be averted, that the rule of law was in jeopardy and needed to be saved. 35 There was no sense that what the Court did in Nixon was business as usual. But the Roberts Court provides cause for caution because its willingness to check political actors across the board has become intricately intertwined with a tendency to aggrandize the judicial role in all areas of political action. While the presence of an active judiciary can be a salutary feature in a vigorous three-branch structure, 36 that is less 34 Based on a LexisNexis search conducted in September 2022. 35 For a discussion of the efforts of the Nixon Court to make itself the "hero of the story," see Chafetz (2021). 36 There is an important and long-standing literature debating the value of a robust Supreme Court committed to pursuing the Constitution's promise of justice by ensuring the conditions of democracy (Brown 1993, Brown 1998, Dworkin 1996 versus a minimalist judicial approach that largely leaves true if the Court is routinely willing to disregard traditional limits on judicial authority, such as appropriate deference to political decision-making and respect for precedent. Indeed, the question hovers: whether the Court may be eyeing Lady Justice's scale less as a humbling aspiration for its work than as an alluring receptacle for its confident thumb. policy choices to the elected branches (Bickel 1962, Sunstein 2001. That debate is not implicated by our analysis here of a Court that takes neither of those approaches. Note: Robust standard errors (clustered on justice) are in parentheses. *p < 0.05. All models include four covariates representing the components of two variables (Ideological Compatibility and Partisan Compatibility). See notes 18 and 20. The models exclude cases in which the party of the president changed between oral arguments and decision. Table 3 with Ideological Compatibility Excluded