Tracking Distinct Coalitions Across the Roberts Court

Shedding empirical light on the question of when and whether the Supreme Court has broken down into blocs of 5-4, 5-3-1, 6-3, 3-3-3, and more

Jul 08, 2025

For two decades, Supreme Court watchers have spoken of “blocs”—liberal, conservative, and sometimes the elusive “swing” justice who could tip the law’s balance. But behind the familiar labels, the Court’s coalitions have ebbed and flowed in ways that confound the soundbites. Since Chief Justice John Roberts took the center seat in 2005, the justices have navigated culture wars, national crises, and historic changes in the Court’s own membership. Yet the real story of who stands together—and who splinters off—on the nation’s highest bench is far more complicated than a simple 5–4 or 6–3 headline. The question of proper coalition has been augmented in recent years by authors and scholars including Sarah Isgur and Dean Jens hypothesizing a 3-3-3 Court where Roberts, Kavanaugh, and Barrett are a separate, institutionalist bloc, and Alito, Thomas, and Gorsuch are the coherent conservative bloc.

This article peels back the conventional wisdom, tracking how Supreme Court coalitions have shifted, fractured, and sometimes unexpectedly held together across nearly twenty terms. Using new data and state-of-the-art empirical tools, I map the hidden alliances, anchor justices, and the rare but revealing moments when ideological lines give way to new, multidimensional divides (for more on the methodology see below1). The result is a fresh portrait of the modern Court: one marked by resilience, subtle evolution, and a surprising amount of doctrinal independence—sometimes in the places you’d least expect.

Roberts Court Early Years: 2005–2009—Coalitions, Anchors, and Multidimensionality

The first five years of the Roberts Court, spanning the 2005 through 2009 Terms, were marked by both striking continuity and subtle evolution in Supreme Court coalition structure. During this era, the data reveal a classic ideological alignment that is at once familiar and more nuanced than the standard 5–4 shorthand often suggests.

I code two types of agreement. “Soft agreement” includes differing views on legal reasoning through concurrences but the same votes on the outcomes while “hard agreement” only includes the same reasoning and vote and so excludes concurrences. Hierarchical clustering of non-unanimous cases each term consistently produces two robust coalitions. The conservative bloc comprises Chief Justice Roberts and Justices Scalia, Kennedy, Thomas, and Alito (the latter joining in 2006), while the liberal bloc consists of Justices Stevens, Souter, Ginsburg, and Breyer. This empirical 5–4 split is confirmed by silhouette analysis, which yields the highest scores (ranging from 0.60 to 0.70) for two clusters, with within-bloc agreement typically between 0.77 (conservatives, “hard” coding) and 0.88 (liberals, “soft” coding including concurrences).

Within these coalitions, anchor justices play a stabilizing role. On the conservative side, Justice Kennedy emerges as the anchor for much of this period: he maintains the highest average agreement with his bloc, almost never dissents from conservative outcomes, and is frequently in the majority without concurrences. For the liberal bloc, Justice Stevens serves as the anchor, with the highest rate of within-bloc agreement and the lowest number of outcome divergences. These anchors not only stabilize their coalitions but also frequently author majority opinions, cementing their leadership roles.

Even in this era of stable blocs, the analysis identifies justices who serve as the coalitions’ weakest links. For the conservatives, Justice Thomas frequently stands out, with a lower “hard coalition” percentage and a disproportionate share of separate concurrences and dissenting votes. On the liberal side, Justice Breyer occasionally diverges from bloc outcomes, most often in cases that raise complex statutory or federalism questions. It is important to note, however, that these divergences are exceptions, not the rule: the underlying coalition boundaries remain largely intact from term to term. It is equally important to note that by weakest link I do not mean least liberal or conservative. To the contrary, it could equally be the most liberal or conservative justices if they disagree most frequently with the coalition’s decisions.

A central question for the early Roberts Court is whether this period of apparent ideological clarity masked deeper, issue-specific realignments. The analysis shows that in most major areas—such as criminal procedure and economic activity—the term-wide 5–4 coalition structure persists, with only minor softening from concurrences or occasional cross-bloc votes. For instance, in Kansas v. Marsh (2006), the coalitions align with the term-wide split, and anchor justices are solidly with their respective blocs.

The exception, as in later years, is civil rights. In several terms, this area produces modest multidimensionality: Thomas, for example, is more likely to author or join separate opinions, and Breyer or O’Connor (in her final term) may align with the majority in outcome but write separately to explain nuanced reasoning. Even so, these divergences rarely persist across terms or cases; they do not coalesce into stable, third coalitions or “swing” blocs, as confirmed by both clustering and dimensionality reduction (PCA/SVD). The first principal component consistently explains over 80% of the variance in voting, and no secondary axis produces a spatially distinct third group.

The early Roberts Court stands as an era of solid, empirically robust ideological blocs, anchored by Kennedy and Stevens and punctuated by only occasional, issue-specific fractures. The much-cited 5–4 division is not a myth, but it is softened by individual doctrinal independence and a handful of outcome divergences. Only in rare instances—most notably in civil rights cases—does the Court’s coalition structure show meaningful multidimensionality, and even there, the deviations rarely threaten the stability of the core alignments. In both substance and statistical structure, the Roberts Court’s early years are best understood as an institution of two strong, cohesive blocs—remarkably resilient, yet never entirely without internal dissent.

Sotomayor and Kagan Join the Court: 2009–2011 Terms

The arrival of Justice Sotomayor in 2009 and Justice Kagan in 2010 ushered in a period of continuity and subtle realignment on the Supreme Court. Coalition structure remained sharply defined, but the composition and cohesion of the liberal bloc evolved as the new justices established their voting identities.

Term-wide, the classic 5–4 split persisted. Hierarchical clustering of non-unanimous cases (with soft agreement counting concurrences as partial alignment) consistently divided the Court into a five-member conservative bloc—Chief Justice Roberts and Justices Scalia, Kennedy, Thomas, and Alito—and a four-member liberal bloc—Justices Stevens (retiring in 2010), Ginsburg, Breyer, and from 2009, Sotomayor, with Kagan replacing Stevens as a fourth liberal in 2010. The within-bloc agreement remained high: conservatives averaged 0.78 (soft), while liberals were slightly higher at 0.83–0.88, with silhouette analysis confirming two well-separated clusters for all three terms.

Anchor justices continued to define each bloc’s identity. Kennedy anchored the conservatives, rarely breaking with his bloc and authoring the most majorities among the group. For the liberals, Breyer and (after 2010) Kagan shared anchor status: both maintained the highest average agreement with their coalition peers, and their dissents or concurrences were rare.

Weakest links emerged on both sides, but did not disrupt the overall structure. Thomas was again the least cohesive conservative, casting the highest number of dissents and separate concurrences—even as he remained, in outcome, aligned with his bloc more than 75% of the time. Among the liberals, Ginsburg showed the most outcome divergence, sometimes joining the majority with the conservatives in economic activity or federalism cases. Still, these differences were minor, and such departures did not challenge the fundamental coalition boundaries.

Issue-area analysis showed only limited multidimensionality. The Court’s dominant 5–4 split held across most topics, including criminal procedure and economic activity. For example, in McDonald v. City of Chicago (2010), the conservative bloc held together in a high-profile Second Amendment case, with the liberals dissenting as a group. In National Federation of Independent Business v. Sebelius (decided the following year), the pattern of bloc voting persisted, with only Chief Justice Roberts’ unusual cross-bloc vote foreshadowing later developments.

Civil rights cases provided the only context for modest divergence. Here, concurrences became more common, particularly from Sotomayor and Kagan as they established their jurisprudential voices. Nevertheless, the overall coalition structure endured: liberals remained united on outcomes, and conservatives maintained majority control.

Robustness checks with principal component analysis (PCA) and multidimensional scaling (MDS) reinforced the two-bloc story. The first principal component always explained over 80% of voting variance, with the liberal and conservative clusters clearly separated. Neither method revealed a persistent middle bloc, and the silhouette scores for three-cluster or “swing” models were always inferior to the basic 5–4 or 5–3–1 split. Even as Kagan and Sotomayor occasionally authored separate opinions, their positions in the PCA/MDS plots remained closest to the liberal bloc center.

In summary, the “Sotomayor and Kagan Join the Court” years (2009–2011) saw an evolution in voice and leadership on the left, but little change in the Court’s dominant coalition architecture. The conservative and liberal blocs were cohesive, anchored by Kennedy and Breyer/Kagan, and only rarely challenged by doctrinal independence from Thomas or Ginsburg. Multidimensionality was limited to issue-specific moments, never rising to the level of a third, stable coalition. The Roberts Court of this period, like the era before, is best understood as a two-bloc institution—adaptable in the details, but remarkably robust in its structure.

The Middle Roberts Court Years: 2012–2016 Terms

The middle years of the Roberts Court (2012–2016) represent a period of both deepened polarization and increased complexity. As the Court weathered the full force of post-recession political battles, the empirical coalition structure both confirmed the familiar left-right split and, in key areas, hinted at subtle but consequential multidimensionality.

Term-wide, the Supreme Court remained dominated by a two-bloc alignment. Hierarchical clustering consistently produced a five-justice conservative coalition—Chief Justice Roberts, Justices Scalia, Kennedy, Thomas, and Alito—and a four-justice liberal coalition—Justices Ginsburg, Breyer, Sotomayor, and Kagan. Average within-bloc agreement rates hovered at 0.80–0.86 for conservatives (soft) and 0.88–0.93 for liberals (soft), with silhouette scores confirming that no alternative clustering (such as a 3–3–3 or 2–4–3 split) outperformed the basic two-coalition model.

Anchor justices solidified each bloc’s cohesion. For the conservatives, Roberts and Kennedy alternated as the anchor, depending on the issue: Kennedy was consistently the highest-agreement member in civil rights and criminal cases, while Roberts anchored the majority in economic activity and federalism. On the liberal side, Ginsburg was the clear anchor, especially after 2014, consistently aligning with her bloc and authoring numerous majorities.

The weakest links became more visible, but did not fracture the coalitions. Thomas, as in prior years, was the least reliable conservative on hard outcome votes—his concurrences and occasional dissents marked a distinctive doctrinal independence. For the liberals, Breyer and sometimes Kagan played the role of weakest link, especially in business or regulatory cases where they occasionally sided with the conservatives or authored separate concurrences. Even so, these divergences were rare.

Issue-area analysis revealed meaningful, but limited, multidimensionality. The civil rights docket, especially in voting rights and affirmative action, saw the sharpest internal division among the conservatives. In Fisher v. University of Texas at Austin (2013, 2016) saw Kennedy—typically the conservative anchor—join the liberals to uphold affirmative action, illustrating both his swing role and the coalition’s flexibility on closely contested doctrine. Economic activity and regulatory cases, such as King v. Burwell (2015), sometimes produced minor coalition softening: Roberts, generally the conservative anchor in this area cases, crossed bloc lines to preserve the Affordable Care Act, though the basic two-bloc structure persisted.

Robustness checks using PCA, SVD, and MDS confirm the two-bloc model’s dominance. In each year, the first principal component explains over 80% of the variance, separating liberals from conservatives. Kennedy’s score moves closest to the ideological midpoint, but even at his most “swing” moments, he remains on the conservative side of PC1 for most terms. Attempts to fit a three-bloc structure always result in lower silhouette scores and reveal no persistent middle group in the dimensionality plots.

In sum, the Middle Roberts Court years are best understood as a time of stable but nuanced ideological blocs. The anchors—Roberts, Kennedy, and Ginsburg—kept their coalitions cohesive, while the weakest links provided doctrinal nuance but not persistent realignment. Civil rights cases remained the likeliest site of multidimensionality and surprise, but even there, departures from the term-wide split were episodic and issue-specific. For all the ferment in doctrine and the drama of divided cases, the empirical map of the Court remained a two-coalition system, robust to both statistical challenge and real-world change.

Trump’s Appointments: 2017–2020 Terms

The arrival of Justices Gorsuch (2017), Kavanaugh (2018), and Barrett (2020) transformed the Supreme Court’s composition but not its fundamental coalition structure. Empirical analysis of the voting data across these terms confirms that, far from splintering or generating new persistent blocs, the Court remained a two-coalition institution—if anything, the lines grew even sharper.

Hierarchical clustering and silhouette analysis reveal dominant 5-4 and (later) 6–3 splits in all non-unanimous cases. For example, in 2020, the conservative bloc—Roberts, Thomas, Alito, Gorsuch, Kavanaugh, Barrett—faced a unified liberal bloc of Sotomayor, Kagan, and (until her death) Ginsburg. The within-bloc agreement rates in this period are strikingly high: for conservatives, 0.79 (hard) and 0.85 (soft, with concurrences); for liberals, 0.93 (hard) and 0.94 (soft). The average silhouette score for two clusters (“k=2”) is consistently in the 0.60–0.68 range across all Trump-era terms, indicating clear, robust separation between the blocs. Forcing the data into three clusters (3–3–3 or 2–4–3) always lowers the silhouette score (dropping below 0.5) and generates at least one group with poor internal cohesion and no real anchor justice.

Anchor justices and weakest links are sharply defined. Kavanaugh and Barrett emerge as anchor justices for the conservatives, achieving agreement rates of 0.90 or higher with the rest of their coalition and rarely dissenting. On the liberal side, Sotomayor maintains nearly perfect agreement with Kagan and Ginsburg and is the least likely to author separate opinions. The weakest links—Thomas for conservatives and Kagan for liberals—are clear in the data: Thomas’s hard coalition rate, for instance, is consistently 6–10 points lower than his peers.

Issue-area analysis reveals multidimensionality primarily in civil rights. In 2020, civil rights cases showed a significant divergence: the conservative bloc splintered into subgroups, with Thomas, Barrett, and Gorsuch forming a tight “mini-bloc” (within-bloc agreement 0.83), while Kavanaugh, Roberts, and Alito sometimes aligned with liberals or authored separate concurrences. The liberal bloc, by contrast, remained cohesive (agreement 0.87–0.94). Thomas authored several forceful dissents during this period while Roberts and Kavanaugh remained in the majority and occasionally issued concurrences (mainly Kavanaugh), underscoring the internal doctrinal diversity. In economic activity cases, mini-blocs sometimes emerged (Thomas/Gorsuch/Alito agreement: 0.83; larger group: 0.66), but these were softer divergences—coalition boundaries were porous, not broken. Criminal procedure and most other issue areas stuck to the classic 6–3 alignment, with no meaningful split or multidimensionality.

PCA, SVD, and MDS analyses powerfully confirm these coalition patterns. Across all four terms, the first principal component (PC1) explains over 80% of variance, cleanly separating conservatives and liberals. Roberts and Kavanaugh—sometimes characterized as “moderates” or “swings”—cluster tightly with the core conservatives on both PC1 and in the MDS spatial map. There is no justice or set of justices who occupies a persistent “middle” space: all attempts to identify a stable third cluster yield low silhouette scores and weak explanatory power in both SVD and PCA. The few justices who do show spatial outlier status do so only in select issue areas, never as a stable, third coalition.

In summary, the Trump era saw the Supreme Court’s coalitions grow sharper and more internally cohesive. Empirical clustering and robustness checks all support a dominant 6–3 split, anchored by Kavanaugh and Sotomayor, with Thomas and Kagan providing most of the doctrinal independence and outcome divergence. Multidimensionality, when it appears, is confined to select civil rights and economic cases and never rises to the level of a persistent “middle” or third bloc. The Court, even with new faces and contentious cases, remained a fundamentally two-coalition institution.

Six-Three? The Supreme Court in the 2021–2023 Terms

The 2021 through 2023 Terms represent the high-water mark of the so-called “supermajority” Supreme Court. With the bench set by three Trump appointments, expectations ran high that a consistent conservative juggernaut would dominate every divided case. The data, however, tell a more nuanced story—one of remarkable bloc stability, but also of occasional doctrinal fluidity and unexpected alignments.

Term-wide, the empirical coalition structure is unmistakably 6–3. Hierarchical clustering on a weighted agreement matrix consistently produces a six-member conservative bloc (Roberts, Thomas, Alito, Gorsuch, Kavanaugh, Barrett) and a three-member liberal bloc (Sotomayor, Kagan, Jackson). Within-bloc agreement rates remain high: conservatives average 0.80 (hard) and 0.86 (soft, including concurrences); liberals, 0.94–0.96. The silhouette score for the k=2 clustering remains above 0.60 for all three years, while any three-cluster or hybrid model drops to 0.4 or below—demonstrating that the 6–3 split is not just intuitive but empirically superior.

Anchor justices and weakest links are clearly visible. For conservatives, Kavanaugh is the empirical anchor: he has the highest average within-bloc agreement, almost never dissents from the majority, and authors few separate opinions. For liberals, Sotomayor fills the anchor role—her agreement rate with Kagan and Jackson is above 0.95, and she is seldom the lone dissenter. The weakest link among conservatives is Thomas, who combines the lowest hard coalition percentage (often 8–12 points below Kavanaugh or Barrett) with the most dissents and separate opinions. For liberals, Jackson is the most frequent outlier, both in outcome and in opinion authorship, sometimes dissenting solo or concurring separately.

Civil rights cases remain the leading context for multidimensionality. In the 2022 term, for example, 303 Creative LLC v. Elenis produced a classic 6–3 conservative-liberal split, but in Students for Fair Admissions, Inc. v. President and Fellows of Harvard College (2023), the soft conservative bloc fractured: Thomas and Gorsuch each authored or joined separate opinions, while Kavanaugh wrote a concurrence that narrowed the majority rationale. In these cases, internal conservative agreement dropped to 0.78 (soft), while liberal cohesion remained above 0.90. Other issue areas, such as economic activity or criminal procedure, displayed occasional mini-blocs (e.g., Gorsuch, Thomas, and Barrett concurring or dissenting together), but the overall split remained intact.

Robustness checks with PCA, SVD, and MDS underscore the resilience of the two-bloc model. In each term, the first principal component explains well over 80% of the variance, and the MDS plot shows two tightly grouped clusters, separated by a clear gap. There is no persistent “middle” group, and no justice regularly occupies a spatially distinct swing or centrist position. Even Roberts and Kavanaugh—often described as institutionalist or moderating influences—remain firmly embedded within the conservative cluster, both in agreement statistics and spatial plots.

Attempts to force a three-bloc or swing bloc structure are not supported by the data. Clustering into k=3 always lowers the silhouette score and produces at least one incohesive or unstable group. While individual justices may shift for particular cases or issue areas (such as Roberts and Kavanaugh’s concurrences in affirmative action or voting rights), these moves are episodic rather than persistent, and do not coalesce into a new coalition.

In summary, the 2021–2023 Supreme Court was every bit a 6-3 institution as it appeared in public debate, but with important caveats. While the main coalitions were remarkably cohesive and anchored, multidimensionality persisted in civil rights and certain doctrinally fraught cases. These moments of fracture, however, never destabilized the overall 6–3 alignment, which remains the most robust and empirically justified description of the Court’s current structure.

The Most Recent Term (2024)

The 2024 Supreme Court Term continues the recent pattern of a sharply defined ideological divide, with coalition lines drawn almost exclusively along conservative and liberal axes. The data show the Court’s non-unanimous docket is best characterized by a classic 6–3 split. The conservative bloc—Chief Justice Roberts and Justices Thomas, Alito, Gorsuch, Kavanaugh, and Barrett—opposes the liberal bloc of Justices Sotomayor, Kagan, and Jackson. This alignment is not just a product of scholarly convention, but is validated by the Court’s own voting data: within-bloc agreement is 0.79 (hard, majority only) and rises to 0.85 (soft, including concurrences) for the conservatives, while liberals are even more cohesive, at 0.93–0.94.

A silhouette analysis confirms that this two-bloc configuration is the empirically optimal coalition structure for 2024, yielding the highest within-bloc cohesion and the clearest separation between clusters. Attempts to fit the data to a three-bloc or “2–4–3”/“2–4–1” coalition structure result in lower silhouette scores and less cohesive groupings; there is simply no justice or group of justices who consistently function as a stable “middle” bloc. Rather, “swing” justices—most often Roberts or Kavanaugh—are better described as coalition anchors for the conservative bloc, reliably aligning on outcome in nearly 90% of closely divided cases.

The “anchor justice” for each bloc is the one who not only maintains the highest average agreement with all bloc-mates but also shows the lowest rate of outcome divergence. In 2024, Kavanaugh serves as the empirical anchor for the conservatives: he almost never breaks with the bloc, rarely authors dissents, and is frequently in the majority. Sotomayor is the liberal anchor, maintaining the highest agreement with both Kagan and Jackson and seldom dissenting alone.

Yet, not all justices align with equal reliability. Thomas is the weakest link among the conservatives—his rate of hard coalition agreement lags his peers, and he leads the bloc in both authored dissents and actual outcome divergences often signaling that he is interested in moving farther in the conservative direction than his colleagues on the Court. For the liberals, Jackson takes this role, with the highest number of authored dissents and the most frequent outcome divergence, sometimes dissenting solo when Kagan and Sotomayor join the majority.

Issue-area analysis reveals that, as in 2020, civil rights cases produce the most meaningful multidimensional divergence from the term-wide structure. In these cases, the conservative bloc is prone to splintering into mini-coalitions: Thomas, Barrett, and Gorsuch sometimes form a tight subgroup, writing or joining dissents together, while Kavanaugh, Roberts, and Alito occasionally align with liberals or author separate concurrences. The liberal bloc, by contrast, remains solid, with only Jackson’s separate opinions and rare outcome divergences providing any instability. In economic activity cases, the coalition boundaries are looser but still largely map onto the term-wide split; in criminal procedure, the 6–3 alignment persists with high cohesion.

The robustness of these findings is underscored by principal component analysis (PCA) and multidimensional scaling (MDS) of the voting data. The first principal component—the primary “ideological axis”—cleanly separates the Court into the 6–3 blocs, with PC1 alone accounting for more than 80% of the variance. Neither PCA nor MDS reveals a persistent, spatially distinct middle group; Roberts and Kavanaugh, sometimes labeled as moderates or “swings,” cluster with the core conservatives rather than forming a true third bloc. Occasional outlier behavior (such as Thomas or Jackson’s separate opinions) is reflected only on the second principal component, where its explanatory power is minimal. These dimensionality reduction techniques thus robustly confirm that the 6–3 coalition is not only the best statistical fit but also the most accurate substantive characterization of the Court’s decision-making in 2024.

In summary, the 2024 Supreme Court was defined by strong, empirically cohesive ideological blocs, with only isolated and issue-specific divergences—primarily in civil rights cases—hinting at deeper multidimensionality. Attempts to force more granular coalition structures, such as a 3–3–3 or 2–4–3 split, are not supported by either the underlying data or by state-of-the-art statistical techniques. The dominant, and most robust, portrait of the Court remains a sharply divided, but internally cohesive, 6–3 institution.

Graphs Over Time

This line chart tracks the average agreement among justices within each coalition (conservative and liberal) over time. Two lines (one for each bloc) show how tightly each group of justices votes together from term to term—using either “hard” (majority only) or “soft” (including concurrences) definitions.

Periods where lines rise or fall reveal eras of increased or decreased coalition discipline, and points where the lines cross highlight shifts in which bloc was most internally cohesive.

This chart displays the silhouette scores for different numbers of clusters (e.g., k=2 for a two-bloc structure, k=3 for possible three-bloc splits) across terms.

Higher scores mean the coalition split is empirically strong (well-separated, cohesive groups). A consistently higher score for k=2 supports a “classic” two-bloc Court; drops in silhouette for k=3 indicate that swing or middle blocs aren’t persistent.

While the dominant empirical structure across the Roberts Court era is a two-bloc split, a handful of terms—such as 2015 and 2020—exhibited higher silhouette scores for three clusters. In these years, clustering revealed not only liberal and conservative blocs, but also a small, cohesive third group or a lone swing justice. These findings highlight moments of genuine coalition instability or ideological fluidity, with justices like Kennedy or, more recently, Gorsuch and Barrett sometimes occupying this ‘middle’ or doctrinally unique position. Nevertheless, the three-bloc structure was always transitory: by the following term, the data reverted to a stronger two-bloc configuration, as confirmed by renewed silhouette dominance at k=2

If you enjoyed this post please subscribe. Paid subscribers have access to additional articles and keep some of these inquiries possible

Share Legalytics