The Small and Diversifying Network of Legal Scholars: A Study of Co-Authorship from 1980–2020

Introduction

This Essay reports the first comprehensive network analysis of legal scholars connected through co-authorship. If legal scholarship was ever a solitary activity, it certainly is not any longer. Co-authorship has become increasingly common over time, and scholarship is now mostly a collaborative endeavor.1.See infra Figure 1.Show More These collaborations are important for both scholars and for scholarship, and so understanding patterns of co-authorship is crucial for understanding how legal academia functions as a market for intellectual labor and for the product of that labor: legal scholarship.

The labor market for law professors functions in many respects like other markets for skilled labor, and scholarly collaboration creates several channels for professional advancement. Co-authorship can result in greater scholarly productivity,2.See infra Part I.Show More and social networks formed through co-authorship may provide information channels for scholars to learn about hiring opportunities at other schools and for those schools to collect first-hand information about a prospective hire’s value to the law school community. In these ways, co-authorship can help a scholar’s promotion, compensation, and lateral mobility.

But as social networks tend to reflect within-group affinities—along dimensions of race, gender, and class, for example—they may have disparate impacts on the career opportunities of legal scholars. And the differences generated by social networks may exacerbate any inequalities that led to the original underrepresentation of certain groups. For example, recent evidence indicates that female economists collaborate less often and generally within smaller networks than men and that this difference in co-authorship networks explains 18% of the gender research output gap.3.Lorenzo Ductor, Sanjeev Goyal & Anja Prummer, Gender and Collaboration, Rev. Econ. & Statistics 24–25 (2021), https://doi.org/10.1162/rest_a_01113 [https://perma.cc/PS4Z-KXAK].Show More

Beyond the interests of the professoriate and law schools themselves, there are consumers of legal scholarship. For these consumers, patterns of co-authorship matter for how they affect the substance of legal scholarship itself. These consumers include judges, legal practitioners, government officials, journalists, and law students. Law students are sometimes direct consumers of legal scholarship, but they are also frequently indirect consumers through the influence of legal scholarship on classroom pedagogy. Legal scholarship influences the views law professors express in the classroom, and it influences the content of course materials, such as law casebooks.4.In some cases, law school texts are based on methodological approaches that have been developed as legal scholarship. See, e.g., Robert Cooter & Thomas Ulen, Introduction to Law and Economics (5th ed. 2007).Show More

In this context, the factors affecting co-authorship relationships influence the process and outputs of knowledge production and the training of lawyers. The contours of the co-authorship network will affect which scholars influence the trajectory of legal scholarship and what gets taught in the classroom, whether scholarship reflects diverse viewpoints and methodological approaches, whether insights and methods migrate between areas of study, which areas of legal scholarship are active and which grow stale. Thus, the composition of the law professoriate and the legal scholarship it produces are partly a product of the network of legal scholars. The history of legal academia can be partly told in these terms, and the future of law and legal scholarship depends on the evolutionary path of that network.

Research on co-authorship in law has focused on the claim that legal scholarship is more frequently solo-authored compared with other disciplines.5.Tracey E. George & Chris Guthrie, Joining Forces: The Role of Collaboration in the Development of Legal Thought, 52 J. Legal Educ. 559, 561–568 (2002); Michael I. Meyerson, Law School Culture and the Lost Art of Collaboration: Why Don’t Law Professors Play Well with Others?, 93 Neb. L. Rev. 547, 548–549 (2014).Show More Professors Ginsburg and Miles documented an increase in the number of co-authored articles from 2000-2010, which they attribute to the emergence of empirical legal studies.6.Tom Ginsburg & Thomas J. Miles, Empiricism and the Rising Incidence of Coauthorship in Law, 2011 U. Ill. L. Rev. 1785, 1800–1812 (2011).Show More A little over ten years ago, Professors Edelman and George considered the network of legal co-authorship.7.Paul H. Edelman & Tracey E. George, Six Degrees of Cass Sunstein, 11 Green Bag 2d 19, 22–31 (2007).Show More But because there was no database available at the time suitable for computational network analysis, they focused only on identifying the one legal scholar—Cass Sunstein—who they placed at the center of the web of legal co-authorship relationships.8.Id. at 27–30.Show More It is only with the recent availability of large amounts of digitized text data that computational network analyses of legal scholarship has become possible. In general, computational social network analysis is only beginning to get a foothold in legal scholarship, and work on the connectedness of legal scholars is limited.9.See, e.g., Daniel Martin Katz, Joshua R. Gubler, Jon Zelner, Michael J. Bommarito II, Eric Provins, & Eitan Ingall, Reproduction of Hierarchy? A Social Network Analysis of the American Law Professoriate, 61 J. Legal Educ. 76 (2011) (discussing the network of legal scholars and its impact on legal developments); Milan Markovic, The Law Professor Pipeline, 92 Temp. L. Rev. 813 (2019) (exploring how professors’ intellectual and social networks relate to the advancement of their careers); Keerthana Nunna, W. Nicholson Price II & Jonathan Tietz, Hierarchy, Race & Gender in Legal Scholarly Networks, 75 Stan. L. Rev. (forthcoming 2023) (analyzing how hierarchy, race, and gender affect the acknowledgment sections of law review articles).Show More In this Essay, I report the first comprehensive evidence of co-authorship patterns and the network of legal scholars.

I explore how the network of legal scholars and patterns of co-authorship evolved from 1980–2020. I document dramatic growth in the network of legal scholars, and a much greater increase in co-authorship, such that the legal academy can be described as a “small world” that has become smaller over time. Legal scholars are a more diverse group than they were in the 1980s, and although legal scholars tend to coauthor with other scholars of the same gender and minority status, this tendency has declined over time.10 10.See infra Part IV.Show More There is also evidence legal scholars are increasingly finding coauthors outside their own institutions. Finally, I examine the ordering of authors’ names on coauthored papers. Many law journals have recently changed their citation convention for articles with more than two authors. Rather than listing only the first author’s name followed by “et al.”, they will include all the authors’ names in the first citation to the work. A concern with the “et al.” convention was the possibility that underrepresented members in the legal academy were less likely to be listed as the first author on co-authored papers, leading to underappreciation of their scholarly contributions. The evidence for this concern is mixed. I find that racial minorities make up a greater share of first authors on such articles than their proportion of authors overall. By contrast, women and lesbian, gay, and bisexual (“LGB”)11 11.“LGB” is used instead of “LGBT” because the American Association of Law Schools (AALS) Directory of Law Teachers does not include data on transgender faculty.Show More scholars make up a lower share of first authors than their proportion of authors overall.

I. Scholarly Collaboration

Legal scholarship is not a solitary enterprise and professional colleagues intervene at various stages of the writing process. At the very start, informal conversations with other professors can help a scholar identify a project that is both tractable and makes a novel contribution. Once a project is far enough along, it will often be shared with colleagues within the scholar’s home institution for feedback and then presented at seminars or at national or regional conferences. Presentations of this sort allow the author to publicize and receive feedback that improves the work.

Scholarly norms in legal academia make it possible to trace many of the contributions made by one scholar to another’s work. The most obvious norm is citation for crediting related work and identifying authorities for propositions. Recent work by Keerthana Nunna, Nicholson Price II, and Jonathan Tietz excavates the “acknowledgement” footnotes that typically appear on the first page of law review articles.12 12.Nunna et al., supra note 9.Show More The footnotes reveal some of the less formal ways that legal scholars engage in knowledge co-production, and they provide a rich picture of a collaborative network among law professors, as well as the ways that hierarchy, race and gender influence who is acknowledged for their contributions to written work.13 13.Id. at 46.Show More

In this Essay, my focus is on co-authorship. Although citations, acknowledgements, and co-authorship are similar insofar as they all represent linkages between scholars that facilitate knowledge transmission, the co-authorship network is different. Most obviously, for citations and acknowledgements, scholarly influence runs in only one direction. And one doesn’t need permission to cite another author or acknowledge their contributions. As a result, the linkage reflects only the decision-making priorities of one party. By contrast, both scholars must agree to work together on a co-authored project.

Second, citations and acknowledgments are weak linkages between scholars. There may be no personal relationship between the citing and the cited authors at all. And the relationship between acknowledging and acknowledged scholars may amount to nothing more than an email exchange or a brief conversation. By contrast, in most cases, co-authorship will entail a long period—months or even years—of correspondence, conversation, and negotiation over a joint project. It is more likely co-authors will have an intellectual influence on each other. And the thickness of the relationship may create a channel for learning about each author’s personality, talents, circumstances, and ambitions, as well as sharing information about career opportunities at their institutions or in legal academia more generally. All of this can facilitate professional advancement.14 14.On the importance of networks in job matching, see Yannis M. Ioannides & Linda Datcher Loury, Job Information Networks, Neighborhood Effects, and Inequality, 42 J. Econ. Literature 1056, 1061–1062 (2004).Show More

Given the commitment that co-authorship entails, scholars do not enter co-authorship relationships lightly. Why do legal scholars co-author? Professors Ginsburg and Miles offer several reasons.15 15.Ginsburg & Miles, supra note 6, at 1788–90, 1794. For a theoretical model of the co-authorship choice, see Bruna Bruno, Economics of Co-authorship, 44 Econ. Analysis & Pol’y 212 (2014).Show More First, scholars may have complementarities in skills or expertise that allow them to make contributions together that that they couldn’t on their own. A related benefit of co-authorship comes from specialization. Even if two scholars have overlapping expertise, they may be able to write an article more efficiently by specializing their contributions. A second set of reasons relate to professional esteem. Coauthoring with other scholars allows an author to both increase her research output and diversify her scholarship by making contributions outside of her primary area of expertise, all of which will generally increase her professional profile.16 16.Id.Show More An important caveat to this is that co-authors may receive unequal credit. For example, there is evidence that female economists are penalized for coauthoring while male economists are not.17 17.Heather Sarsons, Recognition for Group Work: Gender Differences in Academia, 107 Am. Econ. Rev. 141, 144 (2017).Show More

Of course, there are also costs of co-authorship. Co-authors must agree about the substance of their piece and about the style of writing. They must be content in their working relationship, with the timing of the project, with the process of exchanging drafts and editing, and with spending hours together in conversation. And they must work out the assignment of responsibility and credit for the project, including how their contributions will be recognized through the order of their names on the final paper.18 18.Ginsburg & Miles, supra note 6, at 1790–93.Show More

II. Framework and Data

A. Network Analysis

A network is simply a collection of nodes—representing people, companies or countries, for example—and a set of links or “edges” connecting these individual nodes because of some relationship they have to each other. This very general and abstract characterization of a network can describe an enormous set of social, economic, political, and other kinds of arrangements. The co-authorship network is a collection of nodes representing the authors of law journal articles. The edges are the articles that are co-authored by any two or more legal scholars. Whereas citation and acknowledgment edges have a natural direction, running from the cited to the citing article, the co-authorship relationship does not have a natural direction from one author to the another. The graphs depicting this network are therefore known as “undirected” graphs.19 19.Matthew O. Jackson, Social and Economic Networks 20 (2010).Show More

Network graphs can be complicated mathematical objects. Although some of the most interesting structural features of a network are visible from the graph, it’s also helpful to have quantitative metrics to summarize and describe a network’s important properties that may not be readily apparent on visual inspection, as well as facilitate comparison with other networks. To do this, it’s necessary to introduce a little terminology.

We say that two authors are connected if they are co-authors themselves or if a chain of co-authors connects them to each other. These co-authorship links serve as the channel for information and influence of various kinds, and we are often interested in how far two authors are from each other in the network, since that will influence how effectively information and influence passes between them. An important concept in network analysis is the geodesic between any two connected authors, which is the set of links that form the shortest path between them.20 20.Id. at 32.Show More The distance between these authors is simply the number of those links.21 21.Id.Show More

Sometimes a network is not entirely connected but is composed of several components.22 22.Id. at 26.Show More A set of nodes constitutes a component if each node is connected to each other node. To illustrate, suppose that law professors co-authored only with scholars of the same gender and suppose that the network consisted of men, women, and non-binary scholars. The entire network includes all scholars, but it has three components. And clearly it would be important in describing such a hypothetical network to observe that it has three components, because it means that no scholar is connected to any other scholar of a different gender, and members of each gender are on an island, limited to the information possessed by others of the same gender.

In such a network, some components are likely to be larger than others, with the largest component probably being that of men, followed by the component made up of women and then the component made up of non-binary scholars. There are two common measures of the size of a component. The diameter of a component is the longest distance among all geodesics in the component.23 23.Id. at 32.Show More Finding the diameter involves finding all the shortest paths between nodes, and then identifying the longest of these paths. The component with the longest diameter is known as the giant component.

Note that components with more nodes do not necessarily have a longer diameter. To take an extreme example, suppose that the component of men scholars included Cass Sunstein and 500 other men, and Professor Sunstein co-authored an article with each of the 500 but that none of the 500 co-authored with each other. The graph of this component would look like a star, with Professor Sunstein in the middle. The shortest path between Sunstein and each other scholar would be 1, and the shortest path between any two of the 500 would be two (going through Sunstein). The diameter of this component would be two. Suppose that the component of women scholars had the same star structure, but with Jill Fisch at the center and 100 other women scholars. This component would have the same size as the men’s component, because the diameter would still be two.

Another important measure of the size of the component is average path length.24 24.Id. at 33.Show More This is simply the average of all geodesics—shortest paths—between nodes in the network. This measure captures the fewest authors connecting a typical pair of authors in the network, and the measure corresponds to the intuitive notion of the “degrees of separation” between two people in a network.

In addition to wanting to describe the size of a network, its various components, and how close its nodes are to each other, it’s often important to describe the properties of individual nodes. For example, we may want to know if there are particular legal scholars who are especially well connected—who frequently co-author—and whether they resemble their co-authors in certain ways. One way of measuring the connectedness of a professor in the co-authorship network is simply to count the number of her co-authors. The number of her co-authors is known as the professor’s degree.25 25.Id. at 29.Show More Using this measure, we can get a sense of how connected the typical law professor is by looking at the average degree among all professors in the network.

Given our interest in the role of co-authorship in allocating professional opportunities and facilitating intellectual cross-pollination, we are interested in measuring how the characteristics of individual authors are correlated—positively or negatively—with the characteristics of her co-authors. For example, do professors tend to seek similarity in co-authorship, writing with other scholars of the same gender, race, or generational cohort? Or do they tend to seek complementarity, writing with scholars whose differences might result in a more efficient allocation of labor or a more creative output? To measure the tendency for scholars to co-author with other scholars with similar attributes—a network property known as homophily—I use a measure of correlation known as the assortativity coefficient and calculate it for scholars’ gender, minority status, and degree.26 26.Mark E.J. Newman, Mixing Patterns in Networks, 67 Physical Rev. 2 (2003).Show More The coefficient is positive if scholars tend to co-author with scholars who share the same traits, and it is negative if they tend to co-author with scholars having different traits.

An important characteristic of social networks is the amount of clustering or cliquishness. We may want to know, for example, whether it’s typical for a scholar’s co-authors to co-author with each other or whether certain professors are central in some sense to the scholarly network. A professor with many co-authors may be well connected, but a professor at the center of a dense, interconnected part of the co-authorship network may be important because she facilitates other collaborations and network connectivity. There are two measures of the amount of network clustering around individual authors. The transitivity—also known as the triadic closure or overall clustering—of the network is calculated by determining, for each individual author, how frequently any two of her co-authors are themselves co-authors.27 27.Jackson, supra note 18, at 35.Show More The second measure is the average clustering coefficient.28 28.Id.Show More The measure is calculated for each scholar, taking the number of co-authorship relationships among her co-authors as a fraction of all possible such relationships, and averaging across all authors. Because the average clustering coefficient weights the clustering coefficients of all authors equally (i.e., regardless of how many co-authors an author has) it will give greater weight to authors with fewer co-authors than the transitivity measure of clustering.

B. Data

The raw data used to create the network is all articles published in the top 100 general and specialty law reviews as ranked by Washington and Lee for the period 1980–2019. By limiting my sample to student-edited law journals, I will likely underestimate the amount of co-authorship by legal scholars overall, since much co-authored work with an empirical component appears in peer-reviewed journals such as those at the intersection of law and economics or law and psychology. Including these journals, however, would likely only amplify the increase in co-authorship I observe over time, since the emergence of empirical legal studies is a relatively recent phenomenon. The journals included in the sample are listed in Table 1 in the Appendix. I then removed student notes and comments, articles with more than 10 authors, and articles that did not have at least one author who was a professor. This left a total of 67,472 articles. I then identified the 9,320 authors who published more than one article during the period 1980–2019, who I call “scholars.” Focusing on scholars eliminates articles by practitioners, students, and judges that were not filtered at an earlier stage, allowing me to limit my analysis to the social network of research-active authors.

Data on authors’ gender, minority status, institutional affiliation, and age were collected from the American Association of Law Schools (AALS) Directory of Law Teachers. These data are collected from surveys and are self-reported. The Directory includes lists of minority teachers beginning in 1986 and lists LGB professors beginning in 1996.29 29.For racial minorities, I used lists published in the 1986, 1989, 1990, 1991, 1992, 1996, 2000, 2004, 2007, 2011, 2014, and 2019 editions. For sexual minorities, I used lists from the same years, beginning in 1996.Show More I identify a scholar as a minority or member of the LGB community if they ever appear on these lists. There may be reasons why scholars in different time periods or at different institutions may be less likely to submit this information to AALS, which may lead to underreporting of the number of such scholars. I code any scholar whose name does not appear on the lists of minority or LGB professors as not belonging to those communities. The Directory reported gender information from 1986–2011 and I use these data when they are available. For any author whose gender is not reported in the Directory, I first matched the author’s given name with the 200 most popular baby names by gender as reported by the Social Security Administration for each decade from the 1940s to the 2010s. Finally, a student research assistant supplemented the data by doing internet searches of the remaining scholars’ names and assessing their genders. The gender of any scholar for whom I could not obtain data using the procedure above is coded as missing.

III. The Emergent Small World of Legal Academia

In this Part, I analyze the scholar co-authorship network for each of the last four decades. I describe the growth and evolution of that network alongside the overall rise in co-authorship within legal scholarship and report how the distribution of co-authorship is “fat tailed,” with more scholars doing solo-authored work and more scholars having many co-authors than one would expect if co-authorship relationships were formed randomly. Finally, I evaluate whether the network of legal scholars can be described as a “small world” as that term is used in the network analysis literature, such that even a community of very many law professors is densely connected with relatively few degrees of separation between its members. I find that the law professoriate is, in fact, a small world and that it is getting smaller. Even as the number of legal scholars has grown significantly, they are more closely connected to each other than ever before.

A. Network Growth

There are more legal scholars and more legal scholarship being published than ever before. In the 1980s—the first decade of my sample period—there were 3,409 scholars publishing and there were 9,742 articles published. In the 2010s, by contrast, there were 5,595 scholars and 18,159 published articles. And although the number of both solo-authored works and co-authored articles has increased over time, there has been a much more dramatic increase in co-authored work. The absolute number of articles with only one author increased from 9,139 in the 1980s to 15,648 in the 2010s, but the number of articles with at least two legal scholars quadrupled from 603 in the 1980s to 2,511 in the 2010s. Figure 1 shows the share of all published articles that were co-authored each year by scholars. Not only do the data show that this share has tripled from roughly 5% in the early 1980s to over 15% by 2019, but the trend is striking for its steady increase over the entire time frame. The increase in co-authorship over the last forty years is clearly not just a blip, but it reflects a persistent and growing shift in how legal scholarship is produced.

The long time period of my sample provides some context for earlier studies on legal co-authorship and raises questions about the underlying dynamics generating the trend. In earlier work, Professors Ginsburg and Miles reported an increase in the share of co-authored articles published in the top 15 law reviews from 2000 and 2010, and in two peer-reviewed law journals from 1989 to 2011.30 30.Ginsburg & Miles, supra note 6, at 1785.Show More They attributed this increase to the rise of empirical legal studies.31 31.Id. at 1785 (“[R]esults support the view that specialization, and specifically the growth of empirical scholarship, has contributed to the trend of coauthorship in legal academia.”).Show More Legal scholarship that incorporates quantitative analysis is likely to be especially suitable for co-authorship, bringing together scholars with expertise in quantitative analysis with topic matter specialists.

On the one hand, an examination of the scholars who co-author most lends support to this view. Table 2, in the Appendix, lists the legal scholars in each decade with the most co-authors. These lists include several scholars—such as Ian Ayres, Mitu Gulati, and Ted Eisenberg—who do a great deal of empirical scholarship. On the other hand, the lists also include many scholars who do not do empirical work, and it is also notable that the trend in increasing co-authorship had already begun by 1980 and continues unabated through the present. It is difficult to place a specific start date for the contemporary rise of empirical work, but some sociologists place it in the mid-1990s,32 32.Mark C. Suchman & Elizabeth Mertz, Toward a New Legal Empiricism: Empirical Legal Studies and New Legal Realism, 6 Ann. Rev. L. & Soc. Sci. 555, 556 (2010) (“Since the mid-1990s, several groups of scholars have championed a renewed dialog between law and social science.”).Show More and the Journal of Empirical Legal Studies did not begin to be published until 2004.33 33.Journal of Empirical Legal Studies, Wiley Online Libr., https://onlinelibrary.wiley.com/loi/17401461 [https://perma.cc/6GXQ-LCEK] (last visited Oct. 28, 2022).Show More Moreover, the trend toward co-authorship is evident looking at the top 100 law journals and excluding the peer-reviewed journals where much quantitative legal scholarship is published, such as the Journal of Empirical Legal Studies, the Journal of Law & Economics, and the American Law and Economics Review. Including peer-reviewed journals would likely amplify this trend.34 34.See Ginsburg and Miles, supra note 6, at 1785 (“Coauthored articles were far more common in . . . [the Journal of Legal Studies and Journal of Law, Economics and Organization] . . . than in the general interest, student-edited law reviews.”).Show More

Figure 1: Co-Authored Share of Articles

Over the last forty years then, there has been an increase in both the number of legal scholars—the nodes—and the number of co-authorship relationships—the edges. Figures 2–5 show the staggering increase in the size of the largest component of the co-authorship networks for each of the past four decades. In 1980, this component included only 26 scholars, making up 3.4% of the entire professor network. Things look dramatically different in 2010s, where the largest component of the scholarly network included 1,226 scholars and made up 50.9% of the entire network. Thus, a much greater share of the professoriate is now connected to each other through co-authorship linkages. The hubs of each giant component are those with the most co-authors, and they are indicated on each graph by their names. As co-authorship has become more prevalent, the threshold for being one of these hubs has increased as well. In the 1980s and the 1990s, the hubs were scholars with at least four co-authors. In the 2000s and the 2010s, these were scholars such as Ian Ayres and Cass Sunstein with at least twelve co-authors. The ten scholars with the most co-authors by decade are listed in Table 2, in the Appendix. Whereas the most connected professor in the 1980s was corporate and securities law scholar Jonathan Macey with eight co-authors, in the 2000s and 2010s the most connected scholar was IP scholar Mark Lemley with 22 and 26 scholar co-authors in those decades, respectively.

The number of a scholar’s co-authors—their degree—is only one measure of their significance to the co-authorship network, and there are other measures that reflect other concepts of a scholar’s centrality. For example, closeness centrality measures the average distance between a scholar and the other scholars in the network through co-authorship relationships. This measure is conceptually analogous to typical “degrees of separation” between one scholar and all the other scholars. An alternative measure of a scholar’s centrality or importance to the network is her betweenness centrality. This measure captures how often the shortest path between any two other professors runs through the scholar.35 35.For details about the calculation of these centrality measures, see Jackson, supra note 19, at 39.Show More

Lists of the top ten scholars by these two measures of centrality are in Tables 3–4 in the Appendix. There is a fair amount of overlap between the scholars identified as the most closely connected using the closeness centrality measure and the scholars through whom other scholars are linked using the betweenness centrality measure. These lists include prolific scholars who frequently co-author—especially in the areas of corporate and securities law—such as Jill Fisch, Mitu Gulati, and Steven Choi, but also a miscellaneous group of other scholars writing in areas such as IP, behavioral economics, and administrative law.

Figure 2: Legal Scholar Co-authorship Network 1980-1989

Chart

Description automatically generated

Figure 3: Legal Scholar Co-authorship Network 1990-1999

A picture containing schematic

Description automatically generated

Figure 4: Legal Scholar Co-authorship Network 2000-2009

Diagram

Description automatically generated

Figure 5: Legal Scholar Co-authorship Network 2010-2019

Diagram

Description automatically generated

B. Characteristics of the Network

The scholarly network contains many more scholars and is much more connected than it used to be. In this Section, I conduct a closer investigation of the law co-authorship networks that emerged in each of the last four decades, comparing them using the various summary statistics described in Part I to understand how the scholarly network has evolved over time. These summary statistics are reported in Table 1. I constructed a separate co-authorship network for each decade and reported the density, average clustering coefficient, and transitivity of those networks. For the giant component of each network, I also reported the density, the average shortest path (geodesic) between scholars, and the diameter.

Table 1: Co-authorship Network Statistics by Decade

One measure of a network’s connectedness is its density. The network density is simply the share of all possible links between co-authors that are present. As the size of the scholar network has increased over the last forty years—from 760 to 2,410 scholars—its density has declined; there were fewer co-authorship relationships as a share of possible relationships in each successive decade than the one before. This is an almost inevitable consequence of the growth of the network, because adding the th scholar to the network means adding possible links, and so as grows, the set of possible links grows at an increasing rate.

Recall that there are two measures of cliquishness or clustering in the network—transitivity and the average clustering coefficient. Interestingly, these two measures exhibit different patterns over time. The average clustering coefficient has increased over time, indicating that a typical scholar’s co-authors are more likely to co-author now than they were before. Transitivity—or overall clustering—however, declined in the 1990s and then the 2000s before increasing again in the 2010s. What explains these different patterns? Recall that the key difference between average clustering and overall clustering is that overall clustering gives greater weight to scholars with more co-authors. The different temporal patterns from 1980–2010 could be explained by a general increase in co-authorship during this time period among small cliques, while some of the “star” scholars with many co-authors during the early period continued to add co-authors in the 1990s and 2000s who were not themselves well connected. This pattern could arise, for example, if star scholars are more senior members of the academy who chose in the 1990s and 2000s to co-author with relatively junior scholars.

The last row of Table 1 reports the degree assortativity of the four networks. This captures the tendency for scholars with many co-authors to work with other scholars with many co-authors. This statistic follows the same pattern over time as network transitivity. Scholars who were well connected were less likely to co-author with scholars who were also well connected from 1990–2010 than in the 1980s. This statistic is consistent with the idea that “star” scholars in the 1990s and 2000s were more likely to co-author with scholars who were not themselves well connected.

Although the giant component of the scholarly network has grown enormously in terms of its coverage of the entire network—covering 50.8% of the 2010 network and having a diameter of 30—many scholars remain only loosely connected. The density of the giant component has followed the same trend as the entire network and is smaller than it was in the 1980s. Even more importantly for understanding how close scholars are to each other through chains of co-authorship, the average shortest path increased from 3.9 in the 1980s to 10.1 in the 2010s, meaning that, for any two randomly chosen scholars, the expected “degrees of separation” is nine intermediate co-authors.

Since it is hard to know what to make of these statistics in isolation, I have focused on how they have evolved over time to understand the trajectory of legal academia. Another way of contextualizing these numbers is by comparing the network of legal scholars with the co-authorship networks in other academic disciplines. Table 2 reports statistics on the co-authorship networks in Biology, Economics, Math and Physics and compares them with statistics for the 2010s co-authorship network of legal scholars.36 36.The data come from Jackson, supra note 19.Show More

Table 2: Comparison of Co-authorship Networks

Even as the number of published legal scholars has grown over time, there are only a small fraction as many law professors publishing in U.S. law journals as there are scientists, economists and mathematicians publishing in their fields. The second row reports the average degree—the average number of co-authors—among scholars in each field. Researchers in the natural sciences typically have many more co-authors than scholars in other fields, due to the collaborative nature of work that is often funded by outside grants, cooperation between theorists and experimental researchers, and norms for giving co-authorship credit within research labs. The model of knowledge production in law is much closer to that in economics, and so comparisons between law and economics will be especially useful.

It is noteworthy that, notwithstanding the historical scarcity of co-authorship in law37 37.Meyerson, supra note 5, at 548.Show More and the trend toward increasing co-authorship in economics,38 38.Andy H. Barnett, Richard W. Ault & David L. Kaserman, The Rising Incidence of Co-Authorship in Economics: Further Evidence, Rev. of Econ. & Stat. 539 (1988).Show More the average number of co-authors among legal scholars has overtaken that among economists. Focusing on the giant component of each scholarly network, the giant component in law has a longer diameter than in any of these other fields but makes up a much smaller share of the network of scholars than the giant components in biology, physics, and math. Virtually all biologists, physicists, and mathematicians are connected to each other through a single co-authorship network. By contrast, the giant component includes only 41% of economists and 59% of legal scholars. But these numbers do show that that there has been greater consolidation of the legal scholar network than among economists.

The average path length—of 10.1—between legal scholars is longer than in the other networks, indicating that legal scholars are typically more degrees removed from each other than scholars in other fields. The biology and physics networks are particularly dense. But the difference between the average path length in economics and in law is modest, only a difference of half a degree, on average. This suggests that economists and law professors are situated similarly close to each other, as measured by co-authorship relationships. Interestingly, transitivity in the legal scholar network is considerably greater than in all other fields except physics. This means that a legal scholar’s co-authors are more likely to co-author with each other than the co-authors of scholars in the other fields, suggesting a higher degree of cliquishness in law.

C. Co-author Distributions

In the 1980s, the average number of scholar co-authors was only 1.44, and by the 2010s, the average number of scholar co-authors had increased to 2.2. But focusing solely on the average number of co-authors ignores considerable nuance in how co-authorship has evolved. For example, we should want to know whether this increase in the average number of co-authors is being driven by an increase in co-authorship across the board—whether all legal scholars are co-authoring more—or whether the increase has been driven by a relatively small number of scholars. We can see this by looking beyond the average number of co-authors to examine the entire distribution of co-authors among scholars. Let be the share of scholars in the network that have co-authors ( stands for “degrees”). Examining the distribution of scholars by number of co-authors, and how that distribution has changed over time, can shed light on how co-authorship relationships form and why the average has increased over time. But rather than plot graphs showing the share of scholars on the vertical axis against the number of co-authors on the horizontal axis, I transformed these numbers by taking the natural logarithm of each, for ease of visualization and for reasons discussed below. As a result, Figures 6–9 show graphs plotting the natural log of against the natural log of for each decade of the sample period.

Figure 6: Co-author Distribution 1980s Figure 7: Co-author Distribution 1990s
Figure 8: Co-author Distribution 2000s Figure 9: Co-author Distribution 2010s

The dots in each graph represent the actual distributions showing the share of scholars with a specified number of co-authors—after taking the natural log of each number. For example, in Figure 8, consider the third blue dot from the left. This dot corresponds to the roughly 13.5% of scholars in the 2000s network (the natural log of which is about -2) who had 3 co-authors (the natural log of which is about 1.1). To better visualize the distribution, the red line shows the straight line that best fits the dot pattern.

What process of co-authorship formation could have led to these distributional patterns? One useful benchmark is a process whereby co-authorship relationships are formed randomly, as though each scholar when deciding who to write with picked a name out of the hat of all legal scholars. A network generated in this way would have a graph in which the distribution of degrees was much more curved than the distributions we observe among legal scholars. Particularly in the 2000s and 2010s, the distribution of co-authorship is reasonably well-represented by the straight red line.

To see the difference, the curved lines in green show the degree distributions that would result from random graphs. For each decadal network, I used the number of scholars and the number of links from the actual network but assigned those links randomly. The relative linearity of the distribution as compared with the random network means that the actual co-authorship distribution has more scholars with very many co-authors and more scholars with very few co-authors than would be expected if co-authorship relationships were formed at random. This is known as a “fat-tailed distribution.”

What kind of process would lead to these fat-tailed distributions? One possible mechanism for the formation of the law-scholar network is known as preferential attachment. The idea of this process is that when someone is choosing who to write with, they are more likely to pair up with someone the more co-authors that person already has. There are a variety of reasons why preferential attachment could arise in the network of legal scholars. For example, it could be that scholars at some law schools are more likely to be co-authors than at others, and that people tend to write with their colleagues. It could be that some scholars work in areas where co-authorship is more common, so that scholars working in empirical legal studies tend to have many co-authors whereas scholars in other areas tend to write solo-authored works. Or it could be that co-authorship is associated with an increased scholarly profile or a personal preference for writing with others, such that people with many co-authors tend to seek more co-authorship projects, or other scholars seek them out.

D. A Small World

As the number of legal scholars has grown, technological innovations such as email, videoconferencing, online platforms, and cheaper travel have made it easier for scholars to identify potential collaborators and work with scholars at other institutions or in other time zones. For whatever reason, the rise of co-authorship has brought the network of legal scholars closer together. But how close? One way of answering this question in the network literature is to ask whether legal scholars inhabit a “small world.” Following the scholarly literature, I say that the network of legal co-authorship has small-world properties if (1) the number of scholars is much larger than the average number of co-authors; (2) the giant component covers a large share of the entire network; (3) the average shortest path in the giant component is small;39 39.Specifically, the average shortest path should be of the same order as .Show More and (4) if there is significant clustering, so that the transitivity of the network is much larger than the average number of co-authors divided by the number of scholars.40 40.In this I follow Sanjeev Goyal, Marco J. van der Leij & José Luis Moraga‐González, Economics: An Emerging Small World, 114 J. Pol. Econ. 403 (2006). Goyal et al. builds on the seminal analysis of Duncan J. Watts & Steven H. Strogatz, Collective dynamics of ‘small-world’ networks, 393 Nature 440 (1998).Show More To evaluate the network of legal scholars along these dimensions, I benchmark its small-world features against the network of economics co-authorship and also against randomly generated small-world networks.

Consider the first property. In the last four decades, the average number of a legal scholar’s co-authors has increased from 1.4 to 2.2 while the number of legal scholars has increased from 760 to 2,410. Not only is the number of legal scholars much larger than the average number of co-authors, but the ratio of the two is increasing over time. Thus, the legal scholar network satisfies the first requirement for a small world, and even more so in recent years. And the second small-world property is satisfied too. The giant component, which covered only a very small part of the scholar network in the 1980s, now covers roughly 51% of the network. By comparison, the giant component of the economics co-authorship network covers 40% of the network, which scholars concluded is still large enough to be indicative of a small world.41 41.Goyal et al., supra note 39, at 408.Show More

The third small-world criterion is that the average shortest path in the giant component is “small.” Although the average shortest path of the giant component has increased modestly over time, it remains relatively small. Specifically, it is conventional to define the average shortest path as small when it is of the same order of magnitude as the natural log of the number of nodes in the network.42 42.Id. at 405.Show More In the 2010s, the average shortest path of the giant component was 10.1 and the log of the number of scholars was 7.8. These are of the same order of magnitude because if we multiply 7.8 by 10 we get a number that is much larger than 10.1. Another way of thinking about the smallness of the average path length is by comparing it to the average shortest path in the economics network. As noted above, the average shortest path in economics is 9.5, so the two have similar average path lengths and the third requirement for the law network to be a small world is satisfied.

Finally, how strong is the clustering in the law network? From Table 1, we see that the transitivity of the law network declined from 0.42 to 0.33 to 0.24 and increased back to 0.39 in the 2010s. Is this a lot of clustering? One way to answer this question is to compare the overall clustering with the clustering that would be expected if the network were randomly generated. In that case, the transitivity of the random network would be approximately the average degree of a scholar divided by the number of scholars. In the last four decades, the randomly generated transitivity would be 0.0019, 0.0012, 0.0010 and 0.0009—far less than the amount of clustering than we observe. The actual amount of overall clustering in the network over the last four decades is 214, 269, 235, and 352 times the amount of clustering predicted in a random network. We can also compare transitivity against the network of economists. The clustering in the law network is not as high as the clustering in the economics network—where the clustering coefficient was 700 times what one would expect in a random network—but the two are not far apart, so we still conclude that the fourth small-world requirement is satisfied. Since all four conditions are met, we conclude that the network of legal scholars is a small world, and one that has become smaller over time.

Rather than use networks from other fields as a benchmark, we can also compare the network of legal scholars against small-world networks simulated using algorithms designed to generate structures with small-world properties (significant clustering and low average path lengths). A Watts-Strogotz small-world graph is generated from a ring lattice structure in which each node is connected to the nearest nodes in the ring, and then proceeds by taking each edge connecting the node to its neighbor and randomly “rewiring” it to another node with some fixed probability. The resulting rewired network is a small world.

The average degrees in the 2010 legal scholars’ network and its giant component are 2.2 and 2.9, respectively. The average degrees in a small-world graph is an even number , so we cannot match the co-authorship graph precisely to a small-world graph with the same average degrees. As a result, I consider two possibilities. Table 3 compares the properties of the 2010 network and its giant component with connected small-world graphs for and The 2010 network has many fewer edges than the small world, so its density and average path length are lower, but the average and global clustering measures are much higher, confirming that the network of legal scholars had become a small world by the 2010s.

Table 3: Comparison to Watts-Strogotz Small-World Graphs

IV. Demographic and Institutional Trends

Not only has the network of legal scholars grown over time, but it has also diversified. The number and the share of female, minority, and LGB scholars has increased along with the increase in the overall numbers of active legal scholars, although the trends in representation across these groups differ. Consider first self-identified minorities. In the 1980s, 4.7% of scholars were minorities, and in the 1990s, the share of minority scholars doubled to 9.5%. But the increase in minority representation since then has been more modest, with 11.7% of active scholars reporting being minorities in the 2010s. The share of active self-identified LGB legal scholars was 1.8% in the 1980s and has increased each decade until reaching 3.9% in the 2010s. Finally, the number of female scholars has increased from 17.0% in the 1980s to 34.5% in the 2010s.

The growing number of female, minority, and LGB scholars in academia overall is reflected in the authorship of legal scholarship. Figure 10 shows that the share of articles authored by female, minority and LGB scholars has increased over time. Each line shows the trends in authorship by one group, smoothing out annual fluctuations to reveal the overall trend.

Figure 10: Share of Articles with Female, Minority or LGB Authorship

Comparing across decades, in the 1980s, minorities authored 4.3% of articles, and in the 2010s, they authored 12.9% of articles. The share of articles authored by LGB scholars was 1.8% in the 1980s and 4.2% in the 2010s. Women scholars authored 13.2% of articles in the 1980s and 32.9% in the 2010s. Comparing the representation of minorities, women, and LGB scholars as a share of scholars and in terms of their share of scholarship shows that minority and LGB scholars generate a modestly larger share of scholarship (12.9% and 4.2%) than their proportion of scholars (11.7% and 3.9%) while women scholars generate a modestly smaller percentage of legal scholarship (32.9%) than their proportion of scholars.

Although the law professoriate and legal scholarship now has more diverse representation, has it become more integrated? Or are minority, LGB, and women scholars cloistered in terms of their scholarly work? Figure 11 shows how the share of co-authored articles with “mixed” co-authorship (i.e., one minority and one non-minority, one LGB scholar and one non-LGB scholar, or one male and one female scholar) has evolved over time. As with Figure 10, the raw data have been smoothed to better see the trend lines.

The data show that the share of mixed co-authored articles has increased in line with the overall increase in female, minority and LGB scholars. Not only is an increasing share of legal scholarship co-authored, but an increasing share of co-authored scholarship reflects diversity along racial, gender, and LGB lines. This trend appears to be starkest for women scholars, who were represented on over 40% of co-authored scholarship by 2019, more than double their representation in 1980. The increase in women’s representation began its steep ascent around 1990 and continued until the end of the sample period.

Figure 11: Share of Co-Authored Articles with Mixed Co-Authorship

The increase in mixed co-authorship over time does not, however, imply that co-authorship relationships are formed without regard to differences in race, gender, or sexual orientation. Members of the same gender or minority group could still be more likely to co-author because of scholarly or personal affinities or geographic clustering in the same region or at the same law school. On the other hand, members of different genders or minority statuses may be more likely to co-author because their differences are complementary and allow them to generate new ideas and make novel contributions. The differential probability for co-authorship links to form between nodes with the same attributes is known as homophily. To measure homophily by minority, gender, and sexual minority status, Table 4 reports the assortativity coefficient for each characteristic and decade for which I have mostly comprehensive data on the characteristic. This measure has a value of 0 when there is no sorting by the attribute. A positive value is evidence of homophily—scholars co-authoring with scholars of the same “type”—while a negative value is evidence that similar scholars are less likely to co-author.

Table 4: Homophily by Attribute

For each decade and each demographic category, the coefficient is positive, indicating that legal scholars have some affinity to co-author with scholars who are of the same gender and minority status. But also, for gender, minority status, and LGB status, the strength of that homophily is declining over time. Thus, not only is legal academia becoming more diverse over time, but it is also becoming less segregated, at least in terms of scholarly collaboration. This is good news, because it increases the likelihood that the fruits of the scholarly network—information sharing and scholarly visibility for professional advancement—are more widely shared than they would be in a more segregated network.

One other aspect of scholarly segregation is the amount of co-authorship that happens between scholars at different institutions. Although technological developments make it easier to collaborate with scholars from other institutions than before by reducing the costs of communication,43 43.See, e.g., Mu‐Hsuan Huang, Ling‐Ling Wu & Yi‐Chen Wu, A Study of Research Collaboration in the Pre‐web and Post‐web Stages: A Coauthorship Analysis of the Information Systems Discipline, 66 J. Ass’n for Info. Sci. & Tech. 778 (2015).Show More co-authorship networks may remain institutionally fragmented. And fragmentation reduces the information flow across institutions, thereby erecting barriers toward the mobility of law professors and the likelihood of fruitful intellectual collaborations. Moreover, co-authorship across law schools may usefully undermine “letterhead bias,” the idea that opportunities for scholarly publication depend on the perceived status of the author.

Figure 12 shows that the fraction of all co-authored articles with authors from different law schools has increased over time. Although the data series exhibits a lot of volatility, it also has a clear upward trend. The dots plot the raw data, and the solid line is a smoothed line to better illustrate the trend. The data series ends in 2011, which is the last year for which electronic access to the AALS directory and its information on faculty affiliation was available.

There is also considerable variation among law schools in how much their faculty co-author and whether those co-authorship relationships cross institutional boundaries or are within-school. Table 5, in the Appendix, lists the top 10 percent of law schools in terms of published articles from 1980–2019. Scholars from Cornell and Northwestern were most likely to co-author during this period, a fact which may be attributable to those faculties’ strength in and emphasis on interdisciplinary work. Among co-authored papers, scholars from Stanford, Texas, and UPenn are represented more than scholars from other institutions.

Figure 12: Co-authored Articles with Authors from Different Law Schools

V. Author Name Ordering

The rise of co-authorship and the diversification of legal academia raises several questions about how scholars are credited for co-authored work. The answers to these questions feed directly into scholars’ prospects for promotion, tenure, and compensation. And there is reason to worry that the allocation of credit may be biased. In economics, for example, there is evidence that co-authorship reduces the probability of tenure for junior female economists but not junior male economists.44 44.See Sarsons, supra note 17; Andrew Hussey, Sheena Murray & Wendy Stock, Gender, Coauthorship, and Academic Outcomes in Economics, 60 Econ. Inquiry 465 (2022).Show More Moreover, the penalty for co-authorship is greater when women co-author with men than when they co-author with other women.45 45.Sarsons, supra note 17.Show More One reason that co-authors may not receive equal credit for a joint publication is the order in which their names are listed on the article.46 46.Boris Maciejovsky, David V. Budescu & Dan Ariely, Research Note—The Researcher as a Consumer of Scientific Publications: How Do Name-Ordering Conventions Affect Inferences About Contribution Credits?, 28 Mktg. Sci. 589 (2008).Show More In some academic disciplines, the order of authors’ names is intended to convey their relative contributions to the work. Legal scholarship—like economics—typically but not exclusively uses alphabetic ordering. Nevertheless, the first-named author on a co-authored work may receive more credit than later authors because theirs may be the only name that appears when the work is cited by subsequent scholars.

In 2020, some student-edited law journals dispensed with the citation convention to identify only the first author by name for articles with more than two authors. The initiative to list all the authors was intended mostly to ensure that all scholars working on the piece receive credit, but also because the “et al.” convention may have disfavored junior, minority or female scholars if they are underrepresented among first authors. There is evidence of gender bias in the assignment of the first author position in other academic disciplines,47 47.Arturo Casadevall, Gregg L. Semenza, Sarah Jackson, Gordon Tomaselli & Rexford S. Ahima, Reducing Bias: Accounting for the Order of Co–first Authors, 129 J. Clinical Investigation 2167 (2019).Show More so perhaps this is true in law as well. In this section, I explore the representation of scholars in the first author position over the last 40 years.

For articles with more than two authors—those for which the “et al.” convention would be used—I calculated the share of first authors and the share of all authors who are women, minority, and sexual minority scholars. Comparing each group’s share of authors with their share of first authors will reveal the effect of doing away with the et al. convention on representation of these scholars’ names in citations. I then focused on articles with mixed authorship and calculated the share of these articles for which the woman or minority was the first author. To evaluate whether representation is meaningfully different among mixed author articles than one would expect if lead authorship were assigned randomly, I simulated random assignment of lead authorship among these articles 1000 times and calculated the expected share of female/minority lead authors and the probability of observing the actual share of female/minority lead authors.

Whereas 26.2% of all co-authors are women scholars, 24.5% of the first authors are women scholars. LGB scholars made up 2.7% of all authors and only 1.6% of all first authors. By contrast, racial minority scholars comprise 7.7% of all co-authors and 8.3% of all first authors. Thus, the et al. convention reduces the share of named scholars who are women and the share of LGB scholars and increases the share of named authors who are racial minorities.

Who occupies the first author position for articles with mixed co-authors? For women, the expected share of first authors based on random assignment is 44.1% and the actual share is 41.6%. The probability of observing female representation among first authors of 41.6% or less under the assumption of random assignment of author ordering is 0.12, a p-value that is suggestive but not generally associated with a rejection of the null hypothesis (of random assignment of author ordering). For LGB scholars, the expected share of first LGB lead authors among mixed articles is 43.7% but the actual share is 24.6%, a dramatically lower number associated with a p-value of close to 0. For racial minorities, the expected share of first authors based on random assignment is 39.7% and the actual share of minority first authors is 43.4%. Under a null hypothesis of random assignment, the probability of observing minority representation among first authors of 43.4% is 0.13, also suggestive but not strong evidence against the null.

Returning to our discussion at the outset about the reasons for co-authorship, there are many explanations for why minority scholars may be “overrepresented” among first authors and women and sexual minorities “underrepresented,” relative to their share of co-authors in general. One possibility is that author ordering reflects relative contributions. Another possibility is that gender, race, and LGB status are correlated with factors such as seniority or professional status among co-authors that influence the assignment of author ordering. Another possibility is that first authorship has a greater professional return for some scholars. If co-authors are altruistic, they may acquiesce to assigning first authorship to the scholar who will benefit the most. Of course, it need not be altruism. This outcome could also result from bargaining among co-authors as they work together to maximize the aggregate boost to their professional reputations. In that case, the other scholars may extract concessions from the first author in the form of greater or more menial contributions to the project in exchange for the benefit of being first author.

VI. Discussion

The network of legal scholars looks dramatically different now than it did forty years ago. There were 63% more scholars publishing and 86% more articles written in the 2010s than the 1980s. The explosion in co-authorship in the last two decades has created a connected network of scholars to which more than 50% of legal scholars belong. As a result, even as the number of legal scholars has grown, their world is smaller than ever, a conclusion that is further supported by the increasing number of collaborations between scholars from different institutions.

As new scholars join the network, their attachment is not random. Certain scholars co-author much more than others, and some scholars tend not to co-author at all. And legal scholars tend to co-author with others of the same gender and minority status, although this correlation has declined over time. As the number of women, minorities, and LGB scholars has increased over time, the increase in representation is most dramatic for women both overall and in terms of participating in mixed co-authorship arrangements. When legal scholars do co-author, their names are listed alphabetically 65% of the time. Racial minority scholars are listed as first authors more frequently—and women and LGB scholars are listed as first authors less frequently—than one would expect based on their proportion in the population of authors. Future work should explore in greater depth how new scholars become incorporated into the network, what other factors help establish co-authorship relationships, and what the consequences of these relationships are for legal scholarship and the distribution of professional opportunities in legal academia.

Appendix

Table 1: Law Reviews in Sample

Table 2: Top Scholars by Number of Scholar Co-authors

Table 3: Top Scholars by Closeness Centrality

Table 4: Top Scholars by Betweenness Centrality

Table 5: Co-authorship by School

  1. See infra Figure 1.
  2. See infra Part I.
  3. Lorenzo Ductor, Sanjeev Goyal & Anja Prummer, Gender and Collaboration, Rev. Econ. & Statistics 24–25 (2021), https://doi.org/10.1162/rest_a_01113 [https://perma.cc/PS4Z-KXAK].
  4. In some cases, law school texts are based on methodological approaches that have been developed as legal scholarship. See, e.g., Robert Cooter & Thomas Ulen, Introduction to Law and Economics (5th ed. 2007).
  5. Tracey E. George & Chris Guthrie, Joining Forces: The Role of Collaboration in the Development of Legal Thought, 52 J. Legal Educ. 559, 561–568 (2002); Michael I. Meyerson, Law School Culture and the Lost Art of Collaboration: Why Don’t Law Professors Play Well with Others?, 93 Neb. L. Rev. 547, 548–549 (2014).
  6. Tom Ginsburg & Thomas J. Miles, Empiricism and the Rising Incidence of Coauthorship in Law, 2011 U. Ill. L. Rev. 1785, 1800–1812 (2011).
  7. Paul H. Edelman & Tracey E. George, Six Degrees of Cass Sunstein, 11 Green Bag 2d 19, 22–31 (2007).
  8. Id. at 27–30.
  9. See, e.g., Daniel Martin Katz, Joshua R. Gubler, Jon Zelner, Michael J. Bommarito II, Eric Provins, & Eitan Ingall, Reproduction of Hierarchy? A Social Network Analysis of the American Law Professoriate, 61 J. Legal Educ. 76 (2011) (discussing the network of legal scholars and its impact on legal developments); Milan Markovic, The Law Professor Pipeline, 92 Temp. L. Rev. 813 (2019) (exploring how professors’ intellectual and social networks relate to the advancement of their careers); Keerthana Nunna, W. Nicholson Price II & Jonathan Tietz, Hierarchy, Race & Gender in Legal Scholarly Networks, 75 Stan. L. Rev. (forthcoming 2023) (analyzing how hierarchy, race, and gender affect the acknowledgment sections of law review articles).
  10. See infra Part IV.
  11. “LGB” is used instead of “LGBT” because the American Association of Law Schools (AALS) Directory of Law Teachers does not include data on transgender faculty.
  12. Nunna et al., supra note 9.
  13. Id. at 46.
  14. On the importance of networks in job matching, see Yannis M. Ioannides & Linda Datcher Loury, Job Information Networks, Neighborhood Effects, and Inequality, 42 J. Econ. Literature 1056, 1061–1062 (2004).
  15. Ginsburg & Miles, supra note 6, at 1788–90, 1794. For a theoretical model of the co-authorship choice, see Bruna Bruno, Economics of Co-authorship, 44 Econ. Analysis & Pol’y 212 (2014).
  16. Id.
  17. Heather Sarsons, Recognition for Group Work: Gender Differences in Academia, 107 Am. Econ. Rev. 141, 144 (2017).
  18. Ginsburg & Miles, supra note 6, at 1790–93.
  19. Matthew O. Jackson, Social and Economic Networks 20 (2010).
  20. Id. at 32.
  21. Id.
  22. Id. at 26.
  23. Id. at 32.
  24. Id. at 33.
  25. Id. at 29.
  26. Mark E.J. Newman, Mixing Patterns in Networks, 67 Physical Rev. 2 (2003).
  27. Jackson, supra note 18, at 35.
  28. Id.
  29. For racial minorities, I used lists published in the 1986, 1989, 1990, 1991, 1992, 1996, 2000, 2004, 2007, 2011, 2014, and 2019 editions. For sexual minorities, I used lists from the same years, beginning in 1996.
  30. Ginsburg & Miles, supra note 6, at 1785.
  31. Id. at 1785 (“[R]esults support the view that specialization, and specifically the growth of empirical scholarship, has contributed to the trend of coauthorship in legal academia.”).
  32. Mark C. Suchman & Elizabeth Mertz, Toward a New Legal Empiricism: Empirical Legal Studies and New Legal Realism, 6 Ann. Rev. L. & Soc. Sci. 555, 556 (2010) (“Since the mid-1990s, several groups of scholars have championed a renewed dialog between law and social science.”).
  33. Journal of Empirical Legal Studies, Wiley Online Libr., https://onlinelibrary.wiley.com/loi/17401461 [https://perma.cc/6GXQ-LCEK] (last visited Oct. 28, 2022).
  34. See Ginsburg and Miles, supra note 6, at 1785 (“Coauthored articles were far more common in . . . [the Journal of Legal Studies and Journal of Law, Economics and Organization] . . . than in the general interest, student-edited law reviews.”).
  35. For details about the calculation of these centrality measures, see Jackson, supra note 19, at 39.
  36. The data come from Jackson, supra note 19.
  37. Meyerson, supra note 5, at 548.
  38. Andy H. Barnett, Richard W. Ault & David L. Kaserman, The Rising Incidence of Co-Authorship in Economics: Further Evidence, Rev. of Econ. & Stat. 539 (1988).
  39. Specifically, the average shortest path should be of the same order as .
  40. In this I follow Sanjeev Goyal, Marco J. van der Leij & José Luis Moraga‐González, Economics: An Emerging Small World, 114 J. Pol. Econ. 403 (2006). Goyal et al. builds on the seminal analysis of Duncan J. Watts & Steven H. Strogatz, Collective dynamics of ‘small-world’ networks, 393 Nature 440 (1998).
  41. Goyal et al., supra note 39, at 408.
  42. Id. at 405.
  43. See, e.g., Mu‐Hsuan Huang, Ling‐Ling Wu & Yi‐Chen Wu, A Study of Research Collaboration in the Pre‐web and Post‐web Stages: A Coauthorship Analysis of the Information Systems Discipline, 66 J. Ass’n for Info. Sci. & Tech. 778 (2015).
  44. See Sarsons, supra note 17; Andrew Hussey, Sheena Murray & Wendy Stock, Gender, Coauthorship, and Academic Outcomes in Economics, 60 Econ. Inquiry 465 (2022).
  45. Sarsons, supra note 17.
  46. Boris Maciejovsky, David V. Budescu & Dan Ariely, Research Note—The Researcher as a Consumer of Scientific Publications: How Do Name-Ordering Conventions Affect Inferences About Contribution Credits?, 28 Mktg. Sci. 589 (2008).
  47. Arturo Casadevall, Gregg L. Semenza, Sarah Jackson, Gordon Tomaselli & Rexford S. Ahima, Reducing Bias: Accounting for the Order of Co–first Authors, 129 J. Clinical Investigation 2167 (2019).

A Silver Lining to Russia’s Sanctions-Busting Clause?

In 2018, Russia began inserting an unusual clause into euro and dollar sovereign bonds, seemingly designed to circumvent future Western sanctions. The clause worked by letting the government pay in roubles if sanctions cut off access to dollar and euro payment systems. The clause received little scrutiny at the time, perhaps because Russia used a state-owned bank, rather than a global investment bank, as underwriter. But with the invasion of Ukraine and the ensuing sanctions imposed by the United States and other governments, the relevance of the clause has become clear. This Essay examines how the market reacted to the clause before and after the invasion. Our expectation was that the market would charge a premium for bonds with the clause. Investors bought euro and dollar bonds, after all, because they did not want to be paid in roubles. Yet contrary to expectations, investors seemed to prefer bonds that allowed for payment in roubles over bonds that did not. This surprising finding has considerable implications for other countries that may lose access to foreign currency for reasons that are more benign than Russia’s war of aggression. Despite its sordid provenance, Russia’s sanctions-busting clause might turn out to be a positive innovation that could benefit countries facing unexpected crises. Indeed, had Ukraine included such a clause in its bonds, the benefit would have been enormous.

Introduction

After the seizure of Crimea in 2014, the United States and other governments imposed economic sanctions on a range of Russian entities and individuals. Russia’s subsequent invasion of Ukraine in late February 2022 prompted draconian new sanctions, ranging from trade restrictions to a block on the Russian central bank’s ability to access foreign currency reserves.1.For a comprehensive timeline of sanctions imposed against Russia by the United States and other countries, see Chad P. Bown, Russia’s War on Ukraine: A Sanctions Timeline, Peterson Inst. for Int’l Econ., https://www.piie.com/blogs/realtime-economic-issues-watch/russias-war-ukraine-sanctions-timeline [https://perma.cc/MEL7-E7QH].Show More Although unprecedented in scope, it was no surprise that the United States resorted to sanctions to punish a foreign adversary. Countries have long used economic sanctions to achieve foreign policy objectives.2.Benjamin Coates, A Century of Sanctions, Current Events in Historical Perspective, Ohio St. U.: Origins (Dec. 2019), https://origins.osu.edu/article/economic-sanctions-history-trump-global?language_content_entity=en [https://perma.cc/B94U-3X9F].Show More The global dominance of the U.S. payments system, and the dollar’s dominant place among currencies, has made this a tempting practice for the United States.3.See, e.g., America’s Aggressive Use of Sanctions Endangers the Dollar’s Reign, Economist (Jan. 18, 2020), https://www.economist.com/briefing/2020/01/18/americas-aggressive-use-of-sanctions-endangers-the-dollars-reign [https://perma.cc/K2HY-SG4W]; Karen Yeung, How the US Uses the Dollar Payments System to Impose Sanctions on a Global Scale, S. China Morning Post (Aug. 25, 2020), https://www.scmp.com/economy/china-economy/article/3098691/how-us-uses-dollar-payments-system-impose-sanctions-global [https://perma.cc/M5J2-MLXA].Show More Beyond Russia, the U.S. government has used sanctions to target North Korea, Russia, Syria, Venezuela, China, and other countries, as well as institutions and individuals around the world.4.For background on the use of sanctions by the United States, see generally Sheelah Kalhatkar, Will Sanctions Against Russia End the War in Ukraine? New Yorker (Oct. 24, 2022) (documenting the progression of U.S. sanctions against Russia); Nicholas Mulder, The Economic Weapon: The Rise of Sanctions as a Tool of Modern War (2022) (tracing the international development of sanctions through World War II and their shifting scope and purpose).Show More

Foreign states naturally do not relish the prospect of being targeted by sanctions and have looked for ways to eliminate this risk.5.E.g., Frank Tang, China Warned to Prepare For Being Cut Off From US Dollar Payment System as Part of Sanctions Like Russia, S. China Morning Post (June 22, 2020), https://www.scmp.com/economy/china-economy/article/3090119/china-warned-prepare-being-cut-us-dollar-payment-system-part [https://perma.cc/YCC3-AABL]; Karen Brettell, Analysis: As Sanctions “Weaponize” US Dollar, Some Treasury Buyers Could Fall Back, Reuters (Mar. 29, 2022) (describing pressure on foreign governments to diversify their currency use the more the dollar is used as a financial weapon), https://www.reuters.com/business/finance/sanctions-weaponize-us-dollar-some-treasury-buyers-could-fall-back-2022-03-29/ [https://perma.cc/B8A5-HD7U].Show More One option is to reduce reliance on the U.S. dollar, although this requires trade and financial partners willing to deal in other currencies.6.This option inevitably prompts inquiry into whether sanctions might lead to the decline of the dollar as global reserve currency. See, e.g., Michael P. Dooley, David Folkerts-Landau & Peter M. Garber, US Sanctions Reinforce the Dollar’s Dominance (Nat’l Bureau Econ. Rsch., Working Paper No. 29943, 2022), https://www.nber.org/system/files/working_papers/w2994​3/w29943.pdf [https://perma.cc/MGW2-DTU9].Show More More commonly, foreign parties seek ways to bypass or evade U.S. sanctions.7.For a discussion of some of the responses to U.S. financial sanctions, see Pierre-Hugues Verdier, Global Banks on Trial: U.S. Prosecutions and the Remaking of International Finance 109–45 (Oxford Univ. Press 2020).Show More In this Essay, we examine one such strategy, which the Russian Federation adopted after the sanctions prompted by its annexation of Crimea. The Russian government appears to have worried—presciently, as it turns out—that future sanctions would force it to default on its international debt. It sought to prevent this by clever contract drafting.

In 2016, the Russian government was already sanctioned and anticipated that its future conduct might trigger even harsher sanctions. It began to include an “Alternate Payment Currency” (APC) clause in its international bonds—that is, bonds denominated in foreign currency.8.See Lev E. Breydo, Russia’s Bond Roulette, Am. Bar Ass’n (May 20, 2022), https://www.americanbar.org/groups/business_law/publications/blt/2022/06/russia-bond-roulette/ [https://perma.cc/5TEZ-ZGJ9].Show More The clause lets the government pay in an alternative currency if, “for reasons beyond its control,” it cannot pay in the currency specified in the bond. The APC clause first appeared in a U.S. Dollar (USD)-denominated bond and specified alternative currencies over which the Russian government had no control: Euros, Pound sterling, and Swiss francs.9.Prospectus for Russian Federation 4.75% USD bonds due 2026, at 3 (May 26, 2016), https://ise-prodnr-eu-west-1-data-integration.s3-eu-west-1.amazonaws.com/legacy/Prospectu​s+-+Standalone_7cc3442b-cb8c-4394-bc26-33f9ce656e56.pdf [https://perma.cc/YDY6-8A​NP].Show More But, since March 2018, the country’s international bonds have included Russia’s rouble in the list of alternatives. It is the last option; the country must pay in foreign currency if possible. But if not, it can pay in roubles. The meaning of this clause has not been tested in court. At least arguably, the clause lets the Russian government avoid a debt default by paying bondholders in roubles when financial sanctions shut it off from the dollar payment system.

The APC clause is exceptional for multiple reasons. For one thing, we cannot think of other cases in which a country’s sovereign bonds have included an explicit sanctions-busting mechanism. For another, the clause functions as a sort of force majeure clause. In its classic sense, a force majeure clause temporarily excuses a party’s non-performance of its contractual obligations when circumstances outside its control make performance impossible.10 10.Anthony Michael Sabino, The Force Majeure Awakens, N.Y. L.J. (May 16, 2022), https://www.law.com/newyorklawjournal/2022/05/16/the-force-majeure-awakens/ [https://perma.cc/3KBJ-QYRX]. These clauses received considerable attention as a result of the COVID-19 pandemic, since there was considerable debate as to whether the pandemic was a force majeure event. See e.g., Andrew A. Schwartz, Contracts and COVID-19, 73 Stan. L. Rev. 48, 56–58 (2020).Show More But sovereign bonds generally do not include force majeure clauses; the sovereign’s payment obligation is unconditional. In sovereign debt markets, the closest thing to a force majeure clause is the so-called Natural Disaster clause, which allows the issuer to defer payments in the event of a qualifying natural disaster.11 11.Sui-Jim Ho & Stephanie Fontana, Sovereign Debt Evolution: The Natural Disaster Clause, 11 Emerging Mkts. Restructuring J. 5, 5 (2021).Show More Rather than provide for a suspension of payments, the APC clause allows the debtor to use its domestic currency when it cannot access foreign currency.

The formulation of the APC clause as an option to pay in domestic currency has implications beyond the sanctions context. The clause applies to all circumstances where the sovereign cannot access foreign currency for reasons beyond its control. Historically, a loss of such access is a primary reason why sovereigns encounter financial distress.12 12.Barry Eichengreen, Asmaa El-Ganainy, Rui Pedro Esteves & Kris James Mitchener, Public Debt Through the Ages, in Sovereign Debt 7, 33–36 (S. Ali Abbas ed., 2019).Show More The APC clause thus provides a contractual escape hatch from what economists refer to as the problem of “original sin.”13 13.For the classic work on this topic, see Barry Eichengreen, Ricardo Hausmann & Ugo Panizza, The Mystery of Original Sin, in Other People’s Money: Debt Denomination and Financial Instability in Emerging Market Economies (Barry Eichengreen & Ricardo Hausmann eds., 2005); Ricardo Hausmann & Ugo Panizza, Redemption or Abstinence? Original Sin, Currency Mismatches and Counter Cyclical Policies in the New Millennium, 2 J. Globalization & Dev. 1 (2011).Show More A country that cannot borrow abroad in its own currency exposes itself to exchange rate volatility and other risks that can undermine economic stability. High interest rates or weak commodity prices can prompt a debt crisis, which can be avoided if the borrower can temporarily resort to payment in domestic currency. Emerging market borrowers have gradually shifted more of their borrowing into domestic currency, but many have substantial debts denominated in foreign currencies. It may be that Russia, despite causing global chaos and a humanitarian disaster in Ukraine, has inadvertently created a contractual innovation that could benefit other borrowers.

However, the benefit of the APC clause depends on its price. One reason countries borrow in foreign currency is that it is cheaper to do so. A key question, then, is what the pricing implications of using the APC clause have been. How much did the markets charge Russia for the clause? And, once events occurred that brought the clause into play, how did the markets react? In this Essay, we use data on Russian bond prices to analyze these questions.

Our prediction was that the market would charge Russia a premium for using the APC clause. This is because investors who buy international bonds generally do not wish to be repaid in the borrower’s domestic currency, and also because investors might interpret the mere presence of the clause as a signal that Russia anticipated engaging in conduct that would prompt additional sanctions. Conceivably, this premium would not be apparent at the time of issuance. The financial press covered the introduction of the rouble APC clause in 2018, so it could hardly have escaped investors’ notice.14 14.E.g., Jonathan Wheatley, Russia Bond Sales Allow Payments in Alternative Currencies, Fin. Times (Mar. 17, 2018), https://www.ft.com/content/69da000c-2915-11e8-b27e-cc62a39d57a0 [https://perma.cc/UVN8-8Y9J].Show More At the time, however, investors may have been sanguine about the prospect of future sanctions. Certainly, Russia was viewed as a strong creditor, and the bonds were heavily oversubscribed. But during late 2021 and early 2022, the prospect of an invasion of Ukraine became increasingly real. We expected that the market penalty for APC bonds would increase once the invasion (and, thus, additional sanctions) grew increasingly certain.

Markets did not react as we expected. Initially, the market seemed indifferent to the clause, even as Russian forces massed on the Ukrainian border. Once the invasion occurred, the market seemed to wake up to the presence of the APC clause. But rather than charge a premium for APC bonds, as we had expected, the market seemed to value them more highly than comparable bonds without the clause. Although there are dynamics unique to Russian sovereign debt, these findings also suggest that, despite its provenance, the Russian APC clause might be a positive innovation from which other countries facing the risk of unexpected crises could benefit. Indeed, Ukraine itself, which likely faces an unpayable debt burden due to the Russian invasion, would have realized an enormous benefit from such a clause.

I. Background on the APC Clause

The core feature of the APC clause is that it allows the issuer of a dollar or euro bond to pay in roubles if neither the designated currency nor a list of stable alternative currencies is available, and the reason is outside the control of the issuer. As an example, here is the relevant text of the APC clause in a 2019 issuance of dollar bonds by the Russian Federation:

Notwithstanding any other provision in these Conditions, if, for reasons beyond its control, the Russian Federation is unable to make payments of principal or interest (in whole or in part) in respect of the Bonds in U.S. dollars (an “Alternative Payment Currency Event”), the Russian Federation shall make such payments (in whole or in part) in the Alternative Payment Currency on the due date at the Alternative Payment Currency Equivalent of any such U.S. dollar-denominated amount…

….

“Alternative Payment Currency” means Euros, Pound sterling or Swiss francs or, if for reasons beyond its control the Russian Federation is unable to make payments of principal or interest (in whole or in part) in respect of the Bonds in any of these currencies, Russian roubles.15 15.Prospectus for Russian Federation 5.10% USD bonds due 2035, at 66–68 (Mar. 26, 2019), [https://perma.cc/3YBJ-8MYV].Show More

Several aspects of the clause are noteworthy. First, of course, is that it lets the government pay in roubles under circumstances in which, but for the APC clause, it would likely be forced into a payment default. Second, the right to pay in roubles is triggered only if the government cannot pay in a preferred currency “for reasons beyond its control.” The contract does not say when circumstances are outside the Russian government’s control, but the clause is broad enough to cover the types of circumstances that get many emerging markets into trouble (for example, currency outflows that create a liquidity crisis). As events have turned out, Russia has not attempted to invoke the clause, and it is an open question whether the present circumstances are outside its control.16 16.The Russian government has indicated that it intends to pay all of its international bonds in roubles. The Associated Press, Russia Says it Will Pay Foreign Debt in Rubles After U.S. Ban, Associated Press News (May 25, 2022), https://apnews.com/article/russia-ukraine-janet-yellen-government-and-politics-65af86cf89023fbc213c54ab11f212ee [https://perma.cc/29H3-2G54]. This is a default for non-APC bonds and at least a technical default under the APC bonds, since the government has not complied with notice and other procedures applicable to rouble payments. But these are relatively recent developments.Show More Western sanctions are the proximate cause of Russia’s inability to make dollar and euro payments, and Russia cannot directly control the sanctions. But it could presumably entice Western governments to remove them by withdrawing from Ukraine and credibly committing to respect the country’s borders. The question would have to be decided by a court applying English law (the law designated in the contract).

However, it is not a stretch to interpret the APC clause to allow payment in roubles. It does seem that the intent of the clause was to address the scenario in which sanctions cut off access to dollar and euro payment systems. The prospectuses for the APC bond issuances explicitly point to the risk that Western sanctions might jeopardize Russia’s ability to pay in foreign currency.17 17.See Mark Weidemaier & Mitu Gulati, Should Investors Who Care About ESG Buy Russian Sovereign Bonds?, Credit Slips (Mar. 15, 2022), https://www.creditslips.org/credit​slips/2022/03/should-investors-who-care-about-esg-buy-russian-sovereign-bonds.html [https​://perma.cc/EN8T-4JF6].Show More For example, the prospectus for sovereign bonds issued in 2020 details the history of Western sanctions, warns that “continued geopolitical tensions” and new sanctions might put downward pressure on the rouble, and cautions that this might “adversely affect … the Russian Federation’s ability to repay its debt denominated in currencies other than the rouble, including amounts due under the Bonds.”18 18.Prospectus for EUR 750 million 1.125% bonds due 2027, at 17 (Nov. 18, 2020), https://www.creditslips.org/files/nov-2020-prospectus.pdf [https://perma.cc/W6JB-TKRF].Show More Courts generally interpret contracts in a manner consistent with the parties’ presumed intent. This language implies that the intent of the APC clause is to allow payment in roubles when sanctions cut off access to other currencies.19 19.E.g., Guy Faulconbridge & Karin Strohecker, Russia Warns Sovereign Bond Holders That Payments Depend on Sanctions, Reuters (Mar. 6, 2022), https://www.reuters.​com/markets/rates-bonds/russia-says-sovereign-bond-payments-will-depend-sanctions-2022-03-06/ [https://perma.cc/B5T8-4JEY] (noting that the APC was put into bonds in the wake of western sanctions that were imposed as a result of Russia’s Crimean incursion in 2014); Wheatley, supra note 14 (same).Show More Moreover, Gazprom, Russia’s state-owned energy company, also has issued APC bonds. Gazprom has an even better argument for paying in roubles, since its conduct did not prompt the sanctions. We do not think the issue is clear cut for either bond issuer. Bondholder lawsuits probably would be heard by courts in London, and the United Kingdom is one of the sanctioning countries. An English court might be unwilling to let Russian parties continue to pay bondholders in roubles.20 20.For a discussion of this rather unclear basis on which courts across different legal settings periodically invalidate contracts, see Farshad Ghodoosi, The Concept of Public Policy in Law, 94 Neb. L. Rev. 685, 711 (2016); Percy H. Winfield, Public Policy in the English Common Law, 42 Harv. L. Rev. 76, 88, 93 (1928) (discussing differences among judges as to when public policy should void a contract).Show More But it is at least plausible that a court would interpret the clause to allow rouble payments.

So interpreted, the APC clause establishes a mechanism for evading sanctions and, at least potentially, avoiding a payment default. That is a clear benefit to Russia. And while investors who receive rouble payments might worry that the Russian government will manipulate exchange rates or impose capital controls, these are not certainties given the government’s apparent determination to maintain good relations with the market.21 21.The bond contracts are not entirely clear, but they appear to give the Russian government room to manipulate the exchange rate by allowing the NSD, an arm of the Russian government acting as its central securities depository, to determine the exchange rate by reference to the rates quoted by banks in Moscow.Show More

II. Empirical Analysis

We base our analysis on a dataset of sovereign and Gazprom bond issuances as reported on Bloomberg. Overall, we have data on 28 issuances. For simplicity, we focus the empirical analysis on bond pairs, although we have replicated our analysis using bond portfolios. We selected eight bonds with similar maturities to study the impact of the APC clause on the yield spread: two sovereign bonds in roubles, two sovereign bonds in U.S. dollars, two Gazprom bonds in U.S. dollars, and two Gazprom bonds in euros.

Figure 1 shows the daily yield to maturity for the pair of sovereign bonds issued in U.S. dollars.22 22.The pair consists of USD bonds with relatively closely matched maturities: June 23, 2027, for the non-APC bond (“no clause”) and March 21, 2029, for the APC bond (“clause”). Figure 1: Panel A shows the yield to maturity from January 1, 2019, to December 31, 2021; Panel B from January 1, 2022, to September 21, 2022, and Panel C from January 1, 2019, to September 21, 2022.Show More First, we observe a big jump in the yield for both bonds on the date of the Russian invasion (February 24, 2022). Second, from the beginning of 2019 through the date of the invasion, the yield on the two bonds was quite close. On average, over the pre-invasion time period, the yield of the bond with the APC clause is slightly higher than the bond without the clause, but the difference is not statistically significant. After the invasion, the relationship flips and increases considerably in both magnitude and significance. This implies that the market perceived APC bonds as less risky than non-APC bonds. This is the opposite of what we expected.

FIGURE 1. Yield to maturity, percentage, USD sovereign pair.

Panel A. January 1, 2019–December 31, 2021. Panel B. January 1, 2022–September 21, 2022.
                                                                                    Panel C. January 1, 2019–September 21, 2022.

We also analyze pairs of dollar- and euro-denominated Gazprom bonds and observe the same dynamics. Indeed, the pattern is even clearer.23 23.The Gazprom USD pair consists of a non-APC bond maturing March 23, 2027, and an APC bond maturing February 25, 2030. The EUR pair includes a non-APC bond maturing November 17, 2023, and an APC bond maturing April 15, 2025.Show More As shown in Figures 2 and 3, there is basically no pricing difference between APC and non-APC bonds until the invasion. 24 24.Figures 2 and 3: Panel A shows the yield to maturity from January 1, 2019, to December 31, 2021; Panel B from January 1, 2022, to September 21, 2022, and Panel C from January 1, 2019, to September 21, 2022.Show More After that, yields spike for both bonds but dramatically more for the non-APC bond. Again, post-invasion investors seem to view APC bonds as less risky than comparable bonds without the clause.

As a contrast, the pair of sovereign bonds issued in roubles shows little change due to the conflict (Figure 4).25 25.Figure 4: Panel A shows the yield to maturity from January 1, 2019, to December 31, 2021; Panel B from January 1, 2022, to September 21, 2022, and Panel C from January 1, 2019, to September 21, 2022.Show More Neither bond shows an abrupt jump, suggesting that investors with rouble bonds were confident of continuing to get paid even after Russia invaded Ukraine and was subject to severe western sanctions.

FIGURE 2. Yield to maturity, percentage, USD Gazprom pair.

Panel A. January 1, 2019–December 31, 2021. Panel B. January 1, 2022–September 21, 2022.
Panel C. January 1, 2019–September 21, 2022.

Continue reading “A Silver Lining to Russia’s Sanctions-Busting Clause?”

Expedient Imprisonment: How Federal Supervised Release Sentences Violate the Constitution

[T]he Bill of Rights which we guard so jealously and the procedures it guarantees are not to be abrogated merely because a guilty man may escape prosecution or for any other expedient reason.

– Kennedy v. Mendoza-Martinez, 372 U.S. 144, 184 (1963)

Introduction

Each year, more than ten thousand people are imprisoned by federal courts without being charged with a crime, indicted by a grand jury, or found guilty beyond a reasonable doubt by a jury of their peers. Those results are authorized by federal statute, federal rule, and multiple appellate court decisions. The proffered justification: those defendants were convicted of a federal crime and, after serving their full term of imprisonment, failed to comply with a condition of their term of supervised release.

In this Essay, we argue that the federal supervised release system violates the fundamental constitutional protections guaranteed in the Fifth and Sixth Amendments. We begin with a brief history of supervised release, created in 1984 to replace federal parole. Like a defendant granted release on parole, one on supervised release is required to comply with certain conditions set by the court. Unlike parole, where a defendant could be released early from prison in exchange for a promise to abide by those conditions, supervised release does not replace time spent in prison. Best described as post-confinement surveillance, a term of supervised release can take effect only after a defendant has served the entire prison term imposed by a sentencing judge. During that term of supervised release, a defendant accused of failing to comply with any of those conditions may be arrested and sentenced to serve up to five years in federal prison—all without the constitutional protections ordinarily afforded those facing incarceration.

Following an overview of the structure of supervised release, we consider decisions of federal courts, which have repeatedly held that the protections of the Sixth Amendment do not protect defendants serving terms of supervised release and facing imprisonment. We argue that those decisions start from the wrong end of the constitutional analysis by failing to account for the role of the right to indictment enshrined in the Fifth Amendment. More specifically, a defendant accused of violating a condition of supervised release nearly always faces more than a year of imprisonment if convicted—an infamous punishment for purposes of the Fifth Amendment. As a result, we contend that the right to indictment applies to revocation proceedings, and subsequently, the Sixth Amendment rights applicable in a traditional criminal prosecution must also apply. Furthermore, we argue that there is no meaningful distinction, for Sixth Amendment purposes, between a revocation proceeding and an ordinary prosecution, and that decisions concluding otherwise rely on unsupportable legal fictions.

Ultimately, we conclude that the statutory scheme governing supervised release creates an unconstitutional workaround to the fundamental constitutional protections designed to limit the government’s power to arbitrarily imprison. Although dodging the demands of the Fifth and Sixth Amendments surely offers a more expedient route to imprisonment, prosecutors and probation officers should not have the discretion to choose a route to imprisonment that bypasses the Constitution.

I. The History

We begin with an overview of federal parole, the system supervised release was created to replace. We also touch on federal probation, a form of community supervision similar to parole, before turning to the legislative history of the Sentencing Reform Act of 1984 (“SRA”), the statute that created supervised release.1.Sentencing Reform Act of 1984, Pub. L. No. 98-473, tit. II, ch. 2, 98 Stat. 1987 (codified as amended in scattered sections of 18 and 28 U.S.C.); Fiona Doherty, Indeterminate Sentencing Returns: The Invention of Supervised Release, 88 N.Y.U. L. Rev. 958, 996–97 (2013). Professor Doherty’s seminal piece provides a thorough overview of the story of supervised release.Show More

A. Parole

Congress enacted the first parole statute in 1910, modeling the federal system after state parole systems that had been on the books in the majority of states since the late 1800s.2.Id.at 983–84.Show More That statute authorized the early release of federal prisoners who had served at least one third of the prison term imposed by the sentencing judge.3.Id.at 984.Show More The decision whether to grant or deny release on parole was committed to the discretion of a federal parole board,4.Id. at 985;see alsoPeter B. Hoffman, History of the Federal Parole System: Part 1 (1910–1972), 61 Fed. Prob. 23, 27 (1997) (explaining that parole decisions were made by the whole board).Show More and the scope of that discretion was broad: the board could grant parole to any individual who had “observed the rules of [the] institution” in which he was incarcerated so long as release was not “incompatible with the welfare of society.”5.Parole Act, ch. 387, §§ 1, 3, 36 Stat. 819 (1910), repealed by Sentencing Reform Act of 1984, § 212(a)(2).Show More Although allowed to leave prison, a parolee was still considered to be “in the legal custody and under the control of the warden of such prison from which paroled . . . until the expiration of the [prison] term or terms specified in his sentence.”6.Id.§ 3.Show More In other words, a grant of parole allowed a term of incarceration to be served in the community but did not affect the length of that sentence.

A suspected violation of any parole condition could be punished swiftly and summarily: if any member of a board had “reliable information that the offender . . . violated his parole,” the board could issue an arrest warrant and require the parolee to be returned to prison.7.Hoffman, supra note 4, at 28.Show More Only then would the parolee be afforded an opportunity to be heard before the board,8.Id.Show More which could decide either to alter parole conditions or to revoke the grant of parole entirely, sending the parolee back to prison to serve the “remainder of the sentence originally imposed.”9.Id.Show More

In the early days of the federal parole system, courts afforded only limited oversight of the board’s decisions to grant or revoke parole.10 10.E.g., Clark v. Stevens, 291 F.2d 388, 389 (6th Cir. 1961); Hiatt v. Compagna, 178 F.2d 42, 45 (5th Cir. 1949) (“A prisoner sentenced to the penitentiary is turned over by the court to the executive, namely the Attorney General, to have the sentence executed. . . . It is very evident that the whole matter of paroles is left to the informed discretion of the Board. Court action is not made a part of it.”).Show More Federal courts set forth a variety of justifications for declining to overturn the board’s decisions: some concluded that a parolee remained in the formal custody of the executive branch,11 11.Comment, The Parole System, 120 U. Pa. L. Rev. 282, 287–88 (1971); see also Anderson v. Corall, 263 U.S. 193, 196 (1923) (“While on parole the convict is bound to remain in the legal custody and under the control of the warden until the expiration of the term, less allowance, if any, for good conduct. While this is an amelioration of punishment, it is in legal effect imprisonment.”); Padilla v. Lynch, 398 F.2d 481, 482 (9th Cir. 1968) (stating that, because parole is a form of custody, a complaint alleging that the denial of parole violated the equal protection clause is insufficient).Show More while others suggested that parole was an act of grace by a “merciful executive” that could not be demanded as a matter of right.12 12.Comment, supra note 11,at 286–87; cf. Ughbanks v. Armstrong, 208 U.S. 481, 488 (1908) (describing a state’s grant of parole as “the granting of a favor to a convicted criminal confined within one of its prisons”).Show More Finally, some courts rejected constitutional challenges to the parole system by reasoning that the parole board’s relationship with a parolee was not adversarial in nature, given the “genuine identity of interest . . . in the prisoner’s desire to be released and the [b]oard’s policy to grant release as soon as possible.”13 13.Hyser v. Reed, 318 F.2d 225, 237 (D.C. Cir. 1963) (rejecting Sixth Amendment challenge to parole revocation procedures); see also Comment, supra note 11, at 288–89 (stating that the board and the parolee have a shared interest in rehabilitating the parolee).Show More Because the relationship between the parolee and the board was not adversarial, the reasoning went, there was simply no need for courts to intervene to safeguard the rights of the parolee.

In 1972, however, the Supreme Court granted certiorari in Morrissey v. Brewer to consider whether revocation of parole without a hearing violated the constitutional requirements of due process.14 14.408 U.S. 471, 472 (1972).Show More Noting that a parolee was technically still “in custody” and could therefore claim no entitlement to release on parole, lower courts had answered that question in the negative.15 15.Id. at 474–75.Show More The Supreme Court, however, disagreed. Relevant for our purposes, the Court began by explaining that because “the revocation of parole is not part of a criminal prosecution,” the “full panoply” of constitutional protections applicable in a criminal prosecution did not apply to a parole revocation proceeding.16 16.Id. at 480.Show More No real analysis was necessary to reach that conclusion: after all, parole “ar[ose] after the end of the criminal prosecution, including the imposition of sentence”17 17.Id.Show More and involved merely a decision about where the balance of that sentence would be served. Revocation of parole deprived an individual of only the “conditional liberty” granted in connection with early release from prison.18 18.Id.Show More Accordingly, it would make little sense to require the state to initiate a new prosecution simply to require the parolee to serve out the remainder of the existing sentence.19 19.Id.at 483; see alsoJacob Schuman, Supervised Release Is Not Parole, 53 Loy. L.A. L. Rev. 587, 624 (2020).Show More

Although a revocation proceeding differed, for constitutional purposes, from a new prosecution, the Court acknowledged that parole revocation nevertheless inflicted a “grievous loss.”20 20. Morrissey, 408 U.S.at 482.Show More As a result, at least some due process protections were required before parole could be revoked.21 21.Id.Show More Specifically, due process mandated:

(a) written notice of the claimed violations of parole; (b) disclosure to the parolee of evidence against him; (c) opportunity to be heard in person and to present witnesses and documentary evidence; (d) the right to confront and cross-examine adverse witnesses (unless the hearing officer specifically finds good cause for not allowing confrontation); (e) a “neutral and detached” hearing body such as a traditional parole board, members of which need not be judicial officers or lawyers; and (f) a written statement by the factfinders as to the evidence relied on and reasons for revoking parole.22 22.Id.at 489.Show More

Those requirements would eventually be incorporated into Federal Rule of Criminal Procedure 32.1, later made applicable to probation revocation and supervised release proceedings.23 23.Fed. R. Crim. P. 32.1 advisory committee’s note to 1979 and 1989 amendments.Show More

B. Probation

Federal probation has a similar origin story. Although the first federal probation statute was not formally enacted until 1925, federal judges, relying on their broad sentencing discretion, would routinely sentence a defendant to a term of imprisonment and then “suspend[]” execution of that sentence “during the good behavior of the defendant.”24 24.Ex parte United States, 242 U.S. 27, 37 (1916); see also Brent E. Newton, The Story of Federal Probation, 53 Am. Crim. L. Rev. 311, 313 (2016) (noting that federal court judges would suspend prison sentences prior to the passing of the Probation Act).Show More If a defendant failed to comply with restrictions imposed in exchange for granting her leniency in the form of probation, probation could be revoked in favor of imprisonment.25 25.Newton, supra note 24,at 312.Show More As was the case with parole, defendants were afforded little-to-no procedural protections before probation was revoked in favor of prison.

Similar to decisions considering parole revocations, federal courts routinely rejected challenges to the lack of procedural protections available before probation was revoked, on the premise that a grant of probation was a privilege or act of grace afforded by a benevolent court, or suggesting that district courts should have broad discretion over all aspects of probation in order to effectuate the “humane” purpose of probation.26 26.Fiona Doherty, Obey All Laws and Be Good: Probation and the Meaning of Recidivism, 104 Geo. L.J. 291, 330, 335 (2016).Show More

In 1973, shortly after deciding Morrissey, the Supreme Court granted certiorari in Gagnon v. Scarpelli to consider whether “a previously sentenced probationer is entitled to a hearing when his probation is revoked and, if so, whether he is entitled to be represented by appointed counsel at such a hearing.”27 27.411 U.S. 778, 779 (1973).Show More Gagnon, like Morrissey, involved a challenge to a probation revocation that had taken place without a hearing or other procedural protections.28 28.Id.at 780.Show More Citing to a 1967 decision, Mempa v. Rhay, the Court first distinguished a probation revocation hearing from a sentencing proceeding.29 29.Id.at 781.Show More Pursuant to the unique state statutory scheme at issue in Mempa, judges had the authority to defer a sentencing decision in favor of probation; however, if a defendant failed to comply with the conditions of probation, the judge was required to sentence her to the statutory maximum for the original crime of conviction.30 30.Mempa v. Rhay, 389 U.S. 128, 135 (1967).Show More Because that proceeding functioned, in essence, as a sentencing proceeding, the Mempa Court concluded that the Sixth Amendment right to counsel applied.31 31.Id.at 137.Show More

Gagnon, by contrast, involved a different type of revocation proceeding. The petitioner in Gagnon had been formally sentenced to fifteen years’ imprisonment, but the judge had then suspended the sentence in favor a seven-year term of probation.32 32.Gagnon, 411 U.S. at 779.Show More Like revocation of parole, revocation of probation under that scheme merely triggered the execution of the sentence already imposed, and the Gagnon Court therefore concluded that revocation was not a “stage of a criminal prosecution.”33 33.Id.at 782.Show More Further distinguishing the proceeding from a prosecution, the Court noted, was the absence of an adversarial relationship between the state and the defendant. In particular, the state was represented not by a prosecutor, but by a parole officer whose interests were, at least in theory, aligned with those of the probationer.34 34.Id.at 789.Show More Rather than the full panoply of rights afforded in a traditional criminal prosecution, the Court reasoned that only the limited procedural protections identified by the Morrissey Court were required before probation could be revoked.35 35.Id.at 782.Show More

In so holding, the Gagnon Court clarified that whether a revocation proceeding implicates Sixth Amendment rights or merely due process concerns turns on the nature of what the court is deciding. Like in Morrissey, the Court in Gagnon considered a term of imprisonment that had been imposed by a judge but from which the defendant had been granted a reprieve, either in the form of early release from prison or the opportunity to avoid prison altogether. Revocation of that grant of leniency resulted merely in execution of that previously-imposed sentence. Mempa, by contrast, involved something akin to an entirely new sentencing proceeding, to which the Sixth Amendment right to counsel applied.

Probation—and the constitutional mandates set forth in Gagnon—ultimately survived the sweeping changes made by the SRA. Although the SRA directed judges to treat probation as “a form of sentence with conditions rather than as a deferral of imposition or execution of a sentence,”36 36.S. Rep. No. 98-225, at 59 (1983). Arguably, then, Mempa is the only Supreme Court decision that affords constitutional guidance with regard to modern federal probation, which is a sentence in its own right and which requires a judge to engage in a full-blown sentencing proceeding when probation is “revoked.”Show More judges still had broad authority to revoke probation and resentence a defendant for up to the statutory maximum term of imprisonment for the crime of conviction if he or she failed to comply with conditions the judge had imposed in exchange for granting a reprieve from prison.37 37.United States v. Labonte, 520 U.S. 751, 769 (1997) (Breyer, J., dissenting).Show More

C. Supervised Release

By the early 1980s, the federal parole system had come under fire from both sides of the political spectrum as creating unwarranted sentencing disparities among federal offenders and sentences of indeterminate length.38 38.Doherty, supra note 1, at 995.Show More As part of an attempt to comprehensively reform federal sentencing and in response to widespread criticism of parole, the sponsors of the SRA elected to simply abolish federal parole entirely rather than attempt reforms. In lieu of parole, they created supervised release, a form of post-confinement monitoring that followed—rather than replaced—imprisonment.39 39.Sentencing Reform Act of 1984, Pub. L. No. 98-473, tit. II, ch. 2, sec. 212(a)(2), § 3583(a), 98 Stat. 1987, 1999 (codified as amended at 18 U.S.C. § 3583(a)) (“The court, in imposing a sentence to a term of imprisonment for a felony or a misdemeanor, may include as a part of the sentence a requirement that the defendant be placed on a term of supervised release after imprisonment.” (emphasis added)). Imposition of a term of supervised release was required as part of every sentence that imposed imprisonment during the mandatory Sentencing Guidelines era. Since Booker, with few exceptions, imposition of supervised release is discretionary. Nonetheless, it is imposed in the vast majority of federal sentences. SeeChristine S. Scott-Hayward, Shadow Sentencing: The Imposition of Federal Supervised Release, 18 Berkeley J. Crim. L. 180, 186 (2013).Show More As the Senate Report accompanying the SRA explained,

the court, in imposing a term of imprisonment for a felony or a misdemeanor, [may] include as part of the sentence a requirement that the defendant serve a term of supervised release after he has served the term of imprisonment. . . . The term of supervised release would be a separate part of the defendant’s sentence, rather than being the end of the term of imprisonment.40 40.S. Rep. No. 98-225, at 123 (1983) (emphasis added).Show More

As originally designed, supervised release was supposed to afford rehabilitation rather than impose punishment. The punitive portion of a sentence would have been served by the time supervised release commenced, and the primary goal of supervised release was therefore “to ease the defendant’s transition into the community after the service of a long prison term for a particularly serious offense, or to provide rehabilitation . . . .”41 41.Id. at 124.Show More Accordingly, judges were to consider “deterrence and rehabilitation,” but not the need for retribution, in determining whether to sentence a defendant to a term of supervised release and in selecting the conditions to impose.42 42.Doherty, supra note 1, at 998.Show More One important “condition,” however, was mandatory—a judge was required to order that a defendant on supervised release “not commit another Federal, State, or local crime during the term of supervision.”43 43.§ 3583(d), 98 Stat. at 1999.Show More

In line with the rehabilitative purpose of supervised release, the SRA initially provided no mechanism for a judge to “revoke” a term of supervised release. Instead, judges were to “treat a violation of the conditions of supervised release as a criminal contempt.”44 44.Doherty, supra note 1, at 999–1000.Show More In other words, anyone accused of violating a condition of supervised release would be entitled to “trial by jury (for all cases involving a sentence of more than six months), along with all the other procedural protections applicable in a criminal proceeding.”45 45.Id.at 1000.Show More

Just before the SRA went into effect, however, Congress passed the Anti-Drug Abuse Act (“ADAA”) of 1986, a tough-on-crime statute that increased the penalties for certain drug offenses.46 46.Id.Show More Most significantly, the ADAA made what was described as a “technical amendment” to the SRA, authorizing courts to “revoke” a term of supervised release and impose a new term of imprisonment, pursuant to the procedures set forth in Federal Rule of Criminal Procedure 32.1, “if it [found] by a preponderance of the evidence that the person violated a condition of supervised release.”47 47.Id. at 1001; see 18 U.S.C. § 3583(e)(3); Fed. R. Crim. P. 32.1.Show More In so doing, the ADAA jettisoned the carefully crafted procedures originally envisioned by the SRA’s drafters in favor of the revocation mechanism—and limited procedural protections—developed in the context of parole.48 48.Doherty, supra note 1,at 1002.Show More

The sparse legislative history of the revocation provision of the ADAA affords little insight into the intent of its drafters; at best, that history suggests that the amendment was not thoroughly considered.49 49.Id. at 1001.Show More The impact of that amendment, however, was dramatic; passage of the amendment restored a key aspect of the old parole system that the SRA’s drafters had worked so hard to excise from federal sentencing practice.

II. Supervised Release is Constitutionally Different

Citing principally to the similarities between the procedural protections afforded by statute for supervised release revocations and those governing parole revocations, federal courts analyzing the SRA quickly concluded that, at least for constitutional purposes, the two systems were virtually indistinguishable. More specifically, courts reflexively invoked Morrissey and Gagnon to conclude that, like revocation of parole or probation, supervised release violation proceedings were distinguishable from a criminal prosecution to which the protections of the Fifth and Sixth Amendments apply.50 50.Schuman, supra note 19, at 615–16, 615 n.200 (collecting appellate cases).Show More

To be sure, supervised release, parole, and probation share certain similarities—each involves, at one point or another, supervision by a parole or probation officer outside the walls of prison. But we argue that, for constitutional purposes, the differences between parole, probation, and supervised release are more important than their similarities. Most significantly, pursuant to the plain text of the statute governing supervised release, unlike probation or parole, supervised release can take effect only after the full term of imprisonment imposed by a judge has been served and thus cannot afford a reprieve from imprisonment. As a result, when a judge purports to “revoke” 51 51.As United States District Judge Jack Weinstein recognized, even the term “revoke” is a “misnomer” in the context of supervised release. Because supervised release does not replace imprisonment, a judge who “revokes” a term of supervised release is not taking back the freedom she previously bestowed. United States v. Trotter, 321 F. Supp. 3d 337, 346–47 (E.D.N.Y. 2018).Show More a term of supervised release in favor of imprisonment, what she is actually doing is imposing a new and additional term of imprisonment, distinct from the original term of imprisonment imposed as punishment for the underlying crime.52 52.Dissenting in United States v. Haymond, discussed further in Part III, Justice Alito posited that supervised release is equivalent to a sentence of parole:[A] defendant sentenced to x years of imprisonment followed by y years of supervised release is really sentenced to a maximum punishment of x + y years of confinement, with the proviso that any time beyond x years will be excused if the defendant abides by the terms of supervised release.139 S. Ct. 2369, 2390 (2019). But that argument ignores the plain text of the statute, which carefully distinguishes between a term of supervised release and imprisonment. It also ignores the legislative history of the SRA, which makes clear that supervised release was designed to serve a different purpose than imprisonment and was not intended to be interchangeable with imprisonment. Finally, there is no evidence that judges in practice exchange supervised release for imprisonment by imposing shorter prison sentences when they elect to include a term of supervised release as part of a sentence. Empirical evidence instead suggests that many judges routinely impose prison sentences within the range contemplated by the Sentencing Guidelines—which, of course, similarly do not anticipate that shorter prison terms should be imposed where a judge elects to impose a term of supervised release. SeeChristine S. Scott-Hayward, supra note 39, at 186 (2013); U.S. Sentencing Commission Guidelines Manual, ch. 5, introductory cmt., pt. A, sent’g tbl; pt. B, introductory cmt., § 5B1.1 cmt. background (2021).Show More

In our view, that distinction is critically important. Consider the structure of a supervised release revocation proceeding. First, a defendant is formally accused—most often by a United States Attorney or United States Probation Officer—of violating a condition of supervised release. After that formal accusation is lodged with a court, a defendant must either plead guilty to the violation conduct, or instead defend against those accusations at a formal hearing where the government attempts to prove her guilt. If the government succeeds, she faces, in some cases, up to five years in prison for each charge. Because revocation of supervised release involves (1) a formal accusation of wrongdoing, (2) followed by an adversarial proceeding to adjudicate a defendant’s guilt, and (3) leads to a new and additional term of imprisonment, we contend that those proceedings are fundamentally different than those considered by the Morrissey or Gagnon Courts. Instead, supervised release revocation proceedings are—for constitutional purposes—indistinguishable from a criminal prosecution to which the protections embodied in the Fifth and Sixth Amendments always apply.53 53.United States v. Peguero, 34 F.4th 143, 167 (2d Cir. 2022) (Underhill, J., dissenting); see also Rothgery v. Gillespie, 554 U.S. 191, 221–22 (2008) (Thomas, J., dissenting) (discussing elements of a prosecution).Show More

To the extent that courts have considered that argument, they have done so through the lens of the Sixth Amendment. On the premise that a revocation proceeding differs from a prosecution, they have concluded that the full panoply of Sixth Amendment rights does not apply. In our view, those decisions fail to account for the role of the right to indictment guaranteed by the Fifth Amendment. More specifically, in nearly all revocation proceedings, a defendant faces a potential sentence of more than a year of imprisonment if convicted—an “infamous” punishment54 54.Green v. United States, 356 U.S. 165, 183 (1958).Show More within the meaning of the Fifth Amendment. As a result, the right to indictment applies to such proceedings. And if the right to indictment applies, then a revocation proceeding—at which a defendant’s guilt or innocence is adjudicated and that may result in a term of imprisonment—necessarily amounts to a prosecution. In other words, the Sixth Amendment “prosecution” inquiry does not resolve the Fifth Amendment right to indictment issue, but the Fifth Amendment right to indictment inquiry does resolve the Sixth Amendment “prosecution” issue. Furthermore, there are no meaningful differences between “criminal prosecutions” and supervised release violation proceedings for purposes of the Sixth Amendment.

A. The Fifth Amendment

The Fifth Amendment provides that “[n]o person shall be held to answer for a capital, or otherwise infamous crime, unless on a presentment or indictment of a Grand Jury.”55 55.U.S. Const. amend. V.Show More Unlike the Sixth Amendment, which protects the “accused,”56 56.U.S. Const. amend. VI.Show More the Fifth Amendment sweeps more broadly—by its terms, it protects “a person,”57 57.U.S. Const. amend. V.Show More one who has yet to become formally ensnared in the justice system.58 58.Michael J. Zydney Mannheimer, Ripeness of Self-Incrimination Clause Disputes, 95 J. Crim. L. & Criminology 1261, 1321–22 (2005) (noting that the Fifth Amendment protections can apply before a formal prosecution has commenced).Show More Indeed, the right to indictment by a grand jury guards against “an open and public accusation of crime, and from the trouble, expense, and anxiety of a public trial, before . . . probable cause is established.”59 59.Hurtado v. California, 110 U.S. 516, 551–52 (1884) (Harlan, J., dissenting) (cleaned up) (emphasis added).Show More

Significantly, the right to indictment is triggered only when one is “held to answer” for a crime.60 60.U.S. Const. amend. V.Show More Although the historical record on the meaning of that term at the time the Bill of Rights was drafted is relatively sparse,61 61.Elizabeth T. Lear, Is Conviction Irrelevant?, 40 UCLA L. Rev. 1179, 1230 n.246 (1993); Wilkerson v. Whitley, 28 F.3d 498, 502 (5th Cir. 1994).Show More we can conclude at the very least that one is held to answer when he is required to appear and “explain and justify his conduct” in response to the levying of formal charges in a court with the power to punish him.62 62.Ex parte Wall, 107 U.S. 265, 298 (1882); see also Mahon v. Justice, 127 U.S. 700, 713 (1888) (noting that defendant was “held to answer the indictment” after formal charges were served upon him); United States v. McIntosh, 704 F.3d 894, 903–04 (11th Cir. 2013) (quoting Answer, Oxford English Dictionary, http://www.oed.com/view/Entry/8146 (last visited Dec. 21, 2012)) (observing that “answer” means “to speak in reply or opposition to a charge,” a definition recognized when the Fifth Amendment was adopted).Show More

By contrast, the definition of “infamous crime,” as the term is used in the Fifth Amendment, is well-settled. The Supreme Court has explained that infamous punishments include “sentences of imprisonment in a penitentiary and sentences to hard labor. They do not include ordinary misdemeanor sentences of no more than a year in jail.”63 63.United States v. Smith, 982 F.2d 757, 761 (2d Cir. 1992) (cleaned up) (citing, inter alia, Mackin v. United States, 117 U.S. 348 (1886)). Under federal law, persons sentenced to one year or less cannot serve their terms in a penitentiary without their consent. 18 U.S.C. § 4083.Show More Pursuant to federal statute, any sentence of more than one year may be served in a penitentiary. Thus, all federal felonies, i.e., crimes punishable by more than one year in prison, expose the defendant to time in a penitentiary and therefore trigger the right to indictment.64 64.See alsoFed. R. Crim. P. 7(a) (designating that crimes punishable by a term of imprisonment of more than one year must be charged by indictment).Show More

By its terms, therefore, the right to indictment attaches to the great majority of revocation proceedings. First, and most significantly, the statute governing supervised release revocations authorizes judges to impose between two and five years of imprisonment for violating a condition of supervised release, depending on the seriousness of the underlying crime giving rise to the term of supervised release.65 65.18 U.S.C. § 3583(e)(3) (“[A] defendant whose term is revoked under this paragraph may not be required to serve on any such revocation more than 5 years in prison if the offense that resulted in the term of supervised release is a class A felony, more than 3 years in prison if such offense is a class B felony, more than 2 years in prison if such offense is a class C or D felony, or more than one year in any other case.”).Show More In nearly every revocation proceeding, a defendant found guilty of committing the violation conduct could therefore be made to suffer an infamous punishment within the meaning of the Fifth Amendment. Further, one accused of violating a condition of release is formally charged with committing the violation conduct and must appear in court and either admit to the violation or present a defense66 66.See Fed. R. Crim. P. 32.1.Show More—in other words, “held to answer” within the meaning of the Fifth Amendment.67 67.U.S. Const. amend. V.Show More Finally, the Fifth Amendment protects a “person,” rather than an accused.68 68.Id.Show More As a result, even assuming, as some courts have done, that a revocation proceeding cannot be “deemed” a “criminal prosecution” for Sixth Amendment purposes (a proposition with which we disagree, as discussed below), the rights set forth in the Fifth Amendment apply to revocation proceedings.69 69.Indeed, courts already acknowledge that the Due Process clause enshrined in the Fifth and Fourteenth Amendments applies to revocation proceedings. See, e.g., United States v. Haymond, 139 S. Ct. 2369, 2391 (2019) (Alito, J., dissenting).Show More

In contrast to the more robust dispute over whether the Sixth Amendment governs revocation proceedings, the role of the Fifth Amendment in revocation proceedings has been largely ignored. Most courts have simply proceeded from the assumption that, because a revocation proceeding is distinguishable from a prosecution, no right to indictment attaches.70 70.E.g., United States v. Peguero, 34 F.4th 143, 157–58 (2d Cir. 2022) (citing exclusively Sixth Amendment cases for the proposition that no right to indictment attaches); United States v. Cordova, 461 F.3d 1184, 1185–90 (10th Cir. 2006) (concluding that there is no Sixth [sic] Amendment right to indictment, and citing Morrissey v. Brewer,408 U.S. 471, 480 (1972)). Although some courts have held that imposing a term of supervised release to follow a prison sentence of less than a year does not transform a misdemeanor conviction into a felony conviction, that does not answer the question whether one facing more than a year of imprisonment upon revocation is entitled to be indicted. See, e.g., United States v. Celestine, 905 F.2d 59, 60–61 (5th Cir. 1990) (per curiam); United States v. Purvis, 940 F.2d 1276, 1280 (9th Cir. 1991).Show More As a general matter, we can imagine two possible reasons why the right would not apply: (1) supervised release violations are not crimes to which the right to indictment attaches; or (2) the indictment (or waiver thereof) supporting the underlying conviction for which the term of supervised release was imposed suffices to meet the Fifth Amendment indictment requirement. We consider both options below.

1. Is an Indictment Required to Charge a Supervised Release Violation?

One possible reason the right to indictment does not apply in revocation proceedings is that violating a condition of supervised release is not a “crime” within the meaning of the Fifth Amendment.71 71.U.S. Const. amend. V.Show More Although there is no clear definition of a “crime” as used in the Fifth Amendment, the Supreme Court has explained that, at the time of framing, a crime was understood to be an “act[] to which the law affixes . . . punishment” or “the wrong upon which the punishment is based.”72 72.Apprendi v. New Jersey, 530 U.S. 466, 510 (2000) (Thomas, J., concurring) (citing 1 Joel Prentiss Bishop, Commentaries on the Law of Criminal Procedure § 80, at 51, § 84, at 53 (2d ed. 1872)); see also Haymond, 139 S. Ct. at 2376 (describing the concept of a “crime” as broad and linked to punishment); Green v. United States,356 U.S. 165, 202 (1958) (Black, J., dissenting) (“How can it possibly be any more of a crime to be convicted of disobeying a statute and sent to jail for three years than to be found guilty of violating a judicial decree forbidding precisely the same conduct and imprisoned for the same term?”).Show More Of particular importance in determining whether a statute creates a “crime” is the measure and type of punishment that the legislature has elected to impose.73 73.Apprendi, 530 U.S. at 501 (Thomas, J., concurring) (observing that a crime “includes every fact that is by law a basis for imposing or increasing punishment (in contrast with a fact that mitigates punishment)”).Show More In fact, it is the nature and severity of a particular punishment that the Supreme Court has repeatedly used to delineate crimes from civil wrongs.74 74.Kennedy v. Mendoza-Martinez, 372 U.S. 144, 167 (1963); see also Hudson v. United States, 522 U.S. 93, 99 (1997) (noting that a court must utilize tools of statutory construction to determine whether a law is sufficiently punitive to transform a civil remedy into a criminal penalty).Show More

Consider that definition in the context of a revocation proceeding. To be sure, Section 3583 is not a traditional criminal statute—for example, it does not identify the elements comprising a particular offense for which punishment can or must be imposed. By its terms, however, Section 3583 does “affix[] punishment”75 75.Apprendi, 530 U.S. at 510 (Thomas, J., concurring) (cleaned up) (citing 1 Joel Prentiss Bishop, Commentaries on the Law of Criminal Procedure § 80, 51 (2d ed. 1872)).Show More to the act of violating a condition of supervised release.76 76.18 U.S.C. § 3583(e).Show More Most importantly, the nature and severity of that punishment—incarceration in a federal prison or penitentiary—is indistinguishable from the punishment imposed for committing a felony.77 77.See, e.g., Mendoza-Martinez, 372 U.S. at 167 (“[F]orfeiture of citizenship is a penalty for the act of leaving or staying outside the country to avoid the draft. This being so, the Fifth and Sixth Amendments mandate that this punishment cannot be imposed without a prior criminal trial and all its incidents, including indictment, notice, confrontation, jury trial, assistance of counsel, and compulsory process for obtaining witnesses. If the sanction these sections impose is punishment, and it plainly is, the procedural safeguards required as incidents of a criminal prosecution are lacking.”).Show More As a result, the right to indictment ought to attach.

2. Can the Indictment Supporting the Underlying Conviction Also Support a Charge of Violating Supervised Release?

A second possibility is that every person serving a term of supervised release for a felony conviction has already been indicted—or waived the right to indictment—in connection with the original prosecution. It is therefore at least theoretically possible that the indictment for that original offense could satisfy the right to indictment for any supervised release violation. After all, some courts (again relying on Morrissey) have held that any new term of imprisonment imposed following supervised release revocation is authorized by the original sentence and, by implication, the original indictment.78 78.See infra Part III.A (discussing the “original penalty” rationale). That rationale overlooks the statutes respectively authorizing a sentence of probation and a sentence of imprisonment. Compare 18 U.S.C. § 3564(e) (“A sentence of probation remains conditional and subject to revocation until its expiration or termination.” (emphasis ours)), with id. § 3582(b) (“Notwithstanding the fact that a sentence to imprisonment can subsequently be [modified due to compassionate release or corrected or appealed,] a judgment of conviction that includes such a sentence [of imprisonment] constitutes a final judgment for all other purposes.”) and id. § 3582(c) (instructing that with rare exceptions, a “court may not modify a term of imprisonment once it has been imposed”).Show More

Of course, it would be metaphysically impossible for an indictment for the underlying conviction to have actually included charges relating to yet-to-be-committed violation conduct. In addition, the idea that the original conviction somehow authorizes any new punishment imposed after revocation is foreclosed by Supreme Court precedent. Specifically, the Supreme Court has repeatedly held that an indictment must charge the elements of the crime for which an individual is ultimately prosecuted and punished. Indeed,

[t]he very purpose of the requirement that a man be indicted by grand jury is to limit his jeopardy to offenses charged by a group of his fellow citizens acting independently of either prosecuting attorney or judge. Thus the basic protection the grand jury was designed to afford is defeated by a device or method which subjects the defendant to prosecution for interference with [conduct] which the grand jury did not charge.79 79.Stirone v. United States, 361 U.S. 212, 218 (1960).Show More

Any substantive charges set forth in that indictment “may not be broadened through amendment except by the grand jury itself.”80 80.Id. at 215–16; United States v. Simmons, 11 F.4th 239, 268 (4th Cir. 2021) (“[I]ncongruity between the indictment and the conviction that a constructive amendment causes destroys the defendant’s substantial right to be tried only on charges presented in the indictment.” (cleaned up) (quoting Stirone, 361 U.S. at 217)).Show More As a result, the original indictment cannot constitutionally cover future conduct unknown to the grand jury at the time it voted to indict.81 81.Some decisions hold that, when an indictment fails to charge one of the elements of a charged offense, but there is overwhelming evidence in support of the missing element, the right to indictment is not violated. E.g., United States v. Nkansah, 699 F.3d 743, 752 (2d Cir. 2012) (citing United States v. Cotton, 535 U.S. 625, 633 (2002)). Of course, supervised release violation conduct is not an element of any offense actually charged in the underlying indictment, but instead entirely new and separate conduct that can occur years (or even decades) after the original indictment.Show More

The constitutional guarantee against double jeopardy also forecloses the argument that the original indictment somehow covers the new punishment. The Double Jeopardy Clause protects not only against multiple prosecutions for the same offense, but also “against multiple punishments for the same offense”82 82.North Carolina v. Pearce, 395 U.S. 711, 717 (1969), overruled on other grounds byAlabama v. Smith, 490 U.S. 794, 798 (1989); see also Ex parte Lange, 85 U.S. 163, 173 (1873) (“[T]he Constitution was designed as much to prevent the criminal from being twice punished for the same offence as from being twice tried for it.”).Show More unless cumulative punishments are specifically authorized by the legislature.83 83.Missouri v. Hunter, 459 U.S. 359, 368–69 (1983) (Marshall, J., dissenting) (“Where, as here, a legislature specifically authorizes cumulative punishment under two statutes . . . the prosecutor may seek and the trial court or jury may impose cumulative punishment under such statutes in a single trial.”).Show More Given that post-revocation penalties are not imposed as part of the original prosecution, any attempt to attribute those penalties to the original conviction runs afoul of the constitutional prohibition against multiple punishments for the same offense.84 84.In the Sentencing Reform Act, Congress “forbade the federal courts from ‘modify[ing] a term of imprisonment once it has been imposed.’” United States v. Jones, 980 F.3d 1098, 1103–04 (6th Cir. 2020) (quoting Sentencing Reform Act of 1984, Pub. L. No. 98–473, tit. II, ch. 2, § 212(a), 98 Stat. 1987, 1998). To avoid constitutional problems, when courts are authorized to modify sentences, they are limited to reducing sentences. See18 U.S.C. § 3582(c).Show More

In sum, the punishment imposed for violating a condition of supervised release cannot be based on the original prosecution—either the original conviction or the original sentence. As a result, courts cannot constitutionally imprison a person for violating a condition of supervised release unless she has been indicted for that conduct by a grand jury.

B. The Sixth Amendment

The Sixth Amendment proclaims that,

[i]n all criminal prosecutions, the accused shall enjoy the right to a speedy and public trial, by an impartial jury . . . [;] to be informed of the nature and cause of the accusation; to be confronted with the witnesses against him; to have compulsory process for obtaining witnesses in his favor, and to have the Assistance of Counsel for his defence.85 85.U.S. Const. amend. VI.Show More

In our view, whether or not the Sixth Amendment applies in revocation proceedings is a question with a straightforward answer—because the right to indictment must apply in revocation proceedings, revocation proceedings necessarily constitute prosecutions. The indictment, in other words, transforms a “person” into “the accused.”86 86.Although the right to a speedy trial may attach before an indictment issues, each of the rights encompassed in the Sixth Amendment attaches by the time an indictment is returned. See United States v. Black, 918 F.3d 243, 256 (2d Cir. 2019).Show More

But even putting aside the implications of the right to indictment, revocation proceedings have all the earmarks of a traditional prosecution within the meaning of the Sixth Amendment. Although historical records do not offer a precise definition of the term “prosecution” at the time the Sixth Amendment was adopted,87 87.See, e.g., Rothgery v. Gillespie County, 554 U.S. 191, 219 (2008) (Thomas, J., dissenting) (citing to 4 William Blackstone, Commentaries *289, to discern the Framers’ intent); see alsoSanjay Chhablani, Disentangling the Sixth Amendment, 11 U. Pa. J. Const. L. 487, 492 (2009) (noting the dearth of early Supreme Court opinions shedding light on the scope of Sixth Amendment protections).Show More we can glean from that record the type of proceedings to which the rights encompassed in the Sixth Amendment were designed to apply and the interests they were designed to safeguard.

First, the Sixth Amendment is distinct from the Fifth in that it protects only “the accused.”88 88.U.S. Const. amend. VI.Show More A prosecution, within the meaning of the Sixth Amendment, necessarily involves the “filing [of] formal charges in a court with jurisdiction to try and punish the defendant.”89 89.Rothgery, 554 U.S. at 223 (Thomas, J., dissenting); see alsoUnited States v. Haymond, 139 S. Ct. 2369, 2376 (2019) (plurality opinion) (describing a prosecution as “the process of exhibiting formal charges against an offender before a legal tribunal” (cleaned up) (quoting Noah Webster, An American Dictionary of the English Language (1st ed. 1828))).Show More The filing of a formal accusation transforms the proceeding into an adversarial one, with the government standing on one side of the courtroom as the accuser and the citizen standing on the other as the accused. That adversarial relationship between citizen and government is a “defining characteristic”90 90.Rothgery, 554 U.S. at 234 (Thomas, J., dissenting).Show More of a prosecution; perhaps above all else, the Sixth Amendment was designed to protect a citizen “faced with the prosecutorial forces of organized society.”91 91.Id. at 198(quoting Kirby v. Illinois, 406 U.S. 682, 689 (1972)); see also Middendorf v. Henry, 425 U.S. 25, 40 (1976) (“Yet the adversary nature of civilian criminal proceedings is one of the touchstones of the Sixth Amendment’s right to counsel.”); Williams v. Florida, 399 U.S. 78, 112 (1970) (Black, J., concurring in part and dissenting in part) (noting that the Fifth and Sixth Amendments were designed “to shield the defendant against state power”).Show More And perhaps most importantly, a prosecution is a proceeding that can result in the infliction of punishment—most significantly, the loss of liberty.92 92.Note, An Argument for Confrontation under the Federal Sentencing Guidelines, 105 Harv. L. Rev. 1880, 1889 (1992) (arguing that the Sixth Amendment was “meant to apply” when “the state decides how to punish an individual for criminal behavior”); 4 William Blackstone, Commentaries *301 (defining a prosecution in part as “the next step towards the punishment of offenders”); Search for Collocate “Punish” Within Four Words of “Prosecute”, Brigham Young University’s Corpus of Founding Era American English, Version 6.1.0, https://lawcorp​us.byu.edu/byucoeme/concordances (last visited Oct. 17, 2022) (thirteen hits) (enter query term “prosecute” in main search bar; then enter “punish” under the “collocates” tab below, with “scope left” and “scope right” set to four; then click “apply changes”).Show More

In our view, a supervised release revocation proceeding has all the “defining characteristics”93 93.Rothgery, 554 U.S. at 234 (Thomas, J., dissenting).Show More of a prosecution within the meaning of the Sixth Amendment. First, one accused of violating a condition of supervised release is formally accused of committing new, wrongful conduct in a court with the power to adjudicate guilt or innocence and impose punishment. Most significantly—and unlike the petitioners in Morrissey and Gagnon—a defendant convicted of violation conduct may be sentenced to a new and additional punishment, distinct from any punishment inflicted for the original offense. As a result, one facing revocation of supervised release is, for Sixth Amendment purposes, an “accused”94 94.U.S. Const. amend. VI.Show More—she has been formally charged with committing an act to which the law “affixes punishment” in the form of an additional prison sentence.95 95.United States v. Haymond, 139 S. Ct. 2369, 2376 (2019) (plurality opinion) (cleaned up); see also id.at 2392 (defining the accused as “an individual ‘[c]harged with a crime, by a legal process’”) (Alito, J., dissenting) (quoting Noah Webster, An American Dictionary of the English Language (1828)). Dissenting in Haymond, Justice Alito argued that one serving a term of supervised release is no longer the “accused,” but has instead been properly convicted of the underlying offense, and that any sanctions imposed for violation conduct merely sanction a breach of the court’s trust. Id.at 2392–93. As a result, he argued, the Sixth Amendment was inapplicable. Id. at 2392.But as Justice Alito also recognized, the original prosecution has ended by the time a revocation proceeding begins. Id.at 2393–95. And as discussed further below, there is no statutory support for the idea that revocation sanctions a defendant’s breach of the court’s trust. Instead, revocation can only reasonably be understood to sanction the new offense conduct—meaning one charged with a supervised release violation stands in the shoes of the “accused,” for Sixth Amendment purposes. And although the underlying criminal conviction may certainly deprive one of certain constitutional rights, the right to be indicted and the right to trial by jury for a new prosecution are not among them. Cf. Strunk v. United States, 412 U.S. 434, 437–38 (1973) (acknowledging that defendant had the right to a speedy trial irrespective of the fact that he was incarcerated in a penitentiary on an unrelated charge). As a result, the mere fact that a releasee stands convicted of the crime giving rise to the term of supervised release does not afford the government the right to deprive her of Fifth and Sixth Amendment rights in a new prosecution.Show More Finally, the proceeding convened to adjudicate her guilt—a revocation hearing—is inherently adversarial in nature. Although revocation proceedings may be initiated by a U.S. Probation Officer, a U.S. Attorney, or even the court itself, it is an Assistant U.S. Attorney, with all of the resources made available for prosecuting federal crimes, who generally takes on the burden of proving guilt by a preponderance of the evidence. And at that proceeding, the government—whether represented by a U.S. Probation Officer or a U.S. Attorney—has undoubtedly “committed itself to prosecute” such that “the adverse positions of government and defendant have solidified.”96 96.Rothgery, 554 U.S. at 233 (Thomas, J., dissenting) (quoting Kirby v. Illinois, 406 U.S. 682, 689 (1972)).Show More In other words, for Sixth Amendment purposes, the proceeding is indistinguishable from a traditional prosecution.

III. The Constitutional Work Arounds

Unlike with the right to indictment, courts have repeatedly wrestled with the question of whether the Sixth Amendment rights applicable to a traditional criminal prosecution govern supervised release revocation proceedings and have uniformly held that they do not. We consider the primary arguments advanced by those courts below.

A. Part of the Original Penalty

Shortly after enactment of the ADAA, Congress amended Section 3583 to make it more punitive. Specifically, amendments enacted in 1987 had the effect of increasing the length of the supervised release term a court could impose, and, more importantly, the length of imprisonment a defendant could serve upon revocation.97 97.Doherty, supra note 1, at 1003.Show More Amendments passed in 1994 required “revocation” and imprisonment if a court found by a preponderance of the evidence that a supervisee had violated certain “conditions” of supervised release.98 98.Id. at 1003–04.Show More

Defendants sentenced pursuant to those amendments quickly began raising ex post facto challenges, forcing appellate courts to grapple, for the first time, with what post-revocation penalties were actually punishing.99 99.Id. at 1004.Show More Relying in part on the absence of constitutional protections prescribed by Section 3583(e) before supervised release could be revoked, a majority of appellate courts concluded that post-revocation imprisonment had to be part of the original penalty for the underlying crime.100 100.Id. at 1006–07.Show More In so holding, those courts acknowledged that the alternative interpretation—that the new prison term was punishing the violation conduct itself—was constitutionally problematic. In particular, if that new prison term was punishment for the new conduct, courts were imprisoning people for conduct that often did not constitute a crime and were doing so without any of the procedural protections available in an ordinary prosecution.101 101.E.g., United States v. Beals, 87 F.3d 854, 859 (7th Cir. 1996) (concluding that revocation had to sanction the original offense because “[c]onduct that violates the terms of supervised release, like that of parole violations, is often not criminal”); United States v. Meeks, 25 F.3d 1117, 1123 (2d Cir. 1994) (relying in part on the lack of “fundamental constitutional procedural protections” in a revocation proceeding to conclude that it must be punishing the original offense).Show More

The question reached the Supreme Court in Johnson v. United States, a case involving yet another ex post facto challenge to the application of a particular amendment to Section 3583.102 102.Specifically, Johnson concerned a challenge to the application of § 3583(h), which authorized a district court to (1) revoke a term of supervised release; (2) impose a term of imprisonment; and (3) impose a new term of supervised release to follow imprisonment. 529 U.S. 694, 696 (2000).Show More Although acknowledging the “intuitive appeal” 103 103.Id. at 700.Show More of the idea that revocation sanctioned the violation conduct itself, the Johnson Court candidly explained that “construing revocation and reimprisonment as punishment for the violation of the conditions of supervised release” would essentially render the statute unconstitutional.104 104.Id.Show More More specifically, if the new term of imprisonment was sanctioning the new conduct, there were at least three major constitutional issues: (1) courts were imprisoning people for conduct that was not criminal; (2) courts were imprisoning people based on a finding by a single judge employing a preponderance of the evidence standard; and (3) where the violation conduct was criminal in its own right, courts were punishing people twice—potentially in violation of the Double Jeopardy Clause.105 105.Id. The double-jeopardy problem the Court identified in Johnson involved punishing violation conduct both as a supervised release violation and as the basis for a new prosecution. The Court did not consider the double-jeopardy problems discussed above—the fact that the imposition of an additional penalty for the original crime also violates the Double Jeopardy Clause.Show More In short, if post-revocation penalties were attributable to the new offense conduct, revocation proceedings looked uncomfortably similar to criminal prosecutions, albeit without affording defendants the requisite constitutional protections.106 106.Johnson did not make that point explicitly; it did, however, cite to Gagnon to suggest that revocation was not akin to a prosecution. Id. at 700–01.Show More Rather than address those constitutional problems, the Johnson Court offered a workaround—by calling the new prison term “part of the penalty for the initial offense,” the Court suggested those constitutional problems could simply be “avoid[ed].”107 107.Id. at 700.Show More

Ultimately, the Johnson Court also dodged the question whether the application of amendments to Section 3583 violated the Ex Post Facto Clause, rendering its discussion of the constitutionality of post-revocation penalties dicta that might reasonably have been disregarded.108 108.Id.at 713; see also Jacob Schuman, Criminal Violations, 108 Va. L. Rev. (forthcoming Dec. 2022) (manuscript at 125) (on file with authors) (noting the Court’s evasion of that constitutional obstacle and its adoption of the “original offense” doctrine could be seen as dicta).Show More Instead, and perhaps because of the dearth of other Supreme Court authority addressed to supervised release, lower courts seized upon that rationale in response to myriad challenges to the constitutionality of Section 3583. Claiming that post-revocation penalties could be “treated” or “deemed” part of the original penalty—rather than new punishment for new conduct—courts purported to distinguish revocation proceedings from prosecutions. As a result, they determined that defendants in those proceedings were not entitled to “the full panoply of rights that criminal defendants generally enjoy.”109 109.United States v. Peguero, 34 F.4th 143, 160 (2d Cir. 2022); see also United States v. Jackson, 559 F.3d 368, 371 (5th Cir. 2009) (“Post-revocation sanctions are not a separate penalty for purposes of the Double Jeopardy clause—they are part of the penalty for the original offense.”); United States v. Gavilanes-Ocaranza, 772 F.3d 624, 627 (9th Cir. 2014) (“The revocation of supervised release is not a criminal prosecution for Sixth Amendment purposes, because the violation ‘simply triggers the execution of the conditions of the original sentence.’” (quoting United States v. Paskow, 11 F.3d 873, 881 (9th Cir. 1993))); United States v. Doka, 955 F.3d 290, 294 (2d Cir. 2020) (noting defendants in supervised release proceedings are not entitled to holding the government to a burden of proof beyond a reasonable doubt or to a trial by jury).Show More

B. Conditional Liberty

The second argument invoked to reject Sixth Amendment challenges to revocation proceedings developed in response to Apprendi v. New Jersey, a Supreme Court decision issued in 2000.110 110.530 U.S. 466 (2000).Show More Apprendi considered the difference between a sentencing factor and an element of a crime for Sixth Amendment purposes and involved a state sentencing scheme that blurred the lines between the two. Building on its prior precedents, the Apprendi Court drew a bright line rule. “Other than the fact of a prior conviction,” the Court explained, “any fact that increases the penalty for a crime beyond the prescribed statutory maximum must be submitted to a jury, and proved beyond a reasonable doubt.”111 111.Id. at 490.Show More

In the wake of Apprendi and its progeny,112 112.The Supreme Court followed up Apprendiwith Blakely v. Washington, in which it clarified that “‘the statutory maximum’ for Apprendi purposes is the maximum sentence a judge may impose solely on the basis of the facts reflected in the jury verdict or admitted by the defendant.” 542 U.S. 296, 303 (2004) (emphasis omitted). Relying in part on that holding, the Court then decided United States v. Booker, in which it held that § 3553(b), which made the Guidelines mandatory and binding on district courts, was unconstitutional. 543 U.S. 220, 245 (2005).Show More defendants began to argue that, if the sanctions for violations of supervised release exceeded the statutory maximum for the underlying offense, the revocation sentence amounted to a violation of Apprendi.113 113.E.g., United States v. McIntosh, 630 F.3d 699, 702–03 (7th Cir. 2011);see also United States v. Huerta-Pimental, 445 F.3d 1220, 1222–25 (9th Cir. 2006) (holding Apprendi has no effect on the imposition of supervised release under § 3583);United States v. Work, 409 F.3d 484, 489–92 (1st Cir. 2005) (rejecting the defendant’s argument that the inclusion of a term of supervised release as a part of his original sentence exposed him to additional punishment beyond the statutory maximum).Show More Others raised more sweeping challenges to the statutory scheme, claiming that imposing a term of imprisonment based solely on a single judge’s findings—made by a preponderance of the evidence—violated the Fifth and Sixth Amendments.114 114.United States v. Cunningham, 607 F.3d 1264, 1266–67 (11th Cir. 2010); United States v. Carlton, 442 F.3d 802, 807 (2d Cir. 2006).Show More

Perhaps because the “original penalty” rationale seemed to run headlong into the holding of Apprendi and its progeny, appellate courts confronting Apprendi-based challenges to revocation fell back on Morrissey, which, of course, is a decision about parole. In particular, courts invoked Morrissey’s discussion of “conditional liberty” to justify revoking supervised release without affording a defendant’s Sixth Amendment rights. One Second Circuit case—United States v. Carlton—offers the strongest example. The petitioner in Carlton, sentenced by a district court to serve thirty-five months’ imprisonment for committing armed bank robbery while on supervised release, challenged his revocation sentence as violative of the Fifth and Sixth Amendments.115 115.442 F.3d at 807.Show More Recognizing that simply labeling that new punishment part of the original penalty did not satisfactorily resolve the Sixth Amendment problem, the Carlton court advanced a new theory. Reasoning that a defendant on supervised release—like one on parole—enjoys only a pared down version of liberty, the Carlton court explained that the petitioner was not entitled to the full complement of rights afforded one “to whom the presumption of innocence attaches.”116 116.Id. at 809.Show More In other words, by virtue of the fact that he had been convicted of a crime and sentenced to a term of supervised release, the petitioner had “surrender[ed] . . . certain constitutional rights” such as his “rights to trial by jury and to having accusations against him proved beyond a reasonable doubt.”117 117.Id. at 809–10.Show More In the wake of Carlton, other courts rejected challenges to the revocation scheme premised on Apprendi by invoking similar reasoning (and often by citing to Carlton).118 118.E.g., Cunningham, 607 F.3d at 1267–68.Show More

C. Breach of Trust

Finally, courts have deemed the Sixth Amendment inapplicable in revocation proceedings by concluding that post-revocation penalties merely sanction a defendant’s “breach of trust” rather than punish the violation conduct itself. Citing principally to the Sentencing Commission’s policy statement on supervision violations,119 119.U.S. Sent’g Guidelines Manual ch. 7, pt. A, introductory cmt. 3(b), at 498 (U.S. Sent’g Comm’n 2021).Show More courts have reasoned that any new term of imprisonment imposed punishes the defendant’s failure to comply with the judge’s order setting conditions of supervised release.120 120.E.g., United States v. Watters, 947 F.3d 493, 497–98 (8th Cir. 2020).Show More Because the punishment imposed is divorced from the new offense conduct, those courts have reasoned, the proceeding is distinguishable from a prosecution.

In United States v. Haymond, the Supreme Court’s most recent tangle with the constitutionality of Section 3583, the Court (specifically, a controlling concurrence) endorsed that rationale. Haymond concerned a challenge to the constitutionality of Section 3583(k), which provided that a judge who finds “by a preponderance of the evidence that a defendant on supervised release committed one of several enumerated offenses, including the possession of child pornography” must impose a sentence of at least five years’ imprisonment.121 121.139 S. Ct. 2369, 2374 (2019).Show More The petitioner in Haymond, who had been sentenced to serve an additional five years in prison for violating a condition of supervised release, challenged his conviction and sentence as violative of the Fifth and Sixth Amendments as interpreted in Apprendi. Noting that the petitioner had been subjected to an increased statutory maximum penalty based solely on a judge’s findings by a preponderance of the evidence, a plurality of the Supreme Court agreed.122 122.Id.Show More

Concurring solely in the judgment, Justice Breyer disagreed with the idea of “transplant[ing] the Apprendi line of cases to the supervised-release context.”123 123.Id. at 2385 (Breyer, J., concurring).Show More Nevertheless, he agreed that Section 3583(k) was unconstitutional. Unlike an ordinary revocation proceeding, which properly sanctioned “the defendant’s ‘breach of trust’—his ‘failure to follow the court-imposed conditions’ that followed his initial conviction,” Section 3583(k) was “more like punishment for a new offense, to which the jury right would typically attach.”124 124.Id. at 2386.In reaching that conclusion, Justice Breyer relied in part on a policy statement from the U.S. Sentencing Guidelines Manual. In our view, such policy statements cannot shed any light on the reach of constitutional rights.Show More In other words, the dispositive question, for constitutional purposes, was the purpose of the new term of imprisonment: punishment for a breach of trust did not implicate the Sixth Amendment, whereas punishment for new criminal conduct certainly did.125 125.Although Justice Breyer’s concurrence appears designed to affirm on the narrowest possible ground, it is arguably a broader holding—instead of merely recognizing that revocation proceedings can constitute sentencing proceedings, it recognizes that they can constitute prosecutions.Show More Following Haymond, courts of appeals rejected a number of Sixth Amendment challenges to the statute by citing to that concurrence, concluding that because revocation punished a defendant’s breach of trust—rather than the violation conduct itself—it was constitutionally sound.126 126.United States v. Peguero, 34 F.4th 143, 159 n.13 (2d Cir. 2022) (collecting cases).Show More

IV. The Constitutional Work Arounds Do Not Work

In our view, each of the primary arguments advanced to distinguish revocation proceedings from prosecutions to which the Sixth Amendment applies is without merit. Turning first to the rationale that revocation is part of the original penalty, the idea that, because “a term of supervised release is imposed as part and parcel of the original sentence,” post-revocation sanctions are “part of the penalty for the initial offense” certainly has intuitive appeal: supervised release was imposed as part of the original sentence, and the defendant was warned at sentencing of the consequences for violating the conditions set by a judge.127 127.Id. at 160–61 (quoting Johnson v. United States, 529 U.S. 694, 700–01 (2000)) (citing United States v. McNeil, 415 F.3d 273, 277 (2d Cir. 2005)).Show More

That rationale, however, ignores the plain text and legislative history of the SRA, which carefully delineates a term of supervised release from a term of imprisonment.128 128.18 U.S.C. § 3583(a).Show More As the Supreme Court has explained, “[s]upervised release fulfills rehabilitative ends, distinct from those served by incarceration,”129 129.United States v. Johnson, 529 U.S. 53, 59 (2000); see also United States v. Aldeen, 792 F.3d 247, 252 (2d Cir. 2015) (“Supervised release is not, fundamentally, part of the punishment; rather, its focus is rehabilitation.”).Show More and “prison time is ‘not interchangeable’ with supervised release.”130 130.Mont v. United States, 139 S. Ct. 1826, 1833 (2019) (quoting Johnson, 529 U.S. at 59).Show More Under Section 3583(e), supervised release and imprisonment also serve different goals; Section 3583(e) excludes the need to “provide just punishment” from the factors a judge must consider in deciding whether to impose or to “revoke” supervised release.131 131.18 U.S.C. § 3583(e)(3) (cross-referencing 18 U.S.C. § 3553(a)(2)(B)–(D)); see also Tapia v. United States, 564 U.S. 319, 326 (2011) (noting that a court imposing a term of supervised release “may not take account of retribution . . . when imposing a term of supervised release”).Show More In short, the punitive part of the sentence has been completed by the time a judge is considering revocation. Any new punishment imposed therefore cannot logically be considered part of the penalty for the original offense or even “part of the whole matrix of punishment which arises out of a defendant’s original crimes.”132 132.United States v. Flora, 810 F. Supp. 841, 842 (W.D. Ky. 1993). Further, to the extent that courts have deemed that rationale binding precedent because of Johnson, Johnson neither considered nor decided the general applicability of the Fifth and Sixth Amendments in revocation proceedings; it considered instead whether application of amendments to the SRA constituted an ex post facto violation in certain cases. Accordingly, Johnson is not controlling with respect to the applicability in revocation proceedings of the Grand Jury and Double Jeopardy Clauses of the Fifth Amendment or of Sixth Amendment rights.Show More Instead, it can only reasonably be understood as new punishment for new conduct,133 133.As some courts have pointed out, there is, of course, a statutory connection between the original sentence and the revocation term—§ 3583(e) sets a statutory maximum term of imprisonment based on the class of the original offense. E.g., United States v. Peguero, 34 F.4th 143, 161 n.14 (2d Cir. 2022). As Professor Schuman explains, however, that does not render the term of imprisonment imposed similar to parole because “parole revocation literally restored the defendant’s original prison term” whereas “revocation of supervised release results in a new prison term limited by the class of the original conviction, not the original sentence.” Schuman, supra note 108, at 172.Show More and thus violative of the Grand Jury and Double Jeopardy Clauses of the Fifth Amendment.

Similarly, the argument that revocation merely sanctions a breach of trust is unsupportable. More specifically, a judge who imposes a term of supervised release is not granting a defendant a reprieve from imprisonment in exchange for the defendant’s promise to comply with the conditions the judge sets. Instead, that judge is imposing an additional period of surveillance that begins only after the entire prison term she imposes has been served. Supervised release is therefore best understood as the opposite of trust.134 134.As Professor Schuman recognizes,[There is] one situation in which revocation of supervised release might be considered punishment for a breach of trust. If a judge sentenced a defendant to a shorter term of imprisonment in exchange for imposing a longer term of supervised release, then that substitution of supervision for imprisonment could be considered an act of trust, and violating a condition could be viewed as a breach of that trust.Schuman, supra note 108, at 174. But, as noted above, judges impose sentences of supervised release in the vast majority of cases, and, furthermore, “[n]either the supervised-release statute nor the sentencing guidelines require courts to exchange incarceration for supervision.” Id.; see also Jacob Schuman, Revocation and Retribution, 96 Wash. L. Rev. 881, 908–09 (2021) (arguing mandatory minimum terms of both imprisonment and supervision often make it impossible for judges to reduce one penalty in favor of the other).Show More As a result, any post-revocation term of imprisonment cannot logically be understood to punish anything other than the new violation conduct.135 135.Although Johnson “treat[s]” supervised release sentences as “part of the penalty for the initial offense,” 529 U.S. at 700, that treatment actually creates the double jeopardy problem of twice punishing the same offense described above.Show More

Moreover, even accepting the premise of the argument that revocation merely punishes failure to follow court orders, federal judges are not all-powerful despots with the authority to summarily imprison a person who disappoints them or breaches their trust. To be sure, judges have the statutory authority to punish persons for criminal contempt and to imprison or fine them for that contempt. But as discussed above, contempt proceedings require nearly all of the fundamental rights applicable to an ordinary prosecution. Therefore, even assuming that a post-revocation penalty merely sanctions the failure to follow a court order, the contempt authority does not justify the summary proceedings currently employed to adjudicate supervised release violations.

Finally, the argument that those on supervised release are entitled only to “conditional liberty” elides the fundamental distinctions between supervised release and parole. To be sure, the liberty of one serving a term of supervised release is “conditional” in one sense—failing to comply with conditions set by the sentencing judge may lead to imprisonment. But that is not the “conditional liberty” described by Morrissey or Gagnon. Both of those decisions considered systems in which a defendant received a benefit—freedom—in exchange for her promise to abide by certain restrictions. That grant of liberty was “conditioned” on compliance with those restrictions and could be revoked if the defendant failed to hold up her end of the agreement.136 136.See 18 U.S.C. § 3564(e) (sentence of probation is “conditional and subject to revocation”).Show More Supervised release, by contrast, involves no reprieve from prison granted in exchange for a defendant’s good behavior. In the context of supervised release, the necessary “condition” never exists. As a result, the “conditional” liberty to which a defendant is subject is distinct from the contractual-like liberty described by Morrissey and Gagnon.

Furthermore, Morrissey did not suggest that an individual on parole somehow forfeited constitutional protections by virtue of being on parole; instead, the Court reasoned that, once the government had successfully prosecuted an individual for a crime, it should not have to go through that burdensome process once again only to reimpose the same sentence a judge had previously imposed. Supervised release, by contrast, does not authorize a judge or parole board to afford a reprieve from prison in exchange for compliance with certain conditions. As a result, the “revocation” of supervised release is simply a new punishment, triggered by the commission of new wrongful conduct. To suggest that one on supervised release is entitled to lesser constitutional protections by virtue of having been convicted of the original offense, then, is to say that a criminal conviction cancels a person’s right to constitutional protections in any future prosecution. Surely, that cannot be the case.137 137.Under that rationale, a supervisee enjoys constitutional rights when indicted in state or federal court, but not when facing a significant term of incarceration for violating supervised release. The limit or grant of constitutional rights therefore would turn entirely on a prosecutor’s decision regarding how to pursue imprisonment—by going through the trouble of securing a federal indictment and formally charging a crime, or by simply asking the sentencing judge to revoke the term of supervised release.Show More

Conclusion

Those who drafted the Fifth and Sixth Amendments did so with the goal of making it difficult for the government to arbitrarily imprison a person. The current supervised release system offers prosecutors and courts what amounts to an end run around those carefully crafted protections, an expedient route to imprisonment that avoids the inconveniences of obtaining an indictment, affording the right to jury trial, and proving guilt beyond a reasonable doubt. In our view, it is that very expediency that violates the Constitution, harming not only those whose constitutional rights are denied, but also eroding—prosecution by prosecution—the constitutional structure put in place by the Framers to protect ordinary citizens from the misuse of government power.

  1. Sentencing Reform Act of 1984, Pub. L. No. 98-473, tit. II, ch. 2, 98 Stat. 1987 (codified as amended in scattered sections of 18 and 28 U.S.C.); Fiona Doherty, Indeterminate Sentencing Returns: The Invention of Supervised Release, 88 N.Y.U. L. Rev. 958, 996–97 (2013). Professor Doherty’s seminal piece provides a thorough overview of the story of supervised release.

  2. Id. at 983–84.

  3. Id. at 984.

  4. Id. at 985; see also Peter B. Hoffman, History of the Federal Parole System: Part 1 (1910–1972), 61 Fed. Prob. 23, 27 (1997) (explaining that parole decisions were made by the whole board).

  5. Parole Act, ch. 387, §§ 1, 3, 36 Stat. 819 (1910), repealed by Sentencing Reform Act of 1984, § 212(a)(2).

  6. Id. § 3.

  7. Hoffman, supra note 4, at 28.

  8. Id.

  9. Id.

  10. E.g., Clark v. Stevens, 291 F.2d 388, 389 (6th Cir. 1961); Hiatt v. Compagna, 178 F.2d 42, 45 (5th Cir. 1949) (“A prisoner sentenced to the penitentiary is turned over by the court to the executive, namely the Attorney General, to have the sentence executed. . . . It is very evident that the whole matter of paroles is left to the informed discretion of the Board. Court action is not made a part of it.”).

  11. Comment, The Parole System, 120 U. Pa. L. Rev. 282, 287–88 (1971); see also Anderson v. Corall, 263 U.S. 193, 196 (1923) (“While on parole the convict is bound to remain in the legal custody and under the control of the warden until the expiration of the term, less allowance, if any, for good conduct. While this is an amelioration of punishment, it is in legal effect imprisonment.”); Padilla v. Lynch, 398 F.2d 481, 482 (9th Cir. 1968) (stating that, because parole is a form of custody, a complaint alleging that the denial of parole violated the equal protection clause is insufficient).

  12. Comment, supra note 11, at 286–87; cf. Ughbanks v. Armstrong, 208 U.S. 481, 488 (1908) (describing a state’s grant of parole as “the granting of a favor to a convicted criminal confined within one of its prisons”).

  13. Hyser v. Reed, 318 F.2d 225, 237 (D.C. Cir. 1963) (rejecting Sixth Amendment challenge to parole revocation procedures); see also Comment, supra note 11, at 288–89 (stating that the board and the parolee have a shared interest in rehabilitating the parolee).

  14. 408 U.S. 471, 472 (1972).

  15. Id. at 474–75.

  16. Id. at 480.

  17. Id.

  18. Id.

  19. Id. at 483; see also Jacob Schuman, Supervised Release Is Not Parole, 53 Loy. L.A. L. Rev. 587, 624 (2020).

  20. Morrissey, 408 U.S. at 482.

  21. Id.

  22. Id. at 489.

  23. Fed. R. Crim. P. 32.1 advisory committee’s note to 1979 and 1989 amendments.

  24. Ex parte United States, 242 U.S. 27, 37 (1916); see also Brent E. Newton, The Story of Federal Probation, 53 Am. Crim. L. Rev. 311, 313 (2016) (noting that federal court judges would suspend prison sentences prior to the passing of the Probation Act).

  25. Newton, supra note 24, at 312.

  26. Fiona Doherty, Obey All Laws and Be Good: Probation and the Meaning of Recidivism, 104 Geo. L.J. 291, 330, 335 (2016).

  27. 411 U.S. 778, 779 (1973).

  28. Id. at 780.

  29. Id. at 781.

  30. Mempa v. Rhay, 389 U.S. 128, 135 (1967).

  31. Id. at 137.

  32. Gagnon, 411 U.S. at 779.

  33. Id. at 782.

  34. Id. at 789.

  35. Id. at 782.

  36. S. Rep. No. 98-225, at 59 (1983). Arguably, then, Mempa is the only Supreme Court decision that affords constitutional guidance with regard to modern federal probation, which is a sentence in its own right and which requires a judge to engage in a full-blown sentencing proceeding when probation is “revoked.”

  37. United States v. Labonte, 520 U.S. 751, 769 (1997) (Breyer, J., dissenting).

  38. Doherty, supra note 1, at 995.

  39. Sentencing Reform Act of 1984, Pub. L. No. 98-473, tit. II, ch. 2, sec. 212(a)(2), § 3583(a), 98 Stat. 1987, 1999 (codified as amended at 18 U.S.C. § 3583(a)) (“The court, in imposing a sentence to a term of imprisonment for a felony or a misdemeanor, may include as a part of the sentence a requirement that the defendant be placed on a term of supervised release after imprisonment.” (emphasis added)). Imposition of a term of supervised release was required as part of every sentence that imposed imprisonment during the mandatory Sentencing Guidelines era. Since Booker, with few exceptions, imposition of supervised release is discretionary. Nonetheless, it is imposed in the vast majority of federal sentences. See Christine S. Scott-Hayward, Shadow Sentencing: The Imposition of Federal Supervised Release, 18 Berkeley J. Crim. L. 180, 186 (2013).

  40. S. Rep. No. 98-225, at 123 (1983) (emphasis added).

  41. Id. at 124.

  42. Doherty, supra note 1, at 998.

  43. § 3583(d), 98 Stat. at 1999.

  44. Doherty, supra note 1, at 999–1000.

  45. Id. at 1000.

  46. Id.

  47. Id. at 1001; see 18 U.S.C. § 3583(e)(3); Fed. R. Crim. P. 32.1.

  48. Doherty, supra note 1, at 1002.

  49. Id. at 1001.

  50. Schuman, supra note 19, at 615–16, 615 n.200 (collecting appellate cases).

  51. As United States District Judge Jack Weinstein recognized, even the term “revoke” is a “misnomer” in the context of supervised release. Because supervised release does not replace imprisonment, a judge who “revokes” a term of supervised release is not taking back the freedom she previously bestowed. United States v. Trotter, 321 F. Supp. 3d 337, 346–47 (E.D.N.Y. 2018).

  52. Dissenting in United States v. Haymond, discussed further in Part III, Justice Alito posited that supervised release is equivalent to a sentence of parole:

    [A] defendant sentenced to x years of imprisonment followed by y years of supervised release is really sentenced to a maximum punishment of x + y years of confinement, with the proviso that any time beyond x years will be excused if the defendant abides by the terms of supervised release.

    139 S. Ct. 2369, 2390 (2019). But that argument ignores the plain text of the statute, which carefully distinguishes between a term of supervised release and imprisonment. It also ignores the legislative history of the SRA, which makes clear that supervised release was designed to serve a different purpose than imprisonment and was not intended to be interchangeable with imprisonment. Finally, there is no evidence that judges in practice exchange supervised release for imprisonment by imposing shorter prison sentences when they elect to include a term of supervised release as part of a sentence. Empirical evidence instead suggests that many judges routinely impose prison sentences within the range contemplated by the Sentencing Guidelines—which, of course, similarly do not anticipate that shorter prison terms should be imposed where a judge elects to impose a term of supervised release. See Christine S. Scott-Hayward, supra note 39, at 186 (2013); U.S. Sentencing Commission Guidelines Manual, ch. 5, introductory cmt., pt. A, sent’g tbl; pt. B, introductory cmt., § 5B1.1 cmt. background (2021).

  53. United States v. Peguero, 34 F.4th 143, 167 (2d Cir. 2022) (Underhill, J., dissenting); see also Rothgery v. Gillespie, 554 U.S. 191, 221–22 (2008) (Thomas, J., dissenting) (discussing elements of a prosecution).

  54. Green v. United States, 356 U.S. 165, 183 (1958).

  55. U.S. Const. amend. V.

  56. U.S. Const. amend. VI.

  57. U.S. Const. amend. V.

  58. Michael J. Zydney Mannheimer, Ripeness of Self-Incrimination Clause Disputes, 95 J. Crim. L. & Criminology 1261, 1321–22 (2005) (noting that the Fifth Amendment protections can apply before a formal prosecution has commenced).

  59. Hurtado v. California, 110 U.S. 516, 551–52 (1884) (Harlan, J., dissenting) (cleaned up) (emphasis added).

  60. U.S. Const. amend. V.

  61. Elizabeth T. Lear, Is Conviction Irrelevant?, 40 UCLA L. Rev. 1179, 1230 n.246 (1993); Wilkerson v. Whitley, 28 F.3d 498, 502 (5th Cir. 1994).

  62. Ex parte Wall, 107 U.S. 265, 298 (1882); see also Mahon v. Justice, 127 U.S. 700, 713 (1888) (noting that defendant was “held to answer the indictment” after formal charges were served upon him); United States v. McIntosh, 704 F.3d 894, 903–04 (11th Cir. 2013) (quoting Answer, Oxford English Dictionary, http://www.oed.com/view/Entry/8146 (last visited Dec. 21, 2012)) (observing that “answer” means “to speak in reply or opposition to a charge,” a definition recognized when the Fifth Amendment was adopted).

  63. United States v. Smith, 982 F.2d 757, 761 (2d Cir. 1992) (cleaned up) (citing, inter alia, Mackin v. United States, 117 U.S. 348 (1886)). Under federal law, persons sentenced to one year or less cannot serve their terms in a penitentiary without their consent. 18 U.S.C. § 4083.

  64. See also Fed. R. Crim. P. 7(a) (designating that crimes punishable by a term of imprisonment of more than one year must be charged by indictment).

  65. 18 U.S.C. § 3583(e)(3) (“[A] defendant whose term is revoked under this paragraph may not be required to serve on any such revocation more than 5 years in prison if the offense that resulted in the term of supervised release is a class A felony, more than 3 years in prison if such offense is a class B felony, more than 2 years in prison if such offense is a class C or D felony, or more than one year in any other case.”).

  66. See Fed. R. Crim. P. 32.1.

  67. U.S. Const. amend. V.

  68. Id.

  69. Indeed, courts already acknowledge that the Due Process clause enshrined in the Fifth and Fourteenth Amendments applies to revocation proceedings. See, e.g., United States v. Haymond, 139 S. Ct. 2369, 2391 (2019) (Alito, J., dissenting).

  70. E.g., United States v. Peguero, 34 F.4th 143, 157–58 (2d Cir. 2022) (citing exclusively Sixth Amendment cases for the proposition that no right to indictment attaches); United States v. Cordova, 461 F.3d 1184, 1185–90 (10th Cir. 2006) (concluding that there is no Sixth [sic] Amendment right to indictment, and citing Morrissey v. Brewer, 408 U.S. 471, 480 (1972)). Although some courts have held that imposing a term of supervised release to follow a prison sentence of less than a year does not transform a misdemeanor conviction into a felony conviction, that does not answer the question whether one facing more than a year of imprisonment upon revocation is entitled to be indicted. See, e.g., United States v. Celestine, 905 F.2d 59, 60–61 (5th Cir. 1990) (per curiam); United States v. Purvis, 940 F.2d 1276, 1280 (9th Cir. 1991).

  71. U.S. Const. amend. V.

  72. Apprendi v. New Jersey, 530 U.S. 466, 510 (2000) (Thomas, J., concurring) (citing 1 Joel Prentiss Bishop, Commentaries on the Law of Criminal Procedure § 80, at 51, § 84, at 53 (2d ed. 1872)); see also Haymond, 139 S. Ct. at 2376 (describing the concept of a “crime” as broad and linked to punishment); Green v. United States, 356 U.S. 165, 202 (1958) (Black, J., dissenting) (“How can it possibly be any more of a crime to be convicted of disobeying a statute and sent to jail for three years than to be found guilty of violating a judicial decree forbidding precisely the same conduct and imprisoned for the same term?”).

  73. Apprendi, 530 U.S. at 501 (Thomas, J., concurring) (observing that a crime “includes every fact that is by law a basis for imposing or increasing punishment (in contrast with a fact that mitigates punishment)”).

  74. Kennedy v. Mendoza-Martinez, 372 U.S. 144, 167 (1963); see also Hudson v. United States, 522 U.S. 93, 99 (1997) (noting that a court must utilize tools of statutory construction to determine whether a law is sufficiently punitive to transform a civil remedy into a criminal penalty).

  75. Apprendi, 530 U.S. at 510 (Thomas, J., concurring) (cleaned up) (citing 1 Joel Prentiss Bishop, Commentaries on the Law of Criminal Procedure § 80, 51 (2d ed. 1872)).

  76. 18 U.S.C. § 3583(e).

  77. See, e.g., Mendoza-Martinez, 372 U.S. at 167 (“[F]orfeiture of citizenship is a penalty for the act of leaving or staying outside the country to avoid the draft. This being so, the Fifth and Sixth Amendments mandate that this punishment cannot be imposed without a prior criminal trial and all its incidents, including indictment, notice, confrontation, jury trial, assistance of counsel, and compulsory process for obtaining witnesses. If the sanction these sections impose is punishment, and it plainly is, the procedural safeguards required as incidents of a criminal prosecution are lacking.”).

  78. See infra Part III.A (discussing the “original penalty” rationale). That rationale overlooks the statutes respectively authorizing a sentence of probation and a sentence of imprisonment. Compare 18 U.S.C. § 3564(e) (“A sentence of probation remains conditional and subject to revocation until its expiration or termination.” (emphasis ours)), with id. § 3582(b) (“Notwithstanding the fact that a sentence to imprisonment can subsequently be [modified due to compassionate release or corrected or appealed,] a judgment of conviction that includes such a sentence [of imprisonment] constitutes a final judgment for all other purposes.”) and id. § 3582(c) (instructing that with rare exceptions, a “court may not modify a term of imprisonment once it has been imposed”).

  79. Stirone v. United States, 361 U.S. 212, 218 (1960).

  80. Id. at 215–16; United States v. Simmons, 11 F.4th 239, 268 (4th Cir. 2021) (“[I]ncongruity between the indictment and the conviction that a constructive amendment causes destroys the defendant’s substantial right to be tried only on charges presented in the indictment.” (cleaned up) (quoting Stirone, 361 U.S. at 217)).

  81. Some decisions hold that, when an indictment fails to charge one of the elements of a charged offense, but there is overwhelming evidence in support of the missing element, the right to indictment is not violated. E.g., United States v. Nkansah, 699 F.3d 743, 752 (2d Cir. 2012) (citing United States v. Cotton, 535 U.S. 625, 633 (2002)). Of course, supervised release violation conduct is not an element of any offense actually charged in the underlying indictment, but instead entirely new and separate conduct that can occur years (or even decades) after the original indictment.

  82. North Carolina v. Pearce, 395 U.S. 711, 717 (1969), overruled on other grounds by Alabama v. Smith, 490 U.S. 794, 798 (1989); see also Ex parte Lange, 85 U.S. 163, 173 (1873) (“[T]he Constitution was designed as much to prevent the criminal from being twice punished for the same offence as from being twice tried for it.”).

  83. Missouri v. Hunter, 459 U.S. 359, 368–69 (1983) (Marshall, J., dissenting) (“Where, as here, a legislature specifically authorizes cumulative punishment under two statutes . . . the prosecutor may seek and the trial court or jury may impose cumulative punishment under such statutes in a single trial.”).

  84. In the Sentencing Reform Act, Congress “forbade the federal courts from ‘modify[ing] a term of imprisonment once it has been imposed.’” United States v. Jones, 980 F.3d 1098, 1103–04 (6th Cir. 2020) (quoting Sentencing Reform Act of 1984, Pub. L. No. 98–473, tit. II, ch. 2, § 212(a), 98 Stat. 1987, 1998). To avoid constitutional problems, when courts are authorized to modify sentences, they are limited to reducing sentences. See 18 U.S.C. § 3582(c).

  85. U.S. Const. amend. VI.

  86. Although the right to a speedy trial may attach before an indictment issues, each of the rights encompassed in the Sixth Amendment attaches by the time an indictment is returned. See United States v. Black, 918 F.3d 243, 256 (2d Cir. 2019).

  87. See, e.g., Rothgery v. Gillespie County, 554 U.S. 191, 219 (2008) (Thomas, J., dissenting) (citing to 4 William Blackstone, Commentaries *289, to discern the Framers’ intent); see also Sanjay Chhablani, Disentangling the Sixth Amendment, 11 U. Pa. J. Const. L. 487, 492 (2009) (noting the dearth of early Supreme Court opinions shedding light on the scope of Sixth Amendment protections).

  88. U.S. Const. amend. VI.

  89. Rothgery, 554 U.S. at 223 (Thomas, J., dissenting); see also United States v. Haymond, 139 S. Ct. 2369, 2376 (2019) (plurality opinion) (describing a prosecution as “the process of exhibiting formal charges against an offender before a legal tribunal” (cleaned up) (quoting Noah Webster, An American Dictionary of the English Language (1st ed. 1828))).

  90. Rothgery, 554 U.S. at 234 (Thomas, J., dissenting).

  91. Id. at 198 (quoting Kirby v. Illinois, 406 U.S. 682, 689 (1972)); see also Middendorf v. Henry, 425 U.S. 25, 40 (1976) (“Yet the adversary nature of civilian criminal proceedings is one of the touchstones of the Sixth Amendment’s right to counsel.”); Williams v. Florida, 399 U.S. 78, 112 (1970) (Black, J., concurring in part and dissenting in part) (noting that the Fifth and Sixth Amendments were designed “to shield the defendant against state power”).

  92. Note, An Argument for Confrontation under the Federal Sentencing Guidelines, 105 Harv. L. Rev. 1880, 1889 (1992) (arguing that the Sixth Amendment was “meant to apply” when “the state decides how to punish an individual for criminal behavior”); 4 William Blackstone, Commentaries *301 (defining a prosecution in part as “the next step towards the punishment of offenders”); Search for Collocate “Punish” Within Four Words of “Prosecute”, Brigham Young University’s Corpus of Founding Era American English, Version 6.1.0, https://lawcorp​us.byu.edu/byucoeme/concordances (last visited Oct. 17, 2022) (thirteen hits) (enter query term “prosecute” in main search bar; then enter “punish” under the “collocates” tab below, with “scope left” and “scope right” set to four; then click “apply changes”).

  93. Rothgery, 554 U.S. at 234 (Thomas, J., dissenting).

  94. U.S. Const. amend. VI.

  95. United States v. Haymond, 139 S. Ct. 2369, 2376 (2019) (plurality opinion) (cleaned up); see also id. at 2392 (defining the accused as “an individual ‘[c]harged with a crime, by a legal process’”) (Alito, J., dissenting) (quoting Noah Webster, An American Dictionary of the English Language (1828)). Dissenting in Haymond, Justice Alito argued that one serving a term of supervised release is no longer the “accused,” but has instead been properly convicted of the underlying offense, and that any sanctions imposed for violation conduct merely sanction a breach of the court’s trust. Id. at 2392–93. As a result, he argued, the Sixth Amendment was inapplicable. Id. at 2392. But as Justice Alito also recognized, the original prosecution has ended by the time a revocation proceeding begins. Id. at 2393–95. And as discussed further below, there is no statutory support for the idea that revocation sanctions a defendant’s breach of the court’s trust. Instead, revocation can only reasonably be understood to sanction the new offense conduct—meaning one charged with a supervised release violation stands in the shoes of the “accused,” for Sixth Amendment purposes. And although the underlying criminal conviction may certainly deprive one of certain constitutional rights, the right to be indicted and the right to trial by jury for a new prosecution are not among them. Cf. Strunk v. United States, 412 U.S. 434, 437–38 (1973) (acknowledging that defendant had the right to a speedy trial irrespective of the fact that he was incarcerated in a penitentiary on an unrelated charge). As a result, the mere fact that a releasee stands convicted of the crime giving rise to the term of supervised release does not afford the government the right to deprive her of Fifth and Sixth Amendment rights in a new prosecution.

  96. Rothgery, 554 U.S. at 233 (Thomas, J., dissenting) (quoting Kirby v. Illinois, 406 U.S. 682, 689 (1972)).

  97. Doherty, supra note 1, at 1003.

  98. Id. at 1003–04.

  99. Id. at 1004.

  100. Id. at 1006–07.

  101. E.g., United States v. Beals, 87 F.3d 854, 859 (7th Cir. 1996) (concluding that revocation had to sanction the original offense because “[c]onduct that violates the terms of supervised release, like that of parole violations, is often not criminal”); United States v. Meeks, 25 F.3d 1117, 1123 (2d Cir. 1994) (relying in part on the lack of “fundamental constitutional procedural protections” in a revocation proceeding to conclude that it must be punishing the original offense).

  102. Specifically, Johnson concerned a challenge to the application of § 3583(h), which authorized a district court to (1) revoke a term of supervised release; (2) impose a term of imprisonment; and (3) impose a new term of supervised release to follow imprisonment. 529 U.S. 694, 696 (2000).

  103. Id. at 700.

  104. Id.

  105. Id. The double-jeopardy problem the Court identified in Johnson involved punishing violation conduct both as a supervised release violation and as the basis for a new prosecution. The Court did not consider the double-jeopardy problems discussed above—the fact that the imposition of an additional penalty for the original crime also violates the Double Jeopardy Clause.

  106. Johnson did not make that point explicitly; it did, however, cite to Gagnon to suggest that revocation was not akin to a prosecution. Id. at 700–01.

  107. Id. at 700.

  108. Id. at 713; see also Jacob Schuman, Criminal Violations, 108 Va. L. Rev. (forthcoming Dec. 2022) (manuscript at 125) (on file with authors) (noting the Court’s evasion of that constitutional obstacle and its adoption of the “original offense” doctrine could be seen as dicta).

  109. United States v. Peguero, 34 F.4th 143, 160 (2d Cir. 2022); see also United States v. Jackson, 559 F.3d 368, 371 (5th Cir. 2009) (“Post-revocation sanctions are not a separate penalty for purposes of the Double Jeopardy clause—they are part of the penalty for the original offense.”); United States v. Gavilanes-Ocaranza, 772 F.3d 624, 627 (9th Cir. 2014) (“The revocation of supervised release is not a criminal prosecution for Sixth Amendment purposes, because the violation ‘simply triggers the execution of the conditions of the original sentence.’” (quoting United States v. Paskow, 11 F.3d 873, 881 (9th Cir. 1993))); United States v. Doka, 955 F.3d 290, 294 (2d Cir. 2020) (noting defendants in supervised release proceedings are not entitled to holding the government to a burden of proof beyond a reasonable doubt or to a trial by jury).

  110. 530 U.S. 466 (2000).

  111. Id. at 490.

  112. The Supreme Court followed up Apprendi with Blakely v. Washington, in which it clarified that “‘the statutory maximum’ for Apprendi purposes is the maximum sentence a judge may impose solely on the basis of the facts reflected in the jury verdict or admitted by the defendant.” 542 U.S. 296, 303 (2004) (emphasis omitted). Relying in part on that holding, the Court then decided United States v. Booker, in which it held that § 3553(b), which made the Guidelines mandatory and binding on district courts, was unconstitutional. 543 U.S. 220, 245 (2005).

  113. E.g., United States v. McIntosh, 630 F.3d 699, 702–03 (7th Cir. 2011); see also United States v. Huerta-Pimental, 445 F.3d 1220, 1222–25 (9th Cir. 2006) (holding Apprendi has no effect on the imposition of supervised release under § 3583); United States v. Work, 409 F.3d 484, 489–92 (1st Cir. 2005) (rejecting the defendant’s argument that the inclusion of a term of supervised release as a part of his original sentence exposed him to additional punishment beyond the statutory maximum).

  114. United States v. Cunningham, 607 F.3d 1264, 1266–67 (11th Cir. 2010); United States v. Carlton, 442 F.3d 802, 807 (2d Cir. 2006).

  115. 442 F.3d at 807.

  116. Id. at 809.

  117. Id. at 809–10.

  118. E.g., Cunningham, 607 F.3d at 1267–68.

  119. U.S. Sent’g Guidelines Manual ch. 7, pt. A, introductory cmt. 3(b), at 498 (U.S. Sent’g Comm’n 2021).

  120. E.g., United States v. Watters, 947 F.3d 493, 497–98 (8th Cir. 2020).

  121. 139 S. Ct. 2369, 2374 (2019).

  122. Id.

  123. Id. at 2385 (Breyer, J., concurring).

  124. Id. at 2386. In reaching that conclusion, Justice Breyer relied in part on a policy statement from the U.S. Sentencing Guidelines Manual. In our view, such policy statements cannot shed any light on the reach of constitutional rights.

  125. Although Justice Breyer’s concurrence appears designed to affirm on the narrowest possible ground, it is arguably a broader holding—instead of merely recognizing that revocation proceedings can constitute sentencing proceedings, it recognizes that they can constitute prosecutions.

  126. United States v. Peguero, 34 F.4th 143, 159 n.13 (2d Cir. 2022) (collecting cases).

  127. Id. at 160–61 (quoting Johnson v. United States, 529 U.S. 694, 700–01 (2000)) (citing United States v. McNeil, 415 F.3d 273, 277 (2d Cir. 2005)).

  128. 18 U.S.C. § 3583(a).

  129. United States v. Johnson, 529 U.S. 53, 59 (2000); see also United States v. Aldeen, 792 F.3d 247, 252 (2d Cir. 2015) (“Supervised release is not, fundamentally, part of the punishment; rather, its focus is rehabilitation.”).

  130. Mont v. United States, 139 S. Ct. 1826, 1833 (2019) (quoting Johnson, 529 U.S. at 59).

  131. 18 U.S.C. § 3583(e)(3) (cross-referencing 18 U.S.C. § 3553(a)(2)(B)–(D)); see also Tapia v. United States, 564 U.S. 319, 326 (2011) (noting that a court imposing a term of supervised release “may not take account of retribution . . . when imposing a term of supervised release”).

  132. United States v. Flora, 810 F. Supp. 841, 842 (W.D. Ky. 1993). Further, to the extent that courts have deemed that rationale binding precedent because of Johnson, Johnson neither considered nor decided the general applicability of the Fifth and Sixth Amendments in revocation proceedings; it considered instead whether application of amendments to the SRA constituted an ex post facto violation in certain cases. Accordingly, Johnson is not controlling with respect to the applicability in revocation proceedings of the Grand Jury and Double Jeopardy Clauses of the Fifth Amendment or of Sixth Amendment rights.

  133. As some courts have pointed out, there is, of course, a statutory connection between the original sentence and the revocation term—§ 3583(e) sets a statutory maximum term of imprisonment based on the class of the original offense. E.g., United States v. Peguero, 34 F.4th 143, 161 n.14 (2d Cir. 2022). As Professor Schuman explains, however, that does not render the term of imprisonment imposed similar to parole because “parole revocation literally restored the defendant’s original prison term” whereas “revocation of supervised release results in a new prison term limited by the class of the original conviction, not the original sentence.” Schuman, supra note 108, at 172.

  134. As Professor Schuman recognizes,

    [There is] one situation in which revocation of supervised release might be considered punishment for a breach of trust. If a judge sentenced a defendant to a shorter term of imprisonment in exchange for imposing a longer term of supervised release, then that substitution of supervision for imprisonment could be considered an act of trust, and violating a condition could be viewed as a breach of that trust.

    Schuman, supra note 108, at 174. But, as noted above, judges impose sentences of supervised release in the vast majority of cases, and, furthermore, “[n]either the supervised-release statute nor the sentencing guidelines require courts to exchange incarceration for supervision.” Id.; see also Jacob Schuman, Revocation and Retribution, 96 Wash. L. Rev. 881, 908–09 (2021) (arguing mandatory minimum terms of both imprisonment and supervision often make it impossible for judges to reduce one penalty in favor of the other).

  135. Although Johnson “treat[s]” supervised release sentences as “part of the penalty for the initial offense,” 529 U.S. at 700, that treatment actually creates the double jeopardy problem of twice punishing the same offense described above.

  136. See 18 U.S.C. § 3564(e) (sentence of probation is “conditional and subject to revocation”).

  137. Under that rationale, a supervisee enjoys constitutional rights when indicted in state or federal court, but not when facing a significant term of incarceration for violating supervised release. The limit or grant of constitutional rights therefore would turn entirely on a prosecutor’s decision regarding how to pursue imprisonment—by going through the trouble of securing a federal indictment and formally charging a crime, or by simply asking the sentencing judge to revoke the term of supervised release.