Is a Science of Comparative Constitutionalism Possible?

Introduction

Nearly a generation ago, Justice Scalia and Justice Breyer debated the legitimacy and value of using foreign law to interpret the American Constitution.¹ At the time, the matter was controversial and invited the interest of both judges and scholars. Foreign law had, after all, been relied on in significant cases like Roper v. Simmons² and Lawrence v. Texas.³ Many years on, there is still much to be debated — including the purpose and potential benefits of judicial engagement with foreign law — but “comparative constitutional law” has unquestionably emerged as a field of study in its own right. We have seen the publication of scores of articles and books that compare constitutional systems and elaborate reflections by judges over the nature and form of comparative judicial reasoning.⁴ Today, it no longer seems necessary to demonstrate, as Professor Mark Tushnet once did, “the possibilities of comparative constitutional law.”⁵

Though the study of comparative constitutional law may not quite require a defense at present, much remains to be settled. In the late 1990s and early 2000s, the question — crudely put — was whether we could compare the constitutional law of different nations. Could a comparison between rules and developments in country A and country B occur in an intelligible and meaningful fashion? Though comparative constitutional law was not new to the American legal academy, it had declined in importance over the years, thereby requiring the field to be somewhat reborn.⁶ Now that comparisons between constitutional orders are commonplace, greater attention is being devoted to a different question: how is comparative constitutionalism to be conducted?⁷ The question implicates tasks that stretch far beyond the judicial citation of foreign legal materials. The recent crisis of constitutional democracy and the phenomenon of democratic backsliding has, for example, led to an outpouring of comparative literature.⁸ It seems natural, even important, to compare the authoritarian turn in, say, Hungary, where Viktor Orbán was recently reelected as Prime Minister, with developments in countries such as India, Poland, Russia, Turkey, and Venezuela — and perhaps to reflect on President Donald Trump’s term in office in light of the global experience.⁹ But the ease of making such comparisons masks the hard question of precisely how to conduct such inquiries.

A recent wave of writing has offered a powerful and distinct answer to this question. It urges us to consider global patterns that relate to legal phenomena. It has suggested that we should observe variations in behavior and offer generalizations in lawlike terms across cultures. In other words, the aim has been to develop a kind of science of comparative constitutionalism that can, among other things, offer causal narratives that are cross-national. Such efforts — that are typically “large-N” and involve a great many nations — are familiar to social scientists.¹⁰ The potential uses and limitations of cross-country data to test hypotheses, make observations, and present causal theories are also well known to comparative law scholars who work in areas other than constitutional law.¹¹ Yet, in the field of comparative constitutional law, the performance of large-scale statistical work that transcends national boundaries and provides us with findings about the workings of different legal measures is a relatively new if burgeoning phenomenon. This method has, as such, been subject to little interrogation.¹²

This new positivist approach has asked questions of considerable importance. Such questions range from the value of constitutional rights and guarantees, such as a prohibition on torture, to the role played by legal institutions and structures, such as courts. The number and range of countries that are evaluated make the enterprise all the more noteworthy. If its findings are indeed accurate, these findings may well encourage us to revisit some of the normative and descriptive assumptions that have shaped constitutional theory for decades. In general, the positivist work that has emerged is reasonably sensitive to expected methodological concerns, such as controlling for confounding variables, coding with care, and so forth. It is worth noting that I use the term “positivist” because, strictly speaking, the work under consideration need not be quantitative.¹³ Its key feature is a Martian perspective — that is, the external observation of behavior. The Martian looks through the telescope and sees what we do. But does the Martian understand it?

Within this body of scholarship, few contributions have been as ambitious and thought-provoking as Professors Adam Chilton and Mila Versteeg’s recent book How Constitutional Rights Matter. Chilton and Versteeg’s focus is on “whether and how constitutional rights matter” (p. 6). In particular, they hope to shed light on the relationship between the de jure protection for a right and the de facto reality of rights enjoyment (p. 6). For instance, does a prohibition on torture decrease the instances of torture (p. 6)? In pursuing such inquiries, the authors present an overall theoretical framework that places emphasis on formal organizations in the enforcement of rights, considers data from 194 countries over a six-decade period to assess rights practices, and supplements the data with select, if limited, illustrative case studies, involving visits and interviews, and certain survey experiments (pp. 13–14). Although the data-driven work is supplemented in these ways, the authors make it clear that their “primary method is large-N statistical analysis” (p. 102).¹⁴

How Constitutional Rights Matter presents three key findings: rights, by themselves, can achieve little in the face of government action; certain rights “are harder to violate than others” when con-stitutionalized; and, finally, “including rights in a constitution is not a panacea . . . [because] a government [that] is determined to erode the protections provided by certain rights” will usually succeed in doing so (p. 7). A central conclusion of the book is that “what matters is the type of right being threatened, not the type of institutions in the country” (p. 48). As one might guess, the main target here is courts, whose importance has long been underlined by those who make the case for rights.¹⁵ Chilton and Versteeg, by contrast, are keen to shift the focus away from judicial institutions to the role of organizations. Rights that are granted to organizations and rights that operate within organizations are, they argue, better protected than other rights (pp. 6–12).

Whether the formal presence of a right has a bearing on the practical realization of that right is a question whose significance cannot be overestimated.¹⁶ As the authors acknowledge, there are a wide range of arguments that have been posited in favor of rights; and the lived experience of rights may encourage us to revisit some of these arguments (p. 60).¹⁷ It is one of the singular achievements of How Constitutional Rights Matter that it encourages us to question familiar truths within constitutional theory. In the book, the authors present “the most comprehensive dataset on constitutional rights compiled to date” (p. 81). In working with this dataset, they take pains to establish the robustness of their results: they compare countries that have and do not have a given constitutional right; they compare countries before and after the adoption of a right; they create control events that involve other countries that had or did not have the same right during the relevant time period; and, finally, they build a regression model with control variables (pp. 106–10).

Despite these efforts and the care that the authors take, there are key questions about the kind of analyses undertaken in How Constitutional Rights Matter. This Review explores some of these questions. It engages with certain studies and the broader conclusions advanced by Chilton and Versteeg, both in their book and in some of their other joint work, as well as other significant contributions that typify the positivist turn in comparative constitutionalism.¹⁸ How Constitutional Rights Matter is emblematic of this broader turn and should therefore be carefully considered in its own right as well as alongside other notable scholarship. This Review proceeds in three parts. It first explores the issue of causality, with special attention to the phenomenon of redundant causation and to the attribution of causation. It then turns to questions of culture, interpretation, and meaning, and underlines concerns relating to identification. The concerns that arise from both causation and interpretation come together in important ways. Finally, this Review highlights the idea of agency and the promise of constitutionalism. In attending to questions of causation, interpretation, and agency, this Review underscores the importance of context in making causal and interpretive claims and in regarding people as agents who have the power to shape their collective political life. The potential limitations with large-N analyses relate to the degree to which they attribute causation, the extent to which they underemphasize differences in meanings and practices in an attempt to arrive at thin universal descriptions, and the ways in which they minimize how humans are actors and participants in the cultures and communities that they create and inhabit. By studying the positivist approach, this Review hopes to engage not only with expected concerns that might arise with regard to such an approach, but also with the very types of studies that might be possible within comparative constitutionalism.

I. Causation

At the heart of the positivist turn in comparative constitutionalism are claims about causation. More precisely, the approach rests on a theory of counterfactual dependence to present a case of causal relations.¹⁹ A relation between a specific legal development (say, the prohibition on torture) and a phenomenon (in this case, the practice of torture) is presented to make a causal claim. One might argue, for example, that the prohibition on torture has not impacted the practice of torture. This seems simple enough, and we are all familiar with counterfactual reasoning. We routinely use such reasoning in our lives in the course of making decisions and exercising choices. But how should such a theory be deployed in the study of legal phenomena?

Let us stay with the torture example. Chilton and Versteeg note that, “[a]s of 2016, 66 percent of countries constitutionally prohibited torture, while in 1946, only 41 percent of countries did” (p. 136). Given the rise in the prohibition of torture, what might be said about the presence of torture prohibitions and de facto torture? How Constitutional Rights Matter presents us with the following conclusions: first, “between countries with and without the right, countries with constitutional torture prohibitions typically see more torture”; second, “countries that constitutionally ban torture do not torture less in the first five years after [prohibiting torture]”; third, “this trend is similar for the control group of countries that did not change their torture provisions during that same period”; and fourth, “there is no difference in rates of torture between countries with a torture prohibition and those without” (pp. 146–47). The work here is “further corroborated by” a prior study by Chilton and Versteeg that establishes the same conclusion (p. 147).²⁰ The earlier study was, like the one in How Constitutional Rights Matter, performed with caution. The authors acknowledged, for example, that countries with a better record on rights may well be less likely to adopt a prohibition on torture, and thus ratification of specific treaties may well be endogenous to the practices prevalent within states.²¹ They also attended to the problem of false negatives and false positives by focusing on transitioning democracies. The logic behind this diachronic emphasis was that the “combination of not-yet-exemplary rights records, possibilities for local advocacy, and good intentions may make transitioning democracies more likely to change their behavior because of constitutional torture prohibitions.”²²

The conclusion that Chilton and Versteeg present is a dramatic one: a major legal development, they suggest, has made no difference to reality. It is therefore worth reflecting on this conclusion carefully. Does the fact that torture has become more widely prohibited across the world and the fact that torture rates remain the same mean that such prohibitions have had no impact? It is not clear that this is the case because it may well be that during the same period that torture became legally prohibited, it also became more socially acceptable.²³ If explanations of this kind — say, the rising social acceptability of torture — have any plausibility, then it could mean that the practice of torture had far greater acceptability than was previously the case, and, if there had been no legal prohibition on torture, torture rates would have increased substantially. In such a scenario, the prohibitions on torture have made a major difference precisely because they have maintained torture rates and have not allowed them to increase. The issue, in a case like this, seems to be that culture may also be shifting. The authors make a number of efforts to add precision to their analysis, but attempts to mitigate this problem cannot quite provide an assessment of the forces and countervailing forces within each single country. In the absence of such an assessment, the facts as they are presented are compatible with the conclusion that the prohibition on torture succeeded in preventing the practice from rising in the face of dramatic shifts taking place in a given society. We do not observe what torture rates would have been had the prohibition not taken place. As such, the counterfactuals are imputed. Even if we choose not to challenge the premises of the claim presented, the conclusion here does not seem to follow from them.²⁴

Similar concerns arise in the study of “organizational rights,” like the freedom of religion. On this matter, How Constitutional Rights Matter concludes that “countries that constitutionally protect religious freedom show more respect for this right in practice than countries without this right” (p. 231). The authors “attribute this finding to the organizational character of religious rights” (p. 231) — that is, to the power that religious groups enjoy to safeguard their rights. Again, the methodology is carefully constructed. It compares countries that recognize the right with those that do not; it contrasts the situation in countries before and after the adoption of the right; it considers countries that did not protect the right during the same period, using them as controls; and it performs a regression analysis with specific controls that are taken to shape religious freedom. Overall, the results “suggest that the freedom of religion is associated with higher de facto respect for religious freedom” (p. 248). Despite the thoroughness with which the analysis is performed, we are confronted with the question of whether we can capture changes in the internal conception of religious freedom within a country. The respect for religious freedom may well be a consequence of an altered approach to the freedom of religion in society. Perhaps respect for religious freedom was rising within a society, and the juridification of religious freedom in the form of a right interfered with this and reduced the pace by which the respect for religious freedom increased.

We notice this challenge when we consider one of the major conclusions of How Constitutional Rights Matter: the potentially limited role that courts can play in the protection of rights (pp. 50–53). Chilton and Versteeg have developed this insight more fully in other work.²⁵ In such work, they primarily study independent courts and conclude that the presence of such courts does not increase the respect that governments have toward rights. In other words, judicial institutions do not have much ability to protect rights, a conclusion that calls into question much of the literature that makes the case for judicial review.²⁶ The authors posit a range of possible explanations for such a result: political branches might punish courts if they challenge the government, courts lack the institutional capacity to address adequately certain kinds of rights-based challenges, and so forth. Importantly, however, courts in this study are examined qua elements of political culture rather than qua courts. The key control is whether or not the relevant country has a court: “We . . . explore whether this relationship between de jure and de facto rights is different in countries that have an independent judiciary equipped with the power of judicial review (a ‘Constitutional Court’).”²⁷ But such a framing is, of course, compatible with courts actually enjoying a great deal of influence. Countries that establish constitutional courts could be systematically different from those that do not along a set of characteristics that are associated with the future breakdown of constitutional rights. The facts are consistent with the possibility that, in countries where courts are needed, they perform extremely well — perhaps, in such countries, courts make it possible to maintain the gap between a constitutional text and the social reality, and they prevent the gap from widening.

The kind of concern that I have raised with the causal analysis presented in How Constitutional Rights Matter may be borne out by turning to some broader literature within the positivist approach to comparative constitutionalism. Consider a key study by Professors Tom Ginsburg and James Melton on constitutional amendment rules.²⁸ The authors begin with the widely held concern that the American Constitution is, as a formal matter, extraordinarily difficult to amend.²⁹ For comparative scholars, this feature of the American Constitution has long stood out when one notices the procedures in other countries, and Ginsburg and Melton perform a great service in helping us better understand its implications. They wonder whether the formal amendment process in the United States accurately reflects the reality of constitutional practice. Is the formal procedure for amending America’s Constitution the reason its text has not been changed as easily as that of some other constitutions? What is the right way to measure constitutional flexibility? Departing from the conventional answers to this last question — answers that stress the textual requirements for amending a constitution — the authors turn to the idea of an amendment culture. Using a cross-national database, they “develop a proxy for amendment culture and show empirically that this does a better job of explaining patterns of amendment within constitutional systems than do any of the institutional indices or variables on offer.”³⁰

Taking into account a number of variables relating to amendment procedures (such as the stages involved in passing an amendment), the authors are able to predict the probability of amendments. What is important is their interest in the broader political culture. As they put it, “we regress the amendment rate on a set of amendment procedure variables as well as on a host of factors that should predict political reform more generally.”³¹ Thus, it is the defining feature of their study that they “take into account social and political factors that are likely to put pressure on countries to amend the constitution.”³² A measurement of amendment culture is undertaken through a proxy — they “operationalize amendment culture as the rate at which a country’s previous constitution was amended,” with the idea being that “attitudes toward amendment will be expressed through amendment practices, and that these attitudes will endure in the form of norms that outlast any particular set of institutions.”³³ The ultimate interest of the study lies, of course, in determining the relative importance of amendment procedures and amendment cultures, and the conclusion of the study validates the authors’ investment in the latter: “The best predictor of constitutional amendment rates, it turns out, is what we have called an amendment culture, as measured by the frequency of amendment in the country’s previous constitution.”³⁴

Ginsburg and Melton note that “attitudes about amendments matter,” and they try to demonstrate the limitations of schematic studies of constitutional orders that do not pay due attention to the cultural forces that shape political behavior and generate institutional outcomes.³⁵ The inattention toward culture that they identify is, without question, a mistake. It is an achievement of the Ginsburg-Melton analysis that they underline this point. However, it is not entirely clear whether their study helps us to remedy matters. The authors view culture as something that persists (though they often observe that it can change).³⁶ By definition, there must be something other than the structure of a constitutional system that is shaping outcomes; as they observe, “barriers to amendment are not merely institutional.”³⁷ But the potential shortcoming here appears to be that, in trying to understand the relationship between cultural attitudes and procedural rules, we cannot ignore the possibility that the nature of procedural rules may well be part of existing cultural attitudes.³⁸ The study does not seem to show that structural constraints are of little relevance, quite simply because the constraints may shape culture and the dispositions that people hold. In other words, human beings may well absorb the procedural rules into their understanding; structure and culture may not be independent from one another. Because of this interconnection, we cannot confidently conclude that the amendment procedure is of little significance. Both culture and procedure may well be determined by similar factors, and in turn by one another.

These concerns do not arise because of an imperfect analysis. Ginsburg and Melton perform the study as carefully and thoughtfully as anyone can. What cross-national causal studies involve, however, is abstraction from context, and without context it is hard to present a causal account. The challenge arises from the method itself, a matter that is captured by an essay on constitutional efficacy.³⁹ Here, Ginsburg, Melton, and Professor Zachary Elkins observe that the efficacy of constitutions changes over time, and they critique the assumption within constitutional theory that constitutional efficacy is broadly constant.⁴⁰ Their specific focus is on rights — that is, on whether rights become more or less effective with the passing of time.⁴¹ The study considers two possible effects of age. The first is a maturation effect, where “the gap between the demands of the constitution and social reality shrinks over time,” a result that “could be because norms grow to be more venerated with age.”⁴² With veneration, the authors observe, enforcement is likely to increase, and thus we are likely to witness greater compliance.⁴³ In contrast, the second effect of age might be a “drift or even decay of constitutional norms as society shifts away from norms that once made sense,” and it is some shared understanding rather than the content of the provisions themselves that seems to matter.⁴⁴

Elkins, Ginsburg, and Melton acknowledge that both these processes can occur simultaneously.⁴⁵ Over time, some aspects of a constitution may mature while others may decay, and how we assess a constitution would be an aggregate of how its different provisions are performing.⁴⁶ Rather than draw conclusions about entire constitutions, they consider individual provisions to assess maturation and provide an empirical account of the efficacy of constitutional rights over time.⁴⁷ But notice how this framing has the potential of missing the key point that veneration and decay can occur for the same thing at the same time. If this is the case, then one simply cannot get a measurement of veneration versus decay and one cannot assess maturation through the method that is adopted. What the analysis can at best reveal is an overall story — say, over time there was slightly more veneration rather than decay — but it cannot, given its abstracted character, capture the dynamic that is being undertaken because it cannot tell us how the forces influencing veneration and decay interact with one another at any given point. To provide a concrete example, the British monarchy may well be an institution around which there is both veneration and decay. To understand why the monarchy survives, one would need an account of how the competing forces are interacting with one another at any given moment.⁴⁸ Here, even though one is attempting to tell a causal story, there does not quite seem to be a causal account on offer. Consider the following paragraph:

The point is that gaps between text and practice — the usual way of approaching the question of whether rights “work” — are a poor indicator of constitutional efficacy because gaps do not indicate when or how a constitution makes a difference. Frequently, constitutional changes reveal the extent of a problem that previously was thought to be much smaller and so can motivate improvements. Methodologically, then, as an alternative to looking at gaps, we might instead see whether there was political and social mobilization around the constitution, which would indicate a causal mechanism. For example, if interest groups demanded enforcement of the constitutional right to housing, or if the government initiated a housing program in response to a court order, then one might say that the constitution has made a difference, assuming that the government program actually reduced homelessness. Thus, to say a constitution matters, one looks not only to the outcome of a reduction in the gap between text and practice but also to the channels by which it did so. Mechanisms are essential to understand efficacy, as they are for any causal study. Our chapter aspires to push scholars to the level of mechanism analysis, though we recognize our own current gap in promise and delivery.⁴⁹

The goals outlined in this paragraph are laudable ones, except that they depart from the analysis offered. It is precisely the mechanisms regarding the gap between text and practice that become challenging to study through a positivist lens. If one wants an account of the separation between text and practice at any given point, then it seems that one would need an account of the relationship between the forces shaping veneration and decay at a particular point — how those forces interact with one another, and the reasons why some forces triumph over others.⁵⁰ Without this, it would be a little difficult to arrive at the conclusion that Elkins, Ginsburg, and Melton reach. They determine that there is a “relatively small” maturation effect, conditional on regime type and judicial independence, and that this small effect is “remarkably stable across groups of rights.”⁵¹ But such a conclusion is consistent with a scenario where rights have an enormous amount of causal significance, but have, on balance, limited impact because their power is being counteracted by other forces.

Whether legal measures can make any real difference is similarly posed in a Melton and Ginsburg study on judicial independence.⁵² In this case, the authors observe that the past few decades have witnessed greater sensitivity to judicial independence, and that this has manifested in new constitutions containing explicit provisions that intend to protect the independence of the judiciary (such as security of tenure, remuneration, and so forth).⁵³ Given this changed landscape, they ask whether the rise in de jure judicial independence has led to a rise in de facto judicial independence. Does the presence of formal legal protections in the case of judicial independence result in judges in fact being more independent? Here, the conclusion that is offered is slightly less skeptical about the nature of the connection than in the aforementioned studies: “[D]e jure judicial independence is correlated with de facto judicial independence, but the effect is limited to those provisions that are self-enforcing as a result of competition between the executive and legislative branches.”⁵⁴

This study suggests that many forms of de jure protections to enable judicial independence do not lead to de facto changes. It hopes to capture the difference between protections that do make a difference (those that focus on the appointment and removal of judges) and those that do not.⁵⁵ To see the potential limitations in this study, consider two separate observations made by the authors. The first is that countries might find it tempting to formalize judicial independence when the judiciary already has credibility; and, thus, “we might observe constitutionalization of de jure independence after increases in de facto independence.”⁵⁶ In other words, they note that it “would not be surprising to find a reciprocal relationship between de jure and de facto independence,” and they register their interest in observing the correlation at work.⁵⁷ The second observation is the authors’ response to the potential concern that “some unobserved factor may be causing any correlation that we find between de jure and de facto judicial independence.”⁵⁸ Milton and Ginsburg proceed to state that:

For instance, deference toward the judiciary may be a societal norm, which prompts constitutional drafters to create strong protection for the judiciary’s independence and politicians to respect the judiciary’s independence in practice. We cannot rule out this possibility in the analysis below, but our emphasis on the specific mechanisms of appointment and removal suggests that there is something about these particular attributes of de jure judicial independence. If a third factor was causing both stronger formal protections and de facto independence, one might expect that all of the formal protections that we identify would be equally correlated with practice.⁵⁹

Both of these observations are intended to insulate the findings from the potential criticism that there is no correlation to be observed — that is, from the claim that provisions relating to judicial independence in a constitutional text do not make any difference to the reality of judicial independence on the ground. Even though the major part of their conclusions is consonant with this thought, the observations are meant to defend the narrow set of instances where they find that de jure protections do indeed matter (instances, as we observed, relating to the selection/removal of judges). But, we might ask, what if we do not raise this expected objection to the analysis but instead address the large chunk of situations where the authors find that de jure safeguards have no impact? What if we turn our attention to situations where they “find that none of the de jure attributes has an independent and statistically significant effect on de facto judicial independence when control variables are included”?⁶⁰ Here, it does seem that we have an analogous situation to that observed in the aforementioned studies. The impact of de jure protections might well have been extraordinary, but, in specific societies, their presence might have had to contend with factors that made a de facto change difficult. One reason why such a hypothesis is far from implausible is that the factors that enable de facto independence are complex and varied — a matter underlined by the fact that we can have de facto independence with de jure dependence, and we can have de jure independence with de facto dependence. In the United Kingdom, for example, judges were appointed by the Lord Chancellor with no political restrictions until the Constitutional Reform Act 2005,⁶¹ but there were no serious concerns about a lack of judicial independence in the country. Alternatively, as regards de jure powers, an important feature associated with them in many liberal democracies is that they are often not used. The Queen’s veto in the United Kingdom is, of course, a common example.⁶²

In discussions on causal studies in comparative constitutionalism, the concern that is typically expressed is that of an omitted-variable bias.⁶³ The scholarship that I have studied is usually very thoughtful in this regard. In How Constitutional Rights Matter, for example, Chilton and Versteeg are sensitive to this matter. They acknowledge this concern and try to meet it by supplementing their statistical efforts with additional statistical efforts. As they observe, their work relies on “ten different models to estimate the effect of each of the rights.”⁶⁴ But though this effort is of course commendable, it is not entirely clear that it does the required work. The effort to remedy a problem within a model may not quite be resolved by additional modeling if the limitation is arising because of the methodology of modeling and not the particular model in question. If the problems arise because it is only a contextually sensitive understanding — and likely one that cannot be entirely reduced to a quantifiable unit — that can hold some promise of ascribing causation by offering us the dynamics that interact with one another at any given time and space to produce a causal effect, then a new model may not quite help us. One might also separately note that the use of multiple models will have limited value beyond a point because many outcomes are likely to be highly correlated — the social and cultural variables that are missing in one analysis are likely to be missing in others, cautioning us against treating each model as an independent piece of evidence. In large-scale comparative efforts, the potential problems rise to a whole new level because the social and cultural factors at play will, given the variations involved, be irreducible in different ways in different contexts.

To appreciate the kind of problem that I am gesturing at, we must go beyond merely thinking about omitted variables and focus on the ascription of causality. In order to do so, we must interrogate the counterfactual theory that is being utilized in How Constitutional Rights Matter and other positivist work. Simply put, counterfactual accounts rest on the idea that an outcome does not obtain if a particular action is not performed. But such a simple description is not, of course, enough for a causal theory to be successful. If, for example, we assert that the nonperformance of X act by A implies that Y event does not occur, then our assertion is incomplete. This is because, for our account to work, it additionally needs to be the case that A does not perform some other action that brings Y about. As it has been observed, in an example where “Suzy throws a rock at noon, breaking a bottle,” the analysis “must be supposing that, in the relevant counterfactual situation in which Suzy is not, at noon, throwing a rock at the bottle, she is not doing anything else that would lead to a bottle-breaking: she is not starting to run up towards the bottle to level a kick at it; she’s not throwing some other object at it; she’s not shooting her slingshot at it; etc.”⁶⁵ There can be causation without any counterfactual dependence, and there can be a lack of counterfactual dependence without a lack of causation.⁶⁶ As Professor Michael Moore once put the matter in plain and simple terms, “causation is distinct from counterfactual dependence.”⁶⁷

The aforementioned example of Suzy and the bottle appeals to the phenomenon of redundant causation — a situation that is typically understood to be one “where there is more than one event that is, in some sense, enough for the effect that occurs.”⁶⁸ Seen as one of the trickiest issues in the philosophy of causation, the phenomenon of redundant causation calls into question the validity of numerous straightforward counterfactual causal accounts that seem, on the face of it, to be plausible.⁶⁹ There are, as one might imagine, many ways for phenomena to occur. For example, there may be cases involving common causes; there may be situations of preemption, of both early and late kinds; and there may be cases of overdetermination.⁷⁰ And there can also be omissions that result in causation, a fact that reminds us that our interest cannot only lie in whether the performance of X act by A is enough for the occurrence of Y event by itself, but must also lie in actions that can interfere and intervene in the causal account.⁷¹

For present purposes, we may limit our focus to the fact that sometimes an outcome can obtain even if the action that we associate with the outcome is not performed. Professor Richard Tuck provides us with a helpful example in this regard by asking us to consider “the case of a policeman who shoots and kills a bank robber, and does so a split second before one of his colleagues would have done.”⁷² Here, Tuck observes: “The first policeman caused the robber’s death — if his action did not do so, what did? But if this policeman had not fired, the robber would still have been killed.”⁷³ The basic upshot is that, if we are to retain some notion of the idea of causation, then an easy counterfactual story may not do all the work that is required. An act can have a causal role to play in the generation of an outcome even if, in the absence of the act, a counterfactual narrative does not give way to a different outcome.

This reminds us that when we think about actions — when we attribute causality — we are thinking about sufficiency.⁷⁴ A person who performs an action asks whether or not their action will bring about the outcome that they desire — that is, whether their action will be sufficient to bring out this outcome. Our interest is not in necessity. Even if an outcome might have come about in some other way, and therefore the act in question was not necessary to bring about the outcome, a causal relationship is established if the action was enough for the outcome to occur. Simply put, causation takes place when our actions are adequate for something to happen. The emphasis on sufficiency rather than necessity makes it very difficult to explain matters in entirely counterfactual terms. This is because causal theories that rest on counterfactual foundations operate within the framework of necessity.⁷⁵

The examples that we have studied, in How Constitutional Rights Matter and elsewhere, relate to the phenomenon of redundant causation in an interesting way. In ordinary cases of redundant causation, we are dealing with situations where an outcome would have come about even if one did not act — and the key point here, as we have noted, is that despite this fact, we still regard it to be a case of causation.⁷⁶ In some of our examples, though, the facts are somewhat different. Here, the cases do not involve the same outcome being generated regardless of whether or not one acts but instead relate to the same absence of an outcome. In these situations, it does not follow that there is no causal role being played by the law or the legal institution because it may well be that the persistence of the phenomenon is now occurring because of a new phenomenon or because of some additional existing phenomenon. Thus, it may well be that there are other barriers apart from the law to the creation of a society that is free of torture, and it may well be that the legal prohibition on torture has broken down one important barrier.

As Moore observes in Causation and Responsibility, “cases (of more than one sufficient condition for an effect) bedevil the counterfactual theory of causation, because if each of two or more conditions is sufficient for e then none of such conditions is (individually) necessary for e.”⁷⁷ As one can easily notice, situations like this “present no equivalent problem for a sufficiency theorist; that each condition is sufficient is enough to count it as a cause, irrespective of the fact that such condition is not necessary.”⁷⁸ It is not fantastic to suppose that an outcome like the end of torture may well be conditional on additional factors beyond changes in legal rules. Inverting the standard examples, we might say that even if the legal prohibition on torture was not sufficient to end torture, it does not follow that it had no potential impact. The fact that something (such as a legal measure) was not sufficient to bring about an outcome is not enough to establish the absence of a causal relationship. Because an act or event was not sufficient to generate a particular outcome, it does not mean that it was not necessary. Thus, it may not be quite right to suggest that because the prohibition on torture did not end the practice of torture, the practice of torture would have persisted in the same way had there been no such prohibition.

What the phenomenon of redundant causation teaches us is that the challenges involved in making causal claims are not limited to those relating to precise measurements. Rather, the problem is that if we have more than one possible cause for an outcome, then however precise our measurements may be (assuming the possibility and accuracy of quantification and the like), we will at some level be attributing causation. One might suggest, at this stage, that such a problem could exist for causal studies regardless of whether they are cross-national. This is true. As such, the concern posed by the phenomenon of redundant causation would apply to studies of causality as a general matter. A regression technique may not, by itself, give us an account of the forces and factors necessary to provide particular kinds of causal accounts even in the case of single-country studies. Though this may well be correct, the problem of redundant causation becomes far more difficult to mitigate in the case of large-N comparative constitutional studies. As we have observed, attention to sufficiency rather than necessity encourages us to think carefully about the precise mechanisms by which actions take place.⁷⁹ We need to understand how processes occur — how, for example, the level of torture in practice remains the same even after torture has been legally prohibited (assuming that the identification and measurement of torture is accurate). If we are sensitive to context and pay careful attention to the mechanisms that are interacting with one another in the production of a certain outcome, then we hold some chance of mitigating the concern posed by the phenomenon of redundant causation. But such mitigation is a challenge in the case of large-N studies because, in abstracting away from context, it is hard for us to unpack the factors and forces at play at any given moment. In other words, the extent to which we attribute causation in such studies will be far greater not least because redundancy may well vary across contexts.

One might ask, at this stage, why we would care about how processes take place. Why should it matter to us how torture rates are persisting in practice despite changes on the law books? Should we not concentrate on the outcome that is generated, namely the fact that the rate of torture remains the same? This is a fundamental and worthy question, though it invites a relatively simple response: the reason we value knowledge of the processes by which an outcome comes about is because it is required for us to bring about the outcome that we care about. Unless we have some understanding of how torture is working in society, and what legal and extralegal forces are involved in its practice, we cannot hope to eliminate torture. So we should care about the processes — the question of how — if we actually hope to achieve the outcome that we desire.

II. Interpretation

Thus far, our discussion of positivist comparative constitutionalism has largely addressed the problem of causality — how we think about the relationship between inputs and outcomes. In large-scale, cross-national analyses, we are also faced with a further issue: the interpretation of practices in different places. This is an issue to which we shall now turn.

In thinking about issues of collective meaning and shared practices, we may begin with a somewhat different strand of scholarship in positivist comparative constitutionalism. Consider, for example, studies that focus on constitutional texts and the question of how we should read and code constitutions. A valuable illustration of this scholarship is a study by Versteeg and Professor David Law that presents “an empirical account of the global evolution of formal constitutionalism” by using “a comprehensive new data set covering the rights-related content of all national constitutions in the world over the last six decades.”⁸⁰ This study concludes that ninety percent of the difference among constitutions can be explained by two variables: comprehensiveness and ideology.⁸¹ In some cases, “constitutions are succinct and tend only to contain relatively generic rights, while others also encompass less commonly encountered, relatively esoteric provisions.”⁸² Secondly, while some constitutions are “relatively libertarian,” others are “more statist in character.”⁸³ The authors further observe that rights are becoming more pervasive, that the explicit power of judicial review is spreading, and that the character of rights is becoming more generic.⁸⁴ Finally, the authors use the data to place constitutions in a relative position to one another.⁸⁵ Based on the rights and duties that such texts recognize, they can be situated on an ideological spectrum.

The question that arises with this mapping of rights variation — in particular, the forming of a relation of content with ideology — is whether it can be undertaken without actually interpreting the constitutional text. Is it possible to know what content a right has, the kind of role that it envisages for the state, without interpreting the right? Without knowing what a right means, is it correct to assert that though some constitutions “epitomize a common law tradition of negative liberty,” others “presuppose and enshrine a far-reaching role for the state in all aspects of life”?⁸⁶ The variable used during coding is binary (a constitution either contains a particular right, say the right to free speech, or it does not contain a particular right). Though Law and Versteeg undertake their study with much care, this surface-level analysis of constitutional texts may potentially hide more than it reveals. Constitutions can contain similar — even identical — expressions but radically different rights. Indeed, on certain occasions, the specific intention will be to ensure that expressions that appear similar mean something very different.⁸⁷

This concern is captured by the qualifications in the study. The authors do not code limitation clauses, observing that “[l]imitation clauses that purport to limit the scope of rights in a constitution, often in a boilerplate or blanket manner, were not coded for a variety of reasons.”⁸⁸ In How Constitutional Rights Matter as well, Chilton and Versteeg observe that they “did not take account of limitation clauses . . . [as] it is usually impossible to determine the extent to which rights can actually be limited” (p. 82). But limitation clauses — and indeed, the interpretation of limitation clauses — may result in entirely different rights. This dynamic presents a potentially serious concern with such studies: a constitution may grant a right and then contain a range of limitations at both the constitutional and statutory level.⁸⁹ If the drafters of constitutions engage in the task of constitution-making knowing the context within which the text will operate, then they will naturally reflect on how the provisions that they are writing will be interpreted. The textual similarity on the face of two or more constitutional texts hides crucial differences, differences that are both lost and potentially mischaracterized when reduced to a binary variable. As Tushnet has pointed out with reference to How Constitutional Rights Matter: “Ignore limitation clauses and you’re not measuring the presence in a constitution of constitutional rights with substantive content but rather the presence in the constitution of certain words.”⁹⁰ Moreover, this aspect of the methodology will also influence the ideological mapping that the study performs. What is ultimately likely to ensue in the characterization of constitutional texts, given that interpretation is not independent of content, is some degree of circularity: how one codes will determine the outcome.

The question of interpretation is thrown into sharp relief by two further studies. The first is a study of “sham constitutions.”⁹¹ Here, Law and Versteeg proceed from the interesting and important observation that constitutions often guarantee a great deal on paper but deliver very little in practice. Countries can claim several protections and rights in their constitutional texts, but their reality in practice may be hindered by political repression and suppression. Given this fact, can we measure the gap between constitutional guarantees in theory and in practice? The authors draw upon the same dataset in the aforementioned study — one that “covers the rights-related provisions of every constitution in the world over the last sixty years” — and they “assign scores and rankings to countries that reflect the extent to which they actually uphold the rights found in their constitutions.”⁹² Moreover, they “identify the constitutional rights that are most often violated in practice, the regions where sham constitutions are most common, and variables that predict the occurrence of sham constitutionalism.”⁹³

As per the study, the interpretation of a constitution is a different matter to the content of a constitution.⁹⁴ But to measure “sham constitutions,” one needs to know what a constitution says. What the authors appear to miss is that they might well be measuring variances in interpretation rather than gaps in performance. Indeed, a reading of their own data suggests the possibility of such a supposition. In assessing which rights are likely to be violated, they contrast women’s rights, the prohibition against torture, the right to a fair trial, and so on, where they find that compliance rates are low, with guarantees such as “a constitutional bar against the death penalty” which were generally honored.⁹⁵ (To be more specific, the compliance rate of the prohibition of torture in countries with the constitutional guarantee is 12.3% whereas the compliance rate of the prohibition of the death penalty in countries with the constitutional guarantee is 100%.⁹⁶) If we reflect on this data, we may notice that one crucial difference between, say, guarantees against torture and guarantees against the death penalty is that the latter has no interpretive looseness: you are either dead or alive. The fact that the death penalty is an instance of perfect compliance may well indicate that the real problem here is not the gap between theory and practice but rather differences in interpretation. (Of course, what we might also be seeing here are problems with measurement — the measurement of the death penalty is straightforward; that of torture is far more difficult.)

The downplaying of the problem of interpretation is captured even by certain observations in the Law and Versteeg study. Consider the following statement: “In substance, it is clearly a form of sham constitutionalism for a regime to pay lip service to the values of the global community by including the world’s most popular rights in its constitution, only to gut those rights of meaning in the name of constitutional interpretation.”⁹⁷ But, here, it is possible the authors may have the matter backwards. The issue in this example may well not be the gutting of rights in the name of interpretation but rather that global practices themselves might have some kind of interpretive looseness. It is possible for there to be genuine disagreement about the meaning of a legal provision.⁹⁸ To put the point simply, the “values of the global community” are not a social fact.

Interestingly, this issue is brought out by a different study that focuses specifically on interpretation. Observing that there can be divergent understandings over what constitutional documents convey, Melton, Elkins, Ginsburg, and Professor Kalev Leetaru ask: “[T]o what degree do citizens and elites agree about what constitutions say and, assuming some variation, which factors affect relative levels of interpretability?”⁹⁹ The aim of this study is to measure interpretability, which is defined as “the ability to produce inter-subjective agreement about the meaning of a text.”¹⁰⁰ By making multiple coders read a constitutional text, the authors are able to show that though the interpretability of constitutions does vary, this variance is not context dependent. “Constitutions written in bygone eras, in different languages, or in extremely different cultural milieux are,” they observe, “no less interpretable by readers than are those written in closer temporal and cultural proximity.”¹⁰¹

The noticeable feature of a study of this kind, a feature revealed by how the authors define interpretability, is that interpretability is taken to be a technical, factual characteristic of a text. But interpretability is not merely the same thing as agreement, which is what is actually being measured. It is substantially different. Interpretability is a hermeneutic enterprise involving an application of mind rather than a fact about a text or the world that is, in some sense, existing in physical space. There is, plainly put, no metric of “correctness” in the interpretation of a constitution. Claims regarding interpretation are often claims about how the consensus over what a text means is incorrect. The framing of this study makes it impossible for there to be broad-ranging agreement on an interpretation of a text that is the wrong interpretation.¹⁰² As per the study, it would seem to be the case that the correctness or incorrectness of an interpretation depends on how many people hold that interpretation. Interpretation, however, is not about just following what people think a text means — one comes to one’s own interpretation of a text using a range of techniques and methods. In this study, by contrast, the methodology seems to have swallowed the subject of interpretation. The risk would be that the subject matter has been redefined to accommodate the method rather than the other way around.¹⁰³

If we do not attend to constitutional meaning, the potential problem is that we may not know what is being captured. Notice, for instance, that in the study of constitutional amendment rules that we considered previously, we observed that the study does not account for the fact that procedural rules could incorporate cultural attitudes and social norms. It is now possible to see that there might be a further problem with this study, one that underscores how problems relating to causality and interpretation intersect with one another in large-scale, cross-national work — though the study acknowledges that some amendments can involve major changes (a single amendment might replace a great many provisions), the change brought about by an amendment cannot be ascertained by merely seeing the number of provisions that are altered. The problem is not a limitation in the factorial system. We need to understand what the provisions mean. In certain cases, where cultural contexts, philosophical assumptions, and linguistic traditions are sufficiently similar, what these studies might be capturing — to adopt a Rawlsian distinction — are different conceptions of the same concept.¹⁰⁴ Here, there may well be some overlap at work, though it would not quite be the kind of overlap that is reducible to a binary variable. (In principle, of course, it is possible to have a continuous variable rather than a binary variable, but assigning precise point values would open up another interpretive can of worms.) In other instances, however, all that these studies might be documenting is something like the counting of letters in a word or the counting of pages in a book — these are efforts that do not inform us about the meaning of the word or the meaning of the book, respectively, even though they capture similarities across some metric. As we can see, the concern here does not relate to a coding error and cannot be resolved by a different coding structure and new coding rules.

With regard to the genre of positivist comparative constitutionalism that we have been considering, it would seem that there are two possibilities. The first is that the work seems to offer us a thin rather than a thick description of some kind of behavior. As Professor Clifford Geertz emphasizes in The Interpretation of Cultures, thin descriptions do not allow us to understand human activity.¹⁰⁵ The second, potentially more worrisome, critique would be that we are not even arriving at thin descriptions of any kind, because the concepts at work as we understand them in one culture are different from those operating in another culture. How do we know that we are actually comparing torture across two cultures without an assessment of how the cultures understand the infliction of pain and the character of violence? It may well be that we are not comparing different approaches to torture but different things. The use and deployment of controls will not quite address this problem because it is not a problem with the causal account that is being offered — the problem, on this occasion, does not relate to how certain factors interfere with the phenomenon that we are interested in studying, but instead relates to whether we are actually studying the phenomenon we are interested in studying.¹⁰⁶ In other words, the concern here is not measurement and the subsequent attribution of a causal account but the very identification of something. Of course, as we have just observed, the concerns that relate to causation and interpretation are linked with one another in key ways — we are faced with having both to identify and measure inputs and outcomes, as well as the forces and factors that are interacting with one another within the chain of causality.

We can begin to appreciate the coming together of the problems of causation and interpretation if we notice how the positivist approach is a form of behaviorism and seems — by externally observing behavior — to make a science out of comparative constitutionalism. Whether in attempting to assess the impact of torture prohibitions or in judging the gap between rights-based guarantees and violations, the emphasis is on similarities that can be observed from without. The focus, in other words, is on patterns rather than practices. On occasions, like the study on amendment rules by Ginsburg and Melton, the contributions register their interest in patterns and critique other positivist work for the ways in which patterns have been understood.¹⁰⁷ But it is not fully clear whether this approach can resolve matters. The answer to an imperfect correlation may not be a better correlation if something other than a correlation is going on. It may not be a solution if there is a practice that is taking place. A practice is not the same as a pattern; it cannot be reconstructed as a correlation. Because the inputs may not correlate to how people actually think, whether the predictive power of our work is high is somewhat beside the point. In focusing on how matters are represented — as brute facts — this kind of positivism carries the risk of reviving Émile Durkheim’s pre-hermeneutical world, where people can be studied as objects.¹⁰⁸ It is in this kind of deeper way that positivism makes a science out of comparative constitutionalism.

The effort to respond to an imperfect correlation with a better one has its origins in Milton Friedman’s suggestion that “theory is to be judged by its predictive power for the class of phenomena which it is intended to ‘explain.’ . . . [T]he only relevant test of the validity of a hypothesis is comparison of its predictions with experience.”¹⁰⁹ For Friedman, in assessing the assumptions behind a theory, one should not attend to whether such assumptions are realistic but rather “see[] whether the theory works, which means whether it yields sufficiently accurate predictions.”¹¹⁰ But as philosophers of social science have long argued, the Friedmanite “black-box” approach ignores the mechanism by which outcomes are generated.¹¹¹ Without attention to the mechanism at work, we cannot have confidence in the results that emerge. What we have is pure induction rather than an explanation for what has occurred. The concern with not knowing the internal structure by which a phenomenon unfolds has long troubled scholars of causation.¹¹² One of the key principles of statistical inference is that a description of a series does not reveal how a series will unfold — a coin toss being an easy and simple example.¹¹³ As John Maynard Keynes argues in A Treatise on Probability, there is much reason to be careful when assessing the “mere repetition of instances.”¹¹⁴ “Pure [i]nduction,” Keynes suggests, “can be usefully employed to strengthen an argument if, after a certain number of instances have been examined, we have, from some other source, a finite probability in favour of the generalization.”¹¹⁵ The key emphasis here is on “probabilities, however small, derived from some other source.”¹¹⁶ Induction, by itself, cannot do the work for us.

There is now a large body of work that shows that though correlations can observe the regularity of behavior, they cannot capture the internal logic of it. When we think of studying human behavior, we are interested in studying the actions of rational creatures capable of thinking and reasoning: creatures who can act. Other creatures sometimes reveal behavioral patterns that might suggest similarities with our patterns, but we view what they are doing differently from what we are doing.¹¹⁷ A behaviorist analysis reveals how a certain action looks but not how it is. The behaviorist cannot really tell us what people are doing. The assessment of social meanings from the outside poses challenges for our work. Unless we have a sense of the underlying forces that are resulting in an outcome that we note, we cannot know whether the model that is being used for the analysis is reliable.¹¹⁸ Without such knowledge, we are trying to interpret social behavior in some way, but we cannot achieve an understanding of what is occurring. We cannot come up with an explanation, because we cannot locate an event within the actual participants and circumstances that relate to its existence.

By focusing on behavior from the outside, the positivist approach in comparative constitutionalism carries the danger of not seeing human actions as those performed by individual agents who reason and act in a specific context.¹¹⁹ As rational beings, we have some idea of a shared life.¹²⁰ We live and function within a community; and when we follow rules, we are engaged in a practice.¹²¹ When we are outside of the shared agreement that constitutes rule following, it can be very hard to comprehend the behavior of those within the community in question. “If a lion could talk,” as Wittgenstein famously observed, “we wouldn’t be able to understand it.”¹²² There is a kind of argument about what it means to understand a practice; to understand how it has meaning for individuals. This is the reason why one can have the same referent but understand matters in a very different way.¹²³ We cannot do very much with a description of similar patterns without certain background ontological assumptions, and the idea of similarity misses the thought that there can be tremendous behaviorist consensus alongside great disagreement over meaning. Studies in positivist comparative constitutionalism — whether on torture prohibitions, sham constitutions, constitutional interpretability, and the like — do not seem to fully appreciate how an orbit of shared meaning contains concepts within it. These concepts can be grasped only in relation to other concepts within that orbit, and one needs to be under the same orbit of meaning to understand the concepts. It is only then that we can formulate some sense of the actions that people are engaged in.¹²⁴

If we do not see actions as having a cognitive content, if we see them as mere behavior, then it is akin to seeing the physical acts that people perform — raising their hand at a faculty meeting or falling down after walking on ice — and forming an opinion of what they are doing or how they are thinking. There is no accounting for the internal reasons that persons have, their forms of deliberation, or their kinds of motivations. As a result, there can be no appreciation for a person’s reasons for action in any given instance and thereby no explanation for their actions.¹²⁵ Under such accounts, humans are not conceptualized as agents. They are not seen as beings who have motives and purposes, who form intentions and act on them. In other words, humans are not taken to be conscious beings.¹²⁶ As lawyers, we must be especially attentive to this not least because individual agency is so central to our legal world. It is at the heart of our understanding of the rule of law and legal procedures, across their varying conceptions.¹²⁷ It is an idea that also finds articulation in legal doctrine across a range of areas. When, to offer merely one example from public law, Germany’s Federal Constitutional Court refused to permit an aircraft with innocent passengers to be shot down to save other lives, it did so because the impugned law did not take seriously the individuality — the dignity and agency — of the passengers on the aircraft.¹²⁸ Cases such as these sit uneasily within the positivist vision. For the positivist, the ideas of responsibility and agency are hard to acknowledge because the domain of internal reasons is deemphasized. In reality, however, our behavior cannot be assessed from the outside because it does not exist or operate independently of our understanding. Rules that are made by and that apply to human beings — including legal rules — do not operate in a cognitive vacuum.

III. Comparativism and Constitutionalism

The questions that How Constitutional Rights Matter and other positivist scholarship ask are of profound significance. Importantly, however, the reason why such work might concern us is not only because of its methods and findings, but also because it carries fundamental implications for legal rules and institutions in society. One of the key features of the social sciences is that, in studying meaning, we also generate meaning for our own lives.¹²⁹ The projects we pursue, the surveys we undertake, and the data we decipher are themselves sources of meaning. And, in becoming sources of meaning, they become guides to our conduct. Thus, when we conclude that torture prohibitions do not impact the practice of torture, that conclusion shapes our understanding of the value of such prohibitions. When we believe that the correct or incorrect reading of a constitutional text has nothing to do with interpretation, that belief alters how we understand such texts.

There are two reasons why the social sciences pose a special challenge. The first reason is that the enterprise that is being undertaken — the very doing of comparative constitutional law — is absorbed into what provides meaning. When torture is being studied in a country, it means that human beings are studying other human beings. This is the key difference between the physical and social sciences — and one that is crucial to highlight in reflecting on the enterprise of comparative constitutionalism. In the former case, the objects of study do not pick up the methodology and knowledge that is at work.¹³⁰ In physics, for example, the planets themselves do not incorporate an understanding of science. Because human beings act in the world, the principles that they incorporate and the activities that they perform feed into one another.¹³¹

The second reason is that, as human beings, we can act differently. We ourselves are agents, capable of behaving in ways that are different from the ways in which the people that we study have behaved. Legal theorists have long been attentive to this fact. As Professors H.L.A. Hart and Tony Honoré put it in their seminal work Causation in the Law: “The idea that individuals are primarily responsible for the harm which their actions are sufficient to produce without the intervention of others or of extraordinary natural events is important, not merely to law and morality, but to the preservation of something else of great moment in human life.”¹³² Within the law, we acknowledge — across a range of domains — that individuals are separate persons and that they can act distinctly.¹³³ When we think about legal responsibility, we think about how the individual whom we choose to hold responsible exercised their own judgment and performed their own actions.¹³⁴ The emphasis is true even for constitution makers, who often internalize the legal reality that is present in another country and differentiate it as part of a process that is also mimetic — a fact that is hard to capture if we focus on behavior and compare patterns, as if legal events in two separate countries are two entirely independent events and are two natural responses to randomly occurring underlying conditions.¹³⁵

Even those who have been skeptical of the establishment of causal relations have had to confront the reality of human experience. David Hume, who saw that relations of causation had an air of mystery to them, acknowledged that one simply had to accept the limits of philosophy if one needed to acknowledge what it meant to be human. Imagining that we cannot attribute meaning and causal relations to our actions will fill us with “the deepest darkness”;¹³⁶ one will feel “utterly depriv’d of the use of every member and faculty.”¹³⁷ Instead, Hume suggested, one needed to look the other way — “play a game of back-gammon” — when one was burdened by trying to explain the world.¹³⁸ Hume grasped that even if something cannot be fully explained philosophically, it might still play some role in our life because of some kind of human necessity. Even if we could not quite understand how causal relations work in theory, we would need some account of causation to function in the practical world. Without some kind of an explanation to ourselves about how one event leads to another, we would have no way to live. Because the positivist comparative constitutional lawyer seems to eschew explanations of how outcomes emerge, they carry the risk of abandoning the idea of agency.¹³⁹ But, as Hume pressed upon us, without agency, we would struggle to have an account of how to live our own lives.

The concerns that I have identified may have some validity to them, but they raise the question of what is possible within comparative constitutional law. In exploring this, the starting point lies in underscoring the importance of studying societies from within — in moving beyond behavior. If one views external practices and study systems from without, one risks concluding that people are doing the same thing by virtue of the fact that they might appear to be doing the same thing.¹⁴⁰ A turn to context does not only promise us a fuller account of what might be taking place in any country. It also is truer to our commitment to constitutionalism. After all, constitutionalism, in its modern democratic sense, is about finding the legal ways by which we can empower ourselves, exercise agency, and structure our collective life.¹⁴¹ The means by which agency can be enabled — how sovereignty can be expressed; how amendment rules should be written; how the relationship between representation and the people might be framed — is, and rightly so, the central preoccupation of constitutional law scholarship.¹⁴² Indeed, the reality of constitutionalism only heightens the importance of context because the practices at work are normative; they are a form of political agency in action. Recall the study of constitutional amendment rules, where we observed how rules might assimilate extralegal norms. We noted how this study illuminates concerns relating to both causality and interpretation, and their relationship with one another. In recognizing how the actors in a constitutional order might incorporate the constitutional culture that exists — procedural rules, we observed, cannot be examined independently of this culture — we can also notice the force of the agency-related argument that we have been emphasizing, and the mutuality between the considerations pertaining to causality, interpretation, and agency. Rules exist in a space where actors think, change, and act. By adopting a behaviorist lens that is bounded in time and space, we risk losing sight of the idea that humans not only are present in the world but also act in it. To appreciate the latter — to see how humans internalize laws — one needs to study how humans think and understand, how they create and form meaning.

In studying societies from within, there are two possible ways in which we might proceed. The first approach adopts a perspective that is internal to a legal system. It usually focuses on doctrinal research, and attends to the inner logic and structure of legal reasoning and legal systems.¹⁴³ Here, comparative research proceeds on the basis that two or more countries have sufficiently similar legal rules and modes of legal argument such that it is intelligible and reasonable to contrast them. The goal here is to better understand one’s legal system as well as to see how legal doctrine within one’s system can evolve. The emphasis in such work is typically on legal interpretation as we traditionally understand it. That is to say, our attention is on how one should understand the meaning of a particular constitutional provision within one’s legal system.¹⁴⁴ As we know, the judicial invocation of such an approach has not been without controversy, especially in the United States.¹⁴⁵ But such critiques are misguided for they see foreign law as extralegal and therefore illegitimate, missing the distinction that Hart once made between sources of law that are mandatory and those that are nonmandatory but nonetheless legitimate.¹⁴⁶ What is important is not whether a court cites a foreign case but rather what reasons it offers to justify that citation. The assessment of those reasons is itself a matter of reasoning; and this is true not merely for the citation of foreign materials but for all nonbinding sources. A wide body of literature has helped us appreciate how doctrinally oriented research might be performed in the context of comparative constitutional law; how judges, lawyers, and scholars might engage with legal materials from legal systems other than their own.¹⁴⁷ And in the courtroom, such practices continue to flourish globally.¹⁴⁸

A second approach is one that is internal to a particular country or set of countries but external to legal doctrine. This approach begins from the premise that constitutions arise and operate in specific contexts, and considers the relationship between legal developments and historical, sociological, and political changes.¹⁴⁹ Rather than focusing on courts and on case law, such work attends to questions of power and legitimacy, to the role of incentives and ideologies, to the workings of social forces and civil society, and to historical pressures and trends. It will, of course, be important for this work to attend to legal doctrine, though the extent of attention is likely to vary across different studies. Recently, scholars have noted the need to ensure that comparative constitutional law is not entirely subsumed within external accounts of legal developments,¹⁵⁰ and there is of course a long tradition of thought that explores how any truly explanatory account of the workings of the law in society will have to be sensitive to the internal structure of legal argumentation.¹⁵¹ Although doctrine will play some role in external accounts, the character of legal reasoning and legal developments will usually be understood as a subset of, and/or sharing a dynamic relationship with, the broader realities within a society. Instead of focusing exclusively or even primarily on the inner life of the law, the attempt within this approach will be to unmask what are understood to be the real forces that drive change and shape outcomes. Here, as in the case of doctrinally oriented work, the field of comparative constitutionalism has evolved with both single-country studies — which enable the field to grow through comparative judgments that might be formed by studying these different single-country studies — as well as studies where more than a single jurisdiction is considered. The countries that are contrasted may, as is familiar to social scientists, be similar and different in various ways. The key point is that they are considered to be comparable with reference to a particular metric — one identified based on the question under consideration — and that the study is sensitive to the contexts of the jurisdictions under study.

Though the distinction between studying countries from within and without — the latter being the Martian perspective — is relatively underdeveloped in comparative constitutionalism, the distinction between internal “legal-interpretive” approaches and external “causal-explanatory” accounts is well established in the field.¹⁵² Each approach has a long and venerable tradition. With regard to the former approach, the practice of foreign engagement in doctrinal contexts — as Professor Vicki Jackson once reminded us — goes back in the United States to at least the late nineteenth century.¹⁵³ In studying this practice, we hold the promise of furthering our understanding of how participants behave within a legal system. As regards the latter approach, the comparativist turn in the nineteenth century offers us a range of examples stretching from Henry Maine to James Bryce who compared legal systems and governance structures in different historical and institutional ways.¹⁵⁴ There are various means by which external “causal-explanatory” accounts might be developed, and these include the use of empirical, statistical, and quantitative methodologies. What is crucial is not whether one brings, say, an ethnographic or a historical or a statistical approach — each of these will have something to add and each of them will carry their own shortcomings. Similarly, what is important is not whether the research focuses on how nonlegal actors approach the law or on whether one tries to explain social and institutional changes over time. Rather, what is key to such accounts is the effort to understand a country within its context.¹⁵⁵ Consider, for example, four books on judicial power that are each crucial contributions to our comparative understanding of courts: Ginsburg’s Judicial Review in New Democracies, Professor Ran Hirschl’s Towards Juristocracy, Professor Samuel Issacharoff’s Fragile Democracies, and Professor Theunis Roux’s The Politico-Legal Dynamics of Judicial Review.¹⁵⁶ Each work presents an account of the growth and workings of judicial power in different terms — crudely put, Ginsburg and Hirschl focus more on the overall power dynamics in society, whereas Issacharoff and Roux pay more attention to legal practices. We might find the arguments in one work more persuasive than the arguments in another, and we might find each work persuasive and unpersuasive in different ways. What is noteworthy, however, is that each of these works takes the idea of a case study seriously, even when the works offer multiple case studies.

The two approaches — the “legal-interpretive” and the “causal-explanatory” — are not without their methodological concerns and limitations. Those who favor the former will often worry that the latter pays insufficient attention to what people are actually doing when they are making legal claims and delivering legal judgments. Those who are partial to the latter will usually emphasize the importance of seeing the law as a product of its broader sociopolitical context. When it comes to comparative constitutionalism, in particular, scholars must not only negotiate the truths on either side of these approaches but also confront the problem of comparison as a more general conceptual challenge.¹⁵⁷ Comparative constitutionalism is a difficult enterprise because it is not always clear that the modes of interpretation in one jurisdiction have similarity with those of another. There are several concrete particularities — cultural, hermeneutic, institutional, and so on — that are often idiosyncratic to a particular system.¹⁵⁸ This poses challenges for both the identification of a practice and the establishment of causality. But the difficulty of the enterprise need not call on us to deny all possibilities of comparative constitutionalism. If we move away from abstraction and try to understand countries from within, we hold the prospect of forming enough learning that we can exercise a certain interpretive charity, make certain comparisons between countries, and come to certain judgments.¹⁵⁹ There is no objective metric, of course, to determine when one arrives at the confidence necessary to undertake a certain kind of analysis, when one feels satisfied with the account that one is providing.¹⁶⁰ Indeed, judgment calls of such kinds are central to the very enterprise of scholarship. The possible methodologies that we may use include, of course, ones that are quantitative and ones that involve more than a single country. What is required is not perfection in these endeavors, but rather an engagement with the material in ways that are sensitive to specific contexts and an appreciation of which kinds of questions might work for which kinds of methodologies. As Professor Angus Deaton observes: “Context is always important, and we must adapt our methods to the problem at hand.”¹⁶¹

Moving beyond the Martian standpoint will not fully resolve concerns relating to causation, interpretation, and agency. However, the closer our attention to context, the greater is our chance of mitigating some of the problems that these concerns pose. If we are to take the task of such mitigation seriously, there seem to be real challenges with “large-N” analyses that involve an abstraction from context. In principle, “large-N” work is possible when all the data points are generated from the same data-generating process. We can estimate with greater precision when we have a greater “N.” The problem of comparability, however, makes such work hard. We cannot satisfy the assumption that the data under study is produced through the same process, and the requirement that the variables are the same.¹⁶² The countries across the world vary in ways that, however precise our controls might be, make it hard to identify shared practices and attribute causation with reasonable confidence from without — our assumptions and ascriptions are simply too great. Given cross-national differences, it is not entirely clear that a larger sample is better than a smaller one: by entering new countries into the sample, we increase the potential comparability challenges. Every new country brings with it a new context. If we do believe that legal rules, cultural forces, and interpretive practices vary across the world, then “large-N” work may not be all that different from gluing together different sample sets. Perhaps the most compelling argument in favor of large-N work is that by broadening the inquiry, such work can help us in moving beyond selection biases.¹⁶³ Though “large-N” work may well help to reduce selection biases, a better answer to such a concern might lie in separate analyses for different parts.

Though the “large-N” approach — the “primary method” (p. 102) deployed — in How Constitutional Rights Matter may not quite be successful, the book has many virtues. As we can see, it broadens the methodological conversation within comparative constitutionalism. The use of statistical analyses, in particular, has been relatively underdeveloped in contrast with other approaches in the field, and How Constitutional Rights Matter will encourage us to reflect on the uses of such methods — and perhaps use them in contextually grounded ways.¹⁶⁴ Furthermore, the work makes a contribution by offering certain kinds of data and patterns that raise fascinating questions, even if it is not quite successful in the explanatory answers that are offered. By urging us to think more carefully about the impact that laws and institutions have, the book is likely to serve a major generative role.

We may conclude with a rather revealing hypothetical that Hirschl offers in support of “large-N” analyses. Hirschl — who is sensitive to the potential concerns with “large-N” work and its limitations with regard to demonstrating causality — puts forth the following thought experiment:

Consider the study of nutritious, healthy eating. . . . [D]espite all the important differences among them, most human beings feed on one combination or another of grains, dairy products, fruits and vegetables, and protein-rich foods. Thus, no one in their right mind would dismiss credible studies that highlight a general, cross-cultural risk in the frequent consumption of foods rich in trans fats.¹⁶⁵

The striking feature about this hypothetical is that it is about the human body rather than the human mind. For reasons that I have tried to underline, it may be more fruitful for comparative constitutionalism to instead focus on the human mind, to start from the premise that humans are agents and that constitutionalism is an act of collective agency. Working from such premises will allow us to understand the complexity of our legal practices and our legal institutions; and it will recognize that we have the power to change them. Unlike planets, our lives are not yoked to the laws of physics.¹⁶⁶ The study of comparative constitutional law is a study of the laws of society, not the laws of science. The world of institutions and rights is no straightforward one — institutions and rights work and don’t work for a range of reasons, under conditions and factors that are interactive and that vary over time and space. The positivist approach to comparative constitutionalism tries to shed light on legal phenomena, but it leaves wide open the possibility that something else might be going on. A turn to context could help us understand what that might be.

---

* Associate Professor of Law, Columbia Law School. For comments and criticism, I am grateful to Bruce Ackerman, Ashraf Ahmed, Shyamkrishna Balganesh, Aslı Ü. Bâli, Jessica Bulman-Pozen, Sujit Choudhry, Rosalind Dixon, Noah Feldman, Stephen Gardbaum, Jamal Greene, David Singh Grewal, Monica Ha-kimi, Bernard Harcourt, Samuel Issacharoff, Devesh Kapur, Jeremy Kessler, Benjamin L. Liebman, Pratap Bhanu Mehta, Lev Menand, Gillian E. Metzger, Niels Petersen, Katharina Pistor, David E. Pozen, Jede-diah S. Purdy, Theunis Roux, Charles F. Sabel, Wojciech Sadurski, Neelanjan Sircar, Richard Tuck, Mark Tushnet, Matthew C. Waxman, James Q. Whitman, Taisu Zhang, and audiences at Columbia Law School, the UC Berkeley School of Law, and Yale Law School.