As he prepared for the Constitutional Convention in the spring of 1787, James Madison plunged into a thorough study of historical confederacies, from the ancient Greeks to the then-modern Dutch and German federations.¹ His inquiry, which was informed by materials he had requested from Thomas Jefferson in France, was both schematic and thorough: he picked six cases to focus on, and for each confederacy recorded the various rules of representation, the allocation of powers between center and subgovernment, and various facts about the performance of the system. He assessed costs and benefits—or as he put it in the language of the time, “virtues and vices”—and drew lessons for the project of reorganizing the United States into a viable federal government. Madison’s chief conclusion was that weakness of the federal center was highly risky, shaping his proposals at the Philadelphia convention.

Madison’s broader effort to inform constitutional design with comparative data has proved to be enduring. We revere the American founding fathers as masters of applied political theory, who leveraged their understanding of human nature to craft a set of enduring institutions. Committed empiricists, they scoured the data available to them for practical wisdom. They provided the grammar of constitutional design and perfected the idea of a written national constitution—a charismatic fusion of social contract theory, natural law, and popular democracy. More generally, they have given us the idea that institutions could be rationally designed through the application of human reason. While they would be the first to admit that their particular design was imperfect (a point often lost in modern-day debates), they created a system that has lasted and been basically workable on many dimensions.

I work in the field of comparative constitutional studies, an interdisciplinary inquiry that combines positive and normative analysis and utilizes tools from several different academic disciplines including history, law, political economy, and sociology. We seek to understand the origins of constitutional ideas, the process of constitutional formation and change, and the causes and consequences of particular choices made by drafters and interpreters. The field reflects Madison’s confidence that we can utilize the comparative method to produce generalizable knowledge, and to reason our way to normative recommendations. My starting point in this essay is to ask whether or not this confidence is justified. What, if anything, can be known about constitutional design?

This is an ancient and important question for academics and political thinkers. Since Aristotle, Western theorists have ruminated on ideal constitutional design, to understand which institutions will produce desirable social outcomes. (Political thinkers elsewhere asked the same questions but without the term constitution.) We have come to understand that constitutions play both symbolic and functional roles, defining a government and regulating the relationship between a government and its citizens, as well as in some cases contributing to national identity.² These are all metrics against which we can measure “success” in some sense, and there are many others as well: democracy, accountability, wealth generation, and other political values provide us with metrics.³

In a more positive vein, we might also inquire about the origins and effects of particular institutional choices. Why did the Constituent Assembly of India, for example, borrow the idea of a “directive principle” (an instruction to government to privilege certain goals) from the Irish Constitution?⁴ How did the Western concept of rights get translated into Japanese during the American occupation?⁵ What is the economic impact of adopting parliamentarism instead of presidentialism as a form of government?⁶ What determines the frequency of constitutional amendment?⁷ Does it matter if a constitution is highly abstract or minutely detailed?⁸ These and myriad other questions have provoked inquiry in the field in recent years.

But this search for knowledge is not a purely academic enterprise. Drafting a new constitution is typically one of the very first acts of a new nation, and so there is a significant real-world demand for information. Close to a thousand times since 1787, political leaders in more than 220 independent nation-states have come together to produce a constitutional text. The number of texts swells further if we include subunits such as states and provinces in federal systems. Most of these constitutional arrangements do not last very long, meaning that, at any given time, there are a half-dozen or so exercises in constitution-making going on around the world. Do we have anything to say to the drafters of these constitutions?

My own intellectual interest in the academic questions was spurred by the practical ones. In the early 1990s, just out of college, I was working as a program officer for an American foundation in Mongolia, as the country was restructuring fundamental institutions and seeking to constitutionalize democracy. Though nominally independent, the country had been in effect a Soviet satellite until the end of the Cold War in 1989. It had produced several constitutional texts, typically reflecting models borrowed from the Soviet Union (to whose leading role the texts occasionally referred). Suddenly, in 1990, there was a new opportunity to choose institutions independently, without direct international pressure. Mongolians sought to instantiate the goals of creating a constitutional democracy and to reflect a broader set of related ideas including human rights and the rule of law. (A few years later, statues of Lenin were replaced with those depicting Chinggis Khan, to symbolize the newfound independence.) In this period, in many countries, post–Cold War optimism prevailed, and there was confidence in the power of institutional design.

My job was to connect a group of young reformers with global experts who could advise in the drafting process. The process was typical of constitutional drafting exercises in that it involved an exchange of knowledge. The Mongolians—politicians, lawyers, civil society activists, and technocrats, called upon by political circumstances to articulate a fundamental law for their country—formed the “demand side” for external constitutional knowledge. On the supply side, there were various outsiders who had particular expertise to deploy and lobby for: human rights groups, free market advocates, and experts in judicial independence. A few had experience in other constitutional settings; others had specializations that were deemed relevant. None knew much about the country.

This basic setup of supply and demand has progressed in the subsequent three decades of constitution-making around the globe, with the supply side becoming more complicated. It has seen the development of new forms of “expertise” that muscle their way into any constitutional design situation, driven by knowledge practices and normative agendas. For example, members of the Constituent Assembly of Nepal, which completed its work in 2015, had the benefit of input from experts in gender, transitional justice, the rule of law, electoral systems, and counter-corruption. Each of these represents a new “field” of sorts. Some of the underlying ideas and concerns originate in the academy; others reflect agendas pushed by international organizations like the United Nations; and some reflect the concerns of experts themselves. Knowledge practices are, after all, in part a function of entrepreneurship. I suppose I have become an expert in constitution-making. The external involvement reflects the fact that the written constitution is now a global form, in which ideas are borrowed and repurposed across national contexts. And constitution-making provides an arena in which various international networks meet and compete.

One might think that this transnational exchange of knowledge is inappropriate, in that constitutions are documents that are supposed to reflect national values, and to somehow be the product of “we the people.” But, of course, the form of the national constitution comes from abroad, traceable to the era of Madison and Rousseau. And many features of the texts are designed to communicate to foreign audiences. A constitution can signal the particular international values and networks to which a country aspires.

The close connection between external scholarly ideas and the real-world demands of advice-giving has a long and distinguished history. No less a figure than Jean-Jacques Rousseau was called on to advise constitution-making processes in Corsica and Poland in the late eighteenth century. Each of these proto-countries was struggling for independence and thought that a constitution would help to cement its claim to nationhood.⁹ Who could be better than the leading political theorist of the day to help in the process? By inclination, Rousseau was prone to abstraction, and he did not focus much on concrete institutions. Indeed, he was unable to finish his “Plan for a Constitution for Corsica” before the French government took over the island.¹⁰ Rousseau had a bit more influence in Poland, where his essay entitled Considerations on the Government of Poland apparently had some impact on the nobles who negotiated the short-lived Constitution of 1791.¹¹ His advice was modest, noting that “institutions for Poland can only be the work of Poles,” such that a “foreigner can hardly do more than offer some general observations for the enlightenment, but not for the guidance, of the law-reformer.”¹² In keeping with his emphasis on collective identity, Rousseau emphasized the importance of cultural activities like sports in building a national consciousness.

Rousseau’s advice had little direct impact. Three years after Rousseau accepted the assignment for Corsica, France took over the fledgling state and deposed Pasquale Paoli, who had drafted the Corsican Constitution of 1755. Poland was partitioned even before Rousseau finished his essay in 1772. Of course, we should not attribute the failures of these states to the advice they received: no constitutional arrangement could change Poland’s unfortunate location between the expanding empires of Prussia and Russia, or Corsica’s plum Mediterranean location. Many latter-day Rousseaus have had to confront a similar lack of impact, illustrating the challenges of translating the abstract ideas of political theory into practical advice.

The problem is not just one of influence, but one of knowledge. The outcome of institutional design will be the result of the interaction of several factors: the general features or qualities of an institution as they interact with local environmental conditions, in the face of exogenous pressures and shocks that might not be anticipated. How do we isolate the effect of the institution in such circumstances, to produce useful knowledge?

One contemporary advice-giver has analogized the work to the medical sciences, in which the designer is a “doctor” to fix “patients” (otherwise known as countries!).¹³ The analogy is highly imperfect and reflects an unsupportable hubris. In the field of constitutional design, unlike the hard sciences, randomized control trials are impossible. We cannot randomly assign sets of institutions to different countries to see how things turn out. This limits our ability to be truly prescriptive.

What then might we do to answer the demand for knowledge? A basic first step is history. We need to understand the local context and environmental conditions, which can be discerned in part from examining prior institutions and assessing how they worked out. Careful process-tracing can illuminate what happened, with the possibility of identifying certain causal channels and alternatives. But what constitution-makers really want is forward-looking knowledge. What will happen if we adopt institution X instead of institution Y? This requires prediction, which in turn depends on some kind of model.

The discipline of comparative politics is helpful, in that it has devoted a good deal of attention to the average or conditional effect of institutions. For example, one major debate in the field has concerned how to design institutions to ameliorate the divisive effects of ethnic diversity.¹⁴ Another major controversy has been focused on the relative merits of different constitutional systems, particularly presidentialism, parliamentarism, and semi-presidentialism.¹⁵ These academic debates took on particular urgency with the end of the Cold War, and the sudden bull market in constitutional advice among new clients. The method by which scholars initially debated institutional design employed informal reasoning, based on a close reading of recent cases in Latin America and southern Europe. The normative concern underlying the analysis was to provide the institutional bases to insulate democracy from reversal and upheaval. Indeed, the very first issue of the Journal of Democracy launched the debate on forms of government. The informal debate tended to favor parliamentarism, as presidentialism was empirically associated with the failure of constitutional democracy. Presidential systems, it was argued, fostered winner-take-all politics and zero-sum competition, leading losers to sometimes try to overturn the whole system.¹⁶

In the absence of randomized control trials, knowledge in comparative politics is often produced using cross-country regressions, which have come in for a good deal of criticism in recent years. The gist of the critique is that there is too much unexplained country-specific variance to draw firm conclusions. Further, in the field of constitutional design, with only a thousand historical observations and 220 countries, there may too few degrees of freedom to draw firm conclusions.

The problem with rejecting cross-country regressions out of hand is that there are few alternatives. Causal inference is complicated, but at least the econometricians are attacking it rigorously. Furthermore, the regression framework can be helpful when used to test propositions that have been developed informally. Consider the debate about the forms of government in the early 1990s. Once sophisticated econometric techniques were introduced to test these conjectures, and data expanded beyond Latin American countries, it became apparent that the failure of presidential systems flowed not from the nature of the institutions but from the adoption of presidentialism by countries that were prone to instability.¹⁷ The more sophisticated econometric techniques undercut the inferences drawn from informal observation.

To summarize, isolating the effects of institutions is a tricky business. There are precious few natural experiments. We might draw on cross-country regressions to determine general trends, but this is not knowledge that can be easily deployed to individual cases. We are asked to speculate as to what will happen if country X adopts institution Y, but speculate is about all we can do, informed by as much information as we can gather. There is demand for practical knowledge, even as there is little supply.

Reflecting on this situation, one of the major figures in the field has recently offered a critique of modern social science for what he calls “excessive ambitions” toward constitutional or institutional design.¹⁸ The turn toward technical complexity in formal modeling and empirical regressions, argues Jon Elster, has led the social sciences down a path of artificial confidence. Social realities, in his view, can never be predicted. “Post-diction,” or the ex post analysis and explanation of events after they occur, is the only achievable goal for the social sciences.

Elster’s normative takeaway is what we might call Hippocratic—given that we know little about the effects of institutions and their internal interactions, we ought to focus mainly on the prevention of bad results rather than optimal institutional design.¹⁹ Elster goes on to critique much normative political theory. He (correctly in my view) identifies that much political theory ultimately rests on empirical arguments about the effects of chosen institutions on outcomes, which (because of his skepticism about social science) cannot be verified. Many claims about “ought” depend on assumptions about the “is,”²⁰ but our understanding of the world rests on causal theories that are underspecified and poorly worked out.²¹

This critique would lead the members of a constituent assembly to despair. If most social scientists cannot predict the effects of institutions, then how can designers—many of whom are chosen not for expertise but because of power or the need for representation—make reasoned choices? Particularly in situations in which constitution-making takes place during radical political transformation, the status quo is likely to be highly undesirable. Choices may be motivated as much by aversion to the past as by a clear view of the future.²²

“Do no harm” is unobjectionable advice at first blush. But some institution must be chosen. As Donald Horowitz points out, there is often an existing set of arrangements that may have little to recommend it other than that it exists.²³ Further, if it is epistemically challenging to predict good outcomes, how can we be so confident that we are avoiding bad ones? That too would require fully specified causal models with high-quality data to test them. If we take Elster seriously, we are left with the impossibility of deploying social science findings, at the same time that we must act.

What other knowledges might we turn to as a way out of the dilemma? Consider the law. Lawyers study texts for a living, and their knowledge practices involve reasoning from the adoption of texts toward what might go wrong. Good transactional lawyers are always looking for worst-case scenarios, identifying what sort of minefields are left in drafts of contracts, wills, and other legal instruments. Lawyers are also used to reading for contradictions in text, in which one clause might undermine another. Call it the knowledge of pitfalls. This kind of expertise is quite useful for constitutional design, even if the pitfalls will be highly localized, and dependent on local understandings of legal terms and institutions. Legal knowledge is Hippocratic, but actively so: “avoid harm to the greatest extent possible” seems to be the equivalent maxim to that of the doctors. A lawyer may be able to raise concerns about whether the drafting of particular provisions reflects the intentions of the designers, even if the lawyer knows little about the full-blown social science of institutions. Relative to comparative politics, the law may have more to say, but about less, since it is focused on the details rather than the big operation of institutions writ large.

Finally, let us consider another discipline focused on “local knowledge” and the internal understandings of institutions: anthropology.²⁴ One might think that this discipline, which eschews the normative (even as it studies normativity) would have little to add. But in trying to understand how institutional choices might interact with local environments to produce outcomes, some deep awareness of the local environment is critical. By understanding local contexts one might be able to identify which externally identified problems are genuine and which ones are likely to be trivial. This involves seeing what is likely to happen from the insider perspective. Perhaps a better way to conceive of the knowledge exchange is that the “demand” side itself possesses distinct local knowledge that can inform a guess as to what institutional options will best fit local conditions.

An example of an institutional design with perverse effects was the electoral law that was in place in Afghanistan from 2005 to 2010, implementing Article 83 of the Constitution. This law stated that MPs who died in office would be replaced by the candidate who finished second in the balloting. This rule was most likely adopted from an electoral system of proportional representation (PR) with closed list voting. Under such a PR system, voters choose a party, which in turn ranks candidates in a list. If the party wins, say, thirty-four seats out of one hundred, the first thirty-four people named on its list will be selected. Because from the party’s point of view, its members are more or less interchangeable, there would be little harm from handing the seat to the next candidate in the event of a death.

Afghanistan’s electoral system does not use proportional representation, but instead a system called the single non-transferrable vote (SNTV). Under this system a voter chooses one candidate in a district in which multiple members are elected. The top five vote-getters in a five-person district will be elected. This system is easy enough for voters to understand, since they choose their favored candidate. It tends to make parties less important and to reinforce personalistic ties, which are a pervasive feature of the Afghan context.

Whatever the merits and demerits of the SNTV system in general, it combined with the clause about replacing dead candidates to produce perverse effects. In the SNTV system, as opposed to proportional representation, the candidate with the second-highest number of votes may represent a different party than the largest vote-getter. This unusual combination might have worked in some contexts, but Afghanistan was not one of them: the replacement rule became known as the “assassination clause” after several winning candidates were killed so that their political rivals could take office.²⁵ This result might not have been foreseeable to the foreign professor or election commissioner coming from an industrial democracy to provide advice, but someone with knowledge of the real world of Afghan politics might have been able to predict the results with grim certainty.

The suggestion, then, is that constitutional knowledge is highly contextual. This does not mean that it is impossible to say anything general. We can still identify average or conditional effects, using standard tools of social science, and this academic aspiration has an inherent value. The deployment of that information into a specific design situation requires a predictive frame that takes local knowledge into account. And the execution of these ideas into an actual constitutional text requires a lawyerly reading of the landscape for potential pitfalls.

Constitutional documents are supposed to reflect local values as well as universal ones and are supposed to serve as operating manuals as well as blueprints for political systems. The process of producing a text is usually undertaken in less than ideal conditions, during periods of political uncertainty, and under great time pressures. The task is exacerbated by the fact that few constitutional drafters have any experience in the field, and so they often search for ideas from outside. Time is short and knowledge is limited.

My own view is that all these features of the field render it conservative. There are few incentives to innovate, even as it sometimes seems as if our institutional vocabulary is wholly unfit for our current conditions. Madison’s design, and his design principles, have been enduring and wildly successful, but they have also led to a system of government in the United States that has been decried as a “vetocracy” and a “relic.”²⁶ We glorify the framers even as we denigrate the institutions they designed. Of course, this raises the question of how we evaluate what works and what doesn’t. In many cases, like the political leaders they empower, constitutions get credit or failure for things that are unrelated to their design. On a recent visit to Mongolia to celebrate the twenty-fifth anniversary of the Constitution, an old friend described that document as a success because the country now had “many more cars and many more cattle.” To really ascribe these things to the Constitution, and not the fortuitous discovery of massive mineral resources in the country since 1992, we need to engage in counterfactual reasoning for how things would have worked out had there been a different set of arrangements.

Constitutional design, then, rests on relatively thin empirical bases, and this means that designers tend to stick with what they “know.” There are some proposals out there for a more experimental approach: for example, former Vermont legislator Terrill Bouricius has proposed an elaborate constitutional system of interlocking bodies using the ancient Greek idea of sortition, or random selection of citizens.²⁷ Bouricius developed a scheme in which different legislative tasks are assigned to different bodies, some selected for fixed terms and others temporary, with complex interactions. Other scholars have proposed expanding the role for citizen deliberation and for meaningful participation in government.

These proposals are all well and good, and we may look back at some of their authors as the Rousseaus of our time. But the implementation of novel innovations requires a leap of faith from the constitutional designer to try something that is untested. This is rarely a selling point to the member of a constituent assembly, who might liken it to driving a new model of car that had never been used before. Does one leap without knowledge? Usually not, especially when one is responsible for the direction of a nation. We stick with the institutions we think we know, rather than try something whose effects we know that we don’t know.

In this regard, constitutional knowledge is perhaps not so different from other forms. Whether in science or politics or law, ideas shift incrementally. The most innovative sets of designs in recent constitutions have tended to flow from revolutions, such as those that occurred in Iran or China in the twentieth century. But these were also cases in which constitutional texts were more about the expression of political ideals than they were genuine guides to the political operation of the regime. Large innovations require sharp ideological breaks, involving significant aversion to the inherited set of beliefs and understandings. In these cases, and only these, leaping before looking is an attractive option.