Introduction

A main task of the courts, said John Hart Ely (Reference Ely1980), is that of clearing the channels of political change. In this chapter, I consider the task of clearing the channels of constitution making. In addition, I discuss the more constructive task of strengthening these channels. I do not assume that the tasks will be undertaken by any particular agent, such as the convener of the assembly or the assembly itself. Rather I consider in the abstract how an omnipotent designer of a constituent assembly ought to structure the selection of delegates, the organization of the assembly, and the mode of ratification (if any). The advice of Bentham (Reference Bentham and Bentham1788/2002) to Louis XVI at the eve of the convocation of the Estates-General offers an example of this kind of exercise. In practice, of course, constituent assemblies believe themselves to be omnipotent once they come into being; otherwise they would be a pouvoir constitué rather than a pouvoir constituant.

As background, let me first state how the argument I present differs from some of my earlier discussions of constitution making (Elster Reference Elster1995a, Reference Elster2000). Drawing on the French moralists, I distinguished among three generic motivations that can animate framers: interest, reason, and passion. The implication was that an optimal constituent process ought to be guided by reason – an impartial concern for the common good in the long term – and eliminate as far as possible the influence of passion and interest. I do not exactly repudiate this idea, but I have come to see that things are more complicated. First, the idea of reason should be expanded to include informed beliefs about means as well as impartial ends (Elster Reference Elster2008). Second, the role of passion does not need to be exclusively negative (Rudenfeld Reference Rudenfeld2001). Third, the appeal to self-interest is not necessarily ruled out on normative grounds, if one allows other parties to appeal to their self-interest (Mansbridge et al. Reference Mansbridge2010). Lastly, the designer has to overcome the tacit traditional assumption that biases in collective decision making spring mainly from passion, and to acknowledge the importance of cognitive or “cold” biases (Eskridge and Ferejohn Reference Eskridge and John2002).

In this perspective, the negative task – clearing the channels – consists in eliminating the impact of bias, interest, and passion, as far as possible and as far as desirable. One may (1a) prevent the selection of interested, passionate, or biased individuals or organize the process so that it will (1b) weaken or (1c) be less vulnerable to these attitudes. As for limits to possibility, random idiosyncratic effects can never be completely eliminated. A domineering individual may exercise an undue influence on the outcome, and a spoiler may ruin the whole process. Some systematic biases may also be hard to eradicate. As for limits to desirability, I just asserted that passion and interest may have a positive or at least a nonnegative role. In addition, the cost and especially the opportunity cost of distortion-removing machinery might also be prohibitive. Constituent tasks are often urgent.

The positive task – strengthening the channels – consists in enhancing the motivations of the framers and improving the information available to them. One may (2a) select impartial and well-informed individuals or organize the process so that it will (2b) foster or (2c) be responsive to these qualities. We may observe that (2a) is not merely the opposite of (1a), given that the strength of the concern for the public good is an important independent factor. A framer might be not only dispassionate and disinterested, but also uninterested in his task. We may also note that in selecting for information, one should aim at creating an informed group of individuals, not a group of informed individuals (see discussion later in the chapter).

I have listed six intermediate variables, (1a) through (2c). To affect them in desired ways, a designer could, in principle, use any number of institutional variables:

1. the mode of election or selection of delegates

2. the qualifications (age, gender, income, literacy, etc.) of electors

3. the qualifications of delegates

4. the number of delegates

5. the seating of the delegates in the assembly

6. secrecy or publicity of debates and votes during and after the tenure of the assembly

7. allocation of time to speakers

8. allowing or forbidding deputies to read from written speeches

9. the task of the assembly (constituent only or constituent and legislative)

10. the location of the assembly

11. the (minimal or maximal) duration of the assembly

12. the (minimal or maximal) length of the constitutional text

13. the procedures of ratification of the constitution

14. the date of promulgation of the constitution.

Not all of these have been used in designing actual assemblies. Some of them would not (indeed should not) be used today. I nevertheless discuss some of the latter, such as elector and delegate qualifications, when they have been proposed or could be proposed for the purpose of affecting the intermediate variables.

It would now seem natural to ask how the intermediate variables affect the outcome variable – the constitution. Reasoning backward, one would first try to determine the intermediate variables likely to produce a good constitution and then the institutional variables likely to produce the desired intermediate variables. This is not how I proceed, however. Rather, my reasoning will be forward-looking – from favorable conditions to the presumption of a good outcome, or as good as one can expect.

Consider the three forms of procedural justice discussed by Rawls (Reference Rawls1971, §14). Clearly, we cannot understand the task of constitution making as involving perfect procedural justice, exemplified by Harington’s two girls (“I divide, you choose”). Even if we had independently defined criteria for what constitutes a good constitution, no procedure could track that outcome perfectly. Given that we do not, in fact, have any such criteria, the idea of imperfect procedural justice, exemplified by jury verdicts, will not do either.

It might seem more plausible to characterize good constitution making in terms of pure procedural justice, exemplified by Rawls’s own theory, in which “there is a correct or fair procedure such that the outcome is likewise correct or fair, whatever it is, provided that the procedure has been properly followed.” Even this requirement is too strong, however. As virtually all processes of collective decision making, constitution making is subject to incorrigible distortions and to indeterminate trade-offs. In a deliberating group, the susceptibility of participants to vanity and amour-propre may be reduced but not eliminated. Being right will always, for some, loom larger than getting it right. Also, when an institutional variable has two opposite effects on an intermediate variable, one negative and one positive, the designer may not know how to strike the right balance. The relation between the size of an assembly and its epistemic quality illustrates this case, as we shall see. Indeterminacy can also obtain when a given institutional variable, such as publicity of the debates and votes, has the desirable effect of reducing the role of interest and the (sometimes) undesirable effect of enhancing the role of passion.

The best one can hope for, then, may be a variety of good collective decision making that Rawls does not mention – impure procedural justice. When distortions that can be reduced by institutional means have been reduced, and trade-offs have been made on an intuitive basis, the designer cannot do any better than let the chips fall where they may.

In pursuing some of these ideas, I shall proceed as follows. In the second section, I consider whether public-spirited motivations can be enhanced by mechanisms (2a), (2b), or (2c). In the third section, I ask the same question with regard to rational belief formation. In the fourth section, I address the issue of reducing the impact of interest by mechanisms (1a), (1b), or (1c). In the fifth section, I ask the same question with regard to passions. In the sixth section, I briefly consider cognitive bias in the same perspective. The seventh section offers summary concluding comments. Although the most important historical examples will be the eighteenth-century constituent assemblies in Philadelphia and Versailles (later Paris), I also draw on a number of other constituent episodes.

Enhancing Impartiality

The quality of impartiality or public-spiritedness – the concern for the common good in the long run – is more crucial in constituent assemblies than in ordinary legislatures. In the latter, partisan majority decisions can be reversed by a later majority; moreover, the expectation of an alternation in power may cause partisan majorities to pull their punches. In a constituent assembly, a small partisan majority may be able to impose its preferences not only on the current minority, but also on a posterity in which it may itself be in a minority. This future tyranny of the minority is made possible by the fact that the majority required to amend a constitution is typically a qualified one, whereas the document itself is always adopted by a simple majority. At the Federal Convention, for instance, Gouverneur Morris tried to lock in the future when he warned that if you “[g]ive the votes to people who have no property, […] they will sell them to the rich who will be able to buy them. We should not confine our attention to the present moment. The time is not distant when this Country will abound with mechanics & manufacturers who will receive their bread from their employers” (Farrand Reference Farrand1966, II, p. 160).

As the stability of the constitution over time is a large part of its raison d’être, it would be pointless to make the decisions of the constituent assembly as easily reversible as those of a legislature. To counteract the risk of a lock-in by the present on the future, we may look to solutions along the lines of (2a), (2b), and (2c). To enhance the likelihood that public-spirited delegates will be selected, one might require qualifications in the electors or in the delegates themselves. However, there are few if any objective characteristics that are demonstrably correlated with public-spiritedness. Education and age are conducive to a better understanding of means-ends relations but are neutral with regard to the quality of the ends themselves. Income and property mainly induce a stake in the status quo. There is no evidence for the claim, frequently made in the nineteenth century, that the possession of land induces a permanent interest in the well-being of the community. Equally implausibly, one might require that framers have large families to give them a personal inducement to consider long-term consequences.

If direct selection for impartiality is unfeasible, the indirect method of cross-voting might seem more promising (Elster Reference Elster2006). In the elections to the Estates-General of 1789, the large majority of electoral districts chose delegates from the three orders separately, the clergy, the nobility and the third estate each choosing its own deputies. In a handful of districts, notably the Dauphiné, the three orders jointly chose the deputies for each estate, each elector casting three votes. (An even more extreme version was used in the elections in Brittany to the Estates-General in 1576 and 1614: Deputies for the clergy were chosen by the nobility and the third estate, those for the nobility by the clergy and the third estate, and those for the third estate by the clergy and the nobility [Picot Reference Picot1888, vol. V, p. 271–72].) The hope was that the method would favor the election of moderate deputies motivated at least in part by the general interest rather than of extremists motivated only by the interests of their particular orders (Tocqueville Reference Tocqueville2004a, pp. 531, 541). In general, however, the procedure is fragile because it can also create an incentive for the members of one group to vote for the most insignificant members of the other estates – raiding rather than hedging (Cho and Kang Reference Cho and Kang2008). A similar concern was voiced by Madison at the Federal Convention (Farrand Reference Farrand1966, II, p. 114).

It is tempting, then, to look for ways in which the process itself might foster or be responsive to impartial motivations. In an admittedly unrealistic proposal, one might delay the promulgation of the constitution for some years, to create a veil of ignorance behind which framers would have to “put themselves in the place of all,” given that nobody would know exactly how the document would affect them personally. More realistically, publicity of debates and votes would induce, if not impartial motivations, at least verbal and nonverbal behavior consistent with such motivations (“the civilizing effect of hypocrisy”). Publicity does not rule out self-serving uses of impartial arguments (Elster Reference Elster1995b, Reference Elster1999, ch. V), but it can to some extent limit opportunism.

Yet even though hypocrisy can mimic impartiality, it does not provide the strength of motivation required for a sustained participation. If their absence goes unnoticed or unsanctioned, deputies may take time off for their private affairs rather than showing up for work in committees or plenary debates. (Note the difference between this impact of private interest and that discussed in the fourth section of this chapter.) Beginning with the Constituante of 1789, French assemblies have addressed this issue by publishing the names of those not responding to roll calls (Pierre Reference Pierre1893, p. 479–80), in the hope that blame and shame might counteract private interest.

We can take the proportion of deputies’ time devoted to the constituent task as a proxy for strength of motivation. To increase that proportion, one may try to ensure not only that they show up in the assembly, but also that when there, they devote all their time to the task. (During the conclave for electing a new pope, cardinals are not allowed to devote themselves to any other task.) It follows that constituent assemblies ought not to serve as legislative assemblies as well. Among the reasons why a separation of constituent power and legislative power is desirable one can, in fact, cite the sheer scarcity of time and the need to work in a sustained manner. The intense focus of the Federal Convention stands in stark contrast to the constituent efforts of the Assemblée Constituante, which were constantly interrupted by legislative and, increasingly, executive tasks.

Needless to say, strength of motivation – like competence (see discussion later in the chapter) – is to be valued only if deputies are impartial or can be induced to behave as if they were. If the worst are full of passionate conviction, they will neither be impartial nor ashamed to show their partiality, which, in their eyes, is righteousness itself. As explained earlier, cross-voting was intended to prevent election of such extremists.

Enhancing Rational Belief Formation

The efficient promotion of impartial ends requires rational beliefs about ends-means causality. Impartiality and rationality interact multiplicatively rather than additively. The capacity for rational belief formation makes no independent contribution to the goodness of decisions; it can help constitution wreckers as well as constitution makers, extremists as well as moderates, to realize their goals more effectively.

Beliefs are based on the possession and processing of information. At the individual level, competence or expertise is a multiplicative function of acquired information and innate ability to process it. In principle, one might select competent delegates by requiring that they (or their electors) possess a certain level of literacy or education. Yet what matters for present purposes is collective rather than individual competence. The idea of “collective wisdom” in its various manifestation is highly complex (Landemore and Elster, Reference Elster, Landemore and Elsterforthcoming). Here I want to emphasize three aspects: how diversity at the group level might trump competence at the individual level; how larger assemblies might be better at processing information; and how small assemblies might be better at acquiring information.

Page (Reference Page2007) shows that under certain conditions, “diversity trumps ability” with regard to problem solving. As far as I understand, it is impossible to verify whether these conditions are satisfied for actual assemblies. Yet it makes good intuitive sense to think that if members of an assembly represent a cross-section of the population, they might outperform a less representative group of more competent individuals both with respect to identifying the most important problems (where the shoe pinches) and with respect to solving them (how to make good shoes).

There are two main ways of achieving representativeness. In classical Athens, the lawgivers or nomothetai were chosen by lot from the citizens or, more accurately, from the citizens who presented themselves on the day of the meeting. (The element of self-selection may have reduced representativeness.) In modern constituent assemblies, universal suffrage combined with some kind of proportional representation will, in general, yield a representative assembly. In communities where elections to ordinary legislatures have imposed economic qualifications on the voters, these have sometimes been lifted in the election to constituent assemblies. Thus in the elections to the 1780 constituent legislature in Massachusetts, the assembly “enfranchised all free adult male town inhabitants for the duration of the constitution-making process” (Kruman Reference Kruman1997, pp. 30–31; see also Hoar 1917, pp. 203–07). Proportional voting, too, is often seen as especially appropriate for constituent assemblies. In 1990, Vaclav Havel imposed this system in the first free elections, to allow a place for his former communist enemies in the constituent assembly. He paid a high price, however, for his impartiality. The communists, notably the deputies from Slovakia, ended up as constitution wreckers rather than constitution makers (Elster Reference Elster1995c).

The Condorcet jury theorem asserts, in nontechnical terms, that if each member of a jury or an assembly has a better than random chance of “getting it right” and if they form their opinions independently of each other, the chance that the majority will “get it right” increases with the number of members, converging to 100 percent as the number increases indefinitely. The relevance of this theorem for actual assemblies, and the feasibility of verifying that the two conditions obtain, are more than dubious. For present purposes, the important question is whether there is any intuitive reason to think that sheer numbers will enhance epistemic accuracy. Would the addition of identical twins with identical life histories to the assembly improve its capacity to get it right? As far as I can see, there is no reason to think it would.

In fact, as Bentham first showed, adding the twins might make things worse. We know that Bentham was aware of Condorcet’s 1785 Essay, since there is an ironic reference to it in a work from 1808 (Bentham Reference Bentham1808/1843). I do not know whether he was already familiar with it when he wrote his two essays for the Estates-General, but internal evidence suggests that he was. In his first essay, he admits that “It is certain that with a larger number [of delegates] there is also a larger probability of a good rather than a bad decision” (Bentham Reference Bentham and Bentham1788/2002, p. 35). Yet, he goes on to say, experience from the British parliament suggests that “the greater the number of voters the less the weight and the value of each vote, the less its price in the eyes of the voter, and the less of an incentive he has in assuring that it conforms to the true end and even in casting it at all.” In the second essay, he first states the argument for a large assembly: “With the number of members increases the chances of wisdom (sagesse). So many members, so many sources of enlightenment” and then objects that “the reduction that this same cause brings in the strength of the motivation to exercise one’s enlightenment offsets this advantage” (Bentham 1788/2002, p. 122).

Well before Grossman and Stiglitz (Reference Grossman and Stiglitz1980), Bentham here puts his finger on the problem of informational free-riding. The motivation to acquire information to form rational beliefs about how to realize the public good will be diluted when the size of the assembly increases. Now, given that information itself is a public good, this claim might seem incoherent. The question is deeper, however, since in a group of public-spirited deputies, each has a reason to seek out information if and only if others do not. Barring implausible mixed strategies, there is no stable point between the underproduction of information, leading to bad decisions, and the overproduction of information, leading to costly redundancy. As Vermeule (Reference Vermeule2007, p. 228–31) argues, however, the assembly can overcome this collective action problem by setting up a separate information-gathering structure staffed by officials who are paid to determine the facts.

The same solution will also take care of the less esoteric but practically more important problem that delegates will usually not be exclusively concerned with the public interest. When allocating their time between conducting their private affairs and informing themselves about public matters, the amount they invest in assuring that their vote “conforms to the true end” will, for the reasons Bentham stated, decrease with the number of fellow deputies. These considerations may perhaps explain the otherwise puzzling decision by the Belgian parliament in 1875 that no proposition of law could be signed by more than six members (Pierre Reference Pierre1893, p. 724).

Reducing the Impact of Interest

In this section, I consider only how the interest of the deputies might have an impact on the substance of their decisions, leaving behind the impact, discussed earlier, on their participation. I distinguish between personal interest and group interest. The normative status of these interests in the constitution-making process differs widely. Whereas personal interests are intrinsically undesirable influences on the decisions, group interests may or may not be inappropriate.

In modern constituent assemblies, the personal interests of the framers are a minor factor, because group interests, especially party interests, are overwhelmingly important. Although the recent transitions in Eastern Europe offer a few examples of constitutional provisions clauses owing their origin to vested interests of individual framers (Elster Reference Elster1993, p. 191), personal interest usually has little purchase on the decisions. In the two great eighteenth-century assemblies, however, many individual framers had a great deal to gain or to lose under the respective constitutions. No effort whatsoever was made to prevent the election of individuals with strong interests. In America, the burning issue of the assumption of federal and state debts by the new government might have triggered a decision to exclude large holders of such debts from the Convention. This anachronistic thought did not, of course, occur to anyone. Self-selection is another matter. Thus Gouverneur Morris claimed that he had deliberately refrained from acquiring such holdings so “that he might urge with more propriety the compliance with public faith” (Farrand Reference Farrand1966, II, p. 329).

An assembly can reduce the impact of personal interest by exposing its decisions to publicity or by insulating them from that motive. Concerning the first, we may cite Bentham (Reference Bentham1791/1999, p. 37): “Is it objected against the régime of publicity that it is a system of distrust? This is true, and every good political institution is founded upon that base.” In the canonical modern formulation, “Sunlight is the best disinfectant.” The Abbé Sieyes lost much of his prestige in the French Constituante when he objected (with perfectly valid arguments!) to the abolition of the tithe. At the Federal Convention, “four men – Gerry, King, Sherman, and Ellsworth – were obviously working ardently for the interests represented by their own investments” (McDonald Reference McDonald1992, p. 106). If the debates on the assumption of the debts had taken place before an audience, they might have spoken more prudently. Oliver Ellsworth, who owned federal debts only, argued that the new government ought to assume these debts and no others. In the presence of an audience, the civilizing force of hypocrisy might at least have induced him to extend the assumption to state debts, thus satisfying the “imperfection constraint” on self-serving arguments (Elster Reference Elster1999, chapter V). (Sherman and King, both of whom held state debts only, argued for the assumption of all debts.)

In addition, an assembly can try to insulate its decisions from personal interest. When designing a future institution, framers can include features that ensure that they will not benefit from it personally. In the two eighteenth-century assemblies, framers as different as Benjamin Franklin and Robespierre made proposals to this effect. When the question of paying a salary to the future Senators came up, Franklin asserted that he “wished the Convention to stand fair with the people. There were in it a number of young men who would probably be of the Senate. If lucrative appointments should be recommended we might be chargeable with having carved out places for ourselves” (Farrand Reference Farrand1966, I, p. 329). His motion that Senators should not receive a salary failed by a five-to-six vote. By contrast, Robespierre’s motion on May 16, 1791 that members of the constituent assembly should be ineligible to the first ordinary legislature was adopted unanimously. Although some of the deputies may have been “drunk with disinterestedness,” in the words of the biographer of one of them (Lebègue Reference Lebègue1910, p. 261), many were probably afraid of being perceived as insufficiently disinterested (Elster Reference Elster2009a, chapter 10).

Consider now group interests, beginning with party interests. In many cases, these have an undesirable impact on the outcome. A large party may be able to impose a majoritarian electoral system because it favors large parties, just as a coalition of small parties may be able to impose a proportional system. The Spanish constituent assembly of 1931 illustrates the first possibility. The second is illustrated by several postcommunist constitutions. A party with a popular presidential candidate may be able to impose a strong presidency and popular election of the president, unless its opponents in the assembly manage to impose a weak presidency and election of the president by the legislature. The constituent episode in France in 1958 illustrates the first possibility, that of Poland in 1921 the second. Although undesirable, such effects may be inevitable. To avoid them, framers would have to be selected on a non-partisan basis, as were the Athenian nomothetai. In the contemporary world, this idea is unthinkable.

Group interest can also take other forms. At the Federal Convention, the interests of the states operated in a number of ways to shape the outcome, notably by the threat of the small states to leave the Convention unless they achieved an equal representation in the Senate and the threat of the slave states to leave unless slavery was protected. These influences were predictably obnoxious, the first distorting American politics to the present day and the second creating massive and durable injustices. In general, however, to the extent that the interests of states in federal systems reflect the interests of their populations, they are not necessarily disreputable. When the interests of Catholics and Protestants or of Francophones and Anglophones diverge, there is no well-defined public interest that can trump interest-based bargaining and voting. When interests of farmers and manufacturers diverge, there may be a public interest, but no reliable procedure to determine it.

One may nevertheless try to create conditions of fair bargaining or, alternatively, reduce the scope for bargaining. The first aim could be met by imposing a time limit on the constitutional process, so as to prevent some participants from getting their way by virtue of their superior “inside options” (ability to hold out). Similarly, the threat of filibustering can be reduced by limiting the time allocated to a given speaker (or speakers from a given party) on a given issue. Secrecy of the debates may also be seen as a means to the same end, since it prevents the bargainers from using the public as an amplifier on their demands. Also, secrecy of the votes can reduce the scope for bargaining by blocking the credible promises that are a precondition for logrolling.

In any case, even with public votes, the one-off nature of constituent assemblies will tend to weaken the incentive to keep promises. At the Federal Convention, the compromise in which the large states agreed to equal representation for all states in the Senate in exchange for denying the Senate the right to initiate and amend money bills was undone later when the right to amendment crept in. In the West German assembly of 1948, “the Minister President of Bavaria persuaded the SPD to vote for [the institution of] a Bundesrat in exchange for a momentary advantage and concessions which were subsequently all but abandoned” (Merkl Reference Merkl1963, p. 69). During the debates over the Spanish constitution in 1978, the Union of the Democratic Center was accused “of breaking a painstakingly negotiated set of compromises” (Bonime-Blanc Reference Bonime-Blanc1987, p. 56). In a legislature, such reneging would be severely punished. In constituent assemblies, the shadow of the future is less effective. It is an open question whether members of a constituent assembly are aware of the fact that it may be hard to sustain logrolling promises by the fear of punishment. It is also an open question whether a reduced capacity for logrolling is desirable or not (see Riker and Brahms Reference Riker and Brahms1973 ; Stratmann Reference Stratmann and Mueller1997; Mueller Reference Mueller2003, pp. 104–12).

Reducing the Impact of Passion

Let me first enumerate some of the cognitive and motivational mechanisms by which emotions can distort decision-making processes (see also Elster Reference Elster, Hedstrom and Bearman2009b). First, they may induce temporary preference changes that differ from the permanent and stable attitudes of the agents. Second, they may prevent the agents from realizing that these changes are indeed temporary. Third, they may induce insufficient investment in information-gathering. Fourth, they can induce wishful thinking or counterwishful thinking: “Each man believes easily what he fears or what he hopes.” Finally, sheer amour-propre can make an agent stick to his beliefs, in the face of contrary evidence, simply because they are his beliefs. Here I limit myself to mechanisms that can to some extent be neutralized by institutional design.

The passions of framers can be exogenous or endogenous. On the one hand, they come to their task with preexisting emotions of fear, resentment, envy, contempt, and the like. On the other hand, the process itself may generate strong emotions. I consider these in turn.

When, as is often the case, constitutions are adopted under the threat of violent uprisings or as part of turbulent regime transitions, strong exogenous passions are inevitably present. Shays’ Rebellion had strongly impressed many of the framers in Philadelphia. Writing to William Smith on November 13, 1787, Jefferson asserted that “Our Convention has been too much impressed by the insurrection of Massachusetts: and in the spur of the moment they are setting up a kite [hawk] to keep the hen-yard in order.” He is clearly implying that the framers were acting under the impulse of visceral fear and not merely out of prudential fear. Although some scholars minimize the importance of the rebellion for the proceedings of the Convention (Feer Reference Feer1966; Ackerman and Katyal Reference Ackerman and Neil1995), I tend to believe that Jefferson knew what he was talking about. At Versailles, the abolition of feudalism on August 4, 1789 certainly owed something to the visceral fear of the constituants that their castles might be burned and their families slaughtered (see, for instance, Ferrières Reference Ferrières1932, p. 109 ff.). The committee drafting the constitution of the Second French Republic moderated its initially radical proposals after the invasion of the Assembly by the Parisian crowds on May 16, 1848. The constitution of the Fifth French Republic was adopted when the parliamentarians of the Fourth Republic granted full powers to de Gaulle under the pressure of events in Algeria. In his own inimitable words, “In 1958 I had a problem of conscience. I could just let things take their course: the paratroopers in Paris, the parliamentarians in the Seine, the general strike, the government of the Americans: it was written on the wall. Finally a moment would have come when everybody would have come looking for de Gaulle, but at what price? Thus I decided to intervene in time to prevent the drama” (Peyrefitte Reference Peyrefitte1995, p. 262).

These exogenous emotions cannot be fully neutralized by institutional means, but procedural design may prevent them from escalating. When debates and votes take place behind closed doors or in a location far from major urban agglomerations, passions may more easily be contained. The framers in Philadelphia were completely insulated from the rest of the country. The choice of Weimar as the birthplace for the 1919 German constitution was dictated by the need to be away from the fighting in Berlin. During the 2007–2008 constituent process in Ecuador, framers were sequestered in a small village. They could leave it, but visitors needed a special permit, which was hard to get, to pass the police roadblocks (Adam Przeworski, personal communication).

Even if it were possible to neutralize the emotions of framers, it might not be desirable. Hegel claimed that “Nothing great in the world has been accomplished without passion.” Tocqueville (Reference Tocqueville2004b, p. 228) wrote that “a man facing danger rarely remains as he was: he will either rise well above his habitual level or sink well below it.” The Federal Convention, in his opinion, illustrated the first case. According to Bentham (Reference Bentham and Bentham1788/2002, p. 31), “the time of great crises is also the time of great virtues; virtue is a good that, like any other, is multiplied by demand.” One of the most impressive of the constituants of 1789, the Comte de Clermont-Tonnerre, said that “Anarchy is a frightening yet necessary passage, and the only moment one can establish a new order of things. It is not in calm times that one can take uniform measures” (AP, 9, p. 461). To overcome interest, reason may need to form an alliance with passion. If the constituants had deliberated behind closed doors, they might not have been able to make a clean break with the past. Needless to say, however, this effect of publicity is too unpredictable to be used as a basis for institutional design.

Endogenous emotions arise in the interaction among framers and in the interplay between framers and audience. Although the exogenous emotion of fear was a central force behind the decrees of August 4, the endogenous emotions of malice and resentment that arose as framers sacrificed one another’s privileges (Elster Reference Elster2007) also mattered. Thus after the bishop of Chartres had proposed the abolition of exclusive hunting rights, the Duc du Châtelet said to his neighbors, “Ah! he takes our hunting, I’ll take his tithe,” and proposed that the tithe be abolished with compensation (Droz Reference Droz1860, II, p. 308). As this exchange shows, the passions were grounded in the strong material interests of the framers. The idea of selecting framers with no personal interest in the outcome would, of course, have been as unthinkable as in the American case.

An important endogenous emotion is vanity induced by speaking before an audience. In Madison’s classic defense of the secrecy at the Convention, “had the members committed themselves publicly at first, they would have afterwards supposed consistency required them to maintain their ground, whereas by secret discussion no man felt himself obliged to retain his opinions any longer than he was satisfied of their propriety and truth, and was open to the force of argument. … No constitution would ever have been adopted by the convention had the debates been public” (Farrand Reference Farrand1966, III, p. 479). Framers did indeed change their mind in the course of the debates at the Convention. By contrast, I have not come across any publicly expressed changes of opinion at the French Constituante. As the records of this assembly amount to some twenty-odd thousand double-column pages in small print, there may, of course, be examples that I missed.

To reduce the role of vanity in assemblies while also ensuring publicity of the debates, Bentham (Reference Bentham1791/1999, p. 64) proposed to exclude women from the galleries. “It has been found that their presence gave a particular turn to the deliberations – that self-love played too conspicuous a part – that personalities were more lively – and that too much was sacrificed to vanity and wit.” For some time after 1778, this practice was, in fact, observed by the British parliament (ibid., note 1). I do not know whether the reason for excluding women was that adduced by Bentham. When the French Convention in 1795 decided to exclude women from the galleries, it was because they had been among the most active among the invaders of the parliamentary arena (Pierre Reference Pierre1893, p. 826).

As most speeches at the Constituante were written, speakers did not really talk to one another and try to refute one another’s arguments. Rather, they were speaking to the audience and to the readers of the numerous newspapers that would report their speeches the next day. If they had wanted to overcome this problem and create a genuine debate, they might have adopted the British practice of banning prepared speeches. Only Mirabeau, however, proposed this radical step (Pierre Reference Pierre1893, p. 899). In addition to the fact that written speeches paralyze debate, it is worthwhile citing the objection to written speeches made (in a written speech!) by the President of the French Assemblée Nationale in 1862: “If [a written speech] is long and diffuse, it chills the debate and is pronounced before empty benches. If it is bitter and violent, it causes a deep irritation, because nothing produces a more painful impression than calculated bitterness and violence that does not have the excuse of improvisation” (Pierre Reference Pierre1893, p. 900).

Reducing Cognitive Bias

I understand bias as a cognitive or “cold” mechanism that shapes beliefs in normatively inappropriate ways. There is little doubt that these operate in collective decision making, either at the level of individual participants or through their interaction. It is widely argued, for instance, that jurors are deeply shaped by hindsight bias and anchoring bias when awarding compensatory and punitive damages. If subject either to the “sunk cost fallacy” or to the “planning fallacy,” members of an assembly may form unrealistically optimistic beliefs about the economic viability of projects. In addition to these individually based distortions, bias may arise through interaction. The “recency effect” and the “first-impression effect,” for instance, can distort decisions by lending excessive weight to the arguments made by the last and the first speakers in the debate. (The “availability bias” corresponds to the first of these.) Conjecturally, these effects may also arise in voting.

Our knowledge about such biases comes mainly from the studies of juries and other small groups. Little is known about the importance of bias in assemblies in general and constituent assemblies more specifically. Nor, to my knowledge, is much written on how to reduce or prevent assembly bias by institutional design (see Eskridge and Ferejohn Reference Eskridge and John2002 for some skeptical comments). Even though the remedies of bicameralism, executive veto, and judicial review may reduce some bias in legislatures, they are not available for constituent assemblies. In any case, bias reduction (debiasing) is in general less effective than bias prevention (Wilson, Centerbar, and Brekke Reference Wilson, Centerbar, Brekke, Gilovich, Griffin and Kahneman2002).

The political thinker who has devoted the most thought to these matters is, no doubt, Bentham, notably in Political Tactics. Thus if the recency effect and the first-impression effect arise in voting, one could prevent them by votes being public but simultaneous (Bentham Reference Bentham1791/1999, p. 106ff.). To prevent memory-induced biases, Bentham (ibid., p. 45) recommended the construction of “a very simple mechanical apparatus for exhibiting to the eyes of the assembly the motion on which they are deliberating.” When deputies tend to seat themselves with others of the same political persuasion, they should speak from a central tribune rather than from their seats, to relieve “the individual from the association of ideas which would connect him with a given party” (ibid., p. 54). Alternatively, one could follow the practice of the French Convention and have the deputies seated randomly (Pierre Reference Pierre1893, p. 829).

Conclusion

The procedural approach to good collective decision making has been studied and applied in several settings, notably jury decisions, academic hiring decisions, and decisions by central banks, committees, or constitutional courts. By manipulating the mode of appointment to these bodies, their voting rules, the secrecy or publicity of their debates and votes, and many other variables, one hopes to eliminate distorting influences of the kind I have mentioned and to enhance the motivation and information of their members.

These are all small-group decisions. The extension of this approach to political assemblies – including constituent ones – is in its infancy, and may perhaps remain there forever. Although we can draw on a vast amount of historical knowledge, it is mostly anecdotal. It can help us identify possible causal mechanisms that may be at work, without enabling us to assess their net effects and to make positive recommendations. The study of the successes and failures of past constituent assemblies can nevertheless be useful if it makes conveners and members of future assemblies pause and reflect before they make potentially important design decisions.

Introduction

Constitution making is as ubiquitous as it is mysterious. In any given year, by our estimate, some four or five constitutions will be replaced, ten to fifteen will be amended, and another twenty or so proposals for revision will be under consideration (Elkins, Ginsburg, and Melton 2009). These revisions represent potentially far-reaching changes to fundamental political rules. Yet our knowledge of constitutional (re)design processes and their consequences is cloudy at best. More than a decade after Elster (Reference Elster1995, Reference Elster and Hadenius1997) lamented the dearth of theory on constitutional design (and, we would add, systematic empirical evidence), the field retains a frontier quality – exciting but unchartered – notwithstanding Elster’s own valuable contributions. Many of us likely suspect that the conditions and rules under which founders write, deliberate, and ratify will be consequential. We just cannot say with any authority how they matter, or to what extent.

At the same time, there is genuine reason for optimism about the prospects of improved knowledge. If the field is thin with respect to rigorous comparative research, it is becoming thicker and richer in case studies that can conceivably motivate researchers and shape theory. Founding moments are generally historical, if mythical, affairs, and historians have documented classic cases (e.g., Golay 1958; Moore and Robinson Reference Moore and Robinson2004; Rakove 1997) in some depth. More recent cases have also been the subject of detailed accounts (e.g., Bannon Reference Bannon2007; Brandt n.d.; Congleton Reference Congleton2003; Ghai, Lattimer, and Said Reference Ghai, Lattimer and Said2003; Goldwin and Kaufman Reference Goldwin and Kaufman1988; Haysom Reference Haysom2004; Hyden and Venter Reference Hyden and Venter2001; Keogh and McCarthy Reference Keogh and McCarthy2007; Moehler Reference Moehler2006; Rubin Reference Rubin2004; Selassie Reference Selassie2003; for an extensive bibliography of sources relating to post-1975 cases of constitutional design, see Widner Reference Widner2005b.)¹ These accounts offer a rich source of inspiration for a literature whose needs are clear: conceptualization and measurement of the process, rigorous theorizing about the effects of different aspects of the process, and the testing of these theories with suitable empirical designs. More extensive comparative empirical work looms on the horizon, and we can expect steady growth, if not a boom, in research on the topic (see Carey Reference Carey2007; Moehler Reference Moehler2007; Samuels Reference Samuels2006; Widner Reference Widner2005a, Reference Widner2007a).

Our goals in this chapter are largely cartographic. We seek to chart the frontiers of our collective knowledge of constitutional drafting and adoption processes. Our particular focus is on the links between process and outcomes. We begin with a historical review of the literature and a description of different modes of constitutional design processes, including both the typical actors involved and the activities in which they are engaged. We then explore some of the micro-foundational assumptions that undergird theory regarding the consequences of different processes; in particular, we address the motivations of participating actors. The bulk of the essay is devoted to identifying hypotheses (or, more accurately, thoughtful conjecture) that appear in the literature on the relationship between these processes and various outcomes of interest. We describe existing evidence bearing on these hypotheses and suggest promising approaches to testing these claims further. Occasionally we draw on new cross-national data that we have at our disposal (Elkins, Ginsburg and Melton Reference Elkins, Tom and James2007) to sketch some baseline associations. These analyses are not at all meant to be conclusive, but rather to serve as a point of departure for further research.

Literature on Cases

Constitutional compilations in the modern era have existed since at least 1783 when the French ministry of foreign affairs, at the request of Benjamin Franklin, authorized the publication of an anthology of U.S. state constitutions (Blaustein and Sigler Reference Blaustein and Sigler1988). Early scholarly literature (~1890–1945) on constitutions and constitutional design was largely case-driven and responsive to new constitutional events. Not surprisingly, the end of World War I and the wave of new national constitutions that emerged in the breakup of empires spawned quite a bit of scholarly interest. These efforts were largely descriptive, in some cases going through new constitutional texts section by section (Davidson Reference Davidson1925; Pollock Reference Pollock1923) and in others simply providing a brief introduction to the constitutional text (Albert et al. Reference Albert, Lindsay and Rowe1894; Moses Reference Moses1893). These early analyses are generally not concerned with process, but instead focus on describing institutions such as executive-legislative relations and regionalism (Dedek Reference Dedek1921; Quigley Reference Quigley1924). The issues of human rights (or, in the parlance of the time, the rights of man) and democratic theory are also raised in discussions of particular constitutional texts (Bentwich Reference Bentwich1924; Clark Reference Clark1921; Morse Reference Morse1919).

There are exceptions to this general characteristic of atheoretical, institutional description, of course. Kantorowicz’s (Reference Kantorowicz1927) examination of the Weimar constitution is motivated by ascertaining the political goals of the designers and how well they met them by comparing the new republican constitution with the old imperial one. He also contrasts the operation of the constitution with its de jure provisions. Chapman (Reference Chapman1925) attempts a similar undertaking in endeavoring to explain the disjunction between Cuba’s constitution and political reality (a culture of corruption, particularly in the Congress), ultimately concluding that the solution may be beyond the scope of any constitution.

Other works are more centrally focused on issues related to constitutional design processes such as legitimacy and the exercise of constituent power. Arangio-Ruiz (Reference Arangio-Ruiz1895), for example, explicitly addresses the evolution of constituent power in Italy from 1848 onward. Even though plans were made for an elected Constituent Assembly to establish a permanent constitution for Italy, they were never realized, and so the constituent function was never clearly assigned. This constitutional silence allowed parliament progressively to assume a constituent function. Perhaps motivated by institutional self-interest, Italian deputies and senators resisted calls for a constituent assembly. Likewise, they resisted the notion of a constituent power residing in the people, who were deemed incapable of exercising it. Arangio-Ruiz approvingly notes the objections of the parliamentarians that, given Italy’s recent despotic past and lingering conservatism in the state bureaucracy and police services, locating constituent powers in the people would be a grave mistake.

Another historical work that speaks to contemporary themes is Currier’s (Reference Currier1893) analysis of the circumstances surrounding the French constitution of 1875. As in the Italian case, Constituent Assembly elections scheduled for October 1870 were cancelled. The National Assembly elected the following February concluded peace with Germany and asserted implicit authorization to proceed with the drafting and adoption of a new constitution. Although the political and legal circumstances were muddied, the National Assembly was elected by universal suffrage and could justifiably claim to be a constituent body representing the “will of the nation and the sovereignty of the people” (Currier Reference Currier1893: 132). As Saleilles (Reference Saleilles1895) notes, however, such an action was out of line with previous French constitutional history. In an argument that anticipates Elster (Reference Elster, Bauman and Kahana2006), Saleilles maintains that the constituent, or sovereign, power cannot simultaneously lay in a constituted power such as a parliament.

Whereas single-country case studies have been and will continue to be an invaluable source of knowledge about specific episodes of constitutional design, recent decades have seen the emergence of volumes attempting to situate constitutional design process in a cross-national, comparative framework. Goldwin and Kaufman (Reference Goldwin and Kaufman1988) is an invaluable example of this approach, involving a series of papers by contributors, who are, with but one exception, all former constitution writers themselves. The case studies provide insights into different aspects of the process in order to facilitate a better understanding of choices faced and decisions made as they worked to craft a new constitution for their respective countries. In a companion volume, Goldwin, Kaufman, and Schambra (Reference Goldwin, Kaufman and Schambra1989) replicate their previous task but with a specific focus on the issue of ethnic, linguistic, racial, and religious diversity. Banting and Simeon (Reference Banting and Simeon1985) highlight the political, rather than the legal or institutional, conflicts associated with constitutional change in select industrial countries.

More recently, both Reynolds (Reference Reynolds2002) and Hyden and Venter (Reference Hyden and Venter2001) have contributed edited volumes that address various aspects of constitutional design. Reynolds combines single-country case studies with thematic ruminations on institutional design to explore the extent to which constitutions can reduce civil conflict and promote democratic governance. As noted later, the conflict resolution literature has provided a good deal of the recent work on constitutional design, taking an instrumental approach.

Hyden and Venter’s analysis of constitution making in four African countries is among the most theoretical and explicitly comparative studies in this vein. They construct a common theoretical framework and evaluate constitutional design processes as to the representativeness of the process, the mechanisms used to create the document and aggregate interests, and finally, the extent of popular participation in the process. In a similar manner, Samuels’s (Reference Samuels2006) twelve-country study commissioned by the International Institute for Democracy and Electoral Assistance (IDEA) evaluated constitutional design processes on the dimensions of inclusiveness, representativeness, and popular participation, in part to determine to what extent a democratic design process can help generate democratic outcomes.

Constitutional Design: Modes, Actors, and Constraints

Any particular instance of constitutional design must deal with certain basic questions of organization and process. These include designating who is to be involved; when that involvement takes place; and how the actors are to proceed in terms of formulating, discussing, and approving a text. There are conceivably as many variants on the process as there are constitutions, but several common patterns emerge.

Constitution making occurs in discernible stages, some of which resemble an ordinary legislative process familiar to many drafters in consolidated democracies. Jennifer Widner (Reference Widner2007b) has provided a very useful schematic of design processes that should guide researchers and practitioners. Widner identifies the phases of constitution making as including drafting, consultation, deliberation, adoption, and ratification. Banting and Simeon (Reference Banting and Simeon1985) begin even earlier, focusing on the stage of mobilization of interests (and counterinterests) prior to the preparation of a text. They call this the “idea-generating stage” at which large parameters are laid out and the process itself may be determined.

These different stages interact with the possible actors who might fill the roles to create a matrix of options for designers. Afghanistan’s constitution of 2004, for example, was drafted in relative secrecy by a commission with foreign advice and then sent to the president’s office before deliberation and adoption at an inclusive constituent assembly, the Loya Jirga (Huq Reference Huq, Hurwitz, Satterthwaite and Ford. Eagan2009). In this model – which appears to be relatively common – each stage is potentially consequential, although it is likely that inertial forces and the power of agenda setting will apportion disproportionate influence to those actors involved at earlier stages. Still, it is quite possible that early-stage actors will anticipate the preferences and needs of later-stage actors, thus mitigating any sequence effects. Elster (Reference Elster1995: 373–75) develops the vivid distinction between “upstream” and “downstream” constraints in the process: Upstream constraints are imposed by the powers setting up the constitution-drafting body, whereas downstream constraints result from the anticipation of preferences of those involved in later stages. Ratification by public referendum, for example, is a downstream constraint that can hamstring leaders in an earlier stage, who recognize that their document must ultimately obtain public approval.

As this discussion of constraints implies, a critical variable in constitution making has to do with which actors are included in the process. Institutional scholars are used to thinking of actors as “veto players” (Tsebelis Reference Tsebelis2002), and the constitutional design realm is no different. As Widner (Reference Widner2007b) describes, actors involved in constitution making can include expert commissions, legislative bodies or committees, the executive, the judiciary, national conferences, elite round tables, transitional legislatures, specially elected constituent assemblies, interest groups and nongovernmental organizations, foreign advisors, and the public itself. Public involvement, discussed later in the chapter, has become the subject of particular attention in recent years, and is urged by scholars, governments, and international organizations (Banks Reference Banks2008; Commonwealth Human Rights Initiative 1999; Elkins, Ginsburg and Blount, Reference Elkins, Tom and Justin2008; Samuels Reference Samuels2006). But not all constitutions involve the public, and some are drafted by a handful of leaders behind closed doors.

To sketch some patterns of actor involvement, we have gathered data on the process of adoption for 460 of the 899 national constitutions promulgated in the period between 1789 and 2005. (In a larger project – the Comparative Constitutions Project – we are engaged in the collection of data on the content of all 899 constitutions). The information on process reveals a pattern of constrained variation in the choice of actors. The principal actors include constituent assemblies, executives, ordinary legislatures, and the public through ratification referenda. At least one of these actors is formally included in 95 percent of the design processes in the sample.² As Table 3.1 reveals, however, there is some variation in how various design processes utilize the actors. Slightly less than one-half of processes utilize a single actor.

Table 3.1. Actors and processes (N = 460)

Note: Coding rules are described in an online appendix at http://www.comparativeconstitutionsproject.org

This accounting does not reveal anything about the depth or quality of involvement. Executives, for example, were involved in some manner in 51 percent of processes we analyzed. In some cases, however, executive involvement may have been merely formal, such as acting as the last official to sign the constitution, whereas in other cases, it may have been more substantial.³ For some analyses, it is useful to identify the institution, or actor, with the most influential role in the shaping of the document (Elster Reference Elster, Bauman and Kahana2006). We engage in this sort of categorization when we turn to some preliminary analysis questions of “self-dealing” (see discussion later in the chapter).

Certainly, a central dimension on which constitution-making processes differ is the degree of public participation. Because the constitution is the highest level of lawmaking and provides the ultimate rule of recognition for lawmaking processes (Hart Reference Hart1961; Kelsen Reference Kelsen1945), it requires the greatest possible level of legitimation in democratic theory. In an ideal world, one would desire universal consent over the rules of society, a standard that is obviously impractical (Buchanan and Tullock Reference Buchanan and Tullock1962). Our sense is that actual constitutional design processes employ scattered and usually rather anemic forms of popular participation and oversight to substitute for actual consent. Higher levels of participation are presumed to function like supermajority rules, restricting the adoption of undesirable institutions and protecting prospective minorities in the democratic processes that are established. Participation thus legitimates and constrains, substituting process for consent to make effective government possible.

The modal form of participation in constitutional design is the power to approve the charter, usually by referendum on the final document as a whole. Figures 3.1 and 3.2 present historical data on the processes of promulgating constitutions, and on public promulgation in particular. Figure 3.1 plots the percent of constitutions in force, by year, whose text requires public ratification. The plot suggests a significant trend, accelerating in the early twentieth century, toward public ratification. We emphasize that the denominator here includes only those constitutions that specify the promulgation procedure in the text itself. As Figure 3.2 attests, although a majority of modern constitutional texts provide information on promulgation, most nineteenth-century texts were silent on the topic. We thus treat the findings in Figure 3.1 with some caution, although it seems unlikely to us that the shift in norms regarding the appearance of promulgation procedures in the text biases the results significantly. Indeed, our review of extratextual case information for a smaller sample suggests that the trend implied by the cases plotted in Figure 3.1 is fairly representative of the trend within the full universe of cases. Thus, it seems likely that public ratification has been on the rise since the turn of the twentieth century.

Figure 3.1. Proportion of constitutions in force that provide for public ratification

Note: Universe: National constitutions that specify promulgation procedures.

Figure 3.2. Proportion of constitutions in force that specify some promulgation procedure.

Approval by referendum may be an increasingly popular mode of public involvement, but it is clearly a limited one in that it involves only an up-or-down vote over a package of provisions.⁴ Since at least World War II, however, participation in constitutional design has become more direct and has penetrated deeper (or at least earlier) in the process. One common approach is to involve the public in selecting those who will draft or deliberate over aspects of the charter (Widner Reference Widner2005a: 7–8). This sort of voice is possible whether the representative group is a constituent assembly elected expressly for the purpose or a regular legislature that takes on the project in addition to other duties. Some constitutional processes have experimented with more bottom-up methods of direct democracy, such as the citizen initiative, in which ideas can bubble up from civil society. We cannot say much yet about the effect of such methods, although anecdotal accounts, such as the report that citizens submitted 64,000 proposals to the Brazilian 1987–1988 assembly (Benomar Reference Benomar2004), suggest the magnitude of the challenges involved in absorbing public suggestions.

Still another mode of participation involves direct consultation with the public or representative groups at various stages, which might occur before, during, or after the drafting of the initial text (Ghai Reference Ghai2004; Ghai and Galli Reference Ghai and Galli2006; Samuels Reference Samuels2006; Selassie Reference Selassie2003; Widner Reference Widner2007a;). The drafting phase would seem to be especially crucial, because we can expect a fair degree of inertia in the later stages of the process. But the phase is also likely to be the least participatory, given the challenges of writing-by-committee, much less “writing-by-nation.” Indeed, in some well-known cases, the public is excluded from the drafting process and not consulted at all.

Of course, actors and their accompanying constraints may come from outside, as well as inside, a state’s borders. An extreme case is that of the “Occupation Constitution” (Elkins, Ginsburg, and Melton Reference Elkins, Tom and James2008; see also Feldman Reference Feldman2005), a document drafted when a country is under the control of a foreign military power. Such constitutions are usually presumed to have less involvement on the part of local actors, and hence to be less legitimate. Burnell (Reference Burnell2008) believes that international involvement creates disincentives to enforce the constitution locally, as actors will strategically acquiesce to conditions they have no intention of fulfilling simply to remove external oversight. He cites Zimbabwe and the Lancaster House Agreement of 1979 as a prime example of this kind of behavior.

The voluminous literature on policy diffusion reminds us that policy reform is a highly interdependent process. Constitution making – often undertaken during moments of crisis when states are at their most amenable to foreign models and “suggestions” – may be especially interdependent and “networked” (Elkins Reference Elkins, Kahler and Ithaca2009). Certainly, scholars have long noted a high degree of similarity across documents, and nearly anyone privy to the details of a case of constitution making can recount an episode of international borrowing. John Meyer and other sociologists working within the “world society” perspective point to constitutional isomorphism as case in point in their compelling globalization narratives (Go 2003; Meyer et al. Reference Meyer, Boli, Thomas and Ramirez1987). The persistence of presidentialism in Latin America, the use of French and Westminster models of government in former colonies, and the recent use of national conferences in Francophone Africa (Clark Reference Clark1994) are all examples of diffusion, which occur on a subglobal level.

Given the persistent centrality of the U.S. constitution to the American legal academy, there has been a fair amount of interest in documenting the influence of the U.S. constitution over the years (e.g., Billias Reference Billias1990), but other constitutional models have also had some impact (Elkins Reference Elkins2003). In public law, much of this research has found its way into a series of compilations of work on “constitutional borrowing,” which cover a decidedly heterogeneous set of transnational issues involving both constitutional design and interpretation (see, e.g., Choudhry Reference Choudhry2006; special issues on the subject in the International Journal of Constitutional Law (2003: 177–324) and Texas Law Review (2004: 1653–737).

Of course, international constraints on constitution making can range in their intensity and degree of “coordination,” ranging from borrowing to imposition. External influence need not be as blatant as in Occupation Constitutions. Constitutional drafting that occurs concurrently with peace negotiations often attracts international advisers and interests, be they donors, creditors, interested states, or the United Nations. Samuels and Wyeth (Reference Samuels2006) briefly discuss the sometimes unintended but nonetheless negative consequences of such “benign” intervention. Also, some models exhibit pull-through proximity: The prospect of future membership in the European Union, for example, led some Eastern European countries to make modifications to their draft constitutions at the behest of the Council of Europe (Elster Reference Elster and Hadenius1997: 129). Many accounts of foreign borrowing point to the decisive role of influential consultants (e.g., Davis Reference Davis2003), and several scholars have sought to profile these consultants and describe their behavior (Perry Reference Perry1992; Reynolds Reference Reynoldsn.d.).

There are other potentially consequential aspects of process apart from the identity of the actors involved. Some constraints reflect the circumstances that lead to constitution making in the first place. The conventional wisdom is that constitution making is coincidental with a cataclysmic event of some kind, such as war, coup, economic crisis, or revolution (Elster Reference Elster1995; Russell 1993). In fact, the evidence suggests that, although crises do frequently precede constitutional reform, the degree of non-crisis constitution making is probably underestimated (Elkins, Ginsburg, and Melton 2008).⁵ Sweden’s 1972 reform of its 163-year-old constitution is a prominent example of crisis-free reform (Congleton Reference Congleton2003). The various socialist constitutions, such as those in the Soviet Union (1936, 1977) and China (1982), seem to follow the installation of new leaders, a practice that was often justified by the Marxist view of evolution in stages (see Go Reference Go2003). These different patterns, reflecting various degrees of crisis or continuity, will affect the process, creating either an atmosphere of urgency or of deliberation.

The process can also vary in terms of time involved. At one extreme, the secretive process that led to Myanmar’s 2008 constitution took seventeen years.⁶ At the other extreme, a small group of American bureaucrats working for the occupation authorities drafted the basic form of Japan’s 1946 constitution in a little more than a week, and the entire process, including elections, legislative deliberation, and approval by the emperor, took a mere eight months (Moore and Robinson Reference Moore and Robinson2004). Which of these cases is closer to the mean? Japan, it seems, by a long shot. We randomly sampled 150 of the 806 cases of constitution making (new and replacement constitutions) since 1789, and were able to identify the start and end dates for the processes in 148 of those cases.⁷ On average, constitution making took 1.32 years in those 148 cases, with a standard deviation of 1.84. The median process length of the surveyed cases was 0.83 years.

Anecdotally, those constitution-making processes involving either a very short or very long amount of time seem to occur in non-democracies. Speedy processes do not allow sufficient time for mobilization of the public and civil society, whereas extended processes are unlikely to hold public attention for the duration. The record for shortest process, formally at least, may belong to the Romanian constitution of 1938, which instituted a brief period of royal dictatorship. A new cabinet taking power on February 12 of that year announced its intention to appoint a commission to draft a new constitution. A new text was published February 21, a referendum conducted using oral voting under a state of siege was held on February 24, with promulgation by the king occurring three days later. This was an expeditious process, to say the least.

Another dimension on which processes differ is the size of the deliberative body. Bannon (Reference Bannon2007) argues that the Kenyan constitutional design process was too bloated, with more than 600 delegates and two bodies producing drafts. Textual coherence and internal consistency appear to have been casualties of the process. Even Yash Ghai, the former chairperson of the Review Commission, noted that Uganda and South Africa managed the process with 300 delegates. To mitigate this problem in the future, Bannon recommends a smaller deliberative body with a greater focus on public consultation at the expense of extensive representation of all interests.

Recent studies have also begun to examine the institutional structure of constituent bodies. Generalizing models from the area of legislative studies, Proksch (Reference Proksch2007) and Tsebelis and Proksch (Reference Tsebelis and Proksch2007) find evidence in the European Union constitutional convention of both agenda-setting power on the part of convention’s Praesidium and subcommittees. One implication of this is that representative, inclusive constituent bodies may not be a sufficient condition for representative, inclusive texts if the possibility exists for biased committees. In effect, the level of analysis most commonly applied to design processes may be misleading. Continued systematic examination of the inner workings of constituent bodies and application of social science analytic methods to their proceedings hold the promise of leading to a better understanding of how particular provisions make their way into a text, and how participation is channeled.

In sum, processes of constitutional design and adoption vary widely along many dimensions. There is much speculation, but relatively little evidence, about the impact of these processes on different outcomes. The remainder of this essay considers the range of hypotheses regarding this impact and reviews the available evidence.

Connecting Process and Outcome: Micro-Foundational Assumptions

Assumptions about the motives of those participating in the process, be they elites or citizens, turn out to be central to hypotheses about the relation between process and constitutional outcomes. Nearly all the normative and positive work on constitutions proceeds from the assumption that constitutional politics are fundamentally different in character from ordinary politics. This is a central assumption of constitutional political economy (Buchanan and Tullock Reference Buchanan and Tullock1962), but also of much other thinking about constitutions and constitutionalism (Ackerman 1993). The basic idea is that legal or political entrenchment distinguishes choice about rules from choice within rules. Because constitutional designers operate without certain knowledge of their prospective position in postconstitutional governance, they are presumed to pay greater attention to the public rather than the private interest. Furthermore, it is argued, constitutions are typically adopted during moments of crisis, and so likely to produce more attention to the general welfare and less likely to be dominated by special interests.

Assumptions about motive also inform normative work on constitutional design processes. If one believes designers will act in their own self-interest, one might want to ensure maximum participation in the process to counter this tendency. On the other hand, if one believes that designers can take the public interest into account, one might design a process with more limited public involvement so as to facilitate elite deliberation.

Elster (Reference Elster1995) postulates three types of motivations that necessarily are balanced in the constitutional design process: reason, passions, and interests. Reason represents disinterested principles; passions refers to emotional factors such as religious or ethnic animosities or sudden, impulsive desires; and interests may refer to those motives identified with the welfare of drafters, or broader groups or institutions of which they are a part. Elster believes that balance is necessary: A constitution that is too crudely identified with self-interest may fail to be adopted, as might one that is perfect in theory but fails to meet key needs of certain players (see also Ghai Reference Ghai and Galli2006). Elster (Reference Elster and Hadenius1997:130) concludes that direct self-interest of the framers is of less importance than the interests of their constituents, but also (Elster Reference Elster1996) finds that institutional interests dominated in Eastern European constitution making.

In a recent article, Brown (Reference Brown2008) modifies some of Elster’s claims. Noting that even in the best of worlds, there is a certain amount of unpredictability in the process, Brown (Reference Brown2008) argues for the notion of “passionate rationality.” In his conception, actors seek to pursue ends efficiently but do not always recognize what these ends are, much less how to pursue them. As a result, miscalculation and unintended consequences may loom large in constitutional design (Smith and Remington Reference Smith and Remington2001). Elster (Reference Elster, Bauman and Kahana2006) provides several examples of miscalculation and its consequences: the decision by Louis XVI to allow parish priests rather than bishops to represent the clergy in the Estates General; the adoption of proportional voting by the Weimar Assembly in 1919, despite the fact that it was against the interests of the Socialist government to do so; and Vaclav Havel’s insistence on proportional voting that allowed communists into the Constituent Assembly where they became “constitution wreckers rather than … constitution makers” (Elster Reference Elster, Bauman and Kahana2006: 189). Scheppele (Reference Scheppele2008) echoes this point in cautioning that the crucial variables for constitutional success are beyond the ability of designers to control, and so “constitutional luck” plays an important role in the ultimate success or failure of constitutional arrangements.

Assumptions regarding the motives of actors undergird expectations about how process affects outcomes. In large part, these assumptions lead scholars to focus on the negotiators and drafters, with the idea that the identity of the constitution’s authors will go far to explain its content. Constitutions, of course, may not simply be the sum of the interests of those involved; it is likely, for example, that decision rules and other conditions will also prove consequential. Nonetheless, strong theory and predictions regarding the “who” of the constitutional process is a good place to start.

Hypotheses and Evidence Linking Processes to Outcomes

In this section, we identify a set of hypotheses regarding process and outcome and summarize the available evidence. Our focus, in particular, is on expectations regarding the interests and influence of institutional loyalists, international actors, and the public – all of whom are thought to play an increasingly consequential role in constitutional design.

Institutional Self-Dealing

It is common to think of constitutions as the product of various competing interest groups organized along economic goals, ethnic claims, or political ideology.⁸ It also seems plausible that drafters who occupy or seek to occupy government positions may act in the interest of their respective institutions. Once we consider this possibility, the institutional identity of actors in the design process becomes paramount. One theme in the literature is suspicion of legislators as constitution makers. As we described earlier, the legislative model of constitutional design involves electing a legislature to accomplish both ordinary and constitutional rule making, so that choices about constitutional design are bundled with the concerns of ordinary law. One problem with this bundling concerns interest aggregation: the voter’s constitutional preferences may be traded off against other concerns in choosing a representative. It may be that the qualities that make a legislator attractive to a voter (e.g., attention to local interests) are not what that voter would look for in a founding father or mother. More centrally, however, there is a reasonable suspicion that legislators will aggrandize their own institution in designing a governance structure. As summarized by Elster (Reference Elster, Claus and Ulrich1998:117): “To reduce the scope for institutional interest, constitutions ought to be written by specially convened assemblies and not by bodies that also serve as ordinary legislatures. Nor should legislators be given a central place in ratification.”

The skepticism regarding institutional self-interest is certainly not limited to sitting legislatures. The critiques of legislative-centered processes would presumably be even more scathing for constitutions drafted in executive-centered processes, given the distrust of heavy-handed executive rule in democratic governance. Indeed, it is hard to imagine that a constitution such as that recently produced in Myanmar, ruled by a military junta that handpicked the drafters and the deliberative body, would undercut executive power. In another example, the Armed Forces Ruling Council in Nigeria, which reserved for itself a decisive editing role at the end of the constitutional process, rejected as tautological a provision drafted by the Constituent Assembly that would have made coup participants punishable by law (Ehindero Reference Ehindero1991). A general expectation of institutional self-dealing means that we ought to expect that executive-centered processes will lead to stronger executives in the resulting constitution.

The evidence of institutional self-dealing is largely anecdotal (Elster Reference Elster1996; Chesterman Reference Chesterman2005: 952; Ghai Reference Ghai2005; Ghai and Galli Reference Ghai and Galli2006; Samuels Reference Samuels2006). Drawing on our own cross-national data on the content of constitutions, we provide here some preliminary findings on these questions intended only to provoke further inquiry, and certainly not to settle the questions. Recall our sample of 460 constitutions for which we had gathered information on the identity of actors involved in each of the processes. We can categorize these processes, following Elster (Reference Elster, Bauman and Kahana2006) as either executive-centered, constituent assembly, constituent legislating assemblies, constituent legislatures, or a residual category for other cases. The labels can be confusing, but the intuitions are straightforward. The principal contrast is between “constituent assemblies,” which are elected especially to design a constitution and then disband, and “constituent legislatures,” which are legislatures that take on the added task of constitution making. The “constituent legislating assembly” refers to an intermediate category of cases in which assemblies elected specially for constitutional design transform themselves into a sitting legislature. Executive-centered processes include those adopted solely by an executive or by an executive and approved through referendum. We were able to categorize 411 of our 460 cases without difficulty according to these four categories, leaving 49 in the “other” category. Given that representatives in the “constituent legislating assembly” category likely aspire to a career in the legislature, we combine this category with that of the “constituent legislature” into a category of processes that we call “legislature centered.”

We next drew on our own data from the Comparative Constitutions Project to create an index of legislative power, based on a parallel set of items from Fish and Kroenig’s (2009) Parliamentary Powers Index. Fish and Kroenig aggregate thirty-two dimensions of legislative power, equally weighted, into an index representing the level of legislative power in a constitutional system. Their measure, which relies on expert coding, is a de facto measure of legislative power, whereas our parallel measure is a de jure measure based on the formal provisions of the text. We employ a set of variables from the CCP that map onto Fish and Kroenig’s items, and score cases a 1 for each provision present in the constitution. We then calculate the mean of these twenty-one binary variables, resulting in an index that ranges between 0 and 1, with 1 representing the maximum amount of legislative power. If the hypothesis of institutional self-interest has merit, we would expect that legislative-centered processes would feature stronger legislatures than would constituent-assembly-centered process, whereas executive-centered processes would feature weaker legislatures than either. Table 3.2 provides the mean value of our de jure measure of parliamentary power as provided in the constitutions produced by processes in each of the three process categories. Table 3.3 provides additional detail.

Table 3.2. Constitutional design processes and average de jure parliamentary power

Table 3.3. Proportion of constitutional texts containing de jure parliamentary power provisions, by design process

¹ Legislature-centered category includes constituent legislatures and constituent legislating assemblies.

Interestingly, we find no bivariate support for the hypothesis that legislatures produce constitutions with more parliamentary power than do constituent assemblies: The mean value for the parliamentary power index for cases of constituent assembly is actually higher than that of those centered in the legislature, although differences of means tests do not meet standard levels of significance. Executive-centered processes, on the other hand, yield significantly less power for legislatures than do processes in the other two categories.⁹ Elster’s conjecture about institutional self-interest, it seems, is evident only relative to executive-centered processes.

This finding that constituent assemblies are more likely to empower the legislature than are the legislatures themselves is striking. Of course, the finding could still reflect institutional self-dealing in that members of a constituent assembly foresee themselves inhabiting the legislature at some point in the future, even if that career path is not guaranteed. The finding could also reflect the possibility that members of a constituent body – sitting and reflecting as a representative body – are philosophically and politically inclined toward representative government, as opposed to a more hierarchical form that they would identify with executive power. Finally, the bivariate association could be the product of any number of confounds or statistical artifacts. One obvious confound is time. Our sample includes cases dating to the early nineteenth century. To the extent that constitutional processes and legislative power both covary with time, we may be capturing a simultaneous, but unrelated, pair of trends. However, if we de-trend the data (by running a simple regression with year as a covariate), the results remain: Constituent assembly products are indistinguishable from legislative-centered texts with respect to legislative power, but both texts provide significantly more legislative power than do texts written in executive-centered processes.¹⁰

Still, this analysis is plagued by another methodological concern that affects nearly all empirical work on process and outcome: endogeneity. In this case, a state’s predisposition toward strong legislatures might influence both their formal constitutional text and the process they use to produce the text. This sort of endogeneity, however, should produce a bias toward a positive association between legislative power in the text and legislature-centered processes. Given the direction of this bias, and our finding of no difference between constituent assembly and legislature processes, we have rather strong exculpatory evidence that legislatures are not guilty of self-dealing. Nonetheless, our results here represent simply an initial baseline finding: Deeper exploration of this relationship is certainly warranted.

Public Participation and Oversight

Elster (Reference Elster and Hadenius1997: 125) generalizes that constitutions produced in more democratic processes will tend to be more democratic. Given the recent trend toward participation in constitutional design, it is worth inquiring how constitutions produced through participatory processes may be systematically different from other constitutions. A small literature nested in the larger trove of work on political participation more generally has generated a host of hypotheses.

First, participatory constitutional design processes may undermine textual coherence (Voigt Reference Voigt and Aaken2003). As Horowitz (Reference Horowitz and Reynolds2002) notes, even under the best of circumstances, constitutional “design” – a term he reserves for a cohesive process – is quite rare, with some process of incremental construction more the norm. Constitution making frequently consists of a combination of institutional borrowing, wholesale grafting, logrolling, and improvisation. As new, and more, actors become involved in the process, bargaining and negotiation becomes both more extensive as well as intensive. In addition, the populace may be subject to cascades that exacerbate the element of “passion” in constitutional design (Elster Reference Elster1995). The constitution that emerges from this process will almost certainly be an ad-hoc creation, rife with internal inconsistencies and institutional mismatches. The loss of design consistency may be offset by resultant gains in legitimacy (Horowitz Reference Horowitz and Reynolds2002), but it may also render the constitutional scheme unworkable. Additionally, simply increasing the number of actors is no guarantee of a more equitable outcome. The composition of a deliberative body is as important to the ultimate outcome as the number of members; extreme outcomes can emerge from a collective decision-making process (Sunstein Reference Sunstein2001).

A related point is that participation may also lead to more specific and detailed constitutional documents (Elkins, Ginsburg, and Melton 2008; Voigt Reference Voigt2007). Analogizing to the contracts literature, more diverse parties are likely to want to specify their bargain in greater detail because of distrust of counterparties and concerns about strategic nondisclosure of preferences during the bargaining process. Thailand’s 1997 document, for example, was designed to limit political institutions by setting up a large number of watchdogs, all elaborated in excruciating detail in the constitution. Similarly, if the public perceives opportunities for participation to be episodic, there may also be a tendency to seek to constitutionalize various institutions that would ordinarily be left to nonconstitutional politics. An example here is Brazil, whose 1988 process was a model of public participation involving citizen proposals on content. The resulting document is one of the world’s longest, at more than 40,000 words.

We know of no empirical study that has systematically analyzed constitutions for coherence or related concepts. That constitutions contain a complex array of institutions certainly poses a challenge to research design. Undoubtedly, one can find examples of poor drafting, internal contradictions, or errors, but no one has been able to tie these directly to participation. Cross-national approaches might focus on issues of constitutional length and scope, either of which might be construed as indicators of specificity or even incoherence. Even then, it seems likely that these sorts of questions will ultimately be best suited to case-oriented research.

A different line of critique emphasizes the difficulty of reaching agreement. More actors will, ceteris paribus, increase the transaction costs of negotiation, particularly when participants have veto powers over the adoption of new rules (Tsebelis Reference Tsebelis2002). A more open process can also make bargaining and the granting of concessions more difficult (Arato Reference Arato1995; Elster Reference Elster1995; Sunstein Reference Sunstein2001). This is in part because the drafters will feel the need to signal positions to their constituents outside the process, potentially leading to more extreme positions. The drafters may also be interested in using the bargaining process to grandstand, decreasing the possibility of agreement. Open processes of negotiation will tend to hinder tough choices and compromise. This suggests that participatory processes are less likely to produce a constitution, although arguably the documents that do emerge will be more legitimate.

The claim that participatory design processes generate constitutions with higher levels of legitimacy and popular support has been subject to only limited study. We can find case studies that seem to support both the more optimistic and more pessimistic hypotheses. South Africa is rightly celebrated as a case in which participation was extensive, and the resulting document scores well on measures of rights, endurance, and enforcement. On the other hand, Eritrea, Ethiopia, and Thailand utilized broadly participatory processes that had little to no effect on the subsequent political system (Ghai and Galli Reference Ghai and Galli2006; Selassie Reference Selassie2003). Thailand’s 1997 process included a provision disallowing constitutional drafters from running for post-constitutional office for a period – a suggestion approved by the theoretical literature (Voigt Reference Voigt and Aaken2003: 217) – but this did not prevent electoral corruption from reviving. In fact, Thailand’s participatory process appears not to have built a reservoir of support for the constitution, which died in a coup in 2006 without much ado (Kuhonta Reference Kuhonta2008). On the other hand, the constitutional orders in Germany, Japan, and Eastern Europe appear healthy despite the fact that these documents were either imposed by foreign powers or were the result of elite-level round table negotiations (Elster, Offe, and Preuss Reference Elster, Claus and Ulrich1998; Ghai and Galli Reference Ghai and Galli2006; Moore and Robinson Reference Moore and Robinson2004).

In perhaps the most extensive study of the question to date, Moehler (Reference Moehler2006) provides evidence from the highly participatory Ugandan process, and found that citizens who were active in the process were no more likely to support the constitution than other citizens.¹¹ She found that individual-level support for the constitution was influenced more by their support for the National Resistance Movement (NRM) regime and elite opinion than even the respondent’s own participation in the design process. This is a nuanced result, suggesting legitimacy is conditional on factors other than process, particularly the mediating factor of elite opinion as well as other aspects of the context (see also Bannon Reference Bannon2007). Moehler (Reference Moehler2007) also found that participation in constitution making had downstream effects on the process, fostering attachment to democratic principles and closer monitoring of government action downstream.

As Moehler’s studies exemplify, much of the recent emphasis on constitutional design process has emanated from the conflict resolution literature (Hart Reference Hart2001, Reference Hart2003; Samuels Reference Samuels2005; Widner Reference Widner2005a, Reference Widner2005b, Reference Widner2007a, Reference Widner2008). These authors link the successful resolution of (primarily) internal conflict to episodes of constitutional design. The most comprehensive student of the question, Widner, finds a correlation between the representativeness of the main deliberative body and the level of violence five years after ratification. With popularly elected representatives, violence decreased in approximately 42 percent of cases and remained roughly the same in 35 percent of cases. Among executive-appointed bodies, the respective figures were 24 percent and 36 percent.

One of the strongest theoretical claims about popular participation concerns its implications for constitutionalism – that is, a constitution’s ability to constrain government. If citizens are to effectively police the actions of government, it must be sufficiently clear what constitutes a violation of the limits of governmental power so that citizens can mobilize to prevent it. Constitutions help resolve this coordination problem by generating common knowledge about the scope of acceptable government behavior and by providing a focal point for citizens to organize enforcement efforts (Carey Reference Carey2000; Przeworski 1991; Weingast Reference Weingast1997). To the extent that popular participation in a constitutional design process serves to construct focal points, it will facilitate the coordination needed to deter potential constitutional violations by government. In the most optimistic scenario, the presence of a focal point in the written text, when coupled with the more robust civil society that emerged as part of a participatory design process, will ensure that the constitution will be enforced and not serve as a mere parchment barrier (Carey Reference Carey2000).

It follows logically that constitutional endurance, an important criterion of constitutional success (Voigt 2003), will be closely related to enforcement. Public involvement should enhance endurance by making enforcement more likely. Elkins, Ginsburg, and Melton (2009), in a book-length study of constitutional endurance at the national level, find that public involvement in constitutional adoption, as captured in the existence of a referendum or popularly elected constitutional assembly, was positively correlated with constitutional lifespan, at least for democracies. For example, South Africa’s celebrated 1996 document has already lasted longer than the historical mean for constitutions on the African continent. Even though the world’s oldest constitution (that of the United States) was not adopted by referendum, there was a relatively high level of involvement in its approval if not its drafting. The Japanese case, it should be noted, is anomalous in this regard (Elkins, Ginsburg, and Melton Reference Elkins, Tom and James2008; Moore and Robinson Reference Moore and Robinson2004).

We might also speculate on further implications of participation for constitutional design. One influential view of constitutions conceives of them as a social contract among the citizenry, designed to limit demands by the state. In this view, one would expect that more participatory processes work like supermajority rules. As the veto power of minorities increases, one might expect the adoption of more minoritarian institutions, such as judicial review (Ginsburg Reference Ginsburg2003), bicameralism, and, assuming that relevant cleavages are geographically concentrated, federalism. Supermajoritarian processes might produce supermajoritarian rules and institutional configurations, to the extent that a rule-making body will produce others in its likeness. One can also expect that the use of referenda to approve the constitution may be mimicked with direct democracy institutions in the constitution itself.

Voigt (Reference Voigt and Aaken2003) develops a set of hypotheses relating inclusive participation to substantive outcomes. He suggests that inclusive processes will lead drafters to create more independent bodies, delegating powers away from the legislature. This is a corollary, of sorts, to the prediction that the legislative model will concentrate powers in the legislature (Chesterman Reference Chesterman2005: 952; Elster Reference Elster1995). Voigt also believes that participatory documents will be more stable, in that there will be fewer demands for renegotiation down the road, and more legitimate.

We might also expect that as the power of the citizenry in design processes increases, the number and extent of constitutional rights will increase as well. The American case, in which the Bill of Rights was inserted only after public discussion and debate, makes the point quite dramatically (Arato Reference Arato1995: 225). The Anti-Federalists wanted to include a bill of rights in the original bargain, and were able to gain agreement on this during the ratification process as a condition of approval (Rakove 1997). Participation, then, begat a more extensive set of limitations on federal power. In more recent examples, we might expect that participation would be associated with “positive” socioeconomic rights, as the constitution becomes an instrument of redistribution.

These hypotheses regarding content would seem most amenable to analysis. What evidence there is seems to support the prevailing wisdom. IDEA’s survey of twelve constitutional design processes suggests that more participatory processes result in more progressive rights provisions and a higher quality of democracy (Samuels Reference Samuels2006). Their general finding is that “more representative and inclusive constitution building processes resulted in constitutions favoring free and fair elections, greater political equality, more social justice provisions, human rights protections and stronger accountability mechanisms” (Samuels Reference Samuels2006: 668). This finding deserves further testing on a broad set of cases. Ghai (Reference Ghai2001) has also shown that rights provisions emerging from deliberation and negotiation have more of an indigenous character and are more fervently defended and respected. Rights provisions imposed by outsiders such as former colonial masters or handed down by elites are frequently not understood or appreciated. Thus, leaders have little compunction about derogating from them.

We are able to offer some cross-national data relevant to this question, again as an effort to start the conversation rather than end it. We divide all constitution-making processes into two categories, based on whether or not they utilize a public referendum to approve the document. If Samuels’s finding is to be generalized, it would suggest that those processes involving a public referendum would be more likely to have various rights provisions than those processes without a referendum. This is because ratification by referendum forms a downstream constraint shaping the drafting process. The IDEA results also suggest that constitutions in which the public had an approval role would be more likely to involve the public in various decisions thereafter. For example, we might expect that constitutions approved by referendum would be more likely to use direct democracy devices such as the referendum in ordinary governance. We might also expect that the scope of elections would be broader.

We have gathered some descriptive data on these issues as part of the Comparative Constitutions Project. The descriptive data in Tables 3.4 and 3.5 provide partial support to the conjecture about public participation. Processes involving a referendum produce constitutions that are more likely to have virtually every category of right. The anomalous exception is health care: Whereas referendum constitutions are more likely to provide for a right to health care, they are less likely to state that health care is to be provided free of charge. However, given the referendum trend shown in Figure 3.1 and the likelihood that the extent of rights provisions also covary with time, de-trending the data is necessary. A simple-count model of the number of rights provisions in a text with the year of adoption and use of a referendum as the sole covariates yields a statistically significant, negative coefficient for the effect of referenda on rights. The most likely explanation for this startling finding is the frequent pairing of referenda with executive-centered design processes. An additional test using year, executive-only, executive-referendum and non-executive-referenda processes as covariates reveals a positive effect (p = .057) on rights for referenda when not paired with executives that approaches standard levels of statistical significance. Our sense from these preliminary analyses is that referenda may indeed serve as a downstream constraint, particularly for nonauthoritarian drafting processes.

Table 3.4. Proportion of constitutions containing direct democracy provisions, by use of referendum

¹ N=324; ²N=71; ³N=149; ⁴N=44; ⁵N=175; ⁶N=36; ⁷N=113; ⁸N=59; ⁹N=233; ¹⁰N=93; ¹¹N=299.

Table 3.5. Proportion of constitutions containing selected rights provisions, by use of referendum

¹ N=46; ²N=93; ³N=56; ⁴N=148; ⁵N=92; ⁶N=271.

Referendum constitutions also appear more likely to provide for universal suffrage, a secret ballot, a referendum process in ordinary governance, and a public role in approving constitutional amendments. We do not, however, find statistically significant differences between public referendum processes and non-referendum processes in electing or recalling various public officials, although in most cases the direction of the difference is consistent with the predictions.

Still, it seems prudent to reserve judgment about any causal inferences with respect to these results. As with the legislative-constituent assembly analysis, because processes are themselves selected at some upstream point in constitution making, it is likely that the association between public involvement on the one hand and rights and democracy on the other reflects the common impact of an unobserved variable. For example, a set of elites might conclude a private agreement to democratize in which constitution making is part of the process. The elites might then seek to ensure that the process of adoption is more open and democratic, in which case both the process and result reflect a level of antecedent agreement. This problem of endogeneity is endemic in efforts to tie process to outcomes, and hence there is an important role for the careful work of case study literature to try to untangle the causal relationships.

The complement to public participation is public oversight, or the visibility of the design process. There is reason to think that transparency will have decisive effects on the manifestations of self-interest. Constitution making typically, albeit not always, involves discrete moments that occur with great public fanfare. This greater visibility may reduce rent seeking and self-interest, as interest groups seek to exploit the relative anonymity of ordinary politics (Mueller 2000). Appeals to public reason, rather than private interest, are presumed to be prevalent during constitutional drafting. On the other hand, publicity may lead to grandstanding, as political leaders seek to mobilize their own supporters (Brown Reference Brown2008). Along these lines, Stasavage (Reference Stasavage2007) provides a game-theoretic justification for limiting transparency, arguing that rather than generate consensus, open deliberation has the potential to lead to mass polarization.

Analyzing the French and American experiences, Elster (Reference Elster2000) finds that secrecy and transparency matter, and that publicity explains some of the failures of the French constitutional assemblies around 1789. Secrecy, in his view, is amenable to hard bargaining, whereas publicity facilitates arguing. As a solution to this tension between transparency and secrecy, Elster (Reference Elster, Bauman and Kahana2006) employs an hourglass metaphor to describe the optimal role of the public in the process with participation via public hearings at the upstream stage and some form of ratification possible at the downstream stage. The actual writing and deliberation (the neck of the hourglass) should be shielded from the public eye to avoid the pitfalls described earlier. Banting and Simeon (Reference Banting and Simeon1985) cite the Spanish constitution of 1978 as mostly achieving this ideal with the small private working groups that hammered out the final draft being bookended by public scrutiny. The precise need for each form of negotiation may depend on particular contextual circumstances.

The visibility of constitutional design also might affect the ability of certain kinds of interests to organize, particularly those groups focused on the general interest. Public-interest groups that face collective action problems in ordinary politics may be more likely to organize for the relatively infrequent iterations of constitutional politics (Boudreaux and Pritchard Reference Boudreaux and Pritchard1993; but see Sutter Reference Sutter1995: 129). If the profile of the participating interest groups veers toward the public good in this way, constitutional politics might indeed achieve the normative ambition of greater focus on the common good. On the other hand, there is the offsetting consideration of stakes. Private interest groups may invest more energy in playing for rules at the constitutional level precisely because of the presumptively higher stakes in the selection of rules, thus discounting the increased participation of public interest groups.

Conclusion

This chapter has explored the theoretical and empirical relationships between the process of constitutional design and constitutional outcomes. On the theoretical side, we found a broad consensus in the literature about the importance of public involvement as well as an apparent trend in practice. Yet many of the assumptions of proponents of participation remain untested, and the precise relationships between participation and desirable outcomes of interest remain underspecified.

In general, scholars have been far better at generating hypotheses relating process to outcomes than testing them. Individual case studies have provided some insights, but large-n work has been hindered by a lack of data and a need for conceptual refinement. Our own analysis utilizing data from Widner and the CCP suggests an association between processes that involve the public in the adoption of the constitution and the presence of rights and certain democratic institutions in the resulting document. This is consistent with the case study literature, although we are cautious about drawing conclusions about causality. On the other hand, we find little support for the claims about institutional self-interest on the part of legislatures that control constitutional design.

Constitutional design processes are loaded with expectations about endurance, efficacy, the resolution of conflicts, and political reconstruction (Arjomand Reference Arjomand2007). In the real world, however, most constitutions fail (Scheppele Reference Scheppele2008). A key normative question is whether aspects of process can be manipulated to reduce the probability of failures, but this question requires much more positive work on the complex relationships among process, content, and outcomes.