Arab Spring constitution-making: polarization, exclusion, and constraints

ABSTRACT What determines the balance that democratizing constitutions strike between majority empowerment and individual rights? Some constitutions deliberately handicap state power to forestall threats to liberty, while others try to empower the government to hold the country together. We answer this question in the context of post-Arab Spring constitution-making, hypothesizing a U-shaped relationship between polarization among politically significant factions and net majority-empowering provisions in constitutions of new democracies, a relationship mediated by breadth of inclusion in the constitutional drafting process. We test the hypothesis through a controlled case comparison of Tunisia, Libya, and Egypt, the three Arab-majority countries in which protestors successfully toppled authoritarian regimes.


Introduction
Constitution drafters in new democracies face an immense challenge: striking a durable political bargain during unsettled times. As Elster notes: "The task of constitution making demands procedures based on rational argument, but the external circumstances of constitution making generate passion and invite resorts to force". 1 Normative theorists of constitutions urge drafters to set aside party interest in favour of "reason", 2 while positivists investigate drafting and approval processes most likely to yield broad support and guarantee the interests of minorities. 3 This paper investigates the effects of ideological polarization on constitutional drafting processes and, in turn, on the balance struck between government empowerment and minority rights in the constitutions of three post-Arab Spring countries.
In 2011, an unexpected wave of revolutionary protests struck the Middle East, namely, the "Arab Spring". 4 In Syria and Bahrain, authoritarians held on through violent repression. In Tunisia and Egypt, dictators stepped down without great bloodshed; in Libya, rebels overthrew the dictator with foreign military aid; and in Yemen, the dictator stepped down through a negotiated transfer after substantial violence.
Tunisia, Libya, and Egypt stand out as the three cases in which the Arab Spring led to at least an interlude of democratic or semidemocratic politics. 5 All three adopted new constitutions, though Egypt's democratic constitution was revoked after a 2013 military coup, and Libya was governed by two interim constitutions before collapsing into factional warfare. The framers of these constitutions intended to constitute a new, democratic polity, setting the future course of day-to-day politics. Politicians in all three countries faced similar challenges and a daunting trade-off: how to secure peace, stability, and unity in the new polity while preventing a return to authoritarian or illiberal governance. Despite their similar origins, the five constitutionsconsidering the two separate documents drafted in each of Libya and Egyptultimately diverged widely in their degree of liberalism, that is, the extent to which they structure institutions so as to preserve individual rights from future majorities.
Why did they differ? We argue that constitutional framers in each case responded to the challenges posed by political polarization among key factions. Greater polarization among political actors meant more exclusion of ideological opponents from the constitution-deliberation process, and the more emphasis on empowering the government of the day rather than protecting minority interests.
In the next section, we present our theory, which draws on the veto-players approach of conceptualizing institutions but goes beyond it, since at constitutional moments basic institutions are themselves subject to bargaining. We present our hypothesis and consider two alternative hypotheses. The third section presents the evidence on polarization and the constitution-making process. We compare four Arab Spring constitutions to each other and to the post-coup Egyptian constitution in order to test our hypothesis against the alternatives. The fourth section concludes with implications for democratizing constitution-making in other contexts.

Polarization and endogenous institutions
Constitutional scholars have long analysed the extent to which particular institutional forms either empower or hobble legislative majorities. Separation of powers may help to constrain government against abuses by the minority through "ambition … counteract [ing] ambition". 6 Sartori sees presidentialism as unstable due to its penchant for gridlock. 7 Bicameral legislatures with houses of relatively equal power pose a further problem for ensuring legislative majorities. Additionally, Lijphart celebrates the consensus-promoting features of power division and power sharing. 8 Tsebelis conceives of political institutions that incorporate veto players whose consent is required for political change. 9 He derives two major propositions: (1) the addition of a new veto player increases policy stability or leaves it the same, and (2) ideological distance among veto players increases policy stability. One possibility that lies outside the scope of Tsebelis' study is that actors facing extreme gridlock may find ways around formal institutions; in other words, institutions are endogenous. Within the constitutional drafting process, institutions are all the more up for negotiation, since there is no "meta-constitution" determining how decisions in the drafting process must be made. 10 When polarization, which we define as ideological distance among veto players, threatens the ability of powerful actors to secure vital interests, they can find a way around formal rules. For instance, several presidential democracies have developed both written and unwritten procedures for getting around the gridlock that a formal institutionalist theory would predict for such systems. In Brazil, Russia, Venezuela, and elsewhere, presidents have decree powers allowing them to legislate without congressional consent. 11 (2011)). The distribution of veto power is all the more readily endogenous in the constitution-drafting context.
During moments of democratizing constitution-making, rules and procedures are contested. During Egypt's transition, elected president Mohamed Morsi issued a constitutional declaration on 22 November 2012 removing the Supreme Constitutional Court's adjudicatory powers over previous presidential decisions, a decision he later had to reverse in a bargain with the court. 12 The Egyptian military issued its own decree giving itself legislative powers.
Constitutional framers have to decide, often without any binding precedent or law, whom to include in the drafting process and how to structure negotiations and decision-making. Normative constitutional theory often disparages appeals to narrow interests and horse-trading of "normal politics". Yet, constitutions are political documents, and the actors responsible for drafting and approving them are political actors. Constitutions routinely bear the stamp of the interests of the actors who created them.
We assume that framers are instrumentally rational, trying to obtain the best constitutional bargain possible for their faction. As Brown observes: "constitutions are very much products of the same kinds of political maneuverings, partisan bargains, personal rivalries, shortsighted reasoning and ethnic jealousies that make up everyday politics". 13 "Inclusion" at the drafting and approval stages is an important variable in Elkins et al.'s investigation of constitutional duration. 14 For us, inclusion is both a dependent and independent variable. Fearing gridlock, constitutional framers will be more exclusionary the greater their ideological distance from opposing factions. Extremists, which could play a spoiler role, will be excluded. Extremism is relative. In post-revolutionary Iran, Islamists took over the constitution-drafting process and excluded liberals due to their deep philosophical disagreements over Shari'ah. 15 For Tepe, deep societal polarization underlies the ideological divergence among elites. Even if ideologically divergent religious and secular elites form coalitions in the legislative process, their supporters often categorically refuse to endorse such cooperation. 16 The narrower the coalition supporting the new constitution, the more concerned it will be about suppressing extraconstitutional challenges to its future authority, that is, the more it will focus on empowering future governments. Constitutional framers are often remarkably myopic, even writing constitutional provisions for the executive with a particular individual in mind. 17 The coalition that succeeds in drafting the constitution will expect to be in power for the first few years after the constitution comes into force, and the more ideologically unified that coalition is relative to the major opposition factions, the more it will try to ensure that it can rule authoritatively and unconstrainedly.
This dynamic appears in the politics of new constitutions in South America in the 1990s. Colombia, Ecuador, and Venezuela all wrote new constitutions, and Venezuela was the most ideologically polarized case. Colombia and Ecuador enjoyed inclusive constitutional participation and produced constitutions that safeguarded political minorities: for instance, Ecuador's indigenous factions demanded and received a constitutional recognition of the country's "plurinationality". Meanwhile, Venezuela's Hugo Chavez excluded the opposition from constitutional participation entirely and wrote a constitution that unleashed his power. 18

Constitutions and government power
We distinguish the power of the "government of the day" from the authoritative capacity of the state structure that the government leads. To be sure, as a state gains the ability to enforce its commands, it becomes a more attractive target for political factions. Still, there is an important conceptual distinction between the authoritativeness of the state and the power of the ruling majority. Constitutions can bind the governments of strong, authoritative states. For Sartori, constitutions by definition create "free governments", in that they structure and discipline the state's decisionmaking processes. 19 How do constitutions bind governments? They can require supermajoritarian power-sharing, provide for judicial review by an independent judiciary, separate power between the legislature and executive and within the legislature, decentralize power territorially, place the military under civilian supervision, and explicitly constrain the state's reach into personal rights and freedoms. These techniques are imperfect because political actors can find ways around formal institutions, if only at a cost. Nevertheless, once constitutional provisions have been negotiated, they are likely to hold at least among the actors that negotiated them.
We focus on comparative constitutional outcomes that emerge from negotiations among factions at the constitution-drafting stage. The more ideologically polarized factions are in society at large, the narrower will be the range of factions represented at the constitutional bargaining table. When a wide range of factions is represented, they will only be able to agree on a constitution that contains strong guarantees for minorities. Blount and Ginsburg claim that inclusive participation in constituent assemblies "increases a constitution's ability to constrain government … [S]cholars think participatory processes will include more rights provisions and better enforcement mechanisms to protect them, including super-majoritarian institutions, and more public involvement in selecting government agents". 20 We posit a positive relationship exists between the degree of elite polarization and the extent to which the constitution they draft is government-empowering ( Figure  1). When elite factions are extremely polarized, the more powerful among them consolidate control over the constitution-drafting process and exclude their ideological opponents. These dominant factions adopt constitutional provisions that not only entrench their power in the resulting polity, but also strengthen the government's overall power in society so as to defend themselves against present and future political opponents. Exclusion at the constitutional stage fully mediates the causal effect of elite polarization on majority-empowering constitutional terms. Now, one could also imagine small, homogeneous polities in which the degree of ideological agreement is far-reaching enough that providing for an effective government is more important than constraining the state to protect putative minorities from domination. Over the full range of possibilities, then, the relationship between constitutional empowerment of the government of the day and social polarization might be U-shaped ( Figure 2).
New democracies should generally lie to the right of the minimum of Figure 2. During transitions from authoritarianism, major factions are typically characterized by a distinction between hardliners and reformers in the previous regime and between moderates and radicals in the opposition. 21 Even after the regime has fallen, its supporters may make up a sizeable proportion of the electorate, or of those with access to weapons and other instruments of political power. Therefore, these societies almost always experience at least a moderate degree of ideological polarization.
Within the cases we study, then, we expect to find a negative relationship between elite ideological polarization and breadth of inclusion in constitutional drafting (Hypothesis 1). Where constitutional drafting processes are more inclusive, the resulting document will be more government-constraining (Hypothesis 2). Therefore, ideological polarization among factions in the constituent assembly ultimately results in more government-empowering constitutions (Hypothesis 3).
We also consider three alternative explanations for net government-empowering provisions in post-Arab Spring constitutions. One is that constituent assemblies more dominated by relatively liberal factions adopt more liberal, state-constraining constitutions, straightforwardly implementing their ideological preferences. Another explanation is that a larger number of distinct factions presents a more severe collaboration problem in providing collective goods in the future polity. In general, more independent actors in the Prisoner's Dilemma are less likely to reach cooperative equilibria. A solution to the problem might be establishing a strong state as a third-party enforcer. This explanation would suggest that more factions within the constitutiondrafting coalition would lead to a stronger state. Third, the rules for constitutional adoption and factional fragmentation affect the balance of power in the constituent assembly. The leading faction is more likely to compromise when it enjoys a smaller share of seats relative to those needed to adopt a constitution.

Empirical strategy and data
We test our hypotheses on the three post-Arab Spring countries that successfully deposed their dictators and attempted to reconstitute the polity through a new constitution: Tunisia, Egypt, and Libya. The constitutional processes we investigate are those that led to the 2014 Tunisian constitution, 2012 Egyptian constitution, 2014 Egyptian constitution, and two amended, interim constitutions in Libya: that of September 2012 and that of April 2014, which are quite different from each other and were enacted by different bodies. The 2014 Egyptian constitution differs from the others in that it was imposed by the military regime rather than through a democratic process, but we include it because our hypotheses generate clear predictions about it as well.
We follow a "most similar systems" design, deploying cases that are similar along several dimensions, spatial, temporal, and cultural, but differ on the key independent variable of interest, the degree of ideological polarization among major political factions. To be sure, the cases are not identical. Tunisia's uprising and transition were mostly peaceful, while Libya's have been by far the most violent, involving foreign military intervention. Moreover, Libya is an oil-rich state (and therefore richer overall), while Egypt and Tunisia are not.
Consistent with the veto players model, we conceive of ideological polarization in multiple dimensions. The primary dimension of ideological conflict in these cases is the Islamist-secularist continuum. In all three cases, the dominant actors are religious and favour state recognition for Islam, but the debate is over the role that Shari'ah plays in the legal system, the rights of religious minorities, and the extent and harshness of criminal penalties for offenses against morality and religion. Our most important indicator of polarization on this primary dimension is the ideological positioning of the major Islamist factions or parties, relative to liberal or secular factions. But other dimensions of ideological conflict are important as well. In Egypt, the military regime is secularist but also illiberal, dividing them from liberals and socialists. Socialists adopt different economic views from liberals. In Libya, ideological conflict is especially multidimensional because of tribal, provincial, and ethnic conflicts. Research has found that political conflicts over territorial issues are more difficult to resolve than other disputes, often leading to violence. 22 We therefore measure polarization by two criteria. First, if a political environment contains an extreme Islamist party of nontrivial size, it is more polarized than one that does not. Second, if a political environment contains parties committed to territorial autonomy or independence for cultural minorities, it is more polarized than one that does not. We consider an environment polarized only by the second criterion to be more polarized than one polarized only by the first.
Our first dependent and second independent variable is the degree of inclusion at the constitutional drafting stage. In their global study of constitutional duration, Elkins et al. measure inclusion in the constitutional process with two indicators: whether the constituent assembly was elected directly or indirectly by the public, and whether a public referendum was used to ratify the constitution. 23 Four of the five constitutions fit the first criterion, while so far only the two Egyptian constitutions fit the second, and they faced major boycotts. By focusing on just three cases, we develop a more nuanced measure of inclusion at the constitutional drafting stage: the proportion of major, organized factions in existence at the time of the constitutional drafting process that voted for the constitutional draft.
Our second dependent variable concept is the extent to which the constitution privileges government empowerment over protections for political minorities. As a first cut, we develop a crude quantitative indicator of this concept: net government-empowering clauses as a proportion of total articles in the constitution. Nevertheless, this indicator overlooks the relative importance of different government-empowering clauses, and therefore our discussion of the cases gives a more accurate, qualitative picture of the differences among constitutions.

Tunisia
Tunisia has been widely hailed as the success story of the Arab Spring, due to the peaceful overthrow of President Zine El Abidine Ben Ali and mostly peaceful transition to democracy. Tunisia's successful transition can be traced to its pre-Arab Spring roots. Not a major oil producer, Tunisia derives most of its revenues from taxes levied on its own population, 24 providing a more favourable environment for democratization. 25 Prior to the Arab Spring, Tunisia was also one of the most repressive and undemocratic Muslim majority states. 26 In 2010, Tunisia was on par with Libya at the bottom of the Freedom House political rights scale. The paradox of Tunisia as an economically liberalized but politically repressive state may explain both why the uprisings started in Tunisia and why the path to democracy has been smoothest there.
Following Ben Ali's flight, his successor Mohamed Ghannouchi appointed a transitional commission consisting of lawyers and judges, while the revolutionaries set up a parallel transition body. On 18 February 2011, the two commissions merged to form the Commission for the Fulfillment of Revolutionary Goals, Political Reform, and Democratic Transition (HCFRG). 27 The HCFRG held elections to a constituent assembly in October 2011, and disbanded. A coalition government formed by the three largest parties elected to the constituent assembly then took power. Ennahda led the coalition, alongside the left-liberal, non-Islamist Congress for the Republic and social-democratic Ettakatol. Ennahda is considered a relatively moderate Islamist party with a decade-long partnership with the Congress for the Republic. 28 In 2013, the assassinations of two liberal political leaders and a series of protests against the Ennahda-led government temporarily derailed the drafting process. The persisting political deadlock between Islamists and secularists was broken by mediation from four civil society organizations known as the Quartet: the General Union of Tunisian Workers, the Human Rights League, the bar association, and the employers' union. 29 The result was a new government formed largely of nonpartisan technocrats.
During constitutional negotiations, Ennahda insisted on a parliamentary system, while other parties proposed a presidential one, hoping to constrain future Ennahdaled parliamentary coalitions. 30 In the end, both camps compromised on a semi-presidential constitution with a weak president.
On 26 January 2014, the constituent assembly adopted a new constitution by a vote of 220 for, 12 against, four abstaining. 31 Because the vote surpassed a two-thirds majority, the constitution was not required to go to a public referendum. The new constitution took force immediately, and parliamentary and presidential elections took place in October and November 2014.
Tunisia is the least polarized of our three cases. Observers generally agree that Ennahda is more moderate than the Muslim Brotherhood. 32 Even were that not the case, Tunisia lacked a radical Salafi-aligned party that was prominent in Egypt. Tunisia's constituent assembly included a small communist party, now known as the Workers' Party, and civil society organizations like the Quartet enjoyed considerable prominence. Like Tunisia, Egypt's first democratic parliament also included far-left elements, known as Completing the Revolution Alliance. Moreover, Tunisia lacks the territorial and ethnic divisions rife in Libya.

Egypt
The crucial element in Egypt's post-uprising trajectory is the dominant national military. Egypt has been a military regime since Gen. Gamal Abdel Nasser helped lead the military coup of 1952 and gradually wrested sole power for himself. A secular Arab nationalist and socialist, Nasser banned all political parties other than his Liberation Rally, which ultimately became the National Democratic Party under Gen. Hosni Mubarak. 33 The military wields significant social and economic power in Egypt, employing over 100,000 people in the defence industries and accounting for about US$500 million every year in GDP. 34 By 2009, the Egyptian military controlled up to 40% of the national economy, including social clubs, villages, restaurants, tourist businesses, mining, and factories producing televisions, pasta, refrigerators, and cars. 35 The profits from these business endeavours go directly toward the military budget without government oversight, making the Egyptian military more independent from government than it would have been by relying on tax funding. 36 Besides the military, the main political factions are liberals, Marxists, and Islamists. 37 In addition, there is a Coptic Christian minority, about 10% of the population, who are politically aligned with the liberal Free Egyptians Party. 38 The dominant Islamist faction is the Muslim Brotherhood, banned in 1954 under the previous regime. With the introduction of managed multiparty politics in 2005, the Muslim Brotherhood was the most successful opposition party, winning 20% of the seats in parliament. 39 Prior to the Arab Spring, there were a series of small, pro-democracy protests beginning in December 2004. 40 However, it took the spark of the Tunisian revolution to ignite massive demonstrations against the Mubarak regime in Cairo and other Egyptian cities. The regime inconsistently alternated between accommodation and violence towards demonstrators. In February 2011, Mubarak stepped down, and the interim military government announced parliamentary elections that were ultimately held from November 2011 to January 2012.
The Muslim Brotherhood and its allies won 46% of the seats in that election, and the extreme Islamist (Salafist) Alliance won 24% of the seats, yielding a combined 70% supermajority for Islamic parties. 41 Identifying the Muslim Brotherhood as the single most powerful and well-organized civil institution in Egypt, the military temporarily backed their ascent to power. The judiciary, appointed under Mubarak and surviving his fall, intervened to weaken Muhammad Morsi's Islamic government. Still, with their dominant position in the second, ultimately successful constituent assembly of 2012, which had been elected by parliament, the Islamic parties chose to ignore the vociferous opposition of liberals and secularists and impose a conservative constitution. 42 For much of 2012, it appeared that the military's "deep state" and the ascendant Muslim Brotherhood had reached an accommodation, to the exclusion of liberals. Polarization was so deep that Christian, socialist, and liberal members of the two constituent assemblies formed during 2012 walked out in protest at Islamic dominance multiple times. 43 Coptic constituent assembly member George Messiha claimed on resigning from the assembly: "Every section tacitly bolsters Islamic rule in Egypt, whether politically or socially". 44 By November 2012, Morsi's aforementioned constitutional declaration turned tension between Islamists and liberals "into outright animosity … Given Egypt's extreme polarization and the distrust between political forces, the level of resistance the declaration engendered was not surprising". 45 Without accepting the accusations of the constitution's opponents at face value, we nevertheless acknowledge that the constitution presented to citizens for a referendum in December 2012 instilled Islam into the fabric of the state, strengthened the executive, and institutionalized the role of the military in politics: The 2012 struggle over the constitution of Egypt provides a tragic illustration of the consequences of constraining participation: the Muslim Brotherhood rammed through a document without involvement of other groups in the constituent assembly, generating a backlash of bloody protests. 46 Perceptions that Morsi pushed through the Islamist constitution using a presidential decree, coupled with falling living standards, created the circumstances for massive demonstrations similar to those that brought down the Mubarak regime. Though Morsi and his supporters remained defiant against the protestors, the military joined the opposition to oust Morsi and his supporters, appointed the chief justice of the Supreme Constitutional Court Adli Mansour interim president, and suspended the 2012 constitution. The coup d'état had broad support from both the left, including liberals, Christians and labour, and the extreme conservative al-Nour Party. 47 The broad opposition to Morsi accused the regime of using the Muslim Brotherhood majority to monopolize power and exclude other parties.
Post coup, interim president Mansour issued a Constitutional Declaration creating a roadmap for drafting a new constitution. 48 The new constituent assembly known as the "Committee of 50" was comprised of union leaders, prominent public figures, human rights leaders, and religious leaders. Coptic Christians and representatives of Al-Azhar each had four representatives. Conspicuously, Muslim Brotherhood and other Islamist representatives were largely excluded from the constitution drafting process, despite enjoying support from at least half the Egyptian population. Here, our hypothesis about the effect of increased polarization on exclusion derives further confirmation.
Experts have called the resulting 2014 Egyptian constitution an "aspirational document" that has expanded the power of "deep state" elites. 49 The constitution "outlines an ambitious vision of a democratic, liberal welfare state, a Sweden on the Nile, but confirms the dominant position of the military, a highly political judiciary, and the old political establishment that never hesitated in curbing rights and liberties in order to preserve its power". As the "Committee of 50" owed its position to the military, the constitution it drafted strengthened the role of the military. In direct response to the polarization that had dominated the Morsi years, the constitutional committee banned religious parties for fear of another election in which the Muslim Brotherhood and the Salafi Al-Nour Party would win a majority.

Libya
Muammar Gaddafi took power in a coup d'état in 1969 and ruled Libya as a personalist dictatorship until 2011. 50 Gaddafi crushed all political opposition and co-opted tribal leaders into his "Social Leadership People's Committee". He even undermined the military in favour of personal guards, mostly hired from sub-Saharan Africa with oil revenues. Loyal to Gaddafi (up to a point) and without any connection to the Libyan people, the dictator's security forces fired on demonstrators without qualm. After foreign military intervention brought revolutionaries to power, they had no existing national institutions to build upon, and because of the fragmented structure of the rebellion, the new state lacked a monopoly of force.
Libya has more than 140 tribes and clans, about 20 or 30 of which are of significant size. 51 There are also a Berber minority in Tripolitania and Touareg and Tebou minorities in Fezzan and Cyrenaica, respectively. The three provinces of Tripolitania, Cyrenaica, and Fezzan originated before Roman times and were administered separately under Italian colonial rule. The Sanusiyyah order of Cyrenaica ruled it as an independent state in the nineteenth century and even declared unilateral independence in 1949, only to absorb the more populous Tripolitania as the new Kingdom of Libya, under a federal arrangement, in 1951. 52 Federalist and even secessionist aspirations are therefore deeply rooted in the eastern part of the country. As Mezran puts it, Libya's first, "long, expensive, complicated constitution was the product of compromises" among many factions temporarily united by their fear of UN Trusteeship and recolonization. 53 The National Transitional Council (NTC) was formed by representatives of local rebel councils during the civil war that overthrew Gaddafi, consisting largely of non-Islamists in order to attract Western military support. 54 The militant Libyan Islamic Fighting Group (LIFG) had opposed Gaddafi clandestinely in the 1990s and 2000s, allying with the NTC during the revolution. The LIFG long maintained links to al-Qaeda in Afghanistan, which they renounced in the 2000s. Nevertheless, radical Islamic jihadists had a foothold in eastern Libya even before Gaddafi's overthrow. 55 Following Gaddafi's final defeat in October 2011, the NTC governed Libya, promulgating an interim constitution, under which elections were held to a General National Congress (GNC) in July 2012. 56 The most successful party group in this election was the broadly secular National Forces Alliance. The interim constitution remains the document governing Libyan official politics, but it has been comprehensively amended multiple times. From the beginning of its assumption of power, the NTC faced significant domestic opposition. Islamists stormed its Benghazi headquarters in January 2012. 57 Rogue militias have dispensed their own justice, run mafia-like protection rackets, and smuggled goods. 58 Over 100 militia brigades from Misrata engaged in retributive killings against Gaddafi supporters with impunity. 59 March and April 2012 saw Arab-Berber and Arab-Tebou fighting. 60 The NTC was ultimately a selfselected group of rebels sharing a desire to subordinate Islamic politics to the need for Western military intervention to overthrow Gaddafi. They lacked enforcement authority over much of Libyan territory and the myriad militias that sprang up after the collapse of the Libyan state.
After the GNC took power, it faced its own internal power struggle between the liberal National Forces Alliance and the Islamist Justice and Construction Party. By 2013, Islamists gained ascendancy and passed a "political isolation law" at gunpoint, forcing Gaddafi-era officials generally allied with more "secular" forces, including GNC president Muhammed al-Mugariaf, from office. 61 This power shift led to increasing violence with different elements of the GNC allying with different militias on the government's payroll. In 2014, the Islamic-dominated GNC extended its governing mandate and postponed parliamentary elections. In response, liberal-aligned Gen. Khalifa Hifter launched an assault against Islamist militias in Benghazi with the intent to win concessions from the GNC. Hifter's military action, which he labelled Operation Dignity, succeeded in pushing the GNC to hold parliamentary elections in June 2014, though his demand for establishing an emergency government under Libya's Supreme Court was denied. 62 The parliamentary election of June 2014 for a 200-person House of Representatives brought to power a majority of liberal-aligned politicians. Refusing to give up power, the GNC entrenched itself in the west supported by an alliance of Islamic, Berber, and Misratan militias, while the newly elected Parliament, aligned with Arab tribes and Arab militias, took up residence in Beidan in the east. 63 Divided by two separate parliamentary and military bodies and with vast ungoverned land areas, Libya has descended into chaos and civil war largely along identitarian lines.
Of the three cases examined here, Libya's interim constitutions possessed the most exclusive drafting process. Its purpose was to fend off, at various times, radical Islamists, secularists, Gaddafi-era officials, secessionists, and undisciplined militias. Libya's politics were not simply the unvarnished rule of the gun. Elections were so hotly contested precisely because actors believed elections mattered. Indeed, until 2014, some expert commentators believed Libya was progressing towards stable government. 64

Conclusions
Of the three cases examined here, Tunisia had the least ideological polarization and Libya the most, correlating precisely with the degree of inclusiveness in the constitution-drafting process. Tunisia's moderate Ennahda party accommodated the demands of liberals, and the resulting document won virtually unanimous support in the constituent assembly. Egypt's more radical Muslim Brotherhood joined with extreme Salafists to impose a conservative document on a radically opposed liberal and secular minority, who mostly boycotted the process. The subsequent Egyptian constitution excluded Islamists and entrenched military power. Libya has failed to adopt a final constitution and has now held two elections under an interim document. That interim document was initially drafted by the NTC formed of rebel representatives, who intentionally excluded factions that could frighten off Western support or otherwise cause havoc within the council. The current constitution is the handiwork of the GNC, which expelled a number of its own members via the political isolation law.
Our evidence confirms Hypothesis 1: the greater the ideological polarization among major factions, the less inclusive was the constitution-drafting process. We now turn to investigating Hypothesis 2; whether less inclusive constitutional processes resulted in constitutions that privileged empowerment of the government of the day over guarantees for political minorities.

The constitutional process and constitutional design
We evaluate constitutional design through two distinct methods. First, we attempt to score each of the constitutions under consideration quantitatively, by summing up government-empowering and government-constraining provisions, then normalizing by length of constitution. The next subsection describes this process and our results. Then, we focus specifically on the basic institutional design of the future government: executive-legislative relations, role of the military, decentralization, and so on. Here we assume that a greater number of veto players, less autonomous role for the military, more decentralization, and so forth are institutional features that constrain future governments. More exclusive constitution-drafting processes should yield fewer constraints on legislative majorities, according to our second hypothesis.

Document scoring
Post-Arab Spring constitutions, like most constitutions, accomplish several tasks. They define the nature and identity of the polity, set forth guaranteed rights, freedoms, and other policies, and specify how political decisions will be made, thereby structuring relationships among different branches of government.
For the purpose of our quantitative analysis, we sum up all government-empowering and government-constraining provisions in each constitution. An empowering provision increases the score by one, and a constraining provision subtracts one. Then we divide the sum of "net government-empowering provisions" by the number of scored provisions.
Libya's constitution is measured as two distinct documents: the provisional constitution as it stood in September 2012, and the substantially revised constitution of April 2014. Libya's provisional constitution was first issued by the NTC in 2011 during the rebellion. It was not intended to be a working constitution for day-to-day government, but growing instability and violence required amendments in 2012 to allow it to serve this function. The provisional constitution was then radically amended in 2014 as the GNC extended its mandate to rule and refused to hold parliamentary elections. Gen. Hifter's aforementioned attack forced the GNC to hold elections, which the liberals won by an overwhelming majority, in turn causing Islamists to break away and form their own parliament. 65 Because the two Libyan constitutions are so different, we are able to exploit some temporal variance to test whether the security situation in Libya affected the government-empowering provisions of these documents. We approach the two Egyptian constitutions, the 2012 Islamist-drafted constitution and the 2014 military-drafted constitution, similarly.
We count as a government-empowering clause any provision that lays duties on citizens to the state, provides for official languages, religions, or other tools of national unity, prohibits or discourages private competitors to the state, empowers the government to act quickly and decisively, permits or requires the state to regulate or control the economy, or provides exceptions to rights and freedoms enjoyed by citizens. We count as a government-constraining clause any provision that decentralizes power, creates checks and balances within government, limits government power over the individual, or constrains the military or executive. When coding limitations on government power, we code "negative rights" of non-interference but generally ignore "positive rights" to provision (e.g. health care, education, living wage, etc.), unless such clauses explicitly authorize government regulation or direct provision of a good or service, in which case we code the provision as government-empowering. We do this not because such positive rights are unimportant, but because it is unclear whether on balance they constrain the government of the day or empower it.
We expect more exclusionary constitution-drafting processes to correlate with more "net government-empowering" constitutions. Accordingly, Tunisia should have the least government-empowering constitution, Egypt 2012 next, Egypt 2014 third, and Libya should have the most government-empowering constitutions, with the second constitution (amended during a period of chronic instability) more governmentempowering than the first.
This method faces limitations. By simply summing up empowering and constraining provisions, we are implicitly assuming that each provision is equally important. Yet, core features of government structure matter more to government operation in practice than do substantive requirements, which have to be interpreted by some branch of the government itself. A guarantee of freedom of religion may not matter much when the constitution empowers a legislative majority to rule without constraint, and the electorate is religiously intolerant.
The Supplemental data display all constitutional provisions we code for all four constitutions. There are 132 constitutional provisions we code in all, and not every constitution features each provision. Tunisia's net government-empowering provisions as a proportion of all scored provisions are −21.0%, while Egypt's first constitution's are −19.8%, Egypt's second constitution's are −23.4%, Libya's first constitution's are 3.4%, and Libya's second constitution's are 3.0%. Libya's constitutions definitely look less liberal than Tunisia's and Egypt's, which are clumped together on this measure. Egypt 2014 beats Egypt 2012 because it abolished interventionist clauses such as the prohibition of blasphemy, state ownership of water, state control of health care, and university teaching of values and morals. Does Egypt's second constitution really create a "Sweden on the Nile", or are appearances misleading? As a next step, we evaluate the institutional structure each constitution creates in greater detail.

Institutional design
As argued above, more government-empowering constitutions contain fewer veto players allowing for quicker decision-making, more tools for the majority party or coalition to enforce discipline on backbenchers, more tools for parliament to select its own members, a greater, more autonomous, role for the military, and more territorially centralized institutions.
Libya's 2012 constitution, following Amendment 4, provided for a three-fifths majority for ordinary legislation and an undefined "relative majority" for electing a president. This supermajority requirement virtually ensures veto power for an additional legislative faction or party beyond the simple majority coalition needed to elect a government. In that respect, Libya's 2012 constitution was less majority-empowering than the others. However, the 2014 Libyan constitution repealed these requirements and instituted direct election of presidents (like Tunisia and Egypt).
The Tunisian and Egyptian constitutions provided for different sorts of veto players on various issues. For instance, Egypt's constitutions guarantee some independence to the central bank. Both Egypt and Tunisia provide for judicial review by a relatively independent judiciary, though Egypt's constitutions fail to specify the terms of office for constitutional judges.
Unlike Tunisia, Libya, and Egypt since 2014, which are effectively unicameral, Egypt's 2012 constitution had a "semi-strong" form of bicameralism: in the event of disagreement between the Senate and the House of Representatives, a joint sitting of an "absolute majority" of both chambers would take the final decision. Because the House was larger than the Senate, it enjoyed some advantage over the latter, distinguishing Egypt's bicameralism from a full-strength form as in the US. On the other hand, Egypt's president appointed a quarter of the Senate from among a narrow group of qualified individuals. This power increased the possibility of disputes between the House and Senate but also empowered the directly elected president on legislative matters. Egypt's 2014 constitution abolished the Senate. Furthermore, both of Egypt's constitutions allow presidents to hold positive referendums, thus bypassing the legislature to enact new laws, and to dissolve the legislature following a referendum, creating greater possibilities for unified government. All five constitutions we investigate provide for executive emergency powers, limited to 30 days in Tunisia, six months in Egypt 2012, three months in Egypt 2014, and indefinite in Libya.
When it comes to veto players, Tunisia's constitution and Egypt's 2014 constitution are formally the least government-empowering. Both constitutions are unicameral, and while Egypt 2014 allows a presidential veto, it also allows the legislature to force the president to resign, rendering the president almost an agent of the legislature. Tunisia, however, constrains government with negative presidential referendums, a shorter emergency period than Egypt, and the lack of presidential authority to appoint legislators, hold positive referendums, and dissolve the legislature following a referendum. Libya's 2014 constitution is clearly the most government-empowering, granting authority to the legislative majority without constraint and even permitting indefinite emergency declarations. Supermajority requirements made Libya's 2012 constitution less empowering than its later constitution, but emergency powers and lack of presidential and judicial checks on the GNC rendered it more government-empowering than Tunisia's and Egypt's constitutions.
Constitutions can also empower legislative leaders by giving them the ability to call early elections or confidence motions, thus shoring up party discipline. Both Tunisia and Egypt discourage early elections by limiting the conditions under which they may be called. In Egypt's 2012 constitution, a referendum must be called before early elections are permitted, a high barrier, and the 2014 constitution mandates a fixed election term under all circumstances except a successful, presidentially initiated referendum. In Tunisia, early elections may be called only if the government resigns and a new government cannot be formed. Tunisia also limits confidence motions: a successful no-confidence motion must be "constructive", that is must propose an alternative prime ministerial candidate. Tunisia's system is thus similar to Germany's, where party discipline is comparable to that of most other parliamentary democracies. 66 Tunisia's system provides stronger mechanisms for government stability and party discipline than Egypt's. Libya's 2012 constitution provided for a three-fifths supermajority for no-confidence motions, a provision repealed by the 2014 constitution.
Libya's 2012 and 2014 constitutions provide extraordinary mechanisms for legislative self-selection. The first permitted the legislature, the NTC and later the GNC, to enlarge its own membership. The second repealed this provision but authorized a "political isolation law" (under Amendment 5 of April 2013), which excluded Gaddafiera officials from serving in the legislature, and allowed (after Amendment 7 of March 2014) the legislature to dismiss its own members with a three-fifths majority. Article 87 of the 2012 Egyptian constitution allowed the legislature to remove its own members by a two-thirds majority vote. The 2014 constitution has a similar provision. These provisions potentially greatly enhance the power of a legislative supermajority to cement a monopoly on power.
When it comes to the role of the military in politics, Egypt is more governmentempowering than the other two polities. While the Tunisian, 2012 Egyptian, and 2014 Libyan constitutions all specify that the president is commander-in-chief, and the former two specify that the president appoints military officers, both Egyptian constitutions provide for a cabinet post as Minister of Defense for a military officer. The 2014 Egyptian constitution goes further, establishing the head of the military, not the president, as commander-in-chief, allowing the military to appoint the Minister of Defense during a transitional period, and setting up a separate military judicial system for trying its own members or civilians accused of crimes against the military.
None of these constitutions provide for substantial decentralization, let alone federalism, even though the latter is a key demand of powerful Libyan factions. Tunisia, Libya 2014, and Egypt all contain a general mandate of decentralization, without specifying details. Egypt's constitutions do provide for directly elected local councils, though the Egyptian president appoints all governors, who serve at his pleasure and enjoy more power than the elected councils. Tunisia, Libya 2014, and Egypt provide that police is a central government function, and Egypt's constitutions explicitly provide for the indivisibility of the country.
Finally, Tunisia and Egypt both impose a limit of two terms on the president. This provision may limit the potential for presidential "self-coups" and helps ensure the parliamentary nature of the overall system.

Conclusions
Egypt's first constitution was significantly more presidential than Tunisia's. While formally slightly more constraining of legislative majorities, Egypt's first constitution is not necessarily less majoritarian than Tunisia's, given the executive's greater scope of independent action. Within the Egyptian political context, in which the Muslim Brotherhood was expected to win the first few presidential elections under free and fair conditions, liberals saw the constitution as empowering the Brotherhood and its extreme allies. The Islamist framers of Egypt's first constitution clearly worried about the military's coercive threat to the new system they were establishing. They gave the military a role in politics, but the president's appointment power could have transformed the military's upper ranks over time. We can plausibly interpret the military's ultimate overthrow of the Morsi regime as a kind of commitment problem: the military worried about losing its veto over "high politics", and Morsi's allies also worried about leaving the presidency subordinate to the military. 67 The catalysts for Morsi's ouster were his appointment of 17 Brotherhood-affiliated governors, including a member of terrorist group Gamaa Islamiyah, and his repeated attempts to declare parliamentary elections under election law procedures ruled unconstitutional by the Supreme Administrative Court, resulting in protests by millions during June 2013. 68 Even though Tunisia's constitution shores up party discipline and government stability in a unicameral parliament, constitutional framers could foresee that no single party or coalition of parties was likely to obtain a permanent majority. Tunisia's constituent assembly had been directly elected by proportional representation, with the largest party (Ennahda) obtaining only 37% of the vote. Therefore, it was easy for Tunisian framers to forecast a future of two-or three-party coalition governments, providing additional veto players within the parliament and reducing the need for institutional veto points outside parliament.
By contrast, 66 out of 100 members of Egypt's constituent assembly were Islamists, associated with the Muslim Brotherhood and Al Nour Party. Therefore, the demand for institutional checks and balances from non-Islamists was much greater in Egypt than in Tunisia. The meagre supply of such provisions therefore provoked greater opposition in the former than in the latter.
In summary, Egypt's somewhat constraining but decidedly presidential constitution of 2012 interacted with the existing distribution of ideological preferences and political power so as to seem likely to empower the incumbent president and his Islamist allies in parliament. The 2014 constitution is less presidential but even more governmentempowering: ending bicameralism, entrenching the military in politics, and allowing the legislature to force the president to resign.
Libya's two constitutions have been the least government-constraining of the five. While the first constitution was intended to be a temporary set of rules for a constituent assembly, these constitutions governed day-to-day lawmaking from 2012 to late 2014, when parliamentary politics definitively broke down. Over time, supermajoritarian features were discarded, and the assembly's ability to select its own members grew. The constitutions place few explicit limits on the law's ability to reach into citizens' private and social lives.

Alternative hypotheses
One alternative explanation for the constitutional patterns we observe is that more liberal constituent assemblies straightforwardly implemented their preferences. This explanation would imply that Libya's first constitution would be the least government-empowering of all, since the NTC was dominated by relatively liberal factions. That prediction is clearly not borne out. At most, this hypothesis could explain why the rise of Islamists in the Libyan GNC resulted in a law decreeing Shari'ah as the basis of legislation and a constitutional amendment imposing Muslim religion as a test for presidential office, as well as why the Egyptian 2012 constitution was slightly more conservative than its 2014 replacement and the Tunisian one. There is nothing in conservative Islam that would suggest a more expansive executive emergency power, greater centralization of power, or broad presidential appointment powers. These are all majoritarian provisions, but not inherently socially conservative.
Another alternative explanation we consider is that a greater number of political factions, not to be confused with polarization among them, leads to a stronger state and more inclusive constitution-drafting in order to solve Prisoner's Dilemmas. But as we have seen, it was the more exclusionary constitution-drafting processes with input from fewer factions that led to more government-empowering constitutions. Factions in Egypt and Libya, especially, were concerned above all with the distribution of political power, not the positive-sum pooling of political power for collaborative ends.
Finally, the "power balance" hypothesis helps to explain why Ennahda had to compromise in Tunisia. Islamists were far weaker in the Tunisian constituent assembly than the Egyptian one. Yet, Ennahda could have formed a narrow majority with centrist and populist groups and independents against the staunchly secular left. Instead they chose to include almost all political factions in the drafting process. The balance of power determines whether a leading faction must compromise, but polarization determines whether and with whom the leading faction builds a more-than-minimum winning coalition.

Conclusion
We investigated whether greater polarization among factions reduces the willingness of the minimum winning coalition in constituent assemblies tasked with drafting constitutions to seek consensus on the drafts, and whether as a consequence of exclusionary drafting processes, more polarized societies end up with more government-empowering institutions.
We found that in the least polarized country, Tunisia, the constitution-drafting process ultimately achieved near-complete consensus, even though the rules of the process required only a two-thirds supermajority for approval without a referendum. By contrast, in the most politically fragmented society, Libya, initial consensus procedures were successively watered down, and Islamists and anti-Islamists engaged in a war of attrition, at first in a metaphorical, political sense and ultimately through a literal contest of arms. In Egypt, polarization rose through the course of the Muslim Brotherhood's presidency, and after their overthrow, a new, more majoritarian constitution emerged from an exclusionary process dominated by the military's appointees.
The one feature our model did not anticipate is that Egypt's constitutions would be only slightly more government-empowering than Tunisia's, given how much more polarized its parties were. Our revised interpretation of the vehement opposition that the Muslim Brotherhood-drafted constitution faced is that in Egypt's circumstancesa majority of the society favouring conservative, Islamist policies facing off against alienated secular and liberal minoritiesthe demand for constitutional provisions disempowering legislative majorities of the day and protecting individual rights was extremely high, but the supply was modest. To put the point another way, the majority's wish to rule over the minority rather than compromise with them can be offset somewhat by the minority's ability to use protest or violence to sanction the majority. It appears that the Muslim Brotherhood did attempt to compromise with its enemies in its constitutional proposal, but not enough to buy off the opposition and to prevent the Brotherhood's ultimate overthrow.

Disclosure statement
No potential conflict of interest was reported by the authors.