Participation, Inclusion, and the Democratic Content of Constitutions

Abstract

Theories of participatory and deliberative democracy contend that participatory and inclusive constitution-making processes are more likely to generate democratic outcomes than the traditional, elite-led approaches. The empirical evidence, however, has remained inconclusive and the propositions mostly normative. Using an original data from 195 constitutions promulgated in 118 countries since 1974, this study examines the impact of participatory and inclusive processes on the democratic content of constitutions. Building on the recently developed conjectures in the literature, this study introduces two original measures for individual-level public participation and aggregate-level group inclusion in constitution-making processes. The statistical analysis provides compelling empirical evidence that increased public participation is associated with an increased number of democratic provisions in constitutions, indicating that broad participatory processes can improve the democratic content of constitutions. Group inclusion, however, is not a significant predictor of the content of constitutions. The findings offer empirical support for participatory and deliberative theories of democracy and their prediction on democratic outcomes of participatory processes.

Introduction

Constitution-making has been viewed as an elite affair, ever since the Federal Convention deliberated on the American Constitution in 1787. It has been conventionally argued that elites are better informed, more committed to democratic values, and could be trusted more with safeguarding the general good than the masses. Recently developed conjectures in the literatures of participatory and deliberative democracy, however, contend that inclusive constitution-making processes are more likely to generate “smart” and “morally superior” outcomes (see Landemore 2017). This call for “direct democracy” in constitution-making processes, or what Chambers (2004, 153) labels “the democratization of popular sovereignty,” not only emphasizes the legitimacy of the process but also predicts that more participatory and inclusive constitution-drafting processes yield more democratic outcomes (Pateman 1970). Despite these predictions and a global trend toward more participatory and inclusive constitution-making processes in the last few decades (Franck and Thiruvengadam 2010), there is no conclusive empirical evidence that public participation has a positive impact on the democratic content of constitutions (Horowitz 2014, 100).

Using an original dataset and statistical analysis of 195 constitutional reform processes in 118 countries from 1974 to 2015, this article examines the impact of public participation and group inclusion in constitution-making processes on improving the democratic content of constitutions. This study distinguishes between public participation and group inclusion which are often used interchangeably in the extant literature and introduces two original measures for individual-level public participation and aggregate-level group inclusion in constitution-making processes which are premised on the participatory and deliberative theories of democracy. The results from the statistical analysis demonstrate that participatory processes are more likely to result in constitutions with more democratic provisions. Inclusive processes, however, are not a significant predictor of democratic content of constitutions. These findings offer empirical support for participatory and deliberative theories of democracy and their call for giving all citizens an equal opportunity and the right to participate and influence the outcome (Chambers 2004; Pateman 2012). Contrary to skeptics that depict mass public participation as dangerous to democracy (see Gibson and Duch 1991; Walker 1996), this study contends that, given a voice, citizens can in fact inject democratic provisions into the constitution.

In what follows, I first review the representative, participatory, and deliberative theories of democracy and discuss the two hypotheses this study raises. The “participation” hypothesis states that constitutions drafted via broad participatory processes are more likely to secure democratic provisions. The “inclusion” hypothesis posits that constitutions drafted through broad inclusive processes are more likely to secure democratic provisions. The next section introduces the “participation” and “inclusion” variables and discusses the measurement of the democratic content of constitutions. Finally, a discussion of the main findings and their implications for our understanding of theories of democracy and participatory constitution-making will be presented. These findings are associational—they do not demonstrate that public participation causes increased democratic provisions in a constitution. Given the relative infrequency of constitutional events and the cross-national nature of these data, proof of causality is extraordinarily difficult. More importantly, this article does not contend that the inclusion of different interests and groups in the constituent assemblies does not yield democratic outcomes. On the contrary, a recent study shows that group inclusion generates more improvements in post-implementation levels of democracy than public participation (Eisenstadt and Maboudi 2019). Focusing on de jure features of democracy in a constitution (rather than de facto levels of democracy which Eisenstadt and Maboudi (2019) study), this article, however, shows that there is strong evidence that public participation increases the number of democratic provisions in a constitution.

Participation and Inclusion in the Crafting of New Constitutions

Historically, the constitution-writing task was reserved for a select group of highly trained and well-informed elites who, it was argued, were more committed to democratic values than the masses and could better serve the general good. Edmund Burke (1999/1790) famously suggests that the mass lacks a “mature judgment” and “enlightened conscience” to be trusted with complicated issues such as the constitution. James Madison (1961/1787) similarly argues in Federalist No. 10 that “the public voice, pronounced by the representatives of the people, will be more consonant to the public good than if pronounced by the people themselves, convened for the purpose.” The role of representation for Madison, as such, is “to refine and enlarge the public views.” By contrast, John Adams (2000/1776) contends, in Thoughts on Government, that a representative assembly should be “in miniature an exact portrait of the people at large.” This notion of representation as a “mirror image” of the society was an expression of fairness and equality for the Anti-Federalists and an objection to the elitism of the Madisonian “filtering” metaphor (Fishkin 2009).

More recently, Pitkin (1967) identifies four views of representation including formalistic, descriptive, symbolic, and substantive representation. Much of the discussion on representation during the Cold War was informed by Pitkin’s emphasis on formal procedures of authorization and accountability, or what she calls formalistic representation (Plotke 1997). Failure of various democratization efforts, with their emphasis on formalistic representation, during the Second Wave of democracy, called into question the effectiveness of representative processes for democratic outcomes. Third Wave of democracy and the normative concerns related to “crisis of representation” associated with delegative and competitive models of democracy (Avritzer 2012) led to unprecedented innovations in public participation in different political processes—which were once reserved for elites—including participatory constitution-making. As Wampler (2012, 667) notes, “the direct incorporation of citizens into complex policy-making processes is the most significant innovation of the third wave.” Eisenstadt et al. (2017a), for example, show that more than two thirds of constitutions promulgated since the Third Wave of democracy have incorporated some form of direct citizen input from voting in constitutional plebiscites to participating in constitutional workshops, focus groups, national conferences, and other forms of public deliberation.

Parallel to these innovations in direct citizen participation in constitution-making processes, recently developed conjectures in theories of participatory and deliberative democracy have laid the foundation for justifying direct public input in constitution-making. Developed mostly in response to a rising criticism of the traditional democratic institutions (Barber 1984; Avritzer 2012), theories of deliberative democracy contend that a Burkean elite delegation with no public input can threaten the foundations of democracy as rule by the people (Mansbridge et al. 2012, 14). On the contrary, democratic authenticity requires substantive and real public participation and deliberation (Dryzek 2000) which empowers the marginalized in a society (see Chambers 2003).

These recent studies in the literature of participatory and deliberative democracy suggest that the people should be involved in the formulation and writing of the constitution for several reasons. First, inclusive and participatory processes are educational and developmental tools for citizens to practice democratic skills and procedures (Pateman 1970, 42). Increased public participation and group inclusion in decision-making processes can also function as a self-enforcing mechanism, preventing the government’s abuse of power (Smith 2009). Second, public participation and group inclusion can generate smart and epistemically superior outcomes (Landemore 2013). Translated to constitution-making processes, this literature suggests that constitutional provisions proposed by more participatory and inclusive constituent assemblies are more democratic than those generated by a select group of elites (see Landemore 2017).

These participatory and deliberative models, however, have been criticized on several grounds. Some scholars suggest that constitution making is an elite affair (Elster 1995; Horowitz 2008, 2013) that should be protected from the often-disruptive mass involvement (Arato 1995; Elster 2000; Sunstein 2001). Moreover, several studies have warned against the dangers of direct public participation in constitutional processes. “Participatory distortion” can result in self-selected groups with extremist views to dominate the public deliberation domain (Verba et al. 1995).¹ Participation in constitutional reform processes can also run the danger of populism and manipulation of the people by interest groups and elites (Ghai 2012). Furthermore, participation increases the number of veto players which can make reaching consensus more difficult (Tsebelis 2002).

Despite the warnings about the potential dangers of public participation and reservations about the effectiveness of participatory and deliberative processes, a few empirical studies suggest that participatory and inclusive processes resulted in the inclusion of various provisions on social and economic justice and corruption in Kenya (Cottrell and Ghai 2007), human rights in South Africa (Sarkin 1999), civil rights such as right of the disabled people in Colombia (Brett and Delgado 2005), rights of indigenous people in Guatemala (Marulanda 2004), and women’s rights in Uganda (Hart 2003). One comparative study of twelve cases shows that overall increased participation yields more provisions on human rights protection (Samuels 2006). These studies suggest that public participation and group inclusion can increase both the legitimacy and the democratic content of constitutions (Bannon 2007). Although these findings are important, the empirical evidence has remained inconclusive and the propositions mostly normative. Most importantly, however, these studies either lump participation and inclusion together (Eisenstadt et al. 2015) or use them interchangeably (Landemore 2017). This article, however, calls for a clear distinction between public participation and group inclusion in the constitution-making.

The main difference between participation and inclusion is in how public will is translated into policy outcomes. A participatory process calls for the direct involvement of individual citizens to protect their own rights in the constitution (Hart 2003). An inclusive process, however, is not a mere aggregation of individual participation. An inclusive process entails the inclusion of interest group advocates for societal positions as a necessary condition for the fashioning of a new constitutional order in all regimes, whether democratic or authoritarian. As Eisenstadt and Maboudi (2019) argue, in procedural terms, formal advocacy for ideas and issues may in fact be more important than widespread support for those issues. Horowitz (2013, 293), for example, noted the need for inclusion of all major interest groups within the constitution-drafting coalition, as Indonesia’s 2002 “slow, consensual, insider-driven process allowed the careful creation of new institutions and the creation of understandings among legislators.” In addition to this call for inclusion, Moehler (2008) demonstrated, that while participation in the1995 Ugandan constitution-reforming process provided civic education, it also made citizens cynical. Moehler’s findings revealed that views of local leaders, rather than mere public participation, drove citizen views of the constitution’s legitimacy.

Inclusion, as such, is very close to Dahl’s notion of pluralism. Dahl explains pluralism as the array of groups beyond the electorate which contribute positions and resources to the governmental framework. Pluralism assures wide representation to a broad array of groups, but also slows down policymaking, as groups adverse to political processes obstruct implementation of adverse policies (Dahl 2005, 1–8). Inclusion, as such, might slow down policymaking, but the articulation of different interests yields more democratic outcomes. Nonetheless, the breadth of inclusion and the articulation of different ideas is different from public participation and as Horowitz (2013) argues, matters more for democratic outcomes.

This article, thus, distinguishes between public participation and group inclusion and introduces two original measures for individual-level public participation and aggregate-level group inclusion in constitution-making processes which are premised on the participatory and deliberative theories of democracy. Using a cross-national analysis of 195 constitutions, it examines the impact of these two choices of constitutional process on democratic content of constitutions.

Hypotheses

If the argument on the “epistemically superior outcome” of more inclusive constitutional processes is correct, then we should expect that constitutions drafted by an inclusive constituent assembly and with broad public participation have more provisions on (1) democratic political institutions (where the executive is accountable to the legislature, and the judiciary is independent), (2) various individual, social, and political rights and freedoms, and (3) transparent political processes (such as independent electoral commissions) than constitutions without such processes. As such, the following two hypotheses are raised:

• H1 Constitutions drafted via broad participatory processes are more likely to secure democratic provisions than constitutions drafted without participation.

• H2 Constitutions drafted via inclusive processes are more likely to secure democratic provisions than constitutions drafted without inclusion.

These hypotheses speak to the deliberative and participatory theories of democracy which emphasize the importance of participatory and inclusive processes for democratic outcomes (Fishkin 2009). If the participation hypothesis is corroborated, it would provide empirical evidence for normative contentions that participatory processes yield democratically superior outcomes (Widner 2008; Wing 2008). It would show that, after all, the general public is also committed to democratic values. The second hypothesis, on the other hand, speaks to the studies that emphasize the inclusion of different societal groups and interests in constitutional processes rather than mere public participation (Horowitz 2013). Any evidence in support of the inclusion hypothesis and against the participation hypothesis would suggest that any participation must be fully channeled to assemble a constitution that resonates with the participants and that constitution-making is a political process, and as such, requires advocates with credibility derived from powerful entrenched interests, the popular will, or some combination of these.

Data and Analysis

This study builds on a previous work (Eisenstadt, LeVan, and Maboudi 2017b) to quantify the design of constitution-making processes across 195 constitutions that have been drafted since 1974. In their Constitutionalism and Democracy Database (CDD) of 138 constitutions, Eisenstadt et al. (2017b) measure the polyarchic nature of constitution-making processes in three stages of origination, deliberation, and ratification.² Their “process” variable is a measure of inclusiveness of constitutional processes in those three stages and it has three dimensions including the extent to which elites are included, the extent to which the general public is allowed to participate, and the transparency of the process.

This article distinguishes between public participation and group inclusion as two separate (but not mutually-exclusive) measures of constitution-making processes (see Table 1), which should not be lumped together or used interchangeably. I introduce two new measures for individual-level public participation and aggregate-level group inclusion in constitution-making processes which are premised on the participatory and deliberative theories of democracy. Participation, in this article, refers to the degree to which the general public is involved (participates) in different stages of constitution-making processes. Inclusion, on the other hand, refers to the extent to which major political and social groups in the society are included in the process.³ Colombia 1991, Ecuador 2008, and Tunisia 2014 are a few examples of constitutional reform processes that were both participatory and inclusive. China 1982, Saudi Arabia 1992, and Somalia 2012, on the other hand, are a few cases where the process was neither participatory, nor inclusive. In-between, there is a range of cases that are either inclusive (for example Portugal 1976) or participatory (for example Egypt 2012).

Table 1 Participation and inclusion in constitutional reform processes, 1974–2015^a

A constitution-making process can be inclusive without being participatory. An illustrative case in point is the Portuguese constitution-making process (1975–1976) where the process was not participatory but inclusive. Following the April 1974 military coup led by the Movement of the Armed Forces, the Portuguese citizens elected the Constituent Assembly tasked with drafting a new constitution in April 1975. However, the people were not invited to provide feedback on the content of the constitution (during the deliberation stage) and the document was ratified without a public referendum. Although non-participatory, the process was inclusive as all major groups (from the left-wing Communist Party to the central-left Socialist Party and the far-right Social Democratic Center) were part of the constitutional negotiations in all stages of the constitutional process (Magalhães 2015, 438–440).⁴

While estimating the level of public participation is easier, measuring inclusion is more complicated. How, exactly, do we determine whether groups are represented in the process? This article measures the inclusion variable as the absence of exclusion. “Negatively” measuring non-inclusion (or partial inclusion through boycotts) is more straightforward than “positively” measuring inclusion. This is due to the extensiveness of available information on boycotts and other forms of constitution refusals by major political and social groups. For example, since all major non-Islamist groups and political parties in Egypt (2012) boycotted the constitution, the process is coded as non-inclusive.

Here, I distinguish between self-exclusion and forced exclusion as two broad forms of non-inclusive constitutional processes. Self-exclusion, which is when a group voluntarily decides not to participate, although it is allowed to, is usually in the form of boycotts, withdrawal from the process, or other forms of vote-holding. Voluntary exclusion can be viewed as either a strategic move (often to press a matter on the majority in constituent assemblies) or a protest (when a group believes its participation will not yield the desired outcome). The inclusion variable considers both forms of exclusion (i.e., self-exclusion and forced exclusion) and treats them the same. In other words, the inclusion variable does not differentiate between the various functions or degrees of scope and breadth of exclusion. Rather, when any of these conditions exist in a case, this study counts that case as a non-inclusive process. This is because differentiation among causes of exclusion is difficult for some cases. But even without being able to code self-exclusion (as in Egypt’s April 6 Movement), as distinct from authoritarian group exclusion (as in Morocco’s February 20 Movement), the inclusion variable reflects whether constitutional processes are representative of all relevant societal groups and interests. Having explained the empirical difficulties of measuring inclusion and the logic of how this article addresses these, in the section that follows, I discuss in more detail how I operationalize inclusion and participation.

Operationalizing Participation and Inclusion⁵

The participation variable estimates the extent to which the general public is involved in the three stages of origination, deliberation, and ratification. For each stage, the level of public participation is coded 0 (non-participatory), 1 (mixed), or 2 (participatory). The participation variable then aggregates the measures of these three stages. It ranges from 0 (no public participation at all) to 6 (genuine public participation in all three stages). I then recoded the participation variable to create three ordered categories of “participatory,” “mixed,” and “non-participatory.” Table 7 in the Appendix shows the coding rules for components of the participation variable. As this table shows, a participatory process in the origination stage entails a popular election of the constitution-drafting body, and participation in the ratification stage entails a public referendum on the constitution. A participatory process in the deliberation stage, however, requires public deliberation and direct citizen input on constitutional proposals, which reflects the deliberative model of democracy and its emphasis on participation beyond merely voting or delegating authority (Fishkin 2009).

As Fig. 1 shows most constitutions since 1974 were drafted with limited public participation in the first two stages. Only 29 constitutions were written using participatory means in the origination stage. In the deliberation stage, only 23 constitution-making processes used extensive public input. However, almost 50% of the cases (96 constitutions) were participatory in the ratification stage.

Fig. 1

Level of public participation in different stages of constitution-making processes

Next, the inclusion variable measures the degree to which different groups in the society are included in the three stages of origination, deliberation, and ratification. For each stage, the level of inclusion is coded 0 (non-inclusive), 1 (mixed), or 2 (inclusive). The inclusion variable then aggregates the measures of these three stages. It, therefore, ranges from 0 (non-inclusive in all stages) to 6 (inclusive in all three stages). Finally, I recoded the inclusion variable to create three ordered categories of “inclusive,” “mixed,” and “non-inclusive.”⁶ Table 7 in the Appendix shows the coding criteria for this variable. The hallmark of the “inclusive” origination stage is the inclusion of representatives from all major groups, whether elected or appointed, in the constitution-drafting body. This is different from a “participatory” origination stage, which requires a constituent assembly directly elected by citizens, because an inclusive body can also be appointed. A case in point is Zimbabwe (2013), where after the disputed and violent elections of March and June 2008, all major political parties agreed to set up a Parliamentary Constitution Select Committee (COPAC) representing all stakeholders to draft a new constitution. The Zimbabwean COPAC was “inclusive” but not elected by the people (Hodzi 2013). An inclusive origination process, however, does not always follow by inclusion in the deliberation and ratification stages. If a political party or group dominates the assembly, it might act in an exclusionary manner, by imposing its views and failing to engage in constructive deliberation with the opposition. Egypt (2012) is an example of such processes which is coded “inclusive” in the origination stage but “mixed” in the deliberation stage. Egypt is coded “non-inclusive” in the ratification stage because all non-Islamist groups withdrew from the Constituent Assembly, did not ratify the constitution, and boycotted the referendum (Maboudi and Nadi 2016).

As Fig. 2 demonstrates, most constitutions since 1974 were non-inclusive (major political parties or groups were excluded from the process). Only 29 constitutions were inclusive in the origination stage, 37 in the deliberation stage, and 47 in the ratification stage. In other words, constitutional processes were more inclusive in the ratification and deliberation stages than in the origination stage.

Fig. 2

Level of inclusion in different stages of constitution-making processes

Take for example Zambia where economic crisis in the 1980s, the collapse of Soviet Union, and growing popular dissatisfaction with the ruling United National Independence Party (UNIP), forced president Kenneth Kaunda to call for constitutional reform aimed at introducing multiparty elections after 27 years of authoritarian (and mostly single-party) rule (Widner 2004). In September 1990, president Kaunda appointed a twenty-two-member Constitutional Review Commission (CRC) headed by a veteran lawyer, Patrick Mphanza Mvunga. The “non-inclusive” CRC soon toured the country, collected public feedback, and met with different parties, and drafted a report of constitutional recommendations. From the onset, the CRC (also known as the Mvunga Commission) was opposed by the main opposition party, the Movement for Multi-Party Democracy (MMD), which boycotted the commission in protest to UNIP’s domination of the process. A breakthrough came in July 1991 when Zambia’s National Organization of Churches mediated a committee of ten delegates (five from the incumbent UNIP and five from the opposition MMD) to negotiate changes to the constitution. These negotiations led to an agreement over the content of the constitution, and a draft was submitted to the Zambian Parliament which approved the constitution in August 1991 (Widner 2004). Because of the boycotts and the UNIP’s dominance over the process, Zambian constitutional reform process is coded “mixed” in the deliberation stage. However, since the constitution was approved by an elected body, the ratification stage is coded “inclusive.”

The cases of Zambia, Zimbabwe, and Egypt show that participation and inclusion are not interdependent, that is, participatory processes do not necessarily yield inclusion and vice versa.⁷ However, as Table 1 shows, in approximately half of the cases, the overall levels of participation and inclusion are similar. As the table shows, still the majority of constitutions (59 constitutions including China 1982, Saudi Arabia 1992, and Somalia 2012) are drafted via non-inclusive and non-participatory processes. Finally, the only case since 1974 that is participatory but non-inclusive is Egypt 2012 which was discussed above.

Measuring the Democratic Content of Constitutions

This article focuses only on de jure constitutional provisions and on the raw number of democratic provisions specifically; rather than de facto democratic quality of constitutions. To estimate the democratic content of constitutions, I use the Comparative Constitutions Project database (CCP) which records the characteristics of national constitutions from 1789 to 2013 (Elkins et al. 2014).⁸ From this dataset, I use 10 binary variables which illustrate different characteristics of constitutions as a proxy for democratic content of constitutions. Although issues covered in constitutions across the world vary widely, they all include a few main subjects including provisions relevant to government configurations as well as rights and freedoms (Elster 1993, 175). I also included a third category for the “political transparency” provisions in constitutions. As Table 2 shows, the dependent variables reflect these three main categories in the text of constitutions.

Table 2 Description of dependent variables

The choice of these three categories and their proxy variables are based on the scholarship of comparative constitutional studies. First, democratic constitutions limit the power of the executive by empowering other institutions including the legislature and the judiciary (Lijphart 2012). Thus, the first category of outcomes (Political Institutions) estimates whether such independent institutions exist. The executive accountability variable measures whether the legislature has the power to hold the executive accountable. And, the judicial independence variable estimates whether the constitution guarantees the independence of central judicial institutions.

Second, protection of rights and freedoms is another major subject in constitutions (Elster 1993), but there is no standard way for categorizing constitutional rights. For example, Elkins et al. (2013) categorize rights, based on their occurrence in national constitutions, into “common” rights (such as freedom of expression) which occur in more than 50% of national constitutions and “rare” rights (such as the right to affordable housing) which appear in less than 50% of all national constitutions. A better way is perhaps to distinguish rights as political, civil, social and economic, pertaining to their desired goals. Historically, core rights included liberty of conscience, property and personal security rights, and freedoms of expression and association. More recently, however, constitutional rights have assumed two new roles. On the one hand, they are designed to protect minority groups, and on other hand, they guarantee material welfare (Elster 1993).⁹ Therefore, the second category (rights and freedoms) consists of six variables as proxies for constitutional provisions for core civic rights and freedoms (freedom of expression and religion variables), political and minority protection rights (party formation and minority quota variables), and institutions protecting rights (variables for human rights commission and ban on restricting rights during state of emergency).

Finally, transparency is another important aspect for the democratic content of constitutions, especially those crafted after the Third Wave. Transparency of political processes is inherent in democratic values, and it is necessary for government accountability and a democratic polity (Dahl 1971; March and Olsen 1994). Hence, it is important for a constitution to ensure that various political processes are as transparent as possible. The last category (Transparency) uses two proxy variables, election oversight and public minutes, to measure whether a constitution includes provisions regarding transparency.

Using the above variables, I created three indices for the democratic content of constitutions. First, political institutions index is created by aggregating executive accountability and judiciary independence. This variable ranges from 0 where the executive is not accountable to the legislative and the judiciary is not independent to 2 where both conditions exist. Second, rights and freedoms index is created from the sum of expression, religion, party formation, minority quota, human rights commission, and state of emergency variables.¹⁰ This variable ranges from 0 where none of these right and freedom provisions exist to 6 where all six provisions exist. The transparency index is created from the sum of election oversight and public minutes variables. It ranges from 0 where neither of the provisions exist in the constitution to 2 where the constitution provides both. Lastly, the democratic content index is created from the aggregation of all variables. It ranges from 0 (no democratic provision exists at all) to 10 (all democratic provisions exist). On average, each constitution in the dataset has 5.3 democratic provisions (with a standard deviation of 2 provisions).

Control Variables

The tests in this study control for several variables that could have significant consequences for the democratic content of constitutions and/or the design of the process. Certain polities might be more inclined to allow a participatory and inclusive process, and simultaneously end up with a ‘democratic’ constitution. In other words, it can be argued that it is not participation or inclusion that is affecting the constitution, it is some characteristic of the country that fosters an inclusive process and democratic constitutions. Several control variables intend to control for these features. First, I lag the dependent variables using the previous constitution. Studies have shown that constitutions are not written on a blank canvas and that for the most part constitutions resemble their predecessors (Elkins et al. 2009, 57). The normative assumption is that countries with democratic constitutions are more likely to utilize inclusive processes when reforming their constitutions and end up with a new democratic document. Several democratic backsliding cases, however, contradict this prediction. Take for example Turkey’s 2017 constitutional reform process which has replaced a democratic constitution with a semi-authoritarian document that its main purpose is to extend President Erdogan’s tenure in office. In my dataset, 169 constitutions have predecessors and the remaining 26 cases are first constitutions in newly established states. As Table 3 shows, constitutions have become slightly more democratic over time. Overall, while constitutions promulgated since 1974 have on average about 5 of the 10 democratic provisions (first column), their predecessors have about 4 of those provisions (second column).

Table 3 Descriptive statistics of outcome variables

This study also controls for level of democracy which could impact both the process of constitution-making and the content of a constitution. The democracy variable uses the average of Polity IV democracy score for 3 years before the constitution promulgation. This variable ranges from − 10 (strongly autocratic) to + 10 (strongly democratic). The domestic crisis variable controls for recent major civil conflicts including revolutions, strikes, riots, and demonstrations since constitutions are often part of postwar peace building, and this can influence the impetus for inclusiveness as well as the political stakes of public deliberation (Miller and Aucoin 2010; Roeder and Rothchild 2005). A natural log of Banks and Wilson’s (2014) conflict variable is used for measuring average domestic crisis for 3 years before the constitution promulgation. Next, ethnic fractionalization could impede democratization by breeding parochialism (Horowitz 1985) or facilitate it by enabling civil society mobilization (Bessinger 2008). Thus, the ethnic division variable controls for ethnolinguistic fractionalization using Alesina et al.’s (2003) data. This variable ranges from 0 indicating ethnic homogeneity to 1 indicating significant fractionalization. This study also controls for democratic and authoritarian transition using the Polity IV democracy score. A five-point increase or decrease in Polity’s democracy score in any year in the last 3 years before promulgation indicates a democratic or authoritarian transition, respectively. I also control for level of development with a natural log of GDP per capita. Next, as larger countries may be less likely to democratize due to population density or other factors (Teorell 2010), a natural log of population variable is included. These last two variables are based on the World Bank’s World Development Indicators (2013). Lastly, this article controls for time using waves of constitution-making. I do not include year as a covariate, because it assumes a linear effect of time. However, democracy and constitution-making often come in waves. In my dataset, I identified two such waves. The first wave (1974–1979) resulted in 41 new constitutions and the second wave (1990–1996) led to 69 new national constitutions. In other words, 110 constitutions in my dataset of 195 constitutions were written during these waves. Two binary variables are used to control for whether the constitution was promulgated during any of these two waves.

Model Specification and Results

In this article, I test whether overall increased participation and/or inclusion increases the overall democratic content of constitutions, including democratic provisions pertaining to checks and balances, rights and freedoms, and transparency. Since all the indices are ordered categories, an ordered probit model is used for all of the outcomes. The general statistical ordered probit model is:

$$ {Y_i}^{\ast }={\beta}_1{Participation}_i+{\beta}_2{Inclusion}_i+{\beta}_3{Z}_i+{\upvarepsilon}_{\mathrm{i}} $$

(1)

where Y^* is the latent (unobserved) dependent variable (democratic content of constitutions). Participation_i and Inclusion_i are the main procedural features of the constitution writing process and Z_i is a vector of covariates including lag of the dependent variable (content of previously in-force constitutions), democracy, ethnic division, domestic crisis, authoritarian and democratic transition, GDP per capita, population, and constitutional waves. Table 4 shows the results of the statistical analysis.

Table 4 The impact of participation and inclusion on the content of constitutions

In all the models, increased overall participation contributes to the democratic content of constitutions, indicating that as the degree of public involvement in a constitution-making process increases, the constitution is more likely to include provisions on democratic political institutions, various citizens’ rights and freedoms, and transparent political processes. The results support hypothesis 1 and indicate that constitutions drafted via broad participatory processes are more likely to secure democratic provisions. The inclusion variable, however, is not statistically significant (except for rights and freedoms) and shows mixed results in different models. In other words, there is not strong evidence that constitutions drafted using broad inclusive processes are more likely to secure a larger number of democratic provisions.¹¹

Among control variables, the content of previous constitution (predecessor content) is the most significant predictor of the democratic content of constitutions. The results show that countries in which the previous constitution was democratic are more likely to adopt constitutions that secure several democratic provisions pertaining to political institutions, citizens’ rights and freedoms, and transparent political processes. The level of democracy and ethnolinguistic fractionalization also have a statistically significant relationship with the democratic content of constitutions, showing that democratic or ethnically diverse countries are more likely to adopt democratic constitutions.

Lastly, the results show that constitutions written during both constitutional waves have a negative correlation with the number of democratic provisions. For the first constitutional wave (1974–1979), this relationship is significant for rights and freedoms, transparency, and democratic content outcomes. That is, constitutions promulgated in the early years of the Third Wave of democracy have fewer democratic provisions than other constitutions in my dataset. For the second constitutional wave (1990–1997), this relationship is slightly significant only for the transparency outcome.

Endogeneity and Robustness Checks

The results from Table 4 stand several robustness analyses. First, I estimate a pared down model using only the main predictors without any controls. The results show that we observe the same effect for participation as in the model with covariates (see Table 11 in the Appendix). Second, I estimate the main model without lag of the dependent variable. This article lags the dependent variables using the previous constitution. This, however, excludes new states from the analysis. It is empirically important to know whether such states are particularly democratic, since they are unburdened by previous constitutions. Table 12 in the Appendix shows that the results are very similar to the main model; indicating that even when we include new states in the analysis, we observe the same effects. Next, I interact participation and inclusion and test the same model. Table 4 shows that participatory processes, in contrast with non-participatory processes, render more democratic constitutions, independent of how inclusive the process is. However, it could also happen that the positive effects of participation are contingent upon the inclusiveness of the process, and vice versa. To test this, an interaction term is added. Table 13 in the Appendix shows that the interaction term is not statistically significant, indicating that the impact of participation is not contingent upon the inclusiveness of the process, and vice versa.¹²

The results from Table 4 and the additional statistical analyses presented in the appendix show that public participation and the content of previous constitutions are the only statistically significant predictors of the democratic content of constitutions in all models. It can be argued, however, that the same factors that propel a society to adopt more participation or inclusion in a constitutional process may lead to more democratic provisions in the constitution itself. One of these important factors is the country’s level of democracy. More democratic countries, with already democratic constitutions, are more likely to have more participatory and inclusive processes and write more democratic constitutions. While this is true in many cases, several robustness checks show that the results are not driven by levels of democracy. To clarify the relation between the constitutional process and democratic content of the constitution, I first split the data into democratic and authoritarian regimes, estimating whether increased participation and/or inclusion in either regime type increases the overall democratic content of constitutions.¹³ If the results are driven by the level of democracy, we would not expect participation to be a significant predictor of democratic content of constitutions in dictatorships.

The results in Table 5, show that regardless of the regime type, participation remains a significant predictor of the democratic content of constitutions. In both democracies and non-democracies, citizen participation has a positive and significant correlation with the number of democratic provisions in constitutions. Similarly, as Table 5 shows, there is a positive and slightly significant relationship between the content of previous constitutions and the number of democratic provisions only in autocracies.

Table 5 Participation, inclusion, and content of constitutions in democracies and dictatorships

Table 5 shows that despite the normative assumption, level of democracy is not a strong predictor of the content of constitutions. Figure 3 also points to the same conclusion. The graphs show that the most authoritarian countries (score of − 10 in Polity IV index) on average have 5 democratic constitutional provisions (out of 10), while the most democratic states (score of + 10 in Polity IV index) on average have 7 democratic constitutional provisions. In other words, moving from − 10 to + 10 on Polity scale results in only 20% increase in democratic provisions of a given constitution. It also shows that constitutions drafted with participatory or inclusive processes have more democratic provisions than those with non-participatory or non-inclusive processes, regardless of their pre-implementation level of democracy.

Fig. 3

Relationship between democracy, process, and content of constitutions

Lastly, I use changes in the content of constitution as the dependent variable. The content change is an ordered categorical variable which indicates whether the constitution has more, less, or a similar number of democratic provisions, compared to its predecessor. Table 6 shows the results for the impact of constitutional processes on change in the democratic content of constitutions. The results are similar to all other models, showing a positive and significant relationship between participation and change in the democratic content of constitutions. In other words, participatory processes are more likely to result in increased number of democratic provisions.

Table 6 Participation, inclusion, and change in the content of constitutions

These statistical results cannot completely insulate the findings from the problem of endogeneity. This study acknowledges that endogeneity exists and that the estimates are likely to be biased. However, the robustness tests suggest that such biases are unlikely to be large. In many societies level of democracy is an important predictor of both procedural features of constitution writing and the democratic content of constitution. The robustness tests, however, show that plausible alternative explanations such as level of democracy cannot systematically explain all the empirical findings, and that regardless of regime type and level of democracy, increased participation in constitution-making processes is associated with more democratic provisions.

In sum, the results offer strong empirical evidence that participatory constitutions are more democratic in content than constitutions made with limited or no public input. This finding speaks to the democratic theory by offering empirical evidence that public participation in constitution-making processes can help ensure democratic constitutions. The statistical analysis also demonstrates that while increased participation is important for the overall democratic content of constitutions, inclusion is not a statistically significant predictor of the number of democratic provisions.

Discussion and Conclusion

Among 195 constitutions promulgated between 1974 and 2015, at least 70% have incorporated some degree of citizen input, showing a global trend in participatory constitution-making processes (see Fig. 1). Theories of participatory democracy suggest that citizen participation is a necessary condition for democratically legitimate political processes, and that all citizens should be given an equal opportunity to participate and influence the outcome (Pateman 2012). Skepticism, however, remains as to the extent to which these participatory processes are important for the democratic content of constitutions. Critics of such processes sometimes even depict mass public participation as dangerous to democracy (see Gibson and Duch 1991; Walker 1996). Using original measures of participation and inclusion in 195 constitution-making processes since 1974, this study shows that when given the chance, direct public input can improve the democratic content of constitutions. Despite skepticism, public participation is not dangerous for democracy. Rather, direct public input is associated with an increased number of democratic provisions in a constitution.

The statistical analysis points to two major findings. First, there is strong evidence that overall increased participation in constitution-making processes is associated with improvements in the democratic content of constitutions. This, however, does not mean that participation improves every single provision in the constitution or that all participatory processes are successful in generating democratic constitutions. Rather, public participation in general is more likely to generate a democratic outcome. Second, inclusiveness of the process is not a significant predictor of the democratic content of constitutions, which indicates that a process inclusive of different societal groups and interests without meaningful public involvement in the drafting of new constitutions does not necessarily translate into more democratic provisions. However, this should not be interpreted as the ineffectiveness of the inclusion of different groups in constituent assemblies in yielding a democratic outcome. On the contrary, a recent study shows that group inclusion generates more improvements in post-implementation levels of democracy than public participation (Eisenstadt and Maboudi 2019). The reason inclusion is not a significant predictor of democratic content of constitutions can be due to the fact that this article focuses on de jure constitutional provisions, rather than de facto constitutional practices which are more difficult to measure. A few cases illustrate these two conclusions.

Colombia 1991 and Tunisia 2014 are both examples of participatory and inclusive processes leading to more democratic constitutions. After the assassination of Colombian presidential candidate Luis Carlos Galán in 1989 which resulted in a students’ movement calling for a referendum “for peace and democracy,” a constituent assembly election was held in December 1990 after a national plebiscite on the assembly itself, in which the public overwhelmingly supported the election of a constituent assembly. After the plebiscite, the Supreme Court ruled that the assembly could not be limited in its scope and capacity. Although some major insurgents like the Revolutionary Armed Force of Colombia (FARC) and the National Liberation Army (ELN) were excluded, the demilitarized April 19 Movement (M-19) leftist insurgents were able to participate and won 19 (out of 70) seats in the constituent assembly. The constituent assembly was by far more inclusionary than earlier national pacts since the 1958 pact which divided rule among the two conservative political parties, to the exclusion of the left (Eisenstadt et al. 2017a). This inclusionary process which accompanied an active public participation resulted in significant improvements in the democratic content of the Colombian constitution. The 2014 constitution of Tunisia is a more recent example of a participatory and inclusive constitution-making process yielding significant democratic improvements. The constitutional renovation process in Tunisia started as an inclusive process where elected representatives, reflecting the public will and interests, were tasked with drafting a new constitution. However, political conflict and mistrust among different groups soon engulfed the process and an intervention by civil society organizations, most notably the National Dialogue Quartet, led to a more inclusive and participatory process which yielded a constitution highly regarded as the most progressive constitution in the Arab World (Maboudi 2019).

While participation led to more democratic constitutions in Colombia and Tunisia, it did not have a positive impact on the democratic content of the Egyptian and Ecuadorian constitutions. The 2012 Constitution of Egypt was drafted through a mass participatory process, but without political inclusiveness and fair and equal representation. More than 100,000 Egyptians enthusiastically participated in a process which was viewed as unfair, corrupt, and illegal by most non-Islamists. Public participation in Egypt did not result in a democratic outcome, mostly because public demands were not channeled through an inclusive constituent assembly. Rather, it led to a constitution that was viewed as unfair and undemocratic by millions of Egyptians and aggravated a situation that led to a military coup and the revocation of the constitution only 6 months after its promulgation (Maboudi and Nadi 2016). Similar to Egypt, in Ecuador’s constitutional reform process of 2008, public participation played only a “window-dressing” function. Although the constituent assembly was directly elected by citizens after a public referendum on the issue in 2007, President Rafael Correa managed the elections and his PAIS Alliance (Proud and Sovereign Fatherland) won 74 of the 130 seats giving Correa the power to write his constitution. Although this participatory process resulted in a highly popular constitution at the time, it failed to improve the democratic outlook of the country.

As these cases demonstrate, not all participatory or inclusive processes result in successful democratic transitions or democratic constitutions. The cross-national analysis presented in this study shows a broad global trend which is generalizable to most but not all cases. It suggests that participation is a necessary but not a sufficient condition for the democratic content of constitutions. As Horowitz (2013, ix) notes, “there is no universal solvent in choice of constitutional reform processes.” Rather, there are several optimal paths, each appropriate in a particular context. While for Indonesia this path was an inclusive, in-house, and gradual process which took place over several years, for Tunisia and Colombia it was a participatory and inclusive process with equal and fair representation of different political ideologies and social groups. Regardless of an optimal constitutional process, this article shows that when citizens are engaged in the constitution-making process and can impact political processes that govern them, constitutions tend to be more democratic and reflective of the will of the people.

Appendix. Coding rules, descriptive tables, and additional analysis

Table 7 Coding Criteria for Participation Variable

Table 8 Coding Criteria for Inclusion Variable

Table 9 Descriptive Statistics of Participation and Inclusion Variables

Table 10 The Impact of Different Stages of Participation and Inclusion on the Content of Constitutions

Table 11 Large Sample Model

Table 12 The Impact of Participation and Inclusion on the Content of Constitutions (including New States)

Table 13 The Impact of Participation and Inclusion on the Content of Constitutions (with Interaction Term)

Table 14 The Impact of Participation and Inclusion on the Content of Constitutions (OLS and Poisson Models)