Abstract
This paper compares Brazil's (1987–1988) National Constituent Assembly and Chile's (2021–2022) Constitutional Convention to examine how partisanship and institutional design shape constitution-making outcomes. Brazil's party-based assembly, despite internal ideological heterogeneity, achieved an accepted constitution through structured negotiation and leadership mechanisms. Meanwhile, Chile's nonpartisan convention faced fragmentation among independent constituents, had limited coordination tools, and ultimately produced a draft rejected by referendum. Drawing on qualitative comparative case approach, roll-call analysis, and matching techniques, we show that partisan infrastructure can mitigate negotiation challenges by fostering compromise. In contrast, nonpartisan assemblies may enhance inclusiveness but risk fragmentation and instability without robust coordination mechanisms that bridge internal divisions. These findings challenge the presumed superiority of nonpartisan conventions, demonstrating the importance of party-based institutional frameworks for managing collective action problems in constitution-making. In doing so, this study highlights the tension between inclusiveness and the organisational capacity required to achieve compromise-based agreements and lasting constitutional outcomes.
Introduction
The crafting of a constitution is a foundational political act that shapes not only a state's legal framework but also the distribution of power and channels for democratic representation. However, scholars have not reached a consensus on the institutional conditions needed for a constituent assembly to succeed, especially during times of social tension, political crises or regime transition.
This study examines the role of political coordination, particularly among political parties, within the broader institutional framework of constitution-making. On the one hand, parties may be merely self-interested actors that distort public will; on the other hand, they may serve as essential vehicles for negotiation, coalition-building, and institutional stabilisation. Parties can centralise decision-making and reduce negotiation costs (Hudson, 2021; Negretto, 2013), or they may reinforce elite dominance or pork barrel practices, undermining the popular legitimacy of constitutional outcomes (Elkins et al., 2009). This dynamic is illustrated by the contrasting experiences of Brazil (1986–1988) and Chile (2019–2022). Brazil's party-driven constituent assembly produced a widely accepted, policy-laden democratic constitution under intense political pressure. In contrast, Chile's independent-dominated constitutional convention, convened after intense social mobilisation, yielded a detailed and divisive draft that was ultimately rejected by a referendum.
We argue that one possible answer for the political coordination puzzle within constitution-making processes lies in both the partisan composition of an assembly and the institutional infrastructure that shapes its behaviour. Specifically, we demonstrate that party-based assemblies, when embedded in institutional frameworks that empower leadership and channel demands, have essential tools for overcoming collective action problems. Conversely, nonpartisan assemblies that lack such coordination mechanisms are more vulnerable to negotiation failures, diffuse representation, and unstable decision-making outcomes. This is not to suggest that partisanship is a panacea, or that independent constituents lead to failure. We contend that successful outcomes depend on a synergy between political actors’ capability for structuring coordination and cooperation and the institutional designs that enable such capability. When this synergy between actors and institutional design is absent, constituent processes are at risk of fragmentation, gridlock, and, ultimately, collapse.
To assess this argument, we employ a multi-method research design. First, we conduct a comparative qualitative analysis of the political processes, institutional architecture, and constituent composition of the Brazilian and Chilean cases. Second, we analyse roll-call voting data to compare negotiation outcome patterns and coalition cohesion across the two assemblies. Finally, we apply matching techniques to isolate the effects of non-partisanship in Chile's convention, using Brazil's party-based case as a benchmark. This integrated approach enables us to systematically assess how institutional design and the political background of actors shape negotiation outcomes in constitution-making.
The paper is organised as follows. Section “Institutional design in contexts of political transformation” introduces and explores the Brazilian and Chilean cases, providing a qualitative comparison of the institutional designs adopted in each constitution-making process. Section “Partisan influence in constitution-making” develops the theoretical framework, drawing on institutionalist literature to examine the contested role of political parties in constitutional design. Section “Quantitative comparison” presents a comprehensive quantitative analysis, beginning with an overview of the data and methodological approach, and proceeding to a discussion of descriptive results based on spatial estimates of voting behaviour and indicators of party cohesion and agenda-setting capacity. This section concludes with a matching analysis designed to isolate the effects of non-partisanship in the Chilean case. The final section synthesises our findings and reflects on their broader implications for theories of constitution-making, democratic representation, and the role of partisanship.
Institutional Design in Contexts of Political Transformation
In Latin America, institutional transformations often occur when existing institutions no longer align with the distribution of political power (Geddes, 1990). Scholars of democratic transitions have long focused on the emergence of a favourable political culture during the authoritarian period or on efforts to address past institutional flaws (Hagopian, 1993), often overlooking the relation between political parties, constituents, mediation networks, and the institutional framework for political competition.
Constitution-making processes often coincide with cataclysmic events (Ackerman, 1994; Elster, 1993, 1995; Widner, 2008) and can play a role in easing internal tensions (Widner, 2005). Brazil (1986–1988) and Chile (2019–2022) experienced significant civil unrest at different times and contexts. Yet, despite their distinct timelines, the two countries shared similar narratives in their responses to these crises. These narratives framed the political bargaining processes under specific constraints (Chilton et al., 2024; Ginsburg and Bisarya, 2022). Brazil and Chile sought to structure their constitution-making processes as part of their transitions to democracy, to heal the wounds left by authoritarian regimes and mitigate political and social upheavals.
In Brazil, political parties and elected officials acted as social and constitution-making brokers. They forged a pact with social movements and controlled a constituent assembly model composed exclusively of members of the National Congress. Social actors played advocacy and consultative roles in the process. The draft, approved by party actors without a public referendum, remains in place today. In contrast, Chile's constituent convention (Convención Constituyente (CC)) was composed entirely of members elected through popular vote, tasked solely with drafting a new constitution. Nonpartisan social movements and independent actors proved to be electorally effective, securing the majority of constituent seats. These actors served as political brokers, incorporating mechanisms for social participation during the drafting process. However, the constitution drafted was ultimately rejected in a plebiscite.
Brazil's democratic transition was closely tied to its constitutional process in the late 1980s. After being defeated in the campaign for direct presidential elections, marked by the iconic “Diretas Já” movement, the democratic elites shifted their focus to drafting a new constitution. While losing the battle for direct elections, they won the broader war by pushing for a constituent legislative assembly, which became the National Constituent Assembly (Assembléia Nacional Constituinte (ANC)) (Lopes, 2008; Martínez-Lara, 1996; Pilatti, 2008).
In turn, Chile began its transition to democracy following the 1988 referendum that mandated the resignation of dictator Augusto Pinochet. Notably, the country continued to operate under the military regime's constitution, which retained authoritarian residues despite democratic reforms in the 1990s (Heiss and Navia, 2007; Larrain et al., 2023; Siavelis, 2016). Efforts to reform the constitution, such as a failed project by President Michelle Bachelet in the late 2010s, highlight the difficulties in navigating Chile's constitutional landscape. These difficulties exacerbated internal tensions within President Bachelet's coalition (Weeks and Borzutzky, 2012).
In 2019, Chilean President Sebastián Piñera shifted his stance amid widespread social crises and mobilisations, endorsing a new constitutional process. A plebiscite followed, with most Chilean voters supporting the drafting of a new constitution and opting for a constituent convention composed of members elected specifically for this purpose. This process, shaped by various substantive, procedural, and institutional constraints, was influenced by a range of political dynamics (Chilton et al., 2024). Nevertheless, the draft was rejected by more than 60 per cent of voters in the subsequent ratification referendum. This outcome makes Chile an outlier in the world's constitution-making cases – featuring many independent actors but enacting no new constitution. Subsequently, political parties pushed for another round of constitution-making under an ad hoc mixed model, but most of the population again rejected this draft.
Both countries faced the challenge of drafting constitutions amid periods of profound political change. Their respective choices of constituent models reflected different responses to the pressures of democratisation, societal demands, and political constraints. These decisions paved the prevailing political paths and defined the extent to which political parties influenced the constitution-making process. This highlights the reciprocal relation among institutional design, partisan involvement, and constitution-making outcomes.
The following subsections examine the constitution-making processes of Brazil (1987–1988) and Chile (2021–2022), qualitatively comparing the historical, contextual, political, and institutional designs that shaped their approaches. In the forthcoming sections, we aim to demonstrate the consequences of the institutional choices made by Brazil and Chile, particularly how the constituent models’ preference for partisan-based versus independent membership influenced political outcomes. We also examine the role of partisanship in those processes and assess its substantial impact on the behaviour of the assemblies.
Qualitative Methodological Consideration
This qualitative methodological subsection outlines a case-based approach to examining the factors that influence constitution-making processes, settings, and outcomes. Case-based research seeks to identify patterns within and across different cases, offering valuable insights into complex phenomena that are challenging to analyse using purely quantitative methods (Gerring, 2017). Unlike direct causal inference approaches, qualitative cross-case comparisons involve conceptualisation, measurement, and exploration (Seawright and Gerring, 2008). By investigating how multiple potential factors interact to produce the observed outcomes, we can gain a deeper understanding of the meanings and implications of these outcomes and refine theoretical constructs (Slater and Ziblatt, 2013).
The following subsection explores causes-of-effects reasoning, examining how multiple interacting factors – rather than a single causal pathway – influence political outcomes (Gerring, 2017). Brazil and Chile were selected as diverse cases representing contrasting institutional models and varying levels of partisanship among constituent members. This selection provides a solid foundation for cross-case comparison. Through a juxtaposed analysis (Schaffer, 2021), we inductively identify critical mechanisms, driving forces, and contextual factors. Thus, we can bridge theoretical debates concerning the role of partisanship in negotiation and constitution-making.
This qualitative framework explores the guiding question: How do differences in institutional design and levels of partisanship influence constitution-making processes and their outcomes? This case-based comparison reveals the underlying institutional mechanisms and their interaction with partisan and nonpartisan constituents. By linking the cases of Brazil and Chile to broader debates, the qualitative exploration adds depth to theoretical discussions and connects to the subsequent quantitative analysis of voting behaviour.
Qualitative Juxtapositional Comparison
The Brazilian constituent assembly emerged in an energised social context, driven by the expectation of dismantling the authoritarian legacy. It fostered public participation and incorporated diverse political factions and groups, including a heterogeneous and disorganised right wing, which shaped the regime transition (Freitas et al., 2009; Pilatti, 2008). In contrast, Chile underwent unprecedented social mobilisation during President Piñera's centre-right administration. This movement challenged the political establishment and culminated in constitution-making efforts and referendums aimed at approving a new draft. Despite these differences, both countries shared a politically and socially mobilised, left-leaning context paired with a fragmented or weakened institutional right wing. In Chile's case, the coronavirus disease-2019 (COVID-19) health crisis may have also amplified social and digital scrutiny.
Brazil and Chile also shared the same macro-institutional constitution-making design: decentralised committees initially, followed by centralised deliberations in plenary sessions, but with different majority requirements to approve constitutional matters. The Brazilian ANC demanded a simple majority on both. Meanwhile, the Chilean CC established a simple majority on committees but a two-thirds majority on the floor.
Constituent assemblies often carry the expectation of ending or reforming a regime, but they have to do so while still being part of the current regime (Bonime-Blanc, 1987). Brazil and Chile significantly differ regarding this issue. Brazil's ANC looked forward to concluding the transition to democracy. Contrarily, Chile was a consolidated democracy, and the constituent convention was politically framed to eliminate the lingering authoritarian residues decades after Pinochet's 1980 Constitution.
Chile's constituent convention began under President Piñera's centre-right administration, which was elected before the constitutional-making process, and concluded with a left-wing one, President Gabriel Boric, who was elected when the constituent convention entered its final stage. Chileans chose the constituent convention model and elected a majority of party-less and new-left societal players, who emerged from the tumultuous social context and drove the constitutional-making process, alongside a few party-affiliate constituent members. Moreover, the constituents joined coalitions within the constituent convention, and most of these so-called “pacts” were previously designed under parties’ agreements and related to the national political and electoral game. Nevertheless, institutional participation bridges to social movements were built in the constituent convention, led and pushed by most nonpartisan representatives with advocacy or civil organisation rights connections. For instance, the constituent convention's Popular Participation Commission conducted 16,434 meetings and hearings, designing seventy-eight popular law initiatives that garnered 15,000 signatures and involved 980,332 people (Chile, 2022). National-level institutional redesigns created opportunities for quota adoption (Piscopo et al., 2024). Implementing gender quotas in the Chilean constituent convention was also seen as a historic opportunity to address gaps in minority rights (Lambert and Scribner, 2021). For instance, Chile's legal and policy approach to gender matters was considered restricted owing to constitutional provisions, often hindering efforts for women's rights advocacy.
Brazil took a different approach. Following an indirect presidential election, the National Congress approved a constitutional legislature model, in which representatives were chosen through a unified election. This allowed Brazilians to elect individuals who simultaneously served in both the National Congress and the constitutional assembly. The transition to democracy in Brazil was driven by party actors who advocated for this model, formed informal political coalitions within the constituent assembly, and established informal channels with social and organised civil society groups.
In the case of Brazil, the number of institutional veto players was comparatively low owing to the absence of formal and institutional representation, as well as channels for social movements, within the ANC. Additionally, neither ethnic nor gender quotas were enforced, as seen in Chile's case. Despite their political activity, Brazil's social and civil society actors played a consultative rather than a formal role in the assembly. For instance, of the 122 popular amendments submitted to the ANC, only eighty-three met the regimental requirements (Michiles et al., 1989). While public participation influenced the thematic area concerning fundamental rights, it had a lesser impact on major, controversial issues (Hudson, 2021). According to the theoretical framework of Lerner (2011), contentious issues were often deferred to future political processes, allowing for the gradual resolution or evolution of societal views. The ANC's constituents, for instance, chose a deferral strategy on the definition of Brazil's type of government, postponing it for five years and establishing a referendum in 1993 that ultimately favoured a republican and presidential system over monarchy and parliamentarism.
In Chile's constituent convention, independent members, indigenous groups, and leftist parties led the discussions on reforms that sought far-reaching changes to the country's political system. For instance, they proposed parliamentarism and semi-presidentialism models, eliminating the Senate, and introducing mechanisms for direct democracy. However, independent moderate, centre-left, and right-wing parties were either reluctant to support these reforms or opposed them outright, which led to significant internal divisions within the constituent convention. Martínez (2022, 2024) argued that proposals to replace the presidential system with a unicameral legislature and introduce immediate presidential re-election conflicted with the 2019 demands for decentralisation. These changes risked concentrating power and undermining governance and stability. These contentious issues also contributed to the rejection of the draft in the Chilean referendum.
Regarding the complexity of negotiations and potential gridlocks, the number of roll calls and duration of completion can provide significant insights into the constitution-making processes. Compared with the constituent convention, the ANC exhibited a lower voting frequency and a longer timeframe to complete the drafting process.
Finally, each assembly's distinct constraints emphasise the differences in operational and institutional contexts. Unlike the constituent convention, which had its draft subjected to a ratification referendum, the ANC's draft was not subject to this additional layer of scrutiny. This distinction highlights the ANC's relative autonomy from the draft's final public approval process, enabling a potentially thorough negotiation process within the assembly. Table 1 summarises the institutional similarities and differences between cases.
Institutional Framework.
Source: Authors’ elaboration based on Chile (2022), Medeiros (2013), Pilatti (2008), and Unholster (2022).
Considering the output of these processes, summarised in Table 2, highly social and politically mobilised contexts produced long and detailed constitutional drafts with similar unique word counts. The party-dominated ANC approved Brazil's draft internally; a national referendum rejected the Chilean one. A crucial political by-product of the Brazilian case is the emergence of a large centre-wing political group called “Centrão” (Freitas et al., 2009; Gomes, 2006). This existing coalition was born informally within the ANC.
Constitution-Making By-Products Framework.
Source: Authors’ elaboration.
Note: For the draft's word and unique word counts, we used Voyant tools. 4
Despite the widespread belief that the Brazilian party system during that period was weak (Elkins et al., 2009; Hudson, 2021) and parties could not act cohesively (Kinzo, 1990; Lamounier and de Souza, 1990; Mainwaring and Pérez-Liñàn, 1997; Martínez-Lara, 1996), both were undoubtedly pivotal in the constitution-making process (Coelho, 1999; Medeiros, 2013; Pilatti, 2008). Since the ANC's inauguration, party leaders have taken steps to organise and exert control over the process. For example, the internal regulations officially acknowledged the presence of party leadership and delegated specific powers to them. Conversely, the Chilean constituent convention's internal regulations made minimal references to parties or party leaders, thereby granting them no formal power.
After almost two years of intense work, Brazil's Workers’ Party (PT) 1 refrained from endorsing the ultimate text. Nevertheless, the ANC approved the draft constitution internally. Owing to the inscription of numerous civil, political, and social rights, the newborn Brazilian Constitution was immediately recognised as the “Citizenship Constitution” (“Constituição Cidadã”). Hunter (2014) acknowledged that the 1988 Constitution shaped Brazil's social policy landscape and opened the institutional paths to enhance the well-being of low-income Brazilians during the administrations of presidents Cardoso and Lula.
In Chile, political elites deferred a foundational institutional decision rather than resolving it through negotiation by placing two competing constitution-making models on the ballot during the 2020 entry referendum: a fully elected convention and a mixed convention. This postponement intensified elite polarisation and undermined the fragile consensus intended to support the interim constitutional framework (García-Huidobro, 2024). Moreover, party actors, constrained by the challenges of the COVID-19 pandemic context and unable to mobilise their grassroots networks effectively, faced significant vote losses (Belmar et al., 2023).
Moreover, the electoral design of Chile's constituent elections allowed independent candidates to bypass party-list formal requirements, further enabling their success (Nocetto et al., 2024). Consequently, independents secured the majority of constituent seats. Lowering barriers to entry allowed independents to participate but hindered their development into organised political parties, limiting their ability to address societal needs and contributing to the rejection of the constitution draft. Larrain et al. (2023) attributed this referendum result to the dominance of independents, the underrepresentation of the political right, and the decentralised drafting process. Piscopo and Siavelis (2023) stated that the right wing played an illiberal hardball game to influence the outcome. However, Alemán and Navia (2023) emphasised that the inexperience of independents and exclusion of right-wing voices created perceptions of chaos and disconnection. They contended that the rejection stemmed from the draft's radical elements and failure to align with the preferences of ordinary Chileans, rather than a victory for right-wing illiberalism.
The rejection of the first draft in Chile marked a significant political defeat for President Boric's administration, prompting a cabinet reshuffle that included centre-left figures from Bachelet's era. In response, Boric initiated a second constitutional process under an ad hoc mixed model. This model featured a committee of experts, reflecting legislative party representation, tasked with drafting a foundational text based on minimal consensus alongside a newly elected Constituent Assembly Council composed entirely of formal party candidates. However, this second round faced challenges as right-wing and mostly far-right representatives dominated the council after compulsory voting revealed a shift from anti-systemic voters toward these groups (Altman et al., 2024). Right-wing parties, pushed by the far-right, leveraged their majority to shape the draft while lingering tensions from the first convention further complicated the process. Ultimately, the second draft was also rejected in a plebiscite, underscoring deep divisions and challenges in Chile's constitution-making efforts.
This qualitative comparison assessed how the institutional frameworks of the Brazilian and Chilean constitution-making processes influenced the composition of their respective assemblies. The distinct constituent models of cases and each level of partisanship conditioned the negotiation dynamics within the assemblies. The following section discusses the implications of partisanship in the constitution-making processes. Next, we examine the voting patterns by groups, parties, and coalitions to compare how institutional structures and levels of partisan engagement affected variations in negotiation outcomes, as indicated by constituents’ voting positions and winning and cohesion patterns. This transition from qualitative comparison to theoretical discussion and then to quantitative analysis bridges the study's contextual and empirical dimensions, elucidating the interplay among political dynamics, institutional design, and partisanship in each constitution-making process.
Partisan Influence in Constitution-Making
Political parties are fundamental to modern politics, as they help structure competition, aggregate societal interests, and organise collective action (Aldrich, 1995; Cox, 1987; Cox and McCubbins, 1993; Downs, 1957; Sartori, 1976). However, their role in constitution-making is complex and can be seen in two ways. On the one hand, parties can be essential for negotiation, transforming individual preferences into coherent institutional proposals (Negretto, 2013). On the other hand, they can also serve as vehicles for institutional capture, prioritising their own power over democratic deliberation (Elkins et al., 2009). Understanding this duality is crucial for explaining why different constituent models yield vastly different outcomes.
The argument for partisan involvement is based on its ability to address fundamental problems associated with coordination and collective action. Assemblies that operate through party systems tend to demonstrate more structured interest representation, centralising negotiations, and facilitating coalition-building. By providing leadership, discipline, and institutional resources, parties can reduce negotiation costs, lower the risk of gridlock, and guide the process toward stable agreements. For example, in Brazil's constituent assembly, the creation of a council of party leaders – known as the colégio de líderes – was instrumental in coordinating legislative behaviour and fostering consensus, effectively replicating the party dynamics game of a regular legislature (Pilatti, 2008). Moreover, demands from civil society are more likely to be included when partisan elites first reach an agreement, underscoring the role of parties as essential intermediaries (Maboudi and Nadi, 2016).
Conversely, concerns about partisan self-interest often prompt calls for independent, nonpartisan constituent bodies. Scholars have argued that separating the constitution-making process from the pressures of everyday governance enhances legitimacy and reduces the risk of politicians entrenching their power (Ackerman, 1994; Elster, 1995). The primary concern is that dominant parties may manipulate the process to undermine constitutionalism itself, thereby embedding their strategic advantages in the national founding document (Bhandari, 2012; Landau, 2013; Pushparajah and Balamayuran, 2024). Such elite-driven agreements can also marginalise dissenting voices, even if they produce stable outcomes (Elkins et al., 2009; Hudson, 2021). This concern significantly influenced the first constituent process in Chile (2021–2022), which adopted a convention model open to independents, allowing it to distance itself from a discredited party system and enhance social inclusion.
However, the risks of capture are not inherent to partisanship, nor does independence guarantee inclusiveness. The key moderating factor is institutional pluralism: when multiple parties compete, their mutual oversight can provide the necessary checks and balances to prevent any single actor from exerting undue influence over the process (Negretto, 2018). Moreover, as former congressmen and then the country's president, Gabriel Boric, cautioned during the Chilean debate, independent actors are not necessarily more virtuous and may lack the collective accountability and predictability that party structures provide. 2 Without party-based coordination, decision-making can become fragmented, making it even more challenging to translate diverse demands into a cohesive constitutional text.
We also highlight another critical variable: alongside the presence of parties, the institutional design of the constituent body may also allow channels for partisan competition, coalition-building, and negotiation pipelines. The nonpartisan and party-based framework sheds light on the trade-offs faced by Chile and Brazil. Chile's constituent convention prioritised a symbolic break from the past, but the partisan minority and weak institutional infrastructure for negotiation installed within the CC hindered coordination and amplified demands from a fragmented array of interests, complicating stable decision-making outcomes. In contrast, Brazil's party-based legislature model, although borne out of pragmatic constraints, benefited from partisan coordination mechanisms that fostered compromise even within a nascent party system, which was ideologically diffuse and diverse. These cases illustrate the central tension in constituent design and membership: the choice between the perceived legitimacy of nonpartisan openness and the coordinating ability of party-based institutional frameworks has crucial consequences. While nonpartisan models may offer greater inclusion, they risk fragmentation and instability. Partisan assemblies, although susceptible to capture and elite bias in addressing socially legitimate demands, provide the institutional tools necessary to forge durable collective decisions.
Quantitative Comparison
We analysed recorded votes to explore the patterns of constituent behaviour and decision-making within Brazil's ANC and Chile's CC. Roll calls provide a crucial perspective for studying the dynamics of political alignment and the role of individual actors and parties in shaping the outcomes of assemblies. Hence, we could test our expectations regarding the effects of partisanship on bargaining output.
Data and Methods
Although roll calls represent only the final stage of a complex web of negotiations (Ainsley et al., 2020; Carrubba et al., 2006), the aggregation of legislators’ votes ultimately results in a political decision. Furthermore, in the specific context of constituent assemblies, nearly all matters are typically subject to a vote. We collected comprehensive data on all constitutional matters from both assemblies, including roll calls conducted in committees and during floor sessions. Medeiros (2013) and Unholster (2022) served as the primary sources for the roll call data from Brazil's ANC and Chile's constituent convention, respectively.
To assess the ideological positions of the constituents, we utilised W-Nominate, a method for estimating ideal points that provides insights into the distribution of preferences (Poole, 2005; Poole and Rosenthal, 1985). To complement it and gain insight into the mechanics of party unity and coalition success, we introduced an analysis of roll rate, a widely used metric in legislative studies (Cox and McCubbins, 2005; Krehbiel, 2007). This metric captures the frequency with which the plenary majority defeated the majority position of a party. We also computed the Rice Index of cohesion, which measures intra-party agreement by assessing the extent to which members vote with their party's majority in each roll call (Carey, 2007; Rice, 1925). Together, these metrics enriched the W-Nominate estimates by providing a multi-dimensional view of party behaviour, combining ideological positioning, internal unity, and legislative influence.
First, we compared the ideal point estimates and roll rates for Brazil and Chile, analysing voting behaviour in both committee and plenary sessions, as well as the roll rate and intra-party cohesion. Building on those metrics, we used matching techniques to assess the Chilean Constitutional Convention as an institutional laboratory. This enabled us to evaluate the potential effects of a nonpartisan political treatment on CC's voting behaviour. This method allowed us to identify and measure the significant impact of the absence of party affiliations on decision-making dynamics within a constituent assembly.
Matching is a technique used in observational studies to create comparable groups and reduce the influence of confounding variables. Unlike randomised controlled trials, observational studies are more prone to confounding factors because they do not involve random assignment. The primary goal of matching is to compare the outcomes of individuals who received treatment with the outcomes they would have experienced in the absence of treatment, known as the counterfactual. This process involves pairing and clustering each treated individual(s) with a similar individual(s) from the control group based on observed characteristics. The control group member's outcome serves as the counterfactual for the treated individual. By estimating what the outcomes would have been if the treated individuals had not received treatment, researchers can compare actual and counterfactual outcomes to evaluate the causal effect of the treatment. This method provides quasi-experimental evidence within observational datasets, enabling researchers to estimate treatment effects (King and Nielsen, 2019).
Results and Discussion
This section examines voting pattern metrics of Brazil's ANC and Chile's constituent convention members to explore how institutional contexts and political dynamics shape decision-making processes. It emphasises the roles of party structures, coalitional behaviour, and independent actors in shaping constitutional outcomes. By comparing W-nominate scores, roll rates and Rice Index estimates, we explore how Brazil's party-centric assembly contrasts with Chile's nonpartisan landscape. The findings underscore the interplay between party alignment and independent participation, driving cohesion, variability, and political compromise in constitution-making output.
Voting Behaviour in Two Constitutional Assemblies
In the initial graphs of both countries’ committees, we present the means and confidence intervals of ideal points (W-nominate scores). Notably, certain committees had fewer votes than others, potentially leading to broader intervals resulting from lower voting frequency or variations in behaviour among party or group-level constituents. The solid line at zero is a visual aid to delineate left-, centre-, and right-wing political orientations.
Figure 1 shows the parties’ mean positions across various committees in the Brazilian ANC, indicating a general tendency towards the political centre. PMDB, the party heir to the former MDB, which served as the official opposition umbrella to the military dictatorship government, had confidence intervals intersecting at zero in all committees, portraying a considerable dispersion across the continuum while averaging close to a moderate position. Even the right-wing parties that formerly supported the military regime (PFL and PDS) were relatively close to the centre. Similarly, the centre-left PDT orbited around the political centre, whereas the left, represented by PT, had only once crossed the centre of the spectrum.
Parties’ Ideal Points in the Committees of Brazil’s National Constituent Assembly.
Figure 2 exhibits the Chilean constituent convention's committees with a clear division between left and right, contrasting with the broader variability observed in the ANC. Although the Chilean committee has a higher roll call frequency (Table 2), the smaller confidence intervals suggest higher party cohesion. Figure 2 also depicts that the party-less independent actors (“Independiente”) had higher variability in the constituent convention's committees. The Chilean Natives (“Pueblos”) cluster acted more cohesively on the left. Other leftist and centre-left parties positioned themselves close to the political centre, such as DC, CS, PPD, PS, and RD; meanwhile, PC and FRVS remained more on the left. Within the right-wing spectrum, the UDI, RN, and EVOPOLI parties demonstrated disciplined behaviour, with the UDI positioned further to the right and EVOPOLI situated between the right and centre-right.
Parties’ Ideal Points in the Committees of Chile’s Constituent Convention.
The ideal point estimates of the committees correspond closely with the voting patterns observed on the floor. The subsequent graphs in Figures 3 to 7 retain the solid line at zero and introduce a dashed line representing the overall mean position in the ANC and CC. In Figure 3, the average position of framers’ votes during Brazil's plenary sessions tends towards the centre (0.078). PMDB members of the ANC exhibit a mean moderate stance overall, positioned even closer to zero (−0.03), although their ideal points span the entire ideological spectrum. Hence, the largest party in the ANC exhibits significant heterogeneity, reflecting its evolution following the democratic transition. The right-wing Brazilian parties (PDS, PFL, and PTB) offset the influence of the left-wing ones (PDT and PT), with PMDB playing a pivotal role in maintaining a centrist position on floor voting. In other words, the heterogeneity and size of the PMDB meant that no single party could secure a majority without negotiating with others.
Members’ Ideal Points by Party in the Floor of Brazil’s National Constituent Assembly.
Members’ Ideal Points by Party and Group in the Floor of Chile’s Constituent Convention.
Members’ Ideal Points by Coalition in the Floor of Brazil’s National Constituent Assembly.
Members’ Ideal Points by Coalition in the Floor of Chile’s Constituent Convention.
Independent Members’ Ideal Points by Coalition in the Floor of Chile’s Constituent Convention.
Figure 4 shows that the mean position of Chilean constituents during plenary sessions leans clearly to the left (−0.36), whereas independent participants’ estimates are even further to the left (−0.44). However, certain PPD and DC members, and a few independents, occupy a centre-left position on the floor.
We also analysed coalitional behaviour in both constituent assembly plenary sessions. In Brazil, a group called “Centrão” emerged, opposing the leadership exerted by PMDB's prominent politicians, who were aligned with the centre-to-left wing. Composed mostly of centre-right-wing constituents, “Centrão” proposed changes in the ANC's internal rules to reduce the power of the rapporteur in charge of producing the baseline draft of the Brazilian constitution (Freitas et al., 2009; Gomes, 2006; Medeiros, 2013; Pilatti, 2008). As shown in Figure 5, this coalition exhibited a centre-right-wing position (0.39). Owing to its comprehensive dispersion across the political spectrum, “Centrão” cannot act more cohesively. Additionally, because it is not a formal party but rather an informal coalition, it lacks the capacity and resources to control the behaviour of its members. Consequently, its leadership has no mechanism to compel members to adhere to the group line. This dynamic became evident during the first plenary vote when Centrão's proposed preamble for the new constitution failed to gain a majority, prompting cross-party collaboration to draft a consensus alternative. The revised preamble was overwhelmingly approved, setting a precedent for the negotiated and collaborative decision-making process that would define the ANC (Gomes, 2006; Medeiros, 2013).
As illustrated in Figure 6, Chilean political parties and independent actors formed coalitional fronts. The “Vamos por Chile” pact represented centre-right and right-wing parties (0.64), whereas the “Apruebo Dignidad” bloc represented some far-left and left-wing ones (−0.67). Remarkably, the constituent convention began under the former administration connected to Piñera's right-wing party pact (“Vamos por Chile”) and ended with a leftist one aligned to Boric's presidency (“Apruebo Dignidad”). The W-nominate scores suggest that both pacts opposed each other within the constituent convention.
Figure 7 displays the floor voting behaviour of only Chilean independent constituents, categorised by pact and their party identification. We applied Rozas-Bugueño et al.'s (2022) constituent political classification, which considers that although independents might not have formal party affiliation, some might have evidence of party identification. Thus, the constituent independent players that had previously competed electorally (i.e. within political party pacts), participated in or joined some organisation formally linked to parties, or were appointed to any political position in a government, were classified as “partisan independents.” Meanwhile, the “nonpartisan independents” did not have any party cues, links, or government experience registered. The last group of party-affiliated members was classified as “party militants” (Rozas-Bugueño et al., 2022).
The division between independents with and without party identification becomes clear, following the political pact each one has joined. Independents with party identification, referred to as partisan independents, exhibit dispersed voting patterns, aligning themselves with either the current administration pact (Boric's leftist “Apruebo Dignidad”), the previous one (Piñera's right-wing “Vamos por Chile”), or the centre-left parties pact (“Apruebo”). Meanwhile, the nonpartisan independents have displayed more cohesive voting behaviour, especially in the radical left pact (“Del Pueblo”) and others (“Otros”). The only exception is the “Nueva Constitución” pact, which shows a more diffuse behaviour.
The spatial distribution of ideal points reveals patterns of ideological variation, but it does not account for the internal cohesion of parties or the frequency with which their positions prevail in plenary votes. To complement this spatial perspective, we incorporated two additional metrics: roll rates, which capture the frequency with which the overall plenary defeats a party's majority, and the Rice Index, which measures the degree of internal cohesion within each party, regardless of vote outcome.
Brazil's ANC illustrates how compromise emerged through fragmented yet strategically pivotal political forces (Table 3). The PMDB, despite being ideologically diverse and internally divided (Rice Index of approximately 65 per cent), maintained a remarkably low roll rate (approximately 2 per cent), suggesting that a core faction of the party consistently aligned with the winning side. In contrast, the informal cross-party faction Centrão, a large centre-right coalition that emerged in opposition to the PMDB's more progressive leadership, displayed moderate cohesion (Rice Index of approximately 70 per cent) but was still able to shape outcomes (roll rate of approximately 5 per cent). These results suggest that the ANC's effectiveness did not hinge on internal unity but rather on its party-based institutional design and leadership mechanisms. These structures facilitated agenda control and enabled negotiations across parties and factions, helping bridge ideological divisions and coordinate internal cleavages. Institutions such as the Colégio de Líderes played a key role in brokering consensus, sustaining deliberation momentum, and ensuring that even loosely organised coalitions like the Centrão could act as pivotal players.
Roll Rates and Rice Index in Brazil’s National Constituent Assembly.
Source: Authors’ elaboration based on Medeiros (2013).
Note: Only parties that voted on more than 500 roll calls are displayed. Small parties are not shown.
Chile's CC, by contrast, was marked by rigid bloc dynamics and systematic exclusion (Table 4). Right-wing parties UDI and RN were highly cohesive (Rice Indexes of 94.1 per cent and 90.2 per cent, respectively), yet they were frequently defeated (Roll rates of approximately 43 per cent and 37 per cent). At the coalition level, the right-wing “Vamos por Chile” displayed a similar pattern: strong internal cohesion (Rice Index of about 85 per cent) combined with a high incidence of roll rates (approximately 43 per cent), indicating consistent misalignment with the plenary majority. Progressive coalitions with numerous independents, including Del Pueblo, Apruebo Dignidad and Pueblos Originarios, were also cohesive and held dominant positions, with roll rates below 4 per cent and Rice Index over 80 per cent. Thus, the constituent convention was characterised by a consistent progressive majority that systematically outvoted a cohesive but structurally isolated conservative bloc, in the absence of mechanisms capable of mediating between pacts and accommodating dissent.
Roll Rates and Rice Index in Chile’s Constituent Convention.
Source: Authors’ elaboration based on Unholster (2022).
The behaviour of independent members further illustrates this pattern. Their near-zero roll rate reflects a technical artefact: lacking formal party structures, they could not be “rolled” in the conventional sense. However, their moderate cohesion (Rice Index of approximately 60 per cent) underscores significant internal fragmentation. Elected amid an anti-party wave, independents lacked partisan anchorage and failed to consolidate into organised political fronts (Nocetto et al., 2024). Their limited organisational discipline and uncoordinated behaviour, even when integrated in broader coalitions, contributed to widespread perceptions of disarray and inefficiency within the constituent convention.
Although roll rates and Rice indices do not capture the substantive content of decisions, they provide an analytical lens for assessing patterns of cohesiveness, inclusion, and exclusion. These measures help identify which actors exerted consistent influence through internal unity and which were systematically marginalised, complementing the spatial estimates of voting and ideological preferences.
In summary, Brazil's ANC, anchored by the ideologically diverse PMDB and challenged by the emergence of the informal Centrão coalition, demonstrated organisational capacity and relied on institutional mechanisms to facilitate compromise. Its effectiveness stemmed not from eliminating fragmentation but from empowering party leadership and structuring negotiation across factions through institutionalised coordination.
Chile's constituent convention, by contrast, was shaped by the confrontation between cohesive but ideologically polarised and asymmetric blocs. A unified progressive majority consistently outvoted a united conservative minority. Furthermore, the large contingent of independents exacerbated the convention's inability to bridge ideological divides or forge stable cross-coalition agreements. Its design lacked not only strong partisan anchors but also institutional channels capable of structuring deliberation and negotiation.
The effectiveness of a constitution-making assembly hinges not only on ideological cohesion but also on its party-based and leadership-structured institutionalisation, as well as its ability to achieve organised and enduring outcomes. These findings underscore the influence of party-based assemblies on deliberative results. To further assess this dynamic, particularly the behavioural impact of independent, nonpartisan actors within Chile's constituent convention, we applied matching techniques to isolate the effect of non-partisanship.
Chile as an Institutional Laboratory: Nonpartisan Treatment on the CC
Building on the previous comparison, we now turn to Chile's constituent convention to empirically assess how the absence of party structures influences constituent behaviour. Despite the ideological diversity of Brazil's recently formed parties and internal division within the PMDB that led to the Centrão coalition, ANC exhibited voting patterns influenced by mechanisms of party coordination. If such a framework plays a key role in structuring constituent behaviour, then its absence – or the presence of “partylessness” – should leave a detectable imprint. Chile's constituent convention offers a unique opportunity to test this proposition: it featured a large contingent of independent, nonpartisan actors operating within a deliberative setting largely devoid of formal partisan coordination structures. In this sense, the CC functions as an institutional laboratory for evaluating the effects of non-partisanship on constitution-making behaviour and dynamics.
To conduct this analysis, we employed matching techniques based on the political classification proposed by Rozas-Bugueño et al. (2022), who categorised constituent convention members into three groups: nonpartisan independents, partisan independents, and party-affiliated individuals. The treatment group consisted of nonpartisan independents. Control groups included both partisan independents and party-affiliated individuals, allowing for two separate matching exercises. We matched all constituent convention participants based on observable covariates, including age, sex, ethnicity (a dummy variable for Chilean natives), the last coalition joined (with the “Chile Vamos” right-wing pact as baseline), whether they switched coalitions (a dummy variable), and district magnitude data (defined as the number of constituent convention seats per Chilean region).
To estimate the effect of partylessness, we compared the treated group to counterfactual scenarios representing varying degrees of party experience. The matching exercises enabled us to compare individuals exposed to the treatment and those with some partisan experience or party affiliations, isolating the effect of the non-partisanship condition from potential confounding influences. Therefore, the matching approach allowed us to isolate the behavioural influence of partisan embeddedness by comparing nonpartisan independents to individuals with varying partisan experiences counterfactually.
We employed kernel matching (KM) and nearest neighbour matching (NNM) techniques to estimate the average treatment effect on the treated (ATT) while minimising biases and model dependencies associated with other matching tools (King and Nielsen, 2019). These methods optimise covariate balance by implementing cross-validation (CV) in the KM method and tailoring neighbours using the NNM technique. This dual approach enhanced the robustness and reliability of our treatment effect estimates (Appendix B).
This methodological strategy helped us determine the statistical association between the distinct voting patterns observed among nonpartisan independents and their lack of partisan experience. By comparing these actors to matched counterparts with some form of party background or affiliation, we tested the hypothesis that partisan embeddedness acts as a behavioural anchor, aligning individual actions with collective strategies and fostering voting cohesion. In doing so, we evaluated whether the absence of formal party structures in Chile's constituent convention contributed to fragmented or inconsistent voting behaviour, offering a counterpoint to the party-driven dynamics observed in Brazil's ANC.
Figure 8 presents the estimates for ATT. Matching partisan independents with nonpartisan independents revealed no statistically significant differences across most specifications. Thus, the nonpartisan treatment does not consistently influence the behaviour of individuals lacking formal party affiliations. In other words, our estimation of how nonpartisan independents would have behaved if they had been partisan independents revealed no meaningful difference in voting patterns. This suggests that previous, informal party ties, or weak partisan links are insufficient to shape constituent behaviour systematically.
Effect of the Non-Partisanship Treatment in Chile’s Constituent Convention (Average Treatment Effect on the Treated).
However, the second matching exercise (i.e. estimating how nonpartisan independents would have behaved if they had been party-affiliated) revealed significant and robust treatment effects across all KM and NNM estimations. This suggests that formal party membership, supported by party institutional toolkit and organisational discipline, is critical in structuring voting behaviour. The findings indicate that the non-partisanship treatment effect significantly influences how constituents vote, particularly when matched with those formally affiliated with party organisations.
In this context, party membership emerges as a key factor influencing behavioural cohesion in constitution-making voting. These results underscore the crucial role of party membership and the impact of non-partisanship on voting behaviour in constitution-making.
Conclusion
Political parties can play a stabilising role in constitution-making by channelling citizen demands through institutionalised participation. They help mitigate legitimacy crises and reduce the risk of populist backlash (Rhodes-Purdy and Rosenblatt, 2023). Within constitutional processes, political parties serve as intermediaries between public preferences and formal institutions, facilitating the structuring of participation and collective decision-making (Hudson, 2021). When parties are absent, representation becomes fragmented and negotiations lose coherence, as illustrated by Chile's constituent convention (Nocetto et al., 2024). Thus, effective constitution-making relies not only on inclusive participation but also on institutional mechanisms that enable enforceable agreements and structured deliberation (Chilton et al., 2024; Maboudi and Nadi, 2016).
Evidence from Brazil and Chile supports this argument, demonstrating how differing party-based institutional designs impact the capacity for coordination. Brazil's ANC, despite operating within a fragile and newly reconstituted party system and an ideologically diffuse environment, effectively utilised its partisan infrastructure. It included formal leadership roles and agenda-setting mechanisms, such as the Colégio de Líderes, to sustain negotiations and produce a broadly accepted constitutional outcome. In contrast, while Chile's constituent convention exhibited voting cohesion, it lacked the institutional framework necessary to translate preferences into lasting agreements. The constituent convention results point to a deeper structural limitation: the consequences of mainly being nonpartisan. The minority of party organisations, combined with the absence of embedded mechanisms for leadership, discipline, and cross-bloc bargaining, created a fragmented negotiation environment, ultimately restricting the constituent convention's ability to reach a stable consensus.
These findings challenge the common preference for nonpartisan models in constitutional design. Advocates of exclusive conventions often emphasise independence and inclusivity – a perspective associated with Elster. However, our analysis highlights the institutional vulnerabilities of such models. By sidelining political parties, these frameworks risk abandoning the very organisational abilities needed to structure deliberation, resolve conflicts, bridge ideological divides, and maintain negotiation momentum. In short, independence alone is insufficient; institutionalised coordination is essential, and the party organisational toolkit may offer it.
In conclusion, this study demonstrates that partisan institutional frameworks, rooted in legislative experience, may act as critical gatekeepers for effective decision-making and consensus-building. By comparing two contrasting cases, we contribute to ongoing debates about the institutional foundations of constitution-making. Future research should further investigate how various configurations of party engagement and constituent infrastructure influence the outcomes of constituent processes across diverse democratic contexts, particularly in Latin America.
Footnotes
Acknowledgements
The authors would like to thank Adrian Albala, Bruna Fonseca, Juan Carlos Arellano, Alejandro Olivares, Christopher Martínez, José Antonio Cheibub, Jose Miguel Cabezas, and Nicolas Taccone for their valuable comments on earlier drafts of this paper. The paper was presented and received feedback at the International Meeting on Participation, Democracy, and Public Policies (PDPP), the Midwest Political Science Association (MPSA) Conference, and the Catholic University of Temuco (UCT).
Data Availability Statement
Data supporting the findings of this study will be accessible through the journal's Figshare repository platform.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Rodolfo Camargo de Lima receive funding from Millennium Nucleus on Political Crises in Latin America (CRISPOL): NCS2024_065, supported by Chile’s National Agency for Research and Development (ANID). Danilo Medeiros was funded by grant #2019/24091-1, São Paulo Research Foundation (FAPESP), and by the Institutional Research Support Program for Newly Hired Faculty (UFMG, Call No. 10/2024). This work was also supported by Vicerrectoría de Investigación y Doctorados de la Universidad San Sebastián – Fondo USS-FIN-25-FATE-05.
Notes
Author Biographies
Appendix A
List of Parties in Chile’s Constituent Convention.
| Party | Name in Spanish | Name in English |
|---|---|---|
| CS | Convergencia Social | Social Convergence |
| DC | Partido Demócrata Cristiano | Christian Democratic Party |
| EVOPOLI | Evolución Política | Political Evolution |
| FRVS | Federación Regionalista Verde Social | Regionalist Federation of Social Greens |
| PC | Partido Comunista de Chile | Communist Party of Chile |
| PPD | Partido por la Democracia | Party for Democracy |
| PS | Partido Socialista de Chile | Socialist Party of Chile |
| RD | Revolución Democrática | Democratic Revolution |
| RN | Renovación Nacional | National Renewal |
| UDI | Unión Demócrata Independiente | Independent Democratic Union |
Source: Authors’ elaboration.
Appendix B
We implement matching techniques to balance covariates, thus reducing model dependence and potential biases. Scholars have widely used propensity score matching (PSM), but its sensitivity to model specification and assumptions often introduces inefficiencies and bias (King and Nielsen, 2019). These limitations have prompted researchers to explore alternatives like Kernel Matching (KM) and Nearest Neighbour Matching (NNM) (King and Nielsen, 2019; Jann, 2017).
KM leverages kernel functions to match treated and control units based on covariate similarity, employing a weighted scheme optimised by adjusting kernel function parameters and bandwidth to improve covariate balance. It incorporates Cross-Validation (CV) with respect to covariates (X) to ensure balance, and applies it to outcomes (Y) to assess the accuracy of treatment effect estimation (Galdo et al., 2008). NNM, on the other hand, matches units based on proximity in covariate space using metric distances. Its tailoring capacity allows for balancing and trade-offs between match precision (e.g., one neighbour) and broader match pools (e.g., three or five neighbours), enhancing validity and flexibility.
Both KM and NNM mitigate the risks of model misspecification and provide more reliable treatment effect estimates than PSM (Jann, 2017; King and Nielsen, 2019). This study employs these methods to ensure robustness, presenting results from NNM with varying neighbours and KM with pair-matching and CV adjustments for X and Y.
By calculating the Average Treatment Effect on the Treated (ATT), we estimate the average outcome change for treated individuals, accounting for characteristics influencing treatment assignment and outcomes. Coupling ATT with matching allows us to draw more reliable conclusions about the effectiveness of the non-partisanship treatment.
The validity of the matching analysis relies on achieving an adequate balance of covariates between the treated and control groups. Diagnostic data from the kmatch package in Stata indicate a high degree of reliability for the primary models. In both comparisons (nonpartisan vs. partisan independents and nonpartisan vs. party-affiliated), the NNM models with one, three, and five neighbours demonstrated excellent covariate balance. Similarly, the KM pair-matching specification yielded a strong balance across all covariates. By contrast, the CV diagnostics for X and Y showed some imbalance in a few variables, such as district magnitude.
The diagnostics also provide information about the data's structure. The absence of specific pact categories in the output is not a modelling error but rather a reflection of the data itself: no individuals in the “nonpartisan independents” group were affiliated with the “Vamos por Chile” or “Apruebo” coalitions. This underscores the political distance between nonpartisan candidates and some formal party-line blocs. Additionally, Stata excluded the ethnicity variable (pueblo_originario) from several models due to collinearity. This is a standard procedure triggered when variables are perfectly collinear, as with the ethnicity dummy and the ‘Pueblos Originarios’ pact category, and does not introduce bias.
In summary, the covariate balance achieved by the various NNM and KM models reinforces the reliability of the results. While estimates from the CV specifications should be interpreted with caution due to residual imbalance, the overall conclusion remains robust: formal party affiliation, serving as a counterfactual, exhibits consistent and structured voting behaviour, whereas informal partisan ties do not. Consequently, the nonpartisanship treatment, when evaluated against the party-affiliated baseline, confirms the organisational role of parties in structuring preferences.