Intra-party Dynamics and Sub-National Influences in the Emergence of Party Splits in Brazil

Introduction

Party splits are a widespread phenomenon. In systematic studies, Ibenskas and Sikk (2017) have identified 154 cases of party splits in Central and Eastern Europe from 1990 to 2015, while Zons (2015) founds 113 cases in the OECD democracies between 1960 and 2002. Understood as the process by which new political parties are formed by politicians previously affiliated with another party (Ibenskas and Sikk, 2017), the extent of this phenomenon in various national contexts indicates that, in a period characterised by intense regulation of political parties (Van Biezen and Rashkova, 2014), the entry of new players is not an easy task. Although the ability to face these challenges varies by institutional framework, societal cleavages, and specific circumstances, competing for resources and voter support against more established parties poses a significant hurdle. In this sense, parties that are formed by political figures who are already known within the system and have prior affiliations may have a strategic advantage.

The Brazilian case is not different. As Rocha and Pimentel (2024) show, it is not uncommon for new parties in Brazil to have more than half of their founders previously linked to other parties. The authors also reveal that these parties are not only more likely to be launched but also to achieve electoral success. As will be discussed, this phenomenon has significantly shaped the Brazilian party system, especially during its democratic transition.

What does explain the emergence of party splits? This work focuses on identifying the conditions that favour their occurrence in Brazil. In this study, a party split is defined as a collective defection from an existing party, characterised by the migration of a group of affiliated politicians to a newly created party – that is, a party participating in general elections for the first time. This definition delineates the phenomenon under investigation. First, the analysis is not concerned with party switching in general, but with movements that are associated with the appearance of new party organisations. Second, the focus is on collective defection, rather than isolated moves by single actors. Third, unlike studies centred on the emergence of new parties, this research is interested in the conditions within the original party that favour the exits. Thus, the emphasis is on party splits as processes of defection, rather than on the broader phenomenon of party formation.

More specifically, this study examines how intra-party factors relate to the occurrence of party splits. Drawing from the literature on new parties, previous research has shown how institutional, societal, and political conditions foster the emergence of splinter parties, that is, new parties formed by politicians previously affiliated with existing ones (Chiru et al., 2021; Zons, 2015). These studies highlight the role of electoral systems, along with economic and political instability, in creating opportunities for breakaway politicians. However, existing analyses often overlook the internal dynamics of the parties from which these politicians depart, factors that may also help explain their decision to leave and support the formation of a new organisation.

This study explores one of these internal dimensions: intra-party competition. It argues that internal competition can influence the phenomenon of party splits in two ways. First, intense competition within a party may fuel members’ discontent. When multiple candidates from the same party compete for the same seats, co-partisan rivalry can intensify, making it harder for politicians to advance their careers within the party. Second, a party that enters an election with a poorly coordinated candidate list – one overcrowded with viable contenders for the same seats – may signal internal dysfunction. In this sense, heightened intra-party competition may reflect a broader coordination problem, revealing difficulties in accommodating different political groups or elites within the same party.

Drawing on a dataset comprising party-level observations across Brazilian states from 1982 to 2022, this study employs statistical analyses to explore the occurrence of party splits. The findings align with previous research, indicating that party splits are influenced by electoral and party system conditions, such as the level of fragmentation, electoral permissiveness, and inter-party competition. Additionally, the characteristics of the original party from which splinters emerge are significant: factors such as government status, and ideological positioning can influence the likelihood of party splits. Most importantly, the study highlights the role of intra-party competition at the subnational level in influencing party splits. While the overall number of candidates within a party has some effect, it is particularly competition among viable candidates that shows a positive and significant association with the likelihood of splits.

Leaving an Old Party, Creating a New One

The phenomenon of party splits is often triggered by internal dynamics that create incentives for exit. Theories on party defection highlight two main factors explaining this decision: dissatisfaction within the party and the pursuit of better prospects elsewhere. Both arguments are based on the premise that political actors value and aim to achieve three goals: policy influence, winning votes and holding office positions (Strøm and Müller, 1999). When the benefits related to these goals decline in their original parties or when opportunities arise elsewhere to enhance gains in terms of policy, vote, and office, members are inclined to defect (Aldrich and Bianco, 1992; Ceron, 2019; Heller and Mershon, 2009; Kato, 1998).

Empirical evidence supports these theories. Regarding policy, defections can be triggered by disagreements with shifts in the party's ideological stance or programme. As Ceron (2019) demonstrates, the likelihood of party factions breaking away increases as the gap widens between their programmatic preferences and the party's official stance. Defections may also result from perceptions that the party leadership is deviating from its foundational principles (Demirkol, 2015). Additionally, studies indicate that legislators often join parties more aligned with their public policy preferences (Desposato, 2006; Thames, 2007), especially in collective defections (Kemahliolu and Sayari, 2017) and advanced democracies (Klein, 2021).

Regarding electoral support, research indicates that parties experiencing a decline in popularity or electoral performance are more prone to seeing their members defect (Klein, 2018; McMenamin and Gwiazda, 2011; O’ Brien and Shomer, 2013). Politicians also leave due to the party's diminishing reputation among the public following political scandals (Cho and Kruszewska, 2018; Daniele et al., 2020) and poor economic performance in government (Zielinski et al., 2005). Moreover, politicians switch to parties that bolster their chances of re-election and have ascending electoral results and popularity (Desposato, 2006; Desposato and Scheiner, 2008; McMenamin and Gwiazda, 2011; Thames, 2007).

Lastly, defection can arise from discontent with the distribution of office payoffs. A lack of access to electoral candidacies can be a problem (Kerevel, 2014), but also access to positions within the government or the party hierarchy (Ceron, 2019; Cox and Rosenbluth, 1995). Some studies indicate that parties in government are less likely to lose members while attracting more switchers, suggesting that politicians seek higher office payoffs and access to pork-barrel resources (Desposato, 2006; Yinka Fashagba, 2014).

Therefore, dissatisfaction with the party and the quest for greater advantages are two primary motivations for the decision to leave a party. However, these motivations alone do not fully account for party defection. The existing literature presents two key points. First, defection is not without its obstacles. It involves costs, not just related to the lost resources from the original party, but also to the politician's transition to a new position. Second, both the costs and benefits of defection vary depending on the characteristics of the political system, the parties involved, and the individuals themselves.

At the political system level, scholars show that the electoral system (Klein, 2018; O’ Brien and Shomer, 2013; Pinto, 2015; Radean, 2022), the presence of anti-defection laws (Mershon and Shvetsova, 2013; Nikolenyi, 2016), the form of government (Mershon and Shvetsova, 2013), and the stage of the democracy (Klein, 2021) all influence the costs and benefits of defection. At the party system and the party levels, defection varies as a function of the ideological positioning of the party (Volpi, 2019), its internal organisation (Boucek, 2012; Ceron, 2019) and the degree of fragmentation in the party system (Mershon and Shvetsova, 2013). Finally, at the individual level, factors such as legislators’ personal electoral strength (Asano and Patterson, 2022; Nielsen et al., 2019), their ideological compatibility with the original party (Desposato, 2006), and their mode of campaign financing (Arriola et al., 2022; Flório Lima and Bodet, 2023) can affect the decision's utility. The variation in costs and benefits across these different levels explains why party defection is more common in certain countries, as well as why its frequency can change over time and differ between units within the same national context.

In the case of party splits, the costs and benefits not only involve leaving a party, but also those associated with founding a new one. According to the rational choice approach, the creation of new parties also depends on political elites’ cost–benefit analyses (Cox, 1997; Hug, 2001; Tavits, 2006, 2008). The formation of new parties incurs substantial costs, primarily dictated by two factors: the difficulty of registering a new party and the challenges of securing electoral seats, as summarised by Tavits (2008). Studies show that more permissive electoral systems (Bolin, 2007; Bolleyer and Bytzek, 2013; Harmel and Robertson, 1985; Moser, 1999) and party registration rules (Bischoff, 2011; Su, 2015; Zons, 2015) increase the chances of new parties emerging and achieving electoral success.

Furthermore, the attractiveness of establishing a new party is influenced by the ease of accessing political power and policy influence. Tavits (2008) suggests that new parties are more likely to form in presidential systems, for example, where the potential gains in terms of office and policy impact are usually more substantial ¹ . In addition, the creation of a new party can be electorally rewarding in itself. As Sikk (2012) demonstrates, novelty can attract voters, even if the party does not represent a new cleavage or issue within the electorate.

However, these benefits are irrelevant if the emergent parties have negligible prospects of harnessing them. That's why a third element is central: the probability of success. As Tavits (2008) points out, political elites assess the viability of a new party according to their perception at the time, informed by “electoral histories.” In new democracies, given the absence of a track record, political actors are more uncertain about their prospects and, consequently, may have more incentive to launch new parties. While Tavit's evidence supports that the emergence of new parties tends to decrease as democracy matures, Zons (2015) and Chiru et al. (2021) suggest that this factor influences only the emergence of genuinely new parties, not those that originate from party splits.

In addition to the previously mentioned influences of political structure and the extent of democratic consolidation, research on new political parties also indicates that social structure plays a significant role in their emergence. Based on the premise of Lipset and Rokkan's (1967) seminal work – that political parties are the political expression of the social cleavages present in a country – studies have explored how the emergence of new cleavages is associated with the formation of new parties, such as the new left, far-right and green parties (see Ford and Jennings, 2020). Less structuring than cleavages, theories also suggest that parties can emerge to represent issues and ideological positions ignored by the existing political parties in a given system (Hug, 2001; Lucardie, 2000). In this sense, several factors can facilitate the entry of new parties: a lower issue coverage by the existing parties (Zons, 2015), an ideological position left open by a collapse of a party (Laroze, 2019), or a misalignment between parties and citizens’ preferences (Lago and Martínez, 2011).

These factors however are generally more associated with the emergence of genuinely new parties rather than splinter parties. Zons (2015) and Chiru et al. (2021) indicate that programmatic differentiation and issue coverage do not significantly influence the formation of splinter parties. Instead, both studies show that their emergence is more linked to institutional factors, like the permissiveness of the electoral system, and to government and economic instability. These rare studies that compare different types of new parties suggest therefore that the incentives for their emergence and success can vary significantly, highlighting the importance of examining each type in depth.

Party Splits in Brazil

The evolution of Brazil's party system has been profoundly influenced by the phenomenon of party splits. With the enactment of the Organic Law on Parties in 1979, the two-party system imposed by the military regime was dismantled. This legislation opened the political arena, enabling politicians who had been restricted to the only recognised parties, MDB and Arena, to either establish new parties or revive those that had been suppressed by the dictatorship. The resultant fragmentation, notably of the MDB, led to the formation of its direct successor, the PMDB, as well as other key parties like the PDT, PT, and PTB (Soares, 1988). Thus, Brazil's democratic transition coincides with these inevitable initial party splits, given the previously restricted freedom of party organisation.

In the years following, the early stages of Brazil's democracy were characterised by the frequent emergence of new parties, here defined and operationalised as organisations that nominate candidates for the first time in general elections to the country's representative assemblies (Hug, 2001). Unlike genuinely new parties, which are formed without “any help from members of existing parties” (Hug, 2001: 13), many of these emerging parties attracted politicians from established ones. Figure 1 illustrates this evolution between 1982 and 2022. The bars represent the percentage of candidates for state and federal deputy positions who switched parties, moving to newly created parties from one electoral cycle to the next. These candidates are referred to here as splinter candidates. The line tracks the number of new parties that absorbed these candidates, showing how many newly formed parties fielded politicians with prior affiliations in their first electoral contest. Table 1 provides an overview of these parties, listing those in each electoral cycle. OPEN IN VIEWER

OPEN IN VIEWER

Table 1.

New Parties with Splinter Candidates by Electoral Cycle in Brazil.

Electoral cycle	New Parties (Abbreviations)
1982–1986	PFL, PH, PL, PMB, PMC, PMN, PND, PPB, PRT, PS, PSB, PSC, PTN, PTR, PSART, PCN, PDC, and PJ.
1986–1990	PSDB, PSL, PST, PT do B, PAP, PAS, PD, PLH, PNAB, PNT, PRP, PRS, and PSD
1990–1994	PSTU and PTRB
1994–1998	PAN, PGT, PSDC, PSL, PSN, and PTN
2002–2006	PRB and PSOL
2010–2014	PATRIOTA, PPL, PROS, PSD, and SD
2014–2018	REDE and PMB

Source: Superior Electoral Court (TSE). Own elaboration.

The period between 1982 and 1986 was particularly noteworthy, during which parties were allowed to run in the 1986 elections even with only provisional registration ² . In this electoral cycle, 28.8 per cent of the candidates for federal and state deputy who ran in both the 1982 and 1986 elections left their parties to join newly formed ones. Several small parties were created in this period but did not obtain definitive registration and then disappeared. However, this cycle is also marked by splits that gave rise to parties with longer histories, such as the PFL, the PL, and the PSC.

The trend of new party emergence persisted from 1986 to 1990, with the emergence of many minors, short-lived parties, but also with the establishment of more relevant parties like the PSDB. During this electoral cycle, 13.1 per cent of splinter candidates were observed among those who ran in both the 1986 and 1990 elections. Following this initial surge, the incidence of party splits declined, though small parties with brief lifespans continued to appear, attracting relatively few splinter candidates. However, the 2010–2014 cycle saw a surge, with five new parties, most notably the PSD, drawing a significant share of candidates (11.7 per cent).

The temporal variation in party formation is primarily explained in the literature by shifts in institutional incentives. In the years immediately following the democratic transition, relatively permissive party registration requirements, combined with a context of political openness and significant uncertainty about electoral prospects in a still weakly structured party system, created a favourable environment for the proliferation of new political organisations (Rocha and Pimentel, 2024). However, from the 1990s onward, the process of party formation became more constrained. Stricter legal and institutional requirements for registering new parties, coupled with the greater accommodation of political elites within existing organisations, led to a decline in the number of new entrants into the Brazilian party system and a progressive closure of inter-party competition (Braga, 2010; Rocha and Pimentel, 2024).

Nevertheless, the surge in party formation during the 2010–2014 electoral cycle signalled a deviation from the expected trend toward greater institutionalisation, despite a more restrictive party system and increasingly stringent requirements for registering new parties. As Simoni Junior and Ribeiro (2019) argue, the Supreme Court's interpretation of electoral rules, particularly its leniency toward deputies’ switching to new parties without losing their mandates ³ , contributed to this dynamic. Additionally, the favourable allocation of state resources to newly formed parties further incentivised party creation. These factors were particularly evident in the emergence of the PSD in 2011.

In addition to the role of institutions in shaping the costs and benefits of leaving parties to form new organisations, the literature also highlights the strategic behaviour of political elites in understanding the phenomenon of party splits. Roma (2002), for example, challenges the notion that the split from the PMDB that form the PSDB in 1988 was primarily ideological. The author argues that the PSDB emerged because its founders, despite being part of the PMDB, had limited space to influence the Sarney government (1985–1990) and were also excluded from the presidential succession process. Moreover, they perceived an electoral opportunity among voters dissatisfied with the federal government, suggesting that the creation of the PSDB was an accommodation of elites’ strategies and interests.

Other cases illustrate a different but equally strategic use of party formation: creating new parties as a means to maintain proximity to the federal government. The PFL, founded in 1984, is one such example. Corbellini (2005) argues that its founders, originally part of the PDS, sought to dissociate themselves from the military regime while still remaining close to federal power. With the impending victory of the opposition party, the PMDB, in the 1985 indirect presidential election, a faction of the PDS decided to break away and support Tancredo Neves, forming the PFL as a way to ensure their continued presence in government. Similarly, Krause et al. (2017) suggest that the formation of the PSD in 2011 permitted centre-right political elites to strengthen their ties with the federal government, then led by the centre-left party, PT (2003–2016).

The literature then supports the idea that party splits in Brazil stem in great part from political elites’ strategic responses to institutional incentives. The costs and benefits of founding a new party vary according to the rules in play and the characteristics of the electoral and party system. Under the influence of these conditions, political elites may leave their parties and join new organisations, motivated by their dissatisfaction with the parties they belong to and/or the goals they wish to achieve. Existing studies, however, leave certain questions unanswered.

On the one hand, research analysing the formation of new parties in Brazil, even when distinguishing between genuinely new parties and those emerging from splits, tends to focus primarily on external factors that facilitate party formation (Krause et al., 2018; Rocha and Pimentel, 2024). These include rules for party registration, the level of party system fragmentation, economic conditions, and electoral system characteristics. However, such analyses often overlook the internal conditions of the parties from which splinters break away. As previously discussed, a significant part of the literature on party switching, whether collective or individual, suggests that the decision to leave a party is influenced by the internal dynamics of the party of origin.

On the other hand, studies focusing on party splits tend to incorporate these intra-party dynamics to explain why divisions occur but often do not explore variations within the same party (Corbellini, 2005; Krause et al., 2017; Roma, 2002). While acknowledging that the creation of new parties is a pragmatic strategy that can serve the interests of political elites, these studies do not fully explain why some politicians choose to follow this strategy while others remain in their original parties. Therefore, case studies highlight the strategic behaviour of elites, yet more systematic studies are needed to identify the specific intra-party conditions that facilitate or hinder defection.

This study seeks to explore these questions by focusing on the internal dynamics of party splits while also considering the external factors commonly included in explanations of new party emergence. The goal is to provide a more systematic analysis of how intra-party competition influences the likelihood of party splits in Brazil.

Intra-Party Competition and Coordination Failures

Analysing the internal characteristics of the parties seems central to understanding the emergence of party splits. Assuming that political actors seek to satisfy their goals related to votes, policy and office (Strøm and Müller, 1999), it is essential to consider how the original parties either facilitate or obstruct these objectives. As the literature shows, the costs, benefits, and risks of defection can vary significantly from one political party to another. Based on this idea, I argue that one important factor that can influence the emergence of party splits is the internal competition between political elites. This competition can favour party splits through two interrelated mechanisms.

First, intense intra-party competition can drive vote-seeking politicians away from their parties. Where elections are based on open-list proportional representation, like in Brazil, candidates not only compete against members of other parties but also, to some extent, against co-partisans (Ames, 1995; Carey and Shugart, 1995). Although this intra-party competition is generally less intense in reality than theories assume (see Cheibub and Sin, 2020), it is expected to vary significantly between parties and districts (Desposato, 2006). Such variations can influence election success and heighten intra-party tensions, potentially increasing dissatisfaction and prompting defections. I argue that, regardless of their actual electoral prospects, politicians are likely to perceive a higher number of competitors within their party as a negative factor. The more candidates vying for the same pool of party resources and voter support, the greater the sense of internal competition, which can fuel dissatisfaction and increase the likelihood of a split. Based on this idea, I propose the following hypothesis:

H1: The greater the number of candidates competing for each seat within a party, the higher the likelihood of a party split.

Second, intense intra-party competition can be an indicator of coordination problems inside the party, which may drive party splits. As Cheibub and Sin (2020) argue, in open-list proportional systems, both parties and candidates have incentives to limit internal competition. For parties, which are office-seeking, excessive competition within their lists is detrimental, as it wastes votes and resources. To avoid a “war of all against all”, parties attempt to strategically structure their lists and allocate resources efficiently. For candidates, particularly those who consider themselves viable, excessive competition means increased vote fragmentation and reduced electoral chances.

To mitigate such conflicts, intra-party competition should ideally be settled before elections, allowing parties to assemble effective lists. Cheibub and Sin (2020) show through empirical evidence that Brazilian parties generally succeed in managing this process, presenting lists with only a few strong candidates competing against each other for the Chamber of Deputies. However, coordination failures can still occur. One key factor contributing to these failures is party uncertainty about the chances of their candidates. The authors argue, for example, that coordination problems are more frequent in larger districts, as they introduce greater uncertainty about parties’ and candidates’ electoral strength, making it harder to prevent overlapping candidacies, particularly in major urban centres.

However, another source of coordination failures can be the internal competition between factions to gain control over party resources and power. While parties are often treated as unitary actors, numerous studies challenge this assumption by highlighting the existence of factionalism – “the partitioning of a political party (or other organisation and group) into subunits which are more or less institutionalised and who engage in collective action in order to achieve their members’ particular objectives” (Boucek, 2009: 468). As Boucek (2009, 2012) argues, when factions shift from cooperatively managing differences to competing over power and resources, internal dissent intensifies and the chances of party splits increase.

Electoral competition seems to be an important aspect of factionalism in general. Factional exits may be driven by policy disagreements, particularly when factions adopt positions that diverge from those of the party leadership. However, as Ceron (2019) demonstrates, electoral and office-seeking incentives also play a crucial role in party breakaways. Invernizzi (2023) notes that factions not only compete for influence in candidate selection and internal decision-making but may also prioritise their own political survival over collective party campaigning, further weakening party cohesion. Open-list proportional systems can exacerbate this dynamic, as preference voting provides factions with independent electoral access, unlike closed-list systems, where party leadership retains stronger control over candidate selection and resource allocation (Ceron, 2019).

Thus, high intra-party competition in elections may signal coordination difficulties before elections. This coordination problem can be caused by factional conflicts, as their action can prevent centralised and effective party decision-making. However, even in the absence of factional action, coordination problems can arise, given the difficulty of drawing up lists and the party's uncertainty about the viability of its candidates. In both cases, these tensions may prevent elites from being accommodated within the same party, fuelling dissatisfaction and prompting defections. Politicians or factions seeking greater influence may then exit to establish new organisations.

Based on these arguments, I propose that intra-party competition in the electoral arena is an indicator of coordination problems that can lead to party splits. If a party has many viable candidates – those with strong electoral prospects – competing for each seat it is expected to win, this signals a poorly coordinated party list. A high number of strong competitors increases intra-party tensions and weakens cohesion, making a split more likely. I then test the following hypothesis:

H2: The greater the number of viable candidates competing for each seat within a party, the higher the likelihood of a party split.

Data and Methods

To test the above hypotheses, this paper analyses datasets with party-level observations in each Brazilian state on ten electoral cycles from 1982 to 2022 ⁴ . To identify occurrences of party splits, I use data on electoral candidacies for federal and state deputies over the specified period. This involves tracking the party affiliations of candidates at time t and noting changes in their party affiliations by time t + 1 ⁵ .

In this analysis, I employ two operationalisations of party splits. The first adopts a broader definition. A split is recorded if two conditions are met: (a) one politician leaves their original party (the party they were affiliated with at time t) to join a newly formed party by time t + 1 – one that is participating in general elections for the first time – and (b) this move is not an isolated act but occurs alongside at least one other politician from the same state. This distinction is crucial, as for a movement to be classified as a split rather than an individual defection, the destination party must be newly created, and the migration must involve at least two actors, indicating some level of coordinated defection.

The second operationalisation applies a stricter definition of party splits, considering only movements to newly formed parties classified as “co-opted parties” by Rocha and Pimentel (2024) – that is, parties in which more than 50 per cent of their founders had prior affiliations with existing parties before joining the new organisation. This distinction is particularly relevant in the Brazilian context, where genuinely new parties, those formed without any prior partisan affiliations (Hug, 2001), are exceedingly rare. Still, some new parties incorporate more experienced political actors than others. This stricter classification thus allows us to distinguish parties that emerge from important elite migration, those closely aligned with the idea of a party split, from new parties more loosely connected to the existing party system. This stricter criterion ensures that only breakaways involving significant levels of prior party coordination and elite realignment are considered splits.

The occurrence of a split is measured as a dichotomous variable, coded as 0 (no split) or 1 (split occurred). It indicates whether at least one split occurred within a party in a given electoral cycle and state. For example, if at least two candidates who ran for the PMDB in Rio de Janeiro in the 1986 elections joined the PSDB (a new party) in the 1990 elections, this transition is classified as a split within the PMDB for the 1986–1990 cycle in Rio de Janeiro.

The explanatory variables, which measure different types of intra-party competition, take into account the electoral lists of the parties in the states at time t. The first is the Number of Candidates per Party Seat. I adapt the formula for the Effective Number of Parties (ENP) to reflect the effective number of candidates competing for each seat contested by the party (number of seats won plus 1). This adaptation involves considering the proportion of personal votes each candidate received relative to the total number of votes garnered by all candidates from the party in the district. To that extent, the electoral strength of the candidates is taken into account, but only in relation to their own party. The actual electoral viability of the candidates on the list is not considered. The aim here is to understand whether perceived competition, independent of the chances of election, influences the chances of a split.

The second explanatory variable is the Number of Viable Candidates per Party Seat at time t. A candidate is considered viable if they receive at least 10 per cent of the electoral threshold in personal votes. Unlike the first variable, this measure explicitly incorporates electoral viability by capturing the number of viable candidates competing for each seat contested by the party (again, calculated as the number of seats won plus one). A higher number of viable candidates indicates greater intra-party competition for each seat.

As controls, I add three groups of variables. The first assesses the characteristics of the party that could influence party split. As presented earlier, the party's degree of access to office and influence over policy can impact the chances of defection. In this sense, I include the dichotomous variables Governor, Governor Split, Federal Coalition Membership and Federal Coalition Exit. Governor indicates whether the party successfully elected the state governor in the election of time t. The second variable, Governor Split, indicates if the governor of the respective state split to a new party in the electoral cycle. Federal Coalition Membership indicates whether the party was part of the governing coalition at any point during the electoral cycle. Federal government exit indicates whether the party withdrew from the federal governing coalition during the electoral cycle. I expect that parties with greater access to executive office – whether at the state or federal level – will be less prone to splits, as holding power provides resources, patronage opportunities, and institutional leverage that incentivise internal unity. Conversely, governor splits and exiting the federal coalition may signal intra-party conflicts or instability, increasing the likelihood of party split.

Ideology is another party characteristic that can significantly influence changes in party affiliation among political elites (see Volpi, 2019). In Brazil's Chamber of Deputies, analyses suggest that party switching is less frequent among members of left-wing parties compared to right-wing parties (Melo, 2000; Nicolau, 1996). Therefore, Left position is a dichotomous variable that takes a value of 1 when the party is classified as having left-wing or centre-left positions. This classification can be challenging in Brazil, where parties often do not maintain consistent positions over time. However, based on a series of classifications and descriptions of various studies (Bolognesi et al., 2022; Codato et al., 2018; Power and Zucco, 2009; Zucco and Lauderdale, 2011), I have identified the parties that most frequently appear as left-wing across these different sources.

The second group of control variables captures electoral competition dynamics at the district level. District magnitude measures the number of seats contested in the district. As previously discussed, district magnitude plays an important role in shaping a party's ability to coordinate candidates and construct effective electoral lists. District competitiveness accounts for variations in inter-party competition across elections. This variable follows the measure of district competitiveness at time t proposed by Blais and Lago (2009), which calculates the minimum number of additional votes a party needs to secure an extra seat. A higher number of votes required indicates lower district competitiveness. However, I have inverted the index for ease of interpretation, meaning that higher values now indicate greater inter-party competition.

The third group of control variables includes factors typically associated with the emergence of new parties. Party system fragmentation, measured using the ENP in the Chamber of Deputies at time t, based on the Laakso and Taagepera (1979) formula ⁶ . Electoral Cycle Dummies, which control for institutional contexts that have historically facilitated party formation. These include: (a) 1982–1986 and 1986–1990 cycles, when parties faced minimal legal barriers to contesting elections, as discussed earlier; (b) 2010–2014 cycle, when the anti-defection resolution may have altered party elites’ strategic calculations, incentivising the creation of new parties.

Finally, since larger parties tend to field more candidates – potentially intensifying internal competition and increasing the likelihood of party splits – I control for the Number of Candidates each party presents in each state and electoral cycle. I also include a control variable distinguishing whether the candidate is running for the Chamber of Deputies or a State Assembly.

Much of the data used comes from the electoral results and candidate data made available by the Superior Electoral Court from 1982 to 2022 ⁷ . Data on the composition of the government coalition from 1998 onwards is provided by NIPE and CEBRAP (2021). Prior to that date, the parties of the ministers who made up the federal government were checked in the Atlas Histórico do Brasil (FGV, 2023). Descriptive statistics for all variables are provided in the Supplemental Material.

Results

Table 2 presents the results of the logistic regression models at the party level. The four models presented estimate the probability of collective party splits using odds ratios (ORs) to facilitate interpretation. Therefore, values greater than 1 indicate a higher probability of a party split, while values less than 1 indicate a reduced probability. The models differ based on their operationalisation of party splits: Models 1 and 3 adopt the broader definition, while Models 2 and 4 apply the stricter criterion of co-optation parties.

OPEN IN VIEWER

Table 2.

Logistic Regression Estimates of Party Splits in Odds Ratio.

	Model 1 (General)	Model 2 (Co-optation)	Model 3 (General)	Model 4 (Co-optation)
Effective number of candidates	1.075***	1.037
	(0.018)	(0.021)
Number of viable candidates			1.344***	1.640***
			(0.117)	(0.167)
District magnitude	0.994	0.994	0.999	0.997
	(0.004)	(0.005)	(0.004)	(0.005)
District competitiveness	1.042*	0.772	1.040*	0.761
	(0.020)	(0.139)	(0.021)	(0.141)
Federal coalition membership	0.987	1.630**	0.763	1.228
	(0.164)	(0.307)	(0.127)	(0.232)
Federal coalition exit	2.363***	2.560***	2.407***	2.553***
	(0.614)	(0.706)	(0.630)	(0.711)
Governor	1.309	1.615*	1.178	1.554
	(0.295)	(0.390)	(0.263)	(0.371)
Government split	1.253	1.336	1.201	1.343
	(0.284)	(0.323)	(0.273)	(0.327)
Left-wing party	0.634**	0.799	0.669**	0.856
	(0.097)	(0.141)	(0.102)	(0.153)
Party system fragmentation	0.961	0.707*	0.964	0.719*
	(0.044)	(0.100)	(0.044)	(0.099)
Election cycle 2010–2014	23.838***	66.373***	23.172***	65.621***
	(4.440)	(25.308)	(4.296)	(24.689)
Election cycle 1986–1990	19.025***	3.621	22.128***	4.590
	(7.804)	(3.091)	(9.009)	(3.831)
Election cycle 1982–1986	33.933***	7.926*	38.563***	9.801*
	(15.463)	(7.317)	(17.637)	(8.886)
Federal deputy	1.411*	1.253	1.420*	1.461*
	(0.216)	(0.225)	(0.220)	(0.267)
Total candidates in election	1.154***	1.129***	1.138***	1.110***
	(0.014)	(0.015)	(0.014)	(0.016)
Num. Obs.	7799	7799	7799	7799
AIC	1676.8	1223.2	1683.8	1203.9
BIC	1781.3	1327.6	1788.2	1308.4
Log. Likelihood	−823.423	−596.600	−826.897	−586.971
F	43.285	30.588	43.358	31.211
RMSE	0.17	0.14	0.17	0.14

Source: Own elaboration.

* p < 0.05, ** p < 0.01, *** p < 0.001.

The results offer mixed support for Hypothesis 1 regarding intra-party competition. In Model 1, the Effective Number of Candidates per Party Seat is statistically significant, with an odds ratio of 1.075, meaning that each additional effective candidate is associated with a 7.5 per cent increase in the odds of a party split. However, this variable is not statistically significant in Model 2, where a stricter definition of party split is adopted.

In contrast, the Number of Viable Candidates per Party Seat, tested in Models 3 and 4, shows a strong and statistically significant positive effect. In Model 3, each additional viable candidate increases the odds of a party split in 34.4 per cent (OR = 1.344). Even when a strict definition of party split is adopted, as in Model 4, the effect remains substantial: each additional viable candidate is associated with a 64.0 per cent increase in the odds of a split (OR = 1.640). These findings support Hypothesis 2, indicating that intra-party competition among electorally viable candidates significantly increases the likelihood of a split.

These results indicate that competition among electorally viable candidates, rather than internal competition in general, may play a more important role in the emergence of party splits. While a high number of candidates per seat is associated with splits under the broader definition, this effect does not hold when a stricter definition of party split is applied. In contrast, the presence of multiple viable contenders, candidates perceived as having a real chance of winning, significantly increases the likelihood of a split across all models. This suggests that splits are more often a response to coordination failures among strong contenders than to general candidate crowding. In this light, party elites may be less concerned with broad internal competition and more responsive to direct threats to their electoral survival.

Beyond intra-party competition, inter-party competition also plays a role. In Models 1 and 3, more competitive districts are associated with increased odds of a split (ORs of 1.042 and 1.040, respectively). When inter-party competition is high and the distance between parties is low, the chances of a split increase under the broader definition of party splits. This may indicate that as competitiveness increases, greater uncertainty about electoral outcomes emerges, potentially creating more opportunities for new parties to form. Political actors may perceive more competitive elections as a less institutionalised environment, encouraging them to launch new parties to maximise their chances of success. However, this dynamic does not apply to the emergence of co-optation parties, which are more closely associated with coordinated elite movements. Unlike broader splits, co-optation parties may require a level of pre-existing organisation and elite consensus that is less likely to emerge in highly competitive environments with high uncertainty.

Other control variables shed additional light on the conditions under which party splits occur. Federal Coalition Exit has a significant and positive effect across all models. Parties that exit the federal coalition during the electoral cycle are 2.36 to 2.56 times more likely to experience a split. This finding supports the idea that governing instability and shifting political alliances can increase intra-party tensions, leading to breakaways. Meanwhile, left-wing parties are significantly less likely to split, but only under the general definitions (Models 1 and 3), with odds ratios of 0.634 and 0.669, respectively. This corroborates existing literature suggesting that left-wing parties may have stronger ideological commitments that reduce incentives for exits.

Election cycle effects are particularly striking. The 1982–1986 and 1986–1990 periods, characterised by low institutional barriers to party formation, show high odds ratios, reflecting the fluidity and volatility of those transitional years. Similarly, the 2010–2014 period, marked by anti-defection regulations tolerant of exits to new parties, displays an overwhelming increase in split probabilities.

Finally, party system fragmentation shows a negative effect in the co-optation models (Models 2 and 4), with odds ratios of 0.707 and 0.719. This suggests that as the party system becomes more fragmented, the incentives for elites to break away and form new parties diminish. In highly fractionalised environments, the viability and visibility of new splinter groups may be constrained, reducing the strategic appeal of elite-coordinated exits.

Discussion

The findings of this study suggest that variables typically associated with the formation of new parties and political defections play an important role in explaining party splits. As research on new party formation indicates, institutional incentives, district-level competition, and party system characteristics can influence the likelihood of breakaways. Similarly, as studies on party switching highlight, access to government resources also seems to be a factor shaping political elites’ strategies.

Beyond these well-established variables, this analysis underscores the importance of intra-party electoral competition in the dynamics of party splits. Both measures of internal competition (i.e., the effective number of candidates per party seat and competition among viable candidates) significantly increase the likelihood of splits in multiple models, though the latter has a stronger and more consistent effect across specifications. This suggests that elites are sensitive to intra-party competition, particularly when it presents an electoral threat for viable contenders within their own parties.

Theoretically, this finding aligns with the coordination problem framework (Cheibub and Sin, 2020), which posits that parties seek to strategically manage their lists to maximise electoral success. However, the presence of multiple viable candidates signals an internal failure in coordination, increasing the risk of splits.

Despite these findings, this study does not allow for definitive conclusions on whether intra-party competition directly causes party splits. Intense competition may generate dissatisfaction and lead to defections, or it may instead serve as an indicator of deeper pre-existing party conflicts. The lack of stronger support for the first hypothesis leans toward the latter interpretation. If competition itself were the primary source of discontent, a larger number of candidates within a party should have significantly increased the probability of a split in different models, regardless of their electoral viability.

However, the present analysis does not entirely rule out the possibility that intra-party competition itself contributes to defections. The significant support for the second hypothesis – demonstrating that a higher number of viable candidates correlates with increased party splits – suggests that competition, when it directly threatens candidates’ electoral prospects, can become a destabilising force. The perception that viable candidates face a heightened risk of losing their seats may drive defections, as it can reinforce dissatisfaction.

Therefore, in this study, intra-party competition is better understood as an indicator rather than a direct cause of party splits. However, to advance the understanding of these possible mechanisms, future research should further explore how pre-electoral disputes among regional leaders, ideological groups, or factions influence party cohesion and electoral list formation. This is particularly relevant in the Brazilian context, where few studies have systematically examined factionalism within parties.

Moreover, the role of party leadership and the internal party organisation in managing competition and preventing breakaways should not be overlooked. Strong leadership with centralised decision-making authority may mitigate the impact of intra-party competition by enforcing internal discipline, allocating resources strategically, and brokering compromises between factions. However, prohibitionist approaches to internal division can lead to marginalisation of losing factions (Lourenço et al., 2024) and centralisation make it difficult to handle dissent among members (Maor, 1995). Future research could explore how leadership styles, internal governance mechanisms, and party discipline strategies shape the ability of parties to retain their members amid competitive pressures.

In sum, the findings presented here align with broader theoretical arguments regarding the strategic behaviour of political actors and their responses to institutional and political conditions. Politicians motivated by electoral gains may become disillusioned in highly competitive intra-party environments, increasing the likelihood of party splits. Likewise, inefficiently structured candidate lists can indicate underlying conflicts and coordination failures within the party, further incentivising defections. However, these decisions do not occur in isolation. The costs and risks associated with founding a new party are shaped by electoral rules, party registration procedures, inter-party competition, and the degree of party system fragmentation. Therefore, the emergence of party splits is driven not only by internal party dynamics but also by broader institutional and systemic factors.

Supplemental Material