The Proscription Paradox: Banning Parties Based on Threshold Requirements and Electoral Volatility in Latin America

Abstract

Banning political parties is an extreme institutional measure that democracies tend to use sparingly. Nevertheless, Latin American countries frequently proscribe their parties through rules that activate dissolution for not reaching a certain number of votes or seats in an election. Such rules are expected to stabilize and simplify party systems. However, a competing theory suggests that such rules instead promote electoral volatility by injecting political uncertainty into the party system through cyclical refoundation of extinct parties and the mechanical effects of parties’ exits. Attempting to resolve this paradox, this analysis tests the effect of dissolution thresholds on electoral volatility in all Latin American democratic elections since 1980. Party bans based on dissolution thresholds are found to promote electoral volatility, which bears implications for democratic governance.

As unstable and volatile party systems jeopardize democratic governance (Roberts and Wibbels 1999; Weyland 1999; Mainwaring 2018c; Carreras and Acácio 2019), countries have employed a variety of institutional devices to curb the influx of small, new, and ephemeral parties. One of them—the outright proscription of such parties—has mostly escaped analytical attention. Yet a number of democracies have adopted legal thresholds that activate the dissolution of parties for not reaching a certain number of votes in elections. This article identifies a central theoretical paradox. Despite hopes for stabilizing and simplifying party systems by such dissolution thresholds, there are also strong theoretical arguments that suggest the completely opposite effect. Elements of political uncertainty injected into the party system by dissolution thresholds, cyclical refoundation of extinct parties, and the mechanical effects of parties’ exits may actually promote electoral volatility. Does the existence and size of dissolution thresholds make party systems more stable or more unstable?

Attempting to resolve this paradox, the following analysis contributes to two research agendas by relying on the experience of Latin American countries. The first contribution answers calls for theoretical innovation in studying the determinants of electoral volatility (Crabtree and Golder 2017). Reporting a series of weak and null findings on theoretically important variables, recent regional comparisons identify only a few, if any, generalizable causes of volatility (Powell and Tucker 2014; Crabtree and Golder 2017; Mainwaring and Bizzarro 2018; Cohen et al. 2018). Yet given the importance of party system instability to democratic performance, this theoretical state of affairs is not satisfactory. The second contribution aims to fill a lacuna within the growing comparative and theoretical literature on party bans. Neither the voluminous comparative legal literature categorizing legal motives for party dissolution nor the comparative political approach, mapping the incidence, manifestations, and political consequences of party bans has produced any discussion whatsoever of party bans due to threshold rules (e.g., Issacharoff 2007; Moroff 2010; Bligh 2013; Casal Bertoa and Bourne 2017; Bourne 2018). This is a surprising omission, given the ubiquity of this rule in party legislation and the high frequency with which the dissolution threshold rules are used to ban political parties in practice.

The structure of the article is the following. The first section provides a descriptive overview of the evolution of the dissolution thresholds in Latin American legislation since 1980. The second section discusses the theoretical underpinnings of the nexus between the size of dissolution thresholds and volatility and focuses on three principal causal mechanisms that drive this relationship. The operation of these mechanisms is not tested empirically, but they underscore the main hypothesis, that the dissolution thresholds increase party system volatility levels. That section also presents selective qualitative examples of actual party bans and the use of the dissolution thresholds in Latin America.

A systematic test of the relationship is provided in the third section, which relies on regression models of the determinants of electoral volatility. The models again provide support for the principal hypothesis that high thresholds of party dissolution promote electoral volatility. The concluding section emphasizes that the widespread use of dissolution thresholds not only falls short of the stringent normative criteria justifying party bans in modern democracies, but that their destabilizing effect on the wider party system is also problematic.

Banning Parties Through Dissolution Thresholds in Latin America

Threshold-induced party bans are almost unique to the region of Latin America. Whereas laws regulating political parties in other democracies allow for party proscription, failing to reach a threshold in an election is almost never one of them. This is reflected in the comparative scholarly literature on party bans, which has so far not provided any discussion, either normative or empirical, of threshold-induced bans (Moroff 2010; Casal Bertoa and Bourne 2017). Nor have explanations of party proscription legislation included this particular legal motive (Bourne 2018).

Much more is known about other legal motives for banning parties, especially as presenting a danger to core democratic principles (Issacharoff 2007; Bligh 2013), espousing antidemocratic or racist ideologies (Finn 2000; Niesen 2002; Mareš 2012), challenging territorial integrity (Celep 2014; Weill 2018), promoting violent or terrorist actions (Navot 2008; Bligh 2013), lacking internal democracy (Mersel 2006), or acting exclusively on behalf of a specific ethnic or particularistic identity group (Moroff 2010). The emphasis on these causes naturally reflects the variety of legal motives for proscribing parties in countries under comparison, which has mostly excluded Latin American countries.

Latin American party legislation provides a completely opposite comparative panorama. Not reaching thresholds constitutes the most frequent legal grounds for proscribing parties in Latin American countries (Bareiro and Soto 2007, 609). Of course, there are also other grounds set in electoral and party laws, but these are less frequent in comparative terms. They include failure to fulfill the requirements of internal democracy (holding elections for party authorities and candidates), military training of their affiliates, committing electoral fraud, participating in military coups, or lacking financial transparency (Bareiro and Soto 2007; Fernández 2007).

Dissolution thresholds are defined here as legal barriers that set a minimum number of votes or elected representatives that a political party needs to obtain in order to remain allowed to field candidates and run in elections. If parties do not exceed this barrier in elections, they are dissolved (dissolution is used here interchangeably with party banning, extinction, proscription, or cancellation). The precise legal definition takes several forms in Latin American legislation—laws stipulate parties’ extinction (extinción), cancellation (cancelación), loss of registration (cancelación de inscripción), or expiration of political-juridical personhood (caducidad de la personalidad política-jurídica). In some cases, party dissolution does not necessarily imply the effective liquidation of the organization itself. In Colombia, for example, parties not reaching the 3 percent threshold lose their juridical personhood: they are not forced to disappear in organizational terms altogether, but they lose their right to nominate candidates for elections and to receive public funding (Lozano 2015). The former is crucial, as parties unable to run their candidates in elections cannot fulfill their main function and effectively cease to be parties. Such threshold requirements are therefore understood here as equal to any other legal concepts of dissolution, as long as they imply that the party is disqualified from participating in elections.

Table 1 provides a descriptive overview of the threshold size and threshold type in each Latin American country and traces the evolution of these rules during all years between 1980 and 2018. The basis for identifying and coding these dissolution thresholds was the constitutions and party and electoral laws and their reforms. The online appendix lists the corresponding laws (and their reforms) and the sources of information for each country. ¹

Table 1. Party Dissolution Thresholds in Latin American Countries, 1980–2018

Source: Compiled by the author. For details of coding and sources see the online appendix.

Notes: RV, registered voters; VV, valid votes; VC, votes cast; V, votes; MP, member of parliament.

Blank cells identify either the existence of authoritarian regimes where party legislation was not applicable, or missing data.

Two types of dissolution thresholds are used in Latin America: thresholds that specify a minimum vote share (or a minimum absolute number of votes) and thresholds that require a minimum number of elected legislators. ² Some countries use a combination of both rules. Some laws (e.g., Ecuador, Paraguay) require that the threshold be exceeded in either of two consecutive elections in order to avoid dissolving the party, although the vast majority of laws derive these thresholds from the results of a single election. As is evident from table 1, dissolution thresholds are subject to frequent changes in most Latin American countries. All but one country have reformed their dissolution threshold rules at least once during this period. El Salvador holds the record with five substantial reforms of this rule during this period. Ecuador and Panama have done so four times.

While thresholds were absent during only very few periods after 1980, the size of thresholds has varied substantially, ranging from very low in Argentina, Costa Rica, and Uruguay to very prohibitive in Guatemala, Ecuador, Panama, Nicaragua, and Peru. The latter group of countries ban or have banned their parties for not reaching 4 percent or 5 percent of votes in an election. Some soften this requirement by mandating that parties may still continue their existence if they receive a certain number of seats in an election (typically at least one). By 2018, three countries—Bolivia, Colombia, and Mexico—used an intermediate threshold of 3 percent valid votes. The general tendency has been to increase the size of threshold over time. While seven countries did not apply any dissolution thresholds at the time of their first competitive elections after 1980, only one country (Brazil) did so by 2018.

Little is known about the justification for the adoption and increasing size of the dissolution thresholds in individual countries. However, it is certain that this pattern fits the overall evolution of party regulation in Latin America, which has come to conceive of parties as semistate institutions (Scherlis 2021). The earlier permissive approach of the 1980s and 1990s gave way to a prescriptive approach, in which parties are managed by the state, dependent on state resources, and seen as public utilities, and in which the state is allowed to regulate parties with almost no limits, in order to secure a functioning, stable democratic system (Scherlis 2021).

Therefore, the adoption of more restrictive party dissolution thresholds seems to be a symptom of a larger reconfiguration of party regulation across the region, which has also witnessed similarly restrictive reforms in the areas of candidate selection, party finance, or party registration. Future research should also reveal the extent to which this trend responds to strategic calculations by political elites. Powerholders in Latin America frequently use electoral reform to pursue and protect their interests (Freidenberg 2022). As dissolution thresholds are not neutral to the distribution of political power, it is possible that specific political interests explain their adoption and reform.

The Effects of Dissolution Thresholds on Electoral Volatility

With only two exceptions, the effect of party bans on the wider party system and its stability has not been systematically analyzed. One study found evidence of such an effect using the experience of party proscription in Spain, Turkey, and Germany, theorizing that party bans fuel electoral volatility because they affect the preferences of voters who transfer their votes to other parties (Casal Bertoa and Bourne 2017). Although those parties were banned for their radical ideology or secessionist attempts, there is every reason to expect that this would have an effect similar to that of bans resulting from the failure to reach the dissolution thresholds. Likewise, a historical analysis of Latin American elections between 1912 and 1991 reported a strong effect on the volatility of the vote share of parties that either boycotted the election or were proscribed (Coppedge 1995).

Still, the rich comparative literature on electoral volatility has not addressed the question of whether the presence of an institutional rule to ban parties (and not just the actual proscription of a particular party) affects the party system. It has so far emphasized different explanations. The problem of extreme electoral volatility is particularly acute in Latin America, where not only have party systems failed to stabilize since the transition to democracy in the 1980s, but the average electoral volatility has increased in all decades since then (Cohen et al. 2018; Carreras and Acácio 2019).

A major part of this instability is driven not by vote swings between existing parties but by the emergence and demise of new parties (Cohen et al. 2018). Early analyses of volatility in Latin America with data from the 1980s and 1990s, or even earlier periods, identified important cross-national predictors, in particular economic performance (Remmer 1991), institutional discontinuity and fragility (Roberts and Wibbels 1999), ethnic fragmentation (Madrid 2005; Birnir 2006), historical suffrage extensions (Kellam 2013), and the creation of support parties by preexisting dictatorial regimes (Frantz and Geddes 2016). Important theoretical advances were also achieved by disaggregating volatility into its within-system and extra-system components (Powell and Tucker 2014) and by using such measures in global comparisons with vastly expanded numbers of observations (Mainwaring et al. 2017; Mainwaring 2018b, c; Lago and Torcal 2020).

When it comes to regional comparisons of either Latin American or postcommunist countries, however, analyses including later election waves have somewhat muted this theoretical optimism. None of the variables usually theorized to affect both types of volatility returned significant results in postcommunist Europe (Crabtree and Golder 2017) after values of a variable measuring GDP change were corrected from an earlier comparison (Powell and Tucker 2014). Surprisingly, the same null results apply to analyses of Latin American volatility: party replacement volatility (Mainwaring 2018b, 68), conditioning by economic factors of all types of volatility (Frantz and Geddes 2016; Cohen et al. 2018), and overall levels of volatility. Strikingly, none of 15 variables measuring different theorized institutional rules returned statistically significant results (Mainwaring and Bizzarro 2018, 119).

On the other hand, comparing Latin American elections over a longer timespan (since 1932) has identified important determinants of both types of volatility (Mainwaring and Su 2021). Such discrepancies between the global and regional comparisons might be attributable partly to the larger numbers of observations in global analyses, in which substantively smaller differences become statistically significant. The difference from earlier regional analyses is, however, more difficult to explain. It is possible that theoretical development has so far not produced strong comparative accounts of volatility, so the null and weak findings are perhaps suggestive of missing important variables and explanations (Mainwaring and Bizzarro 2018, 131). A similar suggestion, concluding yet another inconclusive quest for uncovering statistically significant predictors—that “this area of research cries out for new theoretical development” (Crabtree and Golder 2017, 233)—should be fully heeded.

Seeking to answer these calls for theoretical innovation, this analysis theorizes the effect of dissolution thresholds on electoral volatility. Recent scholarship has usefully distinguished between two types of volatility. On the one hand, volatility may result from the exit of established parties from electoral competition and the entry of new ones (type A or party replacement volatility), while shifts in vote shares among established parties indicate type B, or stable party volatility (Powell and Tucker 2014; Mainwaring et al. 2017; Cohen 2018). Dissolution thresholds are expected to affect primarily the entry and exit of parties. Stable party volatility (type B) is not expected to be affected by dissolution thresholds. Two principal hypotheses therefore will be tested:

H1 As the size of the dissolution threshold increases, the total party system volatility increases.

H2 As the size of the dissolution threshold increases, the party replacement (type A) volatility increases.

This study proposes three principal causal mechanisms that mediate the effect of demanding dissolution thresholds on volatility. These mechanisms are not mutually exclusive but mutually reinforcing, with one or more operating at the same time to generate volatility. Qualitative evidence in the next section highlights the actual practice of party bans in Latin America after which systematic comparative evidence tests these hypotheses using regression models.

Exit of Banned Parties

The most direct way that high dissolution thresholds fuel volatility is by forcing the exit of existing parties for not reaching the threshold. The number of such parties and their combined vote share vary but may reach substantial levels. For example, following the 1999 elections in Panama, almost half of the country’s parties (5 of 12) were banned for not reaching the demanding dissolution threshold of 5 percent of the vote (Tribunal Electoral 1999). As 3 of those parties obtained over 3 percent of the vote, the total vote share of the 5 banned parties was 12.7 percent. In Guatemala, with one of the highest thresholds, of the 26 parties that competed in the 2019 legislative election, 6 were banned for neither reaching the 5 percent threshold nor securing the minimum of a single mandate (Barrientos Castañeda 2020). These parties concentrated among themselves 8.3 percent of the valid votes (TSE 2020).

In Chile, 17 of the 28 parties that competed in the 2017 election were eliminated in 2018 for not reaching the 3 percent threshold or the 3 elected seats. (Servel 2018). Among themselves, these parties acquired 10.5 percent of the vote. Six of these parties were banned altogether, while the remaining 11 avoided that fate by fusing with other parties and creating new parties. Bans of parties with substantial electoral followings are typical also of another country with demanding dissolution thresholds: Ecuador. The Ecuadorian rule requires the threshold requirement to be fulfilled in two consecutive elections. Not only general elections but also municipal elections (which take place one year after the general ones) are taken into consideration. This meant that following the 2013 general and 2014 municipal elections, 4 parties with a combined total of 8.4 percent of the vote were banned for not reaching the threshold.

Even with lower thresholds, a substantial part of the electorate may find itself unrepresented by parties it voted for. Two viable progressive parties totaling 5.4 percent of the vote were disbanded for not reaching the 3 percent threshold following the 2014 general election in Bolivia (Mayorga 2016). Movement Without Fear (Movimiento sin Miedo), founded by the popular mayor of the capital city, La Paz, and running as a potential challenger and moderate leftist alternative to the governing MAS party of president Evo Morales, received only 2.72 percent of the vote. The Green Party of Bolivia, receiving 2.69 percent, ran as an alliance (CONAMAQ) of environmental activists and Indigenous interests critical of president Morales (Springerová and Vališková 2017). With the same 3 percent threshold, four parties were canceled in El Salvador following the 1997 legislative elections, in which they totaled over 3 percent of the vote (Artiga-González 1997).

In sum, this purely mechanical effect of thresholds ensures that parties not reaching representational requirements are forced out of the party competition. This directly affects parties’ exit from electoral competition, thereby fueling volatility. The effect on volatility is expected to vary positively with the size of the threshold.

Reinventing Banned Parties

Seeing their party banned as a result of not reaching the dissolution threshold, former party activists may choose to create a new party in its stead. In this way, high dissolution thresholds not only directly fuel volatility by affecting the exit of parties but also by promoting the entry of new ones. Latin American parties are often built around strong personalistic figures who use them for the sole purpose of having organizational vehicles in their striving for power. They may dispose of them rather freely if these organizations stop suiting their immediate interests, and create new ones with similar ease.

Alberto Fujimori, the former Peruvian president, famously created different parties for four out of five elections between 1990 and 2000 (Levitsky and Zavaleta 2016, 413). If such personalistic parties are banned for failing to reach the dissolution threshold, they may be reinvented, leaving their original mission and their core of party loyalists largely intact. The Partido Roldosista Ecuatoriano (PRE) in Ecuador is an example of such reinvention. Founded as a personalist electoral vehicle for the erstwhile president Abdalá Bucaram, its largely informal organization was fully subordinate to its “owner” Bucaram and his family members and business associates (Freidenberg and Levitsky 2007). Unable to consolidate its party organization and with declining electoral fortunes after Bucaram’s impeachment in 1998, the party failed to reach the threshold of 4 percent votes (or 3 MPs) in two consecutive elections and was dissolved in 2014. However, that very year the party reconstituted itself under a different name, Fuerza Ecuador, with the same purpose (El Comercio 2014). Bucaram’s son Abdalá Jr. ran unsuccessfully for president with this party in 2017, receiving only 4.8 percent of votes.

For other caudillos, such reinvention of parties is cyclical, as for Álvaro Noboa, the owner of Ecuador’s largest banana enterprise, the wealthiest Ecuadorian, and a runner-up in the runoffs of three presidential elections (1998, 2002, and 2006). In 1998, Noboa ran for the presidency as a candidate of the PRE (Bucaram was then in exile in Panama after his ouster from the presidency). However, for the 2002 election, Noboa created his own electoral vehicle, PRIAN. After securing the largest congressional bloc in 2006 with 28 out of 100 seats, PRIAN’s electoral fortunes dissipated (Pachano 2010). Failing to reach the threshold in two consecutive elections (in 2013 and 2014), the party was dissolved in 2014. Noboa created another new party and even used the slightly revised slogan of the extinct PRIAN (¡Adelante Ecuador, Adelante!) as the name for this new organization to underscore its symbolic continuation. After failing to pass the threshold in two consecutive elections, this second party too was dissolved in 2020 (El Comercio 2020).

Regardless of whether they target personalistic parties or other types, dissolution thresholds force organizational discontinuity on parties with a strong mission and a survivalist raison d’être whose election result falls short of the legal barrier at some point in time. Thresholds maintain such organizations in a constant cyclical process of (re)foundation, effectively deinstitutionalizing them and thus weakening the party system as a whole. Instead of banning only ephemeral parties, such cyclical reinvention means that even parties capable of gradual organizational consolidation (and perhaps substantial electoral gains) are eliminated from the electoral playing field, and new and weaker ones enter it.

Generating Uuncertainty in the Party System

Two types of uncertainty, both potentially noxious for party system stability, are generated by dissolution thresholds. Both voters and political parties are expected to coordinate and adapt their behaviors to overcome the negative consequences of this uncertainty. The first type concerns the strategic calculation of both parties and voters, who are likely to adjust their behavior when facing possible party bans as a result of dissolution thresholds. Parties that are in danger of not reaching the threshold in upcoming elections are likely to coordinate with other parties to avoid dissolution and to form preelectoral coalitions or fuse together with other parties.

In this vein, research into preelectoral coalitions has shown that higher effective thresholds of electoral systems greatly increase the probability of such coalitions as parties seek to overcome some barrier of representation (Golder 2005), and research into party mergers confirms that it is more likely to occur when their size is close to the electoral threshold (Ibenskas 2016a). Analogously, high dissolution thresholds are expected to foster preelectoral coalitions and party mergers for the same reason. But such mergers and coalitions incur large costs for parties and their electoral performance. These are not only transaction costs stemming from bargains on joint policy, conflict resolution mechanisms, or the division of leadership offices, but also costs related to the uncertainty over the preferences of potential coalition partners and the risk of their opportunistic behavior (Ibenskas 2016a).

Furthermore, party elites face increased uncertainty because they do not know how voters will respond to such coalitions. Voters might feel that their views are not sufficiently represented in such coalitions (Gschwend and Hooghe 2008). Voters who would have voted for individual parties are likely to defect from incompatible coalitions (Ibenskas 2016b). Ultimately, high dissolution thresholds distort the coherence of political alternatives because they make parties enter pragmatic political alliances without many ideological or programmatic commonalities—an evolution that characterized, for example, the Colombian party system after it substantially increased its dissolution threshold (Arévalo et al. 2013, 262). Preelectoral coalitions also dilute party brands and undermine party building. Both these processes directly affect the overall party system volatility.

The second type of uncertainty concerns the “imprecision with which actors are able to predict future interactions” (Lupu and Riedl 2013, 1344). The implementation of the dissolution thresholds for any given party is unpredictable because it may be subject to ex post judicial interpretation or even political considerations. This creates uncertainty among party activists and voters, potentially shifting their votes and resources to other parties and weakening the parties organizationally and electorally. Parties often contest dissolution, with courts deciding whether their juridical personhood should be retained or not. Banning parties is thus not an automatic process but often a long-term litigation exercise that can take years to resolve, with conflicting decisions of different judicial authorities. Furthermore, even if the legal process fails to ban the party, merely the context of those legal proceedings affects the party system, as it increases the party’s public exposure and reinforces its antiestablishment critique (Casal Bertoa and Bourne 2017).

In any case, this protracted uncertainty is likely to affect the calculations of both party activists, who are less likely to invest their time, resources, and efforts into a possible lost cause of party building, and voters, who may be dissuaded from voting for a party facing the imminent threat of dissolution. Both these processes trigger vote shifts to other parties. Instead of supporting the creation and gradual institutionalization of viable parties, threshold requirements undermine them and generate instability.

To understand the extent of such uncertainty and the large size of the electorate potentially affected by it, consider the situation following the 2003 legislative and 2004 presidential elections in El Salvador. The process of the dissolution of three parties was initiated by the Supreme Electoral Court in 2004 following their poor result in that presidential contest. Two of them, Partido Demócrata Cristiano (PDC) and Centro Democrático Unido (CDU), were banned because their common presidential candidate received only 3.9 percent of the vote, falling short of the 6 percent threshold required of party coalitions. The candidate of the third banned party, Partido de Conciliación Nacional (PCN), received 2.71 percent of the vote, only seven thousand votes shy of the 3 percent dissolution threshold for single party candidacies.

None of these were marginal parties at the time, and all were represented in Congress following the 2003 legislative election, which consolidated among them 26.7 percent of the votes and 26 of the 84 (i.e., 31 percent) legislative seats. PCN received 13.0 percent of the votes and 16 MPs, PDC 7.3 percent of the votes and 5 MPs, and CDU 6.4 percent of the votes and 5 MPs. CDU did not appeal this ruling and was effectively disbanded in 2004, after which some of its erstwhile activists founded another party, Cambio Democrático. PDC and PCN appealed the ruling, and the Constitutional Chamber of the Supreme Court of Justice ruled in their favor and stopped the dissolution proceedings the same year.

Both parties were revived in 2005 as a result of a legislative decree, and were therefore allowed to run in both the 2006 and 2009 legislative elections. However, the Supreme Court of Justice declared the 2005 legislative decree unconstitutional in 2011, thereby disbanding both parties, which accordingly could not compete in the upcoming 2012 legislative election (Labrador Aragón 2011). Activists from both parties quickly registered new parties—former PDC members reconstituted themselves in the Party of Hope (receiving 2.74 percent of the vote) and PCN members in the Party of National Coalition (receiving 7.24 percent of the vote). Appealing the dissolution, the Administrative Chamber of the Supreme Court of Justice declared the ratification of the parties’ dissolution illegal on procedural grounds, only to be overturned by a ruling of the Supreme Electoral Court in October 2015 that finally ratified the parties’ proscription (La Prensa Gráfica 2015).

The dissolution threshold and the 11-year judicial process targeted parties supported by a large share of the electorate (over one quarter). What is more, it also ruptured the organizational continuity of the two longest-enduring—by then quinquagenarian—parties present in Salvadoran politics. In this way, dissolution thresholds are likely to propel party deinstitutionalization and electoral volatility.

The political uncertainty caused by such unpredictability is inimical to party development in new democracies. Parties cope with the uncertainty by developing flexible, fluid, and uninstitutionalized organizations (Lupu and Riedl 2013, 1349). Voters, for their part, respond to political uncertainty by developing only weak partisan attachments and perceptions of party reputation. This makes it difficult for voters to make informed electoral choices and for parties to make credible commitments to voters (Brader et al. 2013; Lupu and Riedl 2013). As newly merged and weakly institutionalized parties and pragmatic coalitions emerge and party-voter linkages weaken under conditions of uncertainty created by dissolution thresholds, electoral volatility results.

Testing the Model

To provide a systematic test of the main hypothesis, the regression models in this study rely on a dataset covering the period between 1980 (or transition to democracy) and April 2021 in all 18 Latin American countries with competitive elections. Observations in which countries experienced a fully authoritarian regime were excluded from the dataset, so only competitive regimes (democratic or semidemocratic) are compared. This exclusion follows the operationalization and coding of Latin American political regimes by Mainwaring and Pérez-Liñán (2013, 67–68). ³

The models include three dependent variables: the total electoral volatility and its two constituent dimensions, party replacement volatility and stable party volatility. Data on electoral volatility are taken from the Latin American Presidential and Legislative Elections (LAPALE) database (Cohen et al. 2018). ⁴

The measures used were those that calculate party vote shares from the valid votes received in legislative elections. This constitutes the most extensive and reliable dataset on volatility in Latin America, with the additional advantages of providing calculations based on all parties that contested the election (not only parties above a certain percentage threshold) and of also providing data on the two constituent dimensions of volatility—party replacement volatility (type A) and stable party volatility (type B).The former is calculated as shifts in vote shares among parties (w) that newly entered competition at election in time t and those that exited competition (o) between time t –1 and t (Powell and Tucker 2014).

$$Party\;Replacement\;Volatility = \;{{\left| {\mathop \sum \nolimits_{o = 1}^n {p_{o\left( {t - 1} \right)}}\; + \;\mathop \sum \nolimits_{w = 1}^n {p_{ot}}} \right|\;} \over 2}$$

Stable party volatility, on the other hand, calculates vote swings from vote shares of parties that were contesting the election in both t –1 and t.

$$Stable\;Party\;Volatility = \;{{\mathop \sum \nolimits_{i = 1}^n \left| {{p_{i\left( {t - 1} \right)}} + {p_{it}}} \right|} \over 2}$$

By definition, summing up both types of volatility yields the conventional Pedersen index of total volatility (Pedersen 1983). This ranges between 0 percent (no change between elections) and 100 percent (all votes were distributed across different parties).

Independent Variable

The main independent variable (Size of the dissolution threshold) is operationalized as the minimum percentage of valid votes required of parties to avoid their proscription. The data for measuring the threshold size were coded based on the party and electoral laws listed in the online appendix. Not all countries specify their thresholds in their laws in terms of the share of valid votes. If thresholds were legally specified in some other manner, the operationalization of the threshold magnitude recalculated them into the valid vote shares to offer the best possible estimates, thus achieving a broadly comparable measure. The following four operational decisions had to be taken.

First, valid votes are most frequently used by the party laws in order to determine the threshold, but they are not the only method. Thresholds derived from the number of registered voters (i.e., citizens eligible to vote) are used in Argentina. Furthermore, some laws derive the threshold from all votes cast, which also includes blank and null ballots (Bolivia until 1998, Panama in the 1980s). Since invalid voting is pervasive in Latin American elections (Kouba and Lysek 2016), this distinction is consequential. Furthermore, some laws stipulate surpassing a minimum requirement of a set absolute number of votes (Uruguay, Costa Rica, Honduras, Colombia). When registered voters or votes cast or absolute numbers of votes were stipulated by law, the threshold was recalculated to the percentage of valid votes in the corresponding election in order to make the measure comparable. When the law stipulated that the percentage of valid votes in an election was the sole legal criterion for party dissolution due to thresholds, this was used directly.

Second, some countries set their thresholds to avoid dissolution not by a required vote share but by stipulating that a certain minimum number of legislators from the party be elected. To make such provisions comparable with the measure using valid votes, this threshold was recalculated into the estimate of the number of valid votes necessary for a party to obtain this minimum number of legislators in the corresponding election. This is necessarily an estimate because parties could have received fewer votes and still have received the minimum number of seats, or they could have received more votes but failed to secure those seats. The operationalization recalculates this based on the estimate of an effective threshold in terms of valid votes. The average vote share of parties that received the lowest possible number of MPs to avoid cancellation is used as a proxy for such a threshold. If no parties received exactly the required minimum of legislators, the estimate calculated the percentage share of this minimum number of legislators based on the total number of legislators.

Third, some countries require for one of the two aforementioned types of thresholds to be passed. When this was the case, the lower of the two thresholds (the threshold based on the share of valid votes or the threshold based on the votes for a set number of legislators) was selected to obtain a conservative estimate.

Fourth, three countries, Chile, Argentina, and the Dominican Republic, use thresholds that need to be reached only in a specified number of subnational units. To make this comparable to other nationwide thresholds, the estimates of these thresholds were recalculated into the valid votes shares necessary at the national level. To operationalize the Argentine threshold (at least 2 percent of registered voters in at least one district), the total number of registered voters in each election was divided by the total number of electoral districts (provinces). Two percent of valid voters in such an average-sized district was taken as the dissolution threshold. As a consequence, this threshold varies somewhat between elections but is very low (about 0.1 percent) in general.

For Chile, the threshold of achieving at least 5 percent of valid votes in at least 8 regions or 3 contiguous regions was calculated in the following way. The share of votes in 8 average regions and 3 average regions, respectively, was calculated (there were 13 regions prior to 2007 and 15 regions after the 2007 reform). Both shares were multiplied by the 5 percent threshold, and the average values of both were used as the threshold. This follows the logic that a party may be too small to pass the threshold in all 8 regions but may have enough regional-based support to pass the threshold in 3 contiguous regions.

Control Variables

The models control for economic and political variables whose effects have been identified by existing research. Short-term economic performance was measured using GDP growth. Following the practice of Mainwaring et al. (2017, 628), the measure tracks the average annual change from the year of the first election in the electoral period to the year before the second election. The variable is based on data from the Quality of Government dataset (Teorell et al. 2021).

Overall economic development is proxied by a logged measure of the GDP per capita (IMF 2021). Two institutional variables are included. Whether the legislative elections were held concurrently with the presidential ones is measured by a dummy variable, Concurrent election. The number of years separating the two legislative elections is also included in the models (Number of years between elections). Furthermore, the models also control for the Age of democracy (in years) from the birth year of democracy to the election. The birth year was identified as the year of the transition to democracy based on an earlier coding (Schenoni and Mainwaring 2019).

A measure of Ethnic fractionalization with data on Latin American countries from an earlier analysis (Birnir and Van Cott 2007) controls for the effects of ethnic diversity that have been theorized to impact electoral volatility. A potential confounder concerns the approach to banning parties by the preexisting authoritarian regimes. Some of them proscribed parties altogether, while some of them created their own support parties. This legacy of past dictatorships was shown to affect volatility during the democratic period (Frantz and Geddes 2016). At the same time, such an institutional legacy could impact the policies of the new democratic regimes regarding party bans. To account for this effect, a dummy variable, Past party repression, was created based on the coding of how the most recent dictatorial regime approached party banning (Frantz and Geddes 2016, 31). Party repression by the dictatorial regime was coded as 1, but if these dictatorships allied themselves with traditional parties or created new ones, they were coded as 0. Descriptive statistics for all these variables are included in Table 2.

Results and Data Analysis

Inferences based on cross-section time-series datasets are notoriously dependent on modelling choices. It is therefore appropriate to compare results using different estimation techniques. Three principal ones have been used to understand the determinants of electoral volatility in different world regions with comparable data structures, numbers of panels, and time periods.

First, in order to deal with unobserved heterogeneity in the standard errors within countries, researchers have relied on models that cluster robust standard errors by country in Latin American (Cohen et al. 2018) and postcommunist regional comparisons (Powell and Tucker 2014; Crabtree and Golder 2017). Second, other analysts have instead used regressions with panel corrected standard errors to assess determinants of volatility in postcommunist Europe (Tavits 2008) and Africa (Kuenzi et al. 2019). This addresses heteroskedasticity and contemporaneous correlation of the error terms in pooled datasets. Third, scholars have estimated the determinants of volatility through generalized estimating equations (GEE) which are more appropriate for datasets containing more units than time periods (Mainwaring et al. 2017; Mainwaring and Bizzarro 2018). This makes for a demanding estimation setting in which it is difficult to obtain statistically significant results.

Table 3 reports results of these three estimation procedures. Separate models for all three dependent variables (party replacement volatility, stable party volatility, and total volatility) are presented here. Both the models with panel corrected standard errors and those using GEE are estimated with the AR(1) autoregressive correlation structure because the dependent variables are assumed to be correlated over time and this correlation is expected to be larger for consecutive periods.

Table 2. Descriptive Statistics

Table 3. Models for Party Replacement Volatility, Stable Party Volatility, and Total Volatility

*** p<0.01, ** p<0.05, * p<0.1

Robust standard errors clustered by countries in parentheses (models 1–3 and 7–9)

As expected, the coefficients of the threshold rules in the models of stable party volatility do not reach conventional levels of statistical significance. However, both the party replacement volatility and the total volatility are consistently associated with the size of dissolution thresholds across all three types of models. This suggests that the relationship between dissolution thresholds and the aggregate party system instability operates through the entry of new parties and the wholesale exit of existing ones.

The estimated contribution is not negligible, considering the range of the dissolution threshold variable. For every 1 percentage point increase in the size of threshold, the total volatility (and the party replacement volatility) is predicted to increase by 3.1 points on average. This is a moderate estimated contribution, considering that most reforms to dissolution threshold rules have been incremental and have often increased (or lowered) the threshold by roughly 1 percentage point. However, the total elimination of such thresholds in countries with intermediate or high threshold sizes would probably be felt throughout the party system. For example, the elimination of dissolution threshold rules in countries with an intermediate threshold size—such as the 3 percent rule in Bolivia, Mexico, and Colombia—would help the stabilization of the party system commensurate to avoiding a sudden electoral disappearance of a party that received 9.3 percent of the vote in the previous election.

The online appendix presents three sets of robustness checks. First, models using an instrumental variable regression framework are presented (A.2). The effect of threshold size on all three types of volatility is instrumented by a variable that measures the frequency with which the threshold size rules were reformed. Second, the original models were rerun using jackknife resampling in order to reveal the sensitivity of results to the presence of potentially influential cases (A.3). The online appendix also identifies and discusses such influential observations noting that their exclusion does not change the interpretation of the main findings. Third, the original models were rerun with the addition of a dummy variable indicating the presence of a subnational dissolution threshold in some countries (A.4). The presence of subnational thresholds may create different incentives for parties to stay in competition, as opposed to rules that do not require the parties to pass the threshold only on a part of the territory. In general, these robustness checks support the main findings from the original models.

Conclusions

Concluding that institutional design has had a limited impact on party-building outcomes in Latin America, Levitsky et al. (2016, 9–10) suggest looking for other types of explanations. This skepticism is echoed in findings of a lack of meaningful association between crucial electoral-institutional rules and volatility (Mainwaring and Bizzarro 2018). The relationship between dissolution thresholds and volatility, however, suggests an effect of a specific institutional rule. Institutional design might influence party building and party demise more powerfully than is assumed by existing research.

The evidence presented here suggests that electoral volatility is promoted by bans, or the threat of bans, of parties, activated by high dissolution thresholds. Stable party volatility is unaffected, but the effect on the overall party system stability operates through party replacement volatility. The size of the threshold regulates the percentage share of votes received by new parties, as well as parties that exit the electoral arena. This argument bears a straightforward institutional prescription: in order to help the stabilization of Latin American party systems, countries should significantly lower or outright eliminate party dissolution thresholds. Not only do such legal barriers not help stabilize and simplify the party universe, they actually imperil its stabilization and institutionalization. To the extent that unstable party systems imperil democratic governance (Roberts and Wibbels 1999; Weyland 1999; Mainwaring 2018c; Carreras and Acácio 2019), high dissolution thresholds are a problematic feature of party legislation.

Banning parties is a grave act, “one of the harshest measures a democracy can take to protect itself” (Bourne 2018, 43). Banning a party is an authoritarian precaution used in the name of democracy because it severely curtails the rights of political association and expression (Celep 2014, 373). To the extent that such a measure is justifiable on democratic grounds, it should be used solely to protect the core principles of democracy itself. This constitutes the basic tenet of the “militant democracy paradigm,” which has justified the proscription of antidemocratic parties in order to prevent them from coming to power, implementing their antidemocratic agenda, and abolishing democracy altogether (Bligh 2013; Bourne 2018, 30). Banning parties for not reaching a certain quantitative threshold of electoral support hardly passes any of the stringent tests developed to justify party proscription in this paradigm.

Party dissolution thresholds are a blunt instrument of electoral engineering. If instituting them at all, policymakers should carefully balance them with other basic requirements of democratic representation on which these thresholds impinge. There can be little doubt that, by itself, the small and medium electoral strength of parties that are banned for not reaching some arbitrary vote threshold does not constitute a threat either to the constitutional order or to the citizens’ fundamental rights and freedoms. If reformers aim to avoid the representation of such parties, they should consider manipulating other electoral system parameters that are known to affect the disproportionality of electoral systems, such as district magnitude or representation thresholds. Contrary to the operation of dissolution thresholds, such rules do not eliminate the organizational continuity of minor parties, which are allowed to continue functioning and may succeed in the following election. In this way, electoral system disproportionality does not necessarily harm party system stability to the degree that party dissolution does.

Raising legal thresholds of representation (not dissolution) might reduce (or be neutral to) party system instability while at the same time might avoid some of the normatively unappealing consequences of dissolution thresholds, such as diminished accountability and pluralism. Paradoxically, Latin American electoral legislation in general sets higher party dissolution thresholds than legal barriers of representation, which are more typical of European electoral legislation.