Although the impact of the district magnitude (DM) on the number of electoral parties and parties with legislative representation has long been established, studies on the incentives posed by the electoral rules have often overlooked the response by party subunits. We study Colombia to assess how the response of parties and factions to different sets of electoral rules. In the second half of the twentieth century, Colombia had one of Latin America's most stable and enduring two-party systems, comprised by the Liberal (PL) and Conservative (PC) parties. In 1957, a political agreement, the National Front (FN) (1958–1974), established that both parties would alternate in the presidency and share government and legislative positions equally. When the FN ended in 1974, new rules allowed for other parties to win seats in the legislature. When the 1991 constitution was enacted, the electoral rules were conducive to a multiparty system. We systematically analyse district level results to evaluate the relationship between the DM and the number of parties and party lists (factions) that competed in each district and obtained seats in the legislature. We postulate two hypotheses linking the DM to the effective number of parties and factions (ENP and ENF, respectively) that compete in elections and that win seats in each electoral unit. After explaining the methodology, we present data on the parties and party factions (or lists) competing in legislative elections and the number of parties and factions represented in Congress. We then present our statistical models, discuss the results and the implications of our findings.
1. The impact of DM on the number of parties and factions
Cox (Reference Cox1997) asserted that the electoral system functions as a straitjacket that establishes an upper limit on the number of parties in each electoral unit. Other studies have qualified those findings by arguing that relationship between electoral rules and the party system is a two-way street (Sartori, Reference Sartori1997; Colomer, Reference Colomer2003). As Benoit (Reference Benoit2007: 387) puts it, most of the time ‘changes to the rules of the game are motivated by the play of the game itself, but in some cases, it might be the referee, the spectators, or even an external body changing the rule book.’
The literature shows that ‘minor changes to electoral laws impact party systems,’ although that effect might be more short-term than lasting (Ward, Reference Ward2019: 403). The DM – the number of seats in an electoral unit (Rae, Reference Rae1967) – impacts the number of parties in competition (Taagepera and Shugart, Reference Taagepera and Shugart1989; Singer, Reference Singer2015; Lucardi, Reference Lucardi2019; Riera, Reference Riera2020; Fiva and Hix, Reference Fiva and Hix2021). The impact of DM on the party system might be mediated by the size of the legislature. As it increases, so does the average DM magnitude and the number of parties with seats in representative bodies (Shugart and Taagepera, Reference Shugart and Taagepera2017; Lewis, Reference Lewis2018).
The DM influences the formation of parties and their coordination (Laakso and Taagepera, Reference Laakso and Taagepera1979; Mainwaring and Scully, Reference Mainwaring and Scully1995; Benoit, Reference Benoit2007, Fiva and Hix, Reference Fiva and Hix2021). The DM affects all proportional systems, including those with open lists (Katz, Reference Katz, Grofman and Lijphart1985) and closed lists (Carey and Shugart, Reference Carey and Shugart1995). In closed-list systems, individual reputation increases in importance as the DM decreases (Carey and Shugart, Reference Carey and Shugart1995; Chang and Golden, Reference Chang and Golden2007). Under open-list rules, competition within parties might allow for the formation – or manifestation of existing – intra-party factions. While open-list arrangements induce the cultivation of a personal vote, single non-transferable vote (SNTV), as it was the case in Colombia under the FN, fosters the formation of factions that can pool votes. When SNTV is used in districts with higher magnitude, each faction can aspire to win more than one seat and, thus, the competition does not solely occur between parties, but also between intraparty factions (Shugart and Taagepera, Reference Shugart and Taagepera2017: 215–235).
The literature, however, has overlooked the effect of DM on party subunits. If the DM impacts the behaviour of parties, it should also impact the behaviour of party factions. As non-monolithic units, political parties involve a plurality of actors between whom there is tension, interaction, and conflict (Kitschelt, Reference Kitschelt1989). Factions, which can be any combination, clique, or group whose members share a sense of identity and purpose and act collectively as a bloc (Zariski, Reference Zariski1960: 33). Others define a faction as a relatively organized group that competes with rivals for power (Beller and Belloni, Reference Beller, Belloni, Belloni and Beller1978). Some stay away from the use of factions and instead refer to party subunits as agents (Morgenstern, Reference Morgenstern2003). The analysis of parties is more accurate when it includes the subunits – the factions – that comprise the party (Sartori, Reference Sartori2005: 93).
Discussing party subunits, Sartori identifies ‘a threefold terminological articulation: fraction (the general, unspecified category), faction (a specific power group), and tendency (a patterned set of attitudes). In this mapping, a pure faction and a pure tendency represent opposite ends of the sub-party continuum’ (Sartori, Reference Sartori2005: 66). By contrast, Boucek identifies three faces of factionalism (Reference Boucek2009: 456). In the cooperative face, a faction can be instrumental in promoting cross-party cooperation and the construction of integrated parties. In its competitive face, factionalism, or the division of parties into opposing factions, serves as a structure for conflict resolution. Finally, in the degenerative face, factions operate primarily as channels for the distribution of power (Boucek, Reference Boucek2009: 479). Frequent divisions within the factions may be sustained by conflicts between prominent figures in which principles often play a strictly secondary role (Zariski, Reference Zariski1965: 33). Moreover, parties may have pathological characteristics, such as an increase in the number of factions or the accentuation of differences of opinion over time (Pasquino, Reference Pasquino1979: 90). In systems where electoral competition takes the form of intraparty competition, faction leaders cultivate a personal reputation that distinguishes them from the label of the party to which they are affiliated (Carey and Shugart, Reference Carey and Shugart1995; Ansolabehere et al., Reference Ansolabehere, Snyder and Stewart2000).
Discussing the concept stretching on the literature of sub-party units, Emanuele et al. (Reference Emanuele, Bruno and Nicola2023) uses the Sartorian notion of fractions as a generic term that includes the variation of within party groups in terms of structure, attitudes, and stability. Electoral rules, especially disproportionality and the incentives to cultivate a personal vote, might foster the development of fractions. Open-list arrangements might induce the emergence of more fractions within parties as open lists, or preferential voting, give more visibility to candidates and foster the cultivation of a personal vote. In turn, closed lists induce within-party discipline as voters must choose between parties and thus, dissidents have incentives to stay loyal to the party leadership, especially during candidate nomination and election periods (Emanuele et al., Reference Emanuele, Bruno and Nicola2023).
Studies on party subunits have focused on the cases of Japan, Italy, and Uruguay. Corruption scandals and the deterioration of the dominant party brands facilitated the exit of legislators from Japan's Liberal Democratic Party (LDP) and Italy's Christian Democratic Party (DC). This contributed to the entry of new competitors – including splinters from the traditional parties (Boucek, Reference Boucek2009). In Italy, factions exercised significant and growing influence (Zariski, Reference Zariski1965; Pasquino, Reference Pasquino1979), as the electoral system promoted the cultivation of a personal or faction vote (Boucek, Reference Boucek2012: 124). In 1992, an electoral reform that introduced a single vote open-list system led to the DC implosion in 1994 and the collapse of the party system (Morgan, Reference Morgan2011; Boucek, Reference Boucek2012). In Japan, factions within the LDP were the political manifestation of the traditional hierarchical patron-client relationship (Richardson and Flanagan, Reference Richardson and Flanagan1984). Until 1993, the SNTV system with two-seat districts was used in elections for the lower house. Voters voted for candidates, not political parties, and the two candidates who obtained the most votes were returned (Cox et al., Reference Cox, Rosenbluth and Thies2000). The SNTV system fostered factionalism (Christensen and Johnson, Reference Christensen and Johnson1995) which resulted in systematic and organized intraparty competition, with longstanding factions (Cox and Rosenbluth, Reference Cox and Rosenbluth1996; Cox et al., Reference Cox, Rosenbluth and Thies1998, Reference Cox, Rosenbluth and Thies2000).
In Uruguay, factions are institutionalized and stable agents within parties with a legal status and tools and other mechanisms that allow them to nominate candidates to the legislature (Buquet et al., Reference Buquet, Chasquetti and Moraes1998, Morgenstern, Reference Morgenstern2001). The electoral system based on closed lists and Double Simultaneous vote (DSV) gives faction leaders control the nomination process within parties (Moraes, Reference Moraes, Siavelis and Morgenstern2008). Because they control the candidate nomination process and the elections use closed lists, faction leaders effectively influence the behaviour of legislators in roll key call votes, as they can punish those legislators that deviate from the faction official positions (Buquet et al., Reference Buquet, Chasquetti and Moraes1998).
In the three cases, changes to the electoral rules altered the behaviour of party factions. In Italy, when the dominant party disappeared, its factions became new parties, producing a realignment of the party system. By contrast, in Japan, the dominant party learned to coexist with the factions that split off from it. In Uruguay, factions have been central to the operating of the party system and parties themselves see factions as central to their own existence.
Since factions allow for different political expressions to compete within a party, the mechanical effect of DM should be reflected in the number of factions in competition, not just in the number of parties. Under a system whose election laws restrict the formation of new parties, the mechanical effect of DM should be manifested in the number of party factions competing in elections but also with seats in the legislature. Thus, we formally, we postulate two hypotheses:
H1: The higher DM at the electoral unit level, the higher the number of parties in competition and the number of within-party lists in competition.
H2: The higher the DM at the electoral unit level, the higher the number of parties and factions with legislative representation.
2. The party system and party factions in Colombia
The case of Colombia underlines a grey area on the studies of party systems, the role of factions. On the surface, there was a stable two-party system with electoral competition under closed lists. Any faction could run under the dominant Liberal (PL) and Conservative (PC) party labels. Party leaderships had limited power to control within-party competition. In short, what could have otherwise been a fragmented party system was straitjacketed into a two-party system with multiple intra party factions fiercely competing under the umbrella labels of the PL and PC and given the quasi-SNTV arrangement in place.
The PL and PC had dominated the political scene since the mid-nineteenth century and throughout most of the twentieth century. Yet, studies on the Colombian party system underlined the absence of social cleavages identified in institutionalized party systems, although those studies mention the presence of a centre-periphery cleavage (Dix, Reference Dix1989: 24–25) and of a ‘partly ideological, at least in a loose sense’ religious cleavage (Dix, Reference Dix1980: 304). Others highlight the territorial dimension of the liberal/conservative divide (Pinzón de Lewín, Reference Pinzón de Lewín1989; Losada et al., Reference Losada, Giraldo and Muñoz2004).
From the beginning, factionalism was a feature of the party system (Dix, Reference Dix1967; Leal, Reference Leal Buitrago1989). In the 1853 elections, General José María Obando, a liberal, defeated another liberal, Tomás Herrera (Bushnell, Reference Bushnell1994: 154). Similarly, in 1946, reformist presidential candidate Jorge Eliecer Gaitán lost the PL nomination to Gabriel Turbay, a moderate. The division within the PL facilitated the victory of the PC's Mariano Ospina (Bushnell, Reference Bushnell1994: 271). In 1948, the assassination of Gaitán unleashed a violent period, known as La Violencia. In 1954, the conflict between the PL and the PC triggered a coup d’état by General Gustavo Rojas Pinilla. Eventually, the PL and PC signed the Benidorm Pact in 1956 and the Sitges Pact in 1957, paving the way for the FN, which established that the two parties would rotate in the presidency every four years and each hold half of government positions and seats in Congress (Hartlyn, Reference Hartlyn1993). The impact of political parties in the shaping of electoral rules has also been noted by works that use Colombia as one of their cases (Remmer, Reference Remmer2008: 13–14).
Decree 0247, issued in 1957, was the electoral law that marked the start of the FN. The decree established that, until the 1966 elections, all popularly elected positions would be assigned equally in each department to PL and PC candidates (Dix, Reference Dix1967: 134). Other political and social movements from democratic representation could exist, but they could not win seats (Dix, Reference Dix1967). In 1959, Legislative Act 1 extended the FN through to the 1970 elections, implying that it would remain in force until 1974. Under the FN, the number of seats for each department would be proportional to its population, with a minimum of six senators and twelve lower house representatives per unit. In departments with an odd number of seats, an extra seat was added to permit equal division between the two parties. If a party fielded more than one list in an electoral unit, its seats would be distributed among the lists using the Hare quota. The rules fostered competition between factions within parties (Dix, Reference Dix1967: 134; Hartlyn, Reference Hartlyn1988). The factions in each closed list organized themselves around leading figures who determined the order of the candidates on the list and, consequently, each candidate's chances of winning a seat. As each party could field multiple lists competing for the same seats, there was an electoral market where the factions that were already established had advantages (Duque Daza, Reference Duque Daza2019).
As Cox et al. (Reference Cox, Rosenbluth and Shugart1995: 442) argue, the closed-list proportional representation arrangement used in Colombia in the 1960s functioned like Japan's single non-transferable vote (SNTV) where lists run under the same party label and votes were not pooled across lists, as it was the case in Colombia. As parties did not exert control over the use of their label by factions and the largest reminder seat allocation formula allowed factions with a small vote share to clinch seats, intraparty factions multiplied (Cox et al., Reference Cox, Rosenbluth and Shugart1995: 456). This was also due to the relatively high district magnitude in some districts that allowed more factions to aspire to win seats (Wei, Reference Wei2017). National factions were grouped around national leaders while departmental factions functioned as clientelist political machines around members of Congress and governors (Pizarro, Reference Pizarro Leongómez, Alcántara and Ibeas2001; Duque, Reference Duque Daza2006), reflecting a progressive personalization of representation (Gutiérrez-Sanín, Reference Gutiérrez-Sanín2007). The electoral rules were a personal list formula, whereby any candidate could use the party label without requiring the endorsement of party leaders (Carey and Shugart, Reference Carey and Shugart1995: 429). This allowed for factions – some stable and some temporary – to coexist within the PL and PC (Cox et al., Reference Cox, Rosenbluth and Shugart1995; Moreno and Escobar-Lemon, Reference Moreno, Escobar-Lemmon, Siavelis and Morgenstern2008). Ironically, the FN fostered such divisions by eliminating, through its parity and alternation arrangements, the potential penalty of loss of power for the opposition party (Dix, Reference Dix1967: 146). Although the fragmentation ‘offered the party absolutely nothing, dispersion grew very dynamically in both large and small constituencies. […] In 1964, there was hyper-fragmentation also in the small ones’ (Gutiérrez-Sanín, Reference Gutiérrez-Sanín2007: 314–315).
Party affiliation rules were weak. In addition, PL and PC factions tended to be personalist. For example, in 1958, there were Laureanistas, Ospinistas, and Alzatistas in the PC. Moreover, some factions were not confined to a single party. The Liberal Revolution Movement (MRL) and the Popular National Alliance (ANAPO) appeared as dissidences in the PL and PC blocs. In 1960, the MRL, led by Alfonso López Michelsen, joined the PL as a radical dissidence (Ayala Diago, Reference Ayala Diago1995). In turn, ANAPO, led by former dictator Rojas Pinilla, was a mixture of populists, socialists, and figures from the traditional parties, reflecting the imprint of its leader, who was more conservative than liberal (Hartlyn, Reference Hartlyn1993). The leadership of Rojas Pinilla challenged the FN since it brought together various conservatives and liberals (Dix, Reference Dix1980: 315). ANAPO increased its vote share from 3.7% in 1962 to 35% in 1970 (Palacios and Safford, Reference Palacios and Safford2002: 473). In 1964 and 1968, it won 26 seats on both the PC (25) and PL (1) lists in 13 of the 19 departments. In 1966, ANAPO won 37 seats in 15 departments, 33 in the PC and 4 in PL. In 1974, when FN disbanded, ANAPO ran as a unified party.
Some of the factions had a national presence and survived several election cycles. In the PC, the Unionista, Ospinista, Doctrinario Laureanista (followers of Laureano Gómez who died in 1965) and the Alzatistas (followers of Gilberto Alzate Avendaño, who died in 1960) had presence in many departments in several consecutive elections. Although there were fewer factions in the PL, there were clear ideological distinctions between them. The National Revolutionary Movement (MRL) had a soft-liner faction and a hard-liner faction that had communist views. In 1964, when the PL factions were especially active, the dominant PL faction, the Oficialistas, won 58 seats in 18 of the 19 departments, while the MRL soft-liners won 22 seats in 15 departments and the MRL hard-liners won 8 seats in 8 departments. Some legislators often switched allegiances between factions in the same party. ANAPO's Jaime Pidrahita Cardona in the PC and Saúl Charris de la Hoz in the PL were among the legislators that won seats representing different factions within their respective parties. They both ran under the ANAPO party label in 1974 and won seats as well. The increase in personalist factions was already apparent in the 1970 elections when all senators, and 205 of the 210 members of the Chamber of Representatives belonged to some faction. In the 1970 presidential election, Evaristo Sourdis, a conservative running as a dissident and supported mainly by legislators from the Atlantic Coast, challenged the PC's official candidate, Misael Pastrana Borrero.
In discussing the candidate nomination process, highlighted the low entry barriers: “There is no requirement that a candidate be a resident of the department in which he seeks elections; in fact, it is by no means unusual for a prominent individual to run for several offices in various departments at the same time in order to attract votes to the party list, though he may have no intention of serving in all, or even in any of them. The contingency of multiple officeholding is provided for by listing on the ballot alternates (suplentes) equal to the number of candidates. A suplente takes the place of a legislator whenever the latter signifies that he does not wish to occupy his seat, although the legislator can choose at any time to resume his place (Dix, Reference Dix1967: 186). Party leaderships did not exercise strong gatekeeping attributions.
For Gutiérrez-Sanín, ‘each party was a national federation of personalities’ (Gutiérrez-Sanín, Reference Gutiérrez-Sanín2007: 154–155). Discussing the limited capacity of national party leaders to exercise control over the party in the different localities, Hartlyn reported that ‘in some cases, party factions began loosely associated with national party figures, as regional leaders sought to build increasingly more independent power bases of their own, potentially to aspire to national leadership’ (Hartlyn, Reference Hartlyn1988: 88). Low levels of electoral participation ‘enhanced the role of regional party leaders and machine style politics. In that context, it also became increasingly more difficult to develop national political figures’ (Hartlyn, Reference Hartlyn1988: 146). As a result, the elections saw an ‘increase in the number of departments in which uninominal lists, often with strengths in particular regions or municipalities, have been victorious. This phenomenon has reflected an individualistic political strategy of developing a dependable captive electorate and illustrates the strengthening of electoral power by regional party leaders and their increased autonomy from national leaders’ (Reference Hartlyn1988: 162). In that sense, the party was a loose label that allowed almost anybody at the departmental with sufficient resources to field a list of candidates in any given district.
The importance of individual leaders in attracting support to the lists can be demonstrated by the presence of politicians who won seats in different departments, especially in the PL. Alfonso López Michelsen won a seat running on the MRL faction in 1962 in the Department of Valle del Cauca. Two years later, López Michelsen ran again, winning a seat in Cundinamarca. In turn, PL's José Ignacio Vices was elected senator in 1966 in La Guajira. In 1970, running in the Rojista faction of the PL, he won a Senate seat in Magdalena. Darío Echandía won a seat in the Senate, representing Tolima on the PL ticket in 1962 and 1966. In 1970, Echandía won a Senate seat representing Cundinamarca. In the PC, Gerardo Candamil Gómez won a seat in the Chamber of Representatives in 1968 in Cundinamarca as a candidate in the ANAPO faction. In 1970, he won a seat in the Chamber of Representatives representing Putumayo, running as a Rojista candidate in the PC.
The dismantling of the FN began with the Constitutional Reform of 1968 that was implemented in 1974. It strengthened the executive, took functions away from the legislature, and altered the number of seats in both houses (Pérez Rivera, Reference Pérez Rivera1988). Since 1974, each department had two seats in the Senate and one additional seat for every 200,000 inhabitants or fraction over 100,000 and, in the Chamber of Representatives, two seats and one additional seat for every 100,000 inhabitants or fraction over 50,000. The number of seats in the Chamber dropped from 210 in 1968 to 199 in 1974 – that number remained unchanged until 1990. The reform also expanded representation in the lower house from sixteen departmental districts in 1958 to twenty-six as from 1974 while, in the case of the Senate, the number of departments rose from fifteen to twenty-two. The average DM in the Chamber decreased from 8.9 in 1968 to 7.7 in 1974, while in the Senate it remained at 5. The range of DM passed from 2 to 24 in the Chamber to 2 to 29 and, in the Senate, it went from 4 to 12 to 2 to 15 in the same period. The term of members of the Chamber was increased from two to four years and indefinite re-election continued to be permitted in both houses.
The 1968 reform, therefore, meant that, as from 1974, the factions could choose between remaining within the PL or the PC or seek to establish themselves as parties. ANAPO and the leftist National Opposition Union (UNO) became autonomous parties and obtained fifteen and five seats, respectively, in the Chamber of Representatives and seven and two seats in the Senate in 1974.
After 1974, the political system remained restrictive (Bejarano and Pizarro, Reference Bejarano, Pizarro Leongómez, Hagopian and Mainwaring2005; Wills Otero, Reference Wills Otero2011). In theory, the system fostered competition between parties, but competition continued to be within the parties (Hoskin and Pachón, Reference Hoskin, Pachón, Caballero, Pachón and Losada2012) as it had been under the FN. Indeed, Gutiérrez-Sanín (Reference Gutiérrez-Sanín2007) shows that, at the end of the 1970s, ‘the number of factions was increasing, and the liberals were disputing regional dominance with unprecedented ferocity’ (Gutiérrez-Sanín, Reference Gutiérrez-Sanín2007: 172). Until 1991, each party could present one or more closed lists and, basically, each list functioned as a party (Botero et al., Reference Botero, Losada, Wills-Otero, Alcántara and Freidenberg2016: 347).
The institutional changes introduced at the end of the 1980s accentuated intraparty factionalism and encouraged the emergence of political movements with their own legal personality, led by regional leaders who nonetheless maintained ties with the labels of the traditional parties (Duque, Reference Duque Daza2006: 177). Although the end of the FN removed the two-party straitjacket, only some factions opted to become political parties between 1974 and 1990. Others continued to compete within the PL or the PC by forming closed lists that enabled them to maintain their identity and, potentially, benefit from surplus votes obtained by other factions within their respective parties. Since 1991, the electoral law made it easier for factions to obtain legal status as parties. Articles 103 to 112 of the 1991 constitution specify that, to form, a party must gather 50,000 signatures and must receive at least 50,000 votes in the election or win seats in Congress to keep its legal status. With more parties competing, electoral volatility increased.
Under the FN, the competition was limited to two parties, but factions freely operated as such within the respective parties. Drawing on Sartori's (Reference Sartori2005) distinction between factions and fractions, Pizarro Leongómez (Reference Pizarro Leongómez, Alcántara and Ibeas2001) suggests that parties were structured along national fractions and departmental factions, suggesting that while there were clear district-based dynamics that affected the number of lists that competed in each party, there were also national level dynamics that accounted for the existence of national factions (fractions, in Leongómez's words) that had their own national leader. District-level factions were clientelist structured set up by representatives, senators, and governors (Duque Daza, Reference Duque Daza2006) and their presence reflected a progressive personalization of political representation (Gutiérrez-Sanin, Reference Gutiérrez-Sanín2007).
The electoral rules created after 1990 incentives for the creation of electoral microenterprises based on candidates’ personal appeal (Pizarro, Reference Pizarro Leongómez2002). Some authors argue that Colombia moved towards a new (im)balance defined as an attenuated and atomized two-party system (Pizarro, Reference Pizarro Leongómez2002), the moderation of a previously extreme two-party system (Wills Otero, Reference Wills Otero2011; Botero et al., Reference Botero, Losada, Wills-Otero, Alcántara and Freidenberg2016), and even a de-institutionalized party system (Albarracín et al., Reference Albarracín, Gamboa, Mainwaring and Mainwaring2018). Since 1991, the PL and the PC saw a drop in their seats in the legislature and the number of lists increased, presumably deepening multipartism (Pizarro, Reference Pizarro Leongómez, Alcántara and Ibeas2001; Duque Daza, Reference Duque Daza2006) and delegitimizing the system (Wills Otero, Reference Wills Otero2011).
Using the factions as the unit of analysis, we assess the impact that changes in the DM had on the behaviour of factions within the political parties under and after the FN. After all, a larger DM should create more incentives for additional factions to enter the competition in a quasi-SNTV arrangement, despite the fact that votes can only be pooled within factions and not-across different factions in the same party (Wei, Reference Wei2017). Since we cannot assess whether lists were long-term or short-term factions, personalist efforts or organized groups, we assume each list to be a faction. To be sure, not all factions were equally organized and structured, but the same applies to political parties in studies that use parties as the unit of analysis. Table 1 shows the party composition of the Colombian Congress between 1958 and 1990. As is clear, the number of legislators elected for parties other than the PL and the PC increased as from 1974, but the two traditional parties continued to hold a majority of seats, even though, as shown below, the factions within them remained very active.
Table 1. Seats by party in legislative elections in Colombia, 1958–1990
Source: Compiled by authors using Nohlen (Reference Nohlen2005) and Colombia's National Civil Registry.
Note: FN period in bold.
3. Methodology
We build on prior works on Colombia by looking at district level data for each party faction. In his seminal work, Dix (Reference Dix1967) reported the party and faction of affiliation for legislators in both chambers at the district level. Others have shown results at the departmental level (Carey and Shugart, Reference Carey and Shugart1995; Gutiérrez-Sanín, Reference Gutiérrez-Sanín2007; Moreno and Escobar Lemmon, Reference Moreno, Escobar-Lemmon, Siavelis and Morgenstern2008; Duque Daza, Reference Duque Daza2019), but insofar as we are aware, nobody has used departmental level legislative elections during and after the FN to estimate the effect of DM on the number of lists running in each unit.
Our unit of analysis is the electoral unit – departments in the Senate and districts in the Chamber of Representatives. There were up to thirty-three electoral districts (or departments, in most cases) for the twelve elections for the Chamber of Representatives and up to twenty-four for the nine elections for the Senate that took place between 1958 and 1990. In total, we have 444 observations.
The dependent variables for Hypothesis 1 are the number of electoral parties and the number of lists (factions), and, for Hypothesis 2, the number of parties and lists represented in Congress. We used the database of electoral results of the Centro de Estudios sobre Desarrollo Económico (CEDE) of Colombia's Universidad de los Andes at the level of candidate and electoral unit for all elections. We calculated the number of parties and within-party lists for eleven of the twelve elections for the Chamber of Representatives and the nine elections for the Senate. We estimated the ENP and ENF competing in each electoral unit in each election. Table 2 shows the number of parties and lists in competition and the number of factions that won seats under and after the FN.
Table 2. Number of parties and lists in legislative elections in Colombia, 1958–1990
a This database does not include the 1960 election for Representatives. In 1958–1970, Chamber of Representatives elections occurred on a 2-year calendar.
Source: Compiled by authors using data from the Centro de Estudios sobre Desarrollo Económico (CEDE) https://datoscede.uniandes.edu.co/es/catalogo-de-microdata.
Note: FN period in bold.
To test Hypothesis 2, we identified the number of parties and lists with seats. For the 1958, 1960, and 1982 elections in the Senate and Chamber of Representatives, we used Nohlen (Reference Nohlen2005). As from 1974, this data shows all the parties that competed, without registering factions within the PL and the PC. For elections between 1962 and 1990, we used the Electoral Organization and Statistics reports of the National Civil Registry, which provide disaggregated election results at the departmental level, identifying factions that competed within the PL and the PC under the FN. As from 1974, they include not only the party for which legislators were elected, but also whether they belonged to specific lists.
There are some inconsistencies between the Civil Registry data and Nohlen (Reference Nohlen2005). For example, Nohlen mentions an Alzaospinistas faction within the PC in the 1962 Chamber of Representatives election while the official records refer to a Unionista faction. Second, also in the Chamber in 1964, Nohlen records five PL and three PC factions, but the election report shows seven PL and five PC factions. Third, Nohlen reports a Línea Blanda MRL faction in the 1966 election, while the official data only mentions the MRL. Fourth, in the 1970 election, Nohlen (Reference Nohlen2005) mentions five seats for ‘others’ in the PC while the official data associates those seats with the Liberal Integration Front, the Popular Unity Movement, Popular Liberal, and two independents. As from 1974, some of the new parties came from earlier factions of the PL and the PC. Table 2 also shows the number of parties and lists (or factions under the FN) that won seats in the two houses of Congress. Under the FN, each party comprised a variable number of factions that were expressed in different lists. In 1974, the number of parties immediately began to increase, but the number of lists with representation in Congress only began to increase a couple of elections later.
Figure 1 shows the evolution, in the Chamber of Representatives, of the effective number of parties and lists in competition and with seats. While ENP in elections held steady under and after the FN, ENF in elections began to increase under the FN and continued to do so after the end of the FN. The same is true of ENP with seats, which remained steady under and after the FN, while ENF with seats was already rising under the FN and continued to do so after 1974. Naturally, there were more lists competing in elections than with seats in Congress (figures on the left and right, respectively). However, while the FN acted as a straitjacket reducing the number of parties in competition and with seats, the lists or factions in competition did not experience this same straitjacket and their number rose both under the FN and after 1974.
Figure 1. ENP and ENF in elections and with seats in the Chamber of Representatives, Colombia, 1958–1990.
Source: Compiled by authors using data from the Centro de Estudios sobre Desarrollo Económico (CEDE) https://datoscede.uniandes.edu.co/es/catalogo-de-microdata and Colombia's National Civil Registry.
We estimated ordinary least squares (OLS) models. The independent variables are the DM of each district. DM ranged from one to twenty-nine in the Chamber of Representatives and from one to fourteen in the Senate. Information about the DM of each department for the Chamber of Representatives and the Senate can be found in Appendices A1 and A2, respectively. We also include an interaction variable for the DM and the National Front, to assess if the effect of the DM was higher during the years the FN was in effect.
As control variables, we use whether the electoral unit corresponds to the Senate or the Chamber of Representatives, whether the election took place under the FN, whether the outgoing president was from the PL, the number of electoral districts for each chamber, and the number of seats in the Chamber of Representatives in each election – an issue that has been raised as a determinant of the number of parties competing at the national level by recent works (Benoit, Reference Benoit2007; Shugart and Taagepera, Reference Shugart and Taagepera2017). We use as control variables the percentage of urban population and the literacy rate. In some models, we include the number of districts and in others, the number of seats in each chamber as additional controls. This data was obtained from the online library of the National Administrative Department of Statistics (DANE). Table 3 shows the descriptive statistics for ENP, ENF, and the independent and control variables.
Table 3. Descriptive statistics of variables
Source: Compiled by authors using data from the Centro de Estudios sobre Desarrollo Económico (CEDE) https://datoscede.uniandes.edu.co/es/catalogo-de-microdata.
4. Results
Table 4 shows the results of the models for ENP and ENF in competition. For our first hypothesis, all models show that the DM has a positive and statistically significant effect on the effective number of parties (ENP) competing in elections and on the effective number of factions competing (ENF). This suggests that the DM impacts parties, but it also impacts factions within parties. The DM acts as a straitjacket limiting the number of parties and the number of lists in competition.
Table 4. OLS determinants of ENP and ENF in legislative elections in Colombia, 1958-1990
Robust standard errors in parentheses
Source: Compiled by authors using data from the Centro de Estudios sobre Desarrollo Económico (CEDE) https://datoscede.uniandes.edu.co/es/catalogo-de-microdata.
*** P < 0.01, ** P < 0.05, * P < 0.1
In general, both the ENP and ENF were lower under the FN, which seems reasonable given that it was established precisely to avoid party fragmentation. But the effect of the FN is stronger and more systematic on the ENP than on the ENF, which suggests that electoral rules have a more significant impact on parties than on factions under the FN. The effects of the DM and the FN interaction variables are negative on the ENP and on the ENF. That suggests that, during the FN, the effect of the DM was constrained by the straitjacket rules that limited the existence of additional parties. But the restrictive effect was also observable among factions within parties. In short, the models very clearly show an effect of the DM, but that effect was constrained during the FN years.
The FN restricted the number of parties that could run in elections, but not the number of lists that could form within parties. Not surprisingly, political groups that could have otherwise formed political parties were forced to compete as lists within either of the two traditional parties, the PC and the PL. When the FN ended, the number of parties multiplied – as the straitjacket was removed – but the number of lists also increased. That suggests that as more political groups formed their own parties, factionalism began to occur in those new parties as well.
The effect of the control variables is as expected. The Senate dummy has positive and significant impact on the ENP but not on the ENF. As the Senate was the more powerful Chamber, the leaders of political groups were more interested in being in the Senate than in the Chamber. As a result, there were a higher number of parties in districts with a higher DM in the Senate than in the lower house. The ENP was lower when the outgoing president belonged to the PL, but the ENF was higher at the end of the terms of PL presidents. This indicates that party fragmentation was lower under Liberal presidents, but the conflicts that led to the creation of within-party lists were greater under PL-led governments. This is consistent with the literature on the party system in Colombia that points to the PL having more internal conflicts and stronger presence of party subunits than the Conservative Party.
In four models, we used the number of electoral districts as a control variable and, in the other four models, we used the number of seats in each chamber, as both indicators are highly correlated. A higher number of districts are positively associated with the number of ENP and ENF. In turn, a higher number of seats in each chamber is negatively associated with ENP and positively associated with the ENF. This latter result is consistent with the findings reported by Shugart and Taagepera (Reference Shugart and Taagepera2017) and Benoit (Reference Benoit2007) that associate the number of parties at the national level with the size of the assembly. This also suggests that political groups were better expressed in the ENF than in the ENP.
The ENP was higher in rural areas and in areas with higher levels of literacy. This point to more fragmentation in the party system in areas where local bosses could mobilize rural voters but also in areas with larger and more educated population, where leaders could mobilize emerging middle-class voters against the party elites, an observation that is consistent with the literature on the Colombian party system.
The results of the models are consistent with the expectations of the first hypothesis. A higher DM is associated with a larger number of parties and lists in competition. Although the FN limited party fragmentation, the DM has an impact on the number of parties and lists in competition. However, the effect of the interaction variable also indicates a need for caution about the effect of the DM when restrictive electoral laws are in place. The FN worked as a straitjacket that limited the effect of the DM on the number of parties in competition more than on the number of factions in competition.
These results provide strong support to the claim that the study of party systems should pay close attention to the dynamics of within-party lists. When the electoral rules allow for multiple lists to compete under the same party label, studies that look at the number of parties running and overlook the number of lists running might be underreporting the level of party fragmentation in the political system. More restrictive DM arrangements do limit the number of parties running, but they have a more limited effect on the number of within-party lists fielding candidates.
For our second hypothesis, on the DM effect on the ENP and ENF that win seats in the legislature, the models in Table 5 show a consistent positive effect on the ENP and the ENF. The effect of the FN is positive on the ENP and not significant on the ENF. Since the FN forced an equal distribution of seats for the PL and PC during the FN, the ENP under the FN was always two in every electoral unit. After the end of the FN, the ENP and the ENF could be higher, but it could also be lower in places where either the PL or PC was strongly dominant. The FN did not impact the number of lists that managed to win seats. However, as the increase in the number of parties competing in elections once the FN ended also meant an increase in the number of lists competing and, thus, each list's chances of winning a seat decreased. These results again point to the impact of more restrictive electoral rules not just on the number of parties winning seats, but also on the number of intraparty factions, or lists, wining seats.
Table 5. OLS models on the determinants of ENF and ENP with seats in the Colombian Congress, 1958–1990
Robust standard errors in parentheses.
*** P < 0.01, ** P < 0.05, * P < 0.1.
Source: Compiled by authors using data from Nohlen (Reference Nohlen2005) and Colombia's National Civil Registry.
The effect of the interaction variable of DM and the FN is negative on the ENP and the ENF with seats in congress. This again suggests that the existence of the FN constrained the effect of the DM. Although a higher DM increases the chances of winning seats for parties and lists, during the FN, those chances were constrained by the restrictive electoral arrangement. That restriction did not only apply to the ENP – as the legislation stipulated – but it was also manifested by reducing the ENF that won seats. The FN was an agreement that intended to reduce the number of actors in the political arena. Although the RN constrained the number of parties that could run, political groups could survive by fielding their own lists within each traditional party. Yet, the FN established that seats would be equally divided between the PC and PL. As a result, during the FN, the ENP with seats was always two. When the FN was dismantled, political groups could choose to form their own parties or continue running as lists within parties. Although the number of parties grew, in some departments, parties could also become dominant and win more seats than all the other parties combined. That explains why the effect of the FN on the effective number of parties with seats in the legislature is positive. After the FN ended, a party could win more than half of the seats in the legislature, and thus, the ENP could conceivably be lower.
The other control variables have the expected effects, albeit not always significant. All in all, the models in Table 5 confirm that, regardless of the chamber in question and whether the FN was in force or not, the DM has a positive effect on ENP and ENF with seats in Congress. The evidence suggests that when the electoral rules are restrictive, political manifestations can disguise as party subunits to run in elections and, occasionally, can win seats doing so.
The models in Tables 4 and 5 clearly demonstrate the electoral system's effect on how party expressions were manifested in Colombia between 1958 and 1990. Although the FN acted as a straitjacket that reduced political representation, the mechanical effect of DM is clear, both under the FN and after 1974. The ENP and ENF in competition and with seats in Congress all increase with DM. These models show that, although there was de facto bipartism under the FN, the political expressions manifested in the form of factions (ENF) during this period also reflected the incentives created by the number of seats in each electoral unit. Under the FN, these political expressions were manifested in factions (ENF) because that was the only available vehicle. However, once the FN ended, some of these political manifestations were expressed in new parties that competed in elections while others continued to be expressed in lists within the traditional parties, emulating what had happened during the FN.
Figure 2 shows the predicted probabilities from the corresponding models in Tables 4 and 5 for the effect of the DM on the ENP and ENF in elections and in the composition of the legislature for the 1958–1990 period. A higher DM has a positive impact on the presence of electoral and legislative parties at the district level. With the same electoral rules at the national level for every election, in those districts with a larger DM, there were more electoral parties and factions (lists) within parties running for seats. Yet, this difference is also significant for districts with smaller DM. When the DM was 15 or above – albeit that only applies to a handful of districts in the country – the ENP was higher after the FN among the parties that ran in elections, but lower among the parties that had seats in congress. Again, this result might seem odd, but since the FN forced each district to have an ENP equal to 2 (the two leading parties equally split the seats in all districts), after the FN ended, the ENP could be both higher than two or lower than 2 depending on the dominance of any given party in any district.
Figure 2. Predicted probabilities for ENP and ENF in legislative elections in Colombia, 1958–1990. In the estimations of the predicted probabilities, we omit the interaction variable.
Source: Authors with data from Nohlen (Reference Nohlen2005) and Registraduría Nacional del Estado Civil.
Consistent with our claim, the effect of the DM was not different for the FN period and the post FN period for the ENF during and after the FN. Although a higher DM is associated with a higher number of factions competing in elections and winning seats, there is no difference between the FN and the post FN period. The FN constrained the number of parties running and winning seats, but the FN did not affect the presence of factions within parties. Under the FN the number of factions competing and winning seats was similar during the FN and after the FN rules ended.
5. Conclusions
Since electoral rules impact how parties and factions compete in elections and their seat share in the legislature, we use results at the district level in all the elections for both houses of Congress in Colombia between 1958 and 1990 to analyse the effect of the DM on the ENP and ENF that competed in these elections and that obtained seats in either of the two houses. Under the FN, factions manifested in electoral lists that competed within each party. The removal of the straitjacket that, under the FN, had imposed a two-party system meant that some factions effectively became parties while others continued to operate as lists within the PL or the PC until 1990. The number of parties and factions running in elections and winning legislative seats is directly affected by the DM. In districts with larger magnitude, there are more parties – and most importantly, more factions.
The case of Colombia shows that, even when electoral laws are restrictive, factions can still appear. Under the straitjacket of the FN rules, multiple factions continued to exist in the two large parties and their number was affected by the DM in the same way as the DM impacts the number of parties elsewhere. With more strict party affiliation rules, the ability of personalist leaders to mount their own factions within either party during the FN would have been more limited. Yet, as party leaders held limited gatekeeping power on who could field lists of candidates within the party in any department, party affiliation turned into a label that anybody could use for their own personalistic or other type of factions. The case of Colombia suggests that studies of party systems should pay closer attention to the effects of electoral rules not just on parties but also on how different factions are manifested in electoral processes within parties. The case of Colombia underscores the importance of analysing the behaviour of within-party factions or lists when studying the effect of electoral rules on the number of actors – multipartism or bipartism – in party systems.
Studies on the impact of electoral rules on the party system often focus too narrowly on the effect of the DM on the number of parties without paying sufficient attention to the within-party dynamics in closed-lists systems that allow for multiple lists to run under the same party label. When there are low entry barriers to field a list of candidates, the number of within-party lists might multiply in ways that are not captured when studies only look at the number of parties competing in elections. Future studies should not overlook the importance of within-party lists when assessing change and continuity in the party system.
Acknowledgements
We express our gratitude to Victoria Finn and Sebastián Umpiérrez de Reguero for commenting earlier versions and to the three anonymous reviewers for their suggestions.
Funding statement
This work was supported by FONDECYT: [Grant Number 1231627]; ANID-Millennium Science Initiative Programme: [Grant Number NCS2021_063].
Competing interests
None.
Data
The data used for the article was obtained from public sources, but we will be happy to share our Stata do and data files with anyone wishing to replicate our findings.
Appendix
Table A1. District magnitude by Department in the Chamber of Representatives, 1958–1990
Table A2. District magnitude by Department in the Senate, Colombia, 1958–1990
