Abstract
Chile’s parties have been characterized as ‘European’ in their development and institutionalization, but ‘Latin American’ in their high degree of political localism. Yet the specialized literature has not tested these tendencies nor developed a theory as to how they may coexist. Using the concept of party nationalization, we establish the veracity of the claims in the literature and propose a theory of static and dynamic nationalization development. We show that, as in Western European systems, Chile’s political parties exhibited progressive static nationalization until 1973. Since re-democratization, parties have demonstrated low levels of static nationalization, while the coalitions are highly nationalized. Concurrently, both parties and the coalitions have exhibited low levels of dynamic nationalization like parties throughout Latin America. We argue that these conflicting patterns are due to the interactive effect of functional cleavages and electoral institutions.
Introduction
While regional analyzes of parties and party systems highlight contrasts between Latin America and Europe, broad comparisons mask distinctive patterns for particular countries. For example, while Mainwaring and Scully (1995) categorize the Chilean party system as ‘institutionalized’ like those in Europe, Montes et al. (2000: 796) argue that ‘in terms of high electoral volatility, rapid changes in party labels and episodic outbursts of personalism, Chile is closer to most Latin American party systems’. In this article we focus on how Chile’s parties also stand out in terms of another prominent trait: party nationalization (and by extension, party system nationalization).
Party nationalization has two dimensions. The first, which Morgenstern et al. (2009) label ‘static’, measures the territorial homogeneity of a party’s vote distribution at a given point in time. Caramani (2004) shows that the Western European countries have become highly nationalized (i.e. parties display a consistent level of vote across districts) on this dimension, while Jones and Mainwaring (2003) show that levels are low across Latin America. The second dimension is meant to capture the relative influence of national versus local forces. Morgenstern et al. (2009) label this dimension ‘dynamic nationalization’, because it compares the consistency in the district-level vote across elections. They also show that European parties are more nationalized than those in Latin America.
Chile, however, stands out both today and in the pre-authoritarian period. In the pre-authoritarian period, Valenzuela (1977), Mainwaring and Scully (1995) and Gil (1966) explain that Chile’s ‘European’ parties’ influence extended far into the rural reaches of the country. Yet they did not overcome extreme localism and the personal brokering of politics (Valenzuela, 1978, 1999). This suggests that pre-authoritarian Chilean parties were ‘European’ on the static scale, but similar to their Latin American neighbors on the dynamic metric. Our data show that Chile’s static and dynamic nationalization began low and progressively increased from the end of the 19th century until the Pinochet dictatorship. Since the end of the dictatorship, parties’ static nationalization levels have dropped or stagnated while their dynamic nationalization levels remained constant. At the same time, the consolidation of Chilean parties into stable coalitions with high European-like static but low Latin American-like dynamic nationalization has allowed the Chilean party system to remain unique. Consequently, Chilean parties appear to be, in the words of Chilean folk singer Victor Jara, ‘ni chicha ni limoná’, neither European nor Latin American. 1
In this article we establish these distinctive patterns by comparing static and dynamic nationalization for parties and coalitions in Chile with others in Latin America and Europe. We argue that this unique pattern is due to the interactive effect of the electoral system and other institutions with politically salient functional cleavages. Specifically, class cleavages that emerged in the 20th century helped strengthen an ideologically-based party system and create nationally-based parties, supporting high levels of static nationalization. Still, presidentialism and a permissive legislative electoral system created incentives for legislators to pursue personal rather than party interests, hampering dynamic nationalization.
The article proceeds as follows. Section one defines nationalization, clarifying the distinction between static and dynamic nationalization. The second presents our data to establish Chile as an outlier, fitting neither Latin American nor European patterns. In section three we present our hypotheses and theory, emphasizing the role of societal cleavages and electoral rules as the elements structuring the evolution and form of party nationalization. Section four demonstrates institutions’ importance in the rising or lowering of static nationalization with an analysis of puzzling trends in pre- and post-Pinochet Chile. The penultimate section does the same for dynamic nationalization. We conclude with observations on the significance of these findings for Chile and party nationalization in general.
1. Concepts, measurement and case selection
As we will establish empirically, Chile is unique in terms of its levels, trajectory and combination of static and dynamic nationalization. Static nationalization refers to the degree of homogeneity of a party’s support across districts. A party with high support in some districts and low support in others is less statically nationalized than one with support evenly distributed across districts. Dynamic nationalization is analytically distinct and implies the degree of homogeneity in the change of a party’s support in each district across two or more elections. 2 If a party’s support in all districts moves together, then national factors are driving vote patterns and the party is considered to be dynamically nationalized. However, if party support moves up in some districts, down in others, and stays constant in others, then local issues – either candidate qualities or district characteristics – must drive electoral decisions. For these reasons, Morgenstern and Swindle (2005) use the inverse of dynamic nationalization as their measure of the ‘local vote’, arguing that this concept is broader than and distinguished from the personal vote because it focuses on electoral patterns of parties rather than individuals.
There are several potential measures for both concepts. To establish the unique position of Chile on these dimensions, we focus on what we consider to be the most appropriate for each, and apply them to a database of district-level legislative elections that extends back as far as 1832 for several dozen European and Latin American countries. 3 As a supplement, we add data for presidential and municipal elections in Chile across the pre- and post-authoritarian periods. To operationalize the static aspect, we use Bochsler’s (2010) standardized Party Nationalization Score (sPNS). This uses a Gini index to measure the spread of a party’s district level support, and weights that distribution by the number of districts and voters per district to generate a score between 0 and 1, where 0 indicates low static nationalization and 1 indicates high (for ease of interpretation, we multiply these values by 100). Along the static dimension, this dataset combines election results for parties earning more than 5 percent of the vote in consecutive elections, beginning with the earliest possible election (as early as 1832 in the United Kingdom and as late as 1984 in Nicaragua). In this dataset, European observations include both pre- and post-World War II elections, while most Latin American observations are restricted to the post-World War II period due to lack of elections or data. 4
We proxy dynamic nationalization with Morgenstern and Potthoff’s (2005) components-of-variance technique, that parses the vote into its national, state and district components. 5 Morgenstern et al. (2009) interpret the inverse of the model’s residual as a measure of ‘dynamic nationalization’, because it captures that part of the vote that is un-attributable to national trends or variance in the distribution of the party’s vote across districts. It is ‘dynamic’ because it uses change in the party’s vote at the district level as the basis of the calculation, akin to taking the standard deviation of the district-level swing for a party. If the party’s change in support for all districts is similar, the standard deviation will be small, implying that national rather than district-level factors are driving elections (resulting in high dynamic nationalization).
This technique produces values for both static and dynamic nationalization, but because calculations require at least two election cycles with consistent electoral boundaries, we cannot estimate as many elections as the sPNS method due to redistricting. However, it is impossible to avoid the redistricting problem when analysing dynamic nationalization, and we thus apply the components-of-variance model for that part of the analysis. Still, we provide extensive data to test the dynamic dimension for many countries in both the pre- and post-World War II periods. 6 The technique runs on any pair of elections, but it provides more accurate estimates when performed across multiple elections. We have therefore calculated scores for the longest period of continuous electoral boundaries available. Tests for pre-war Switzerland, for example, cover 16 elections over a 45-year period (1872–1917), while they cover just four elections from 1939 to 1945 for Colombia’s parties.
2. Chile’s parties and party system
Given Chile’s long electoral history, its parties and party system have been thoroughly studied (for a review, see Montes et al., 2000). Despite brief interruptions in 1891, 1924 and 1932, formal democracy was the rule from the mid-1800s to 1973. One landmark study of parties in this period gives special attention to nationalization. In it, Valenzuela describes the series of linkages that allowed national parties to penetrate the entire country, arguing that nationally-oriented parties structured competition ‘even in the smallest and most backward municipalities’ (1977: 23). However, the author bases his contentions of increasing European-style nationalization on data from a few key provinces, relying on extensive anecdotes and descriptions, though little numerical data.
A 1973 coup inaugurated a 17-year military regime, which has been followed by a second major democratic period. During the democratic transition, the authoritarian regime imposed a legislative electoral system intended to transform the party system by reducing the number of parties and over-representing the right. The regime abandoned the large magnitude pre-authoritarian proportional representation (PR) system and replaced it with a ‘binomial’ or two-member district system (in 60 districts) which forced competitors to propose two-member electoral lists. This produced a strategically complex system with two main coalitions (the centre-left Concertación and the rightist Alianza por Chile), in which member parties bargain over scarce candidacies, allowing them to counteract the system’s reductive tendencies and permitting the perseverance of a multiparty system (Dow, 1998; Magar et al., 1998; Siavelis, 1997). Through this, the system has sustained three principal parties in the Concertación, the Partido Demócrata Cristiano (PDC), Partido Socialista (PS) and Partido por la Democracia (PPD), while the Alianza coalition has had two principal parties, the Renovación Nacional (RN) and Unión Demócrata Independiente (UDI).
There have been a few studies of nationalization in the pre- or post-dictatorship era, but they fail to provide adequate cross-country or temporal comparisons, explanations or consideration of both dimensions of nationalization (Alemán and Kellam, 2008; Harbers, 2010; Jones and Mainwaring, 2003). Here we combine sociological and institutional approaches with extensive electoral data to more accurately understand levels and trends in Chilean party nationalization.
3. Chile as a nationalization outlier
Static dimension
To establish Chile as an outlier, we compare the evolution of its parties’ static nationalization to that of parties in 15 Western European and 18 Latin American countries. 7 We first calculate the sPNS for each party and then estimate a multi-level model of unconditional growth for each country. This model provides estimates for the average sPNS starting point as well as the slope of parties’ trajectory.
Two methodological features make the data well suited for growth modelling: all panels have three or more waves of observations, producing growth curves and, secondly, the observations include an outcome – sPNS – whose value tends to change systematically over time. Essentially, this model accounts for unconditional variance in the intercepts and slopes of the nationalization trends in the following form: Yit = β 00 + β 10 Election Count +∊ it + ζ 0t, where β 00 is the y-intercept, β 10 is the slope coefficient, ∊ it is the stochastic error that represents within-party variation over time, and ζ it is a random effect representing variation in party intercepts within countries, where subscript i denotes each party and t denotes each election. Although each party has its own growth curve, individual parties are aggregated into country scores to give a more streamlined picture of nationalization tendencies. 8
The magnitude and statistical significance of the slope and intercept are of central importance. Conventional wisdom predicts that European countries have low intercepts and positive slopes in the pre-World War II period, since their parties and party systems began as highly territorialized and then nationalized as functional cleavages manifested themselves in electoral politics, while they should have higher intercepts and flatter slopes in the post-World War II period (Caramani, 2004). By contrast, Latin American parties should show comparatively higher y-intercepts and slopes with little growth, given the inability of parties to transform class-based cleavages into electoral support. Figure 1 plots these starting points and growth rates, with European countries denoted as x’s and Latin American cases as triangles (see our supplemental appendix for all relevant statistics). The regional variation is staggering: in the pre-war period, Europe has an average starting point of 57.5 and growth rate of 0.58, while the parties in the three Latin American countries for which we have data have an average intercept of 68.9 and a slope of –0.04. In the post-war period, European parties’ average intercept increases to 63.57 and the slope to 0.59, while Latin American parties’ mean intercept rises all the way to 77.98 and displays a faster rate of decline (–0.13). 9
Scatter plot of static nationalization starting points and growth rates over time for 33 countries.
Chile’s low starting point (52.4) and positive slope in the pre-war period (1.03 units per election), which place it in the lower right-hand corner of the chart, are closer to the European means than the other two Latin American cases, Colombia and Peru. From the end of World War II to the 1973 dictatorship, the mean for Chile’s parties rises to 71.84 and its slope becomes indistinguishable from 0. For the post-dictatorship period the figure separates Chile’s parties from the coalitions. The parties start with a low intercept, since several did not participate in all districts, while their positive slope reflects increasing geographic coverage. The coalitions have a different pattern, showing high intercepts and slopes of zero that reflect their continued national uniformity.
Again, these patterns are unique. While Chile’s parties look like others (mostly in Europe) in the years between World War II and the dictatorship, afterwards the parties are among the least statically nationalized and reveal an upward trend. 10 Meanwhile, the coalitions are among the most statically nationalized in the world. Of course, parties in several Latin American countries also appear highly statically nationalized, but all save Costa Rica’s show a decreasing pattern not shared by most European parties or Chile’s coalitions.
Dynamic dimension
While Chile’s pre-dictatorship parties (and current coalitions) are similar to European parties in terms of static nationalization, analysis of dynamic nationalization yields evidence of high localism for both the parties and coalitions.
Cross-regional comparisons show strong contrasts. The limited democratic experience in pre-World War II Latin America limits data availability, but Figure 2 shows local forces were much more influential in determining electoral outcomes in the pre-war period for the Chilean parties (represented by white bars) than most European or Latin American parties. The Conservative Party is the most nationalized Chilean party, but its score of 128.7 indicates a high degree of variance in district level support and a significant effect of local factors. By contrast, the least nationalized parties in Uruguay and Colombia have scores under 50. While Europe shows wide variance, its least nationalized parties – in Switzerland between 1872 and 1917 – still have values two-thirds the size of Chile’s least nationalized party.
Pre-World War II Dynamic Nationalization in Latin America and Europe.
More extensive data for the post-war period continue to place Chile’s parties and coalitions in the Latin American norm. Figure 3 shows Latin American parties whose scores range from the single digits (Uruguay’s three parties) to nearly 200 (Colombia’s Liberal Party), but most are between 25 and 75. Chilean parties (post-dictatorship period displayed here due to space considerations) are at the top of the range; the most dynamically nationalized is the UDI, which has a score of 73.5, and the least is the PPD, with a score over 170. The coalitional data also indicate that local factors are highly significant, but lower than Chilean parties and more consistent with the modal Latin American party. The European parties stand in stark contrast with scores that seldom rise above 10. In sum, legislative elections in Chile and the rest of Latin America reflect local politics to a much greater degree than in Europe.
Post-World War II Dynamic Nationalization in Latin America and Europe (Chilean parties denoted in white; coalitions in black).
Chile, therefore, is a puzzling case. On the one hand, static nationalization has grown over time as in Western Europe, but in contrast to Latin American. Conversely, dynamic nationalization has been much lower than Western European cases and on a par with Latin American parties. So what explains these contradictory patterns?
4. An explanation: Social cleavages and institutions
Our explanation focuses on Chile’s sociological cleavage structure in combination with institutional change. To explain increasing static nationalization, we follow Caramani’s (2004) analysis of Europe, focusing on the development of a national cleavage structure in Chile. We argue that this cleavage structure’s impact was reinforced by Chile’s permissive PR electoral system in the pre-dictatorship period, which enhanced parties’ ability to spread support throughout the nation. Yet these factors did not yield high dynamic nationalization, since Chile’s combination of presidentialism and open-list electoral rules provides voters with incentives to dissociate national from local concerns when electing legislators, especially as the number of legislative districts increased from 28 multi-member districts to 60 two-member districts. 11 Furthermore, the post-authoritarian electoral system has set the two dimensions of nationalization in different directions and created a distinction between the parties and coalitions. The electoral system has guaranteed high static nationalization for the coalitions but not for the parties, since coalitions compete everywhere while individual parties cannot. Dynamically, the new electoral system continues to encourage a focus on candidate qualities and local issues, reducing dynamic nationalization for both parties and coalitions.
Social cleavages
National cleavage structures have a theoretical relation to both dimensions of nationalization. Lipset and Rokkan’s (1967) analysis of Western Europe suggests two potential types of politicized cleavages: territorial, in which countries are divided by region, often as the result of concentrations of linguistic, cultural or religious minorities, and functional, which are based on ideology or class and thus cut across the territorial divisions. The first type defined party systems in the 19th century, as early democratic competitors sought support based on commitment to locality, irrespective of economic position, in part because lower classes lacked the franchise. The industrial revolution and expansion of the franchise then yielded new class-based (functional) divisions (Lipset and Rokkan, 1967: 12).
Caramani (2004: 32) hypothesizes that this evolution drove progressive static nationalization in Western Europe. As class-based voter demands displaced territorial ones, parties increasingly sought support from a broad cross-section of territory. There is also a link between cleavage type and dynamic nationalization. Territorial cleavages enhance localism by conditioning votes on narrow local issues instead of broad national ones. By contrast, a dominant functional cleavage should make voters more responsive to national issues, and thus electoral change should be more uniform across territories. Thus the displacement of territorial cleavages by functional ones should lead to increasing nationalization along both dimensions.
This is consistent with the socio-political evolution in Chile. Scully finds that three sets of ‘generative cleavages’ shaped Chile’s party system: (1) the clerical/secular one (common in Latin America) from about 1857–1861, (2) an urban worker/employer class cleavage that took place in the 1920s, and (3) a second rural worker/employer cleavage that developed in the 1950s (Scully, 1992). These overlapping functional cleavages transformed a territorial-based party system into an ideologically-based one, like Western Europe and unlike the rest of Latin America, with the possible exception of Argentina (Dix, 1989; Roberts, 2002). As noted, post-dictatorial society has been cleaved by Pinochet’s contested legacy.
Theory thus suggests that the historical cleavage structure should have produced high or progressive nationalization along both dimensions for Chile in the pre-war period, which would have continued over time. But cleavages alone do not determine nationalization.
Institutions
Institutions also influence the two dimensions of nationalization. Following Cain et al. (1987), who find a much higher ‘personal vote’ in the US than in the UK, Morgenstern et al. (2009) associate presidentialism – whose executive and legislative elections share weaker ties – with higher localism. In contrast, voters indirectly choose the prime minister in parliamentary systems, focusing district-level legislative elections on the national race. Parliamentarism also holds constant the personality of the executive, while candidate qualities and personalities deeply affect legislative elections in presidentialism.
While dynamic nationalization is a function of the executive system, we expect that the electoral system drives static nationalization. Like Western European systems and most of Latin America after World War I, pre-authoritarian Chile had a relatively permissive PR formula (i.e. moderate to large magnitudes, open lists, permissive of pacts) for legislative and municipal elections. 12 This promoted static nationalization as parties vied to expand their organizational infrastructure and compete in every electoral district, with fewer votes needed to gain congressional representation. Under a more constraining (low district magnitude or high threshold) system, such as SMD or the current Chilean two-member districts, static nationalization should be more limited as parties seek to avoid wasting efforts where winning is unlikely.
The binomial system has an additional impact. With just two seats available per coalition but more than two parties in each, 13 bargaining and horse-trading are the norm. Parties cannot place candidates in all districts, even if they have the potential to win, generating a large number of ‘zeros’ in the data. This not only produces high variance in parties' support, but since the parties are not consistent about which districts they compete in, it also produces a large vote swing and low dynamic nationalization.
These effects are not as dramatic for the coalitions, since they can compete in all districts. Consistent participation supports high levels of nationalization relative to parties, and could be bolstered if the coalitions tailor their identities to individual districts by using different combinations of parties on their lists. The lack of zeros would also support higher dynamic nationalization for the coalitions. It may not be as high as a larger magnitude system would generate, however, since voters may still have to choose among ‘second-best’ candidates if their preferred party is kept off the list. Still, those voters can choose a candidate from within their favoured coalition. As a result, dynamic nationalization for coalitions should be considerably higher than that of parties.
The number of districts and the list type are also central. As we noted above, if a country had only one district, both dynamic and static nationalization would be at their theoretical maximum. As the number of districts rises, the potential for heterogeneity and influence of local forces among them also grows. The increase in the number of districts in Chile should then decrease each type of nationalization. Our static nationalization indicator tries to control for the number of districts, while our dynamic indicator does not.
List type and district magnitude also affect both types of nationalization. Unlike Western Europe, which primarily relied on closed-list PR, but similar to much of Latin America, pre-authoritarian Chile used an open-list system, and currently uses a two-member variant. Theoretically, closed lists promote party unity (at least when district magnitude is high) at the expense of personalism, while open-list systems should yield the reverse (Ames, 1995; Carey and Shugart, 1995). The expectation is that dynamic nationalization should suffer under systems that promote a personal vote, although Chile’s two-member version of the open-list system reduces those incentives significantly. Comparisons show large vote margins between the intra-coalition competitors, suggesting that the competition for the last votes is not fierce (Navia, 2005). Personal voting, therefore, is not high, potentially offsetting the expected reduction in the dynamic nationalization that the increased number of districts could generate. These aspects of the electoral system have no theoretical impact on static nationalization.
The intersection of cleavages and institutions
Cleavages and institutions have both independent and interactive impacts on nationalization. This conditionality is most important for the static dimension, since the electoral system’s impact is dependent on the type of political cleavage structure. Where territorial dimensions predominate, parties have little incentive to expand beyond their territorial strongholds regardless of the electoral formula. When the functional class cleavage is dominant, a PR formula lowers entry barriers, favouring the spread of homogenous left–right alignments, while majoritarianism inhibits the spread of parties (Caramani, 2004: 228). The predominance of the functional dimension allows parties to build support across regional divides, which is reinforced by a permissive electoral system that encourages parties to expand their geographic base.
Presidentialism and open-list electoral laws work against dynamic nationalization, though a cleavage structure that ties district interests together may counter this. Therefore, the effects of presidentialism and an open-list formula should be stronger when territorial cleavages dominate functional ones. Still, the emergence of functional cleavages is no guarantee of increased dynamic nationalization, because the number of parties is not constant. In Chile, the emergence of class-based party blocs was accompanied by an increase in the effective number of parties for much of the pre-authoritarian period. This mollified potentially concomitant increases in parties’ dynamic nationalization since voters developed links to party families rather than individual parties. Thus, despite stronger sociological and ideological divisions, the multiplicity of parties gave voters wide choices within party families.
In sum, a number of factors have conspired to produce ‘Janus-faced’ parties (Shugart and Carey, 1992: 182) in the pre- and post-war eras that are at once relatively uniform in their territorial coverage yet strongly local. As class-based political cleavages gradually emerged and replaced territorial ones, the permissive PR electoral system allowed static nationalization to increase. At the same time, presidentialism, open-list electoral rules and multipartism also caused low levels of dynamic nationalization. In this context, the introduction of the binomial electoral system at the legislative level generated lower levels of static nationalization for the individual parties, but not their coalitions. And while the new electoral system may have reduced the incentives for a personal vote at a legislative level, other aspects of the system have had the reverse effect. In the next section we test these institutional hypotheses, comparing nationalization levels across electoral systems for local, presidential and legislative races considering both parties and coalitions.
5. The effect of institutions on static nationalization
The wealth of district-level electoral data for Chile (legislative results from 1891, municipal from 1942 and presidential from 1938) permit a test of the institutional theory within the context of ingrained social cleavages. Furthermore, the variations in electoral systems over time and across offices create a natural experiment to observe how nationalization levels and trends react to electoral system change. We cannot vary cleavage structures, but the comparative data shown in Figure 1 provide evidence of the impact of cleavage structure on static nationalization.
Caramani’s model predicts a low starting point of static nationalization, with a gradual rise that reflects parties’ increasing ability to nationally represent voters in different corners of the country. The data in Figure 4, which show trends in sPNS for legislative, municipal and presidential elections for all major parties from 1891 to 2009, support the predictions of growth in the pre-authoritarian period. 14 The figure is broken into pre- and post-Pinochet charts, with a plethora of parties at the municipal and legislative levels, and fewer data in presidential elections (where coalition-making is the norm in both eras). The black lines represent party scores, whereas the dotted and dashed lines represent the two dominant coalitions in the post-1989 era; as the bottom middle graph illustrates, municipal candidates compete on party rather than coalition tickets. The finding of increasing static nationalization for parties competing in municipal elections is telling, since local issues or politicians have the potential to dominate those elections. As the top three graphs show, parties generally followed expectations, progressively nationalizing up to the dictatorship at all three levels. In the post-1989 period, the trends are flat, but most clearly the coalitions – which do not compete as such for local elections – show much higher scores than the parties.
Legislative, Municipal, and Presidential Static Nationalization over time in Chile.
These data also allow us to test the institutional hypotheses. Because presidential elections are national in scope, they should generate higher levels of static nationalization than other elections. Comparisons among the graphs support this expectation, though there is no clear distinction between legislative and municipal levels. This results in part from parties forming coalitions for the presidential elections, and explains why the coalitions are also highly statically nationalized after 1989.
The restrictive, but open-list, electoral system implemented after the dictatorship is also related to the low static nationalization for parties. A comparison of the two legislative graphs (on the far left) shows that the levels dropped and stagnated after the return to democracy. 15 The scores for the PDC, for example, fell from over 0.90 before the dictatorship to between 0.66 and 0.84 after re-democratization. Likewise, the Socialist Party’s sPNS was 0.81 in the 1973 elections, yet has oscillated between only 0.30 and 0.58 in the post-dictatorship years. Municipal and presidential elections, by contrast, substantiate the importance of the electoral system, since systems at those levels did not significantly change and their static scores remain almost uniformly high after 1989.
The open-list binomial system drives this result because it limits the menu of options to voters. The distribution of legislative candidate slates is determined by negotiations involving trade-offs, pacts of exclusion and veto power for party elites. Since there are three to four main parties and only two slots on the ballot, Concertación elites must award candidate positions to the PRSD, for example, instead of the PS in some districts. In this situation a PS voter has to opt for a second-best candidate. As a result, the PS may seem less statically nationalized, precisely because it did not have the opportunity (as it did during the pre-authoritarian period or at the municipal level in the post-authoritarian period) to present candidates in every district. Note that the coalition’s static nationalization would stay high in this case, as long as the voter’s second-best candidate was a member of the same coalition.
Given this limitation, a second measure of static nationalization analyzes only districts in which each party competed. As Figure 5 illustrates, eliminating ‘zero’ districts from the analysis yields high scores for all parties (between 0.8 and 0.95), with little variation over time. These levels are higher than parties’ respective scores in municipal elections during the same period, and significantly higher than the other legislative election scores, as the left-hand chart on Figure 5 shows. These changes are most evident among the parties of the Concertación. In the early post-dictatorship period, the PPD and PS split one of two seats in each district at the behest of the more powerful PDC. The PDC’s static nationalization decreased dramatically as it was forced to cede more seats to its coalition partners. It is unlikely that PDC voters decided to shift loyalties to other parties; rather, the PDC was not an option and voters opted for another Concertación party.
Post-Pinochet Comparison of Static Nationalization.
A similar pattern of change among parties on the right underscores this point. The UDI and RN trend lines on the left-hand chart in Figure 5 show important variance, with the RN falling from 1989–2001 before rising again, while the UDI gradually increases. The low UDI scores in the first years after the democratic transition reflect its limited candidate placement ability in those years (the UDI competed in about 30 districts in 1989 and 1993, then 47 in 1997 and nearly all thereafter). In the founding elections, the RN was more powerful and could demand more seats from the UDI. As UDI’s national presence grew, it was able to extract more seats from the RN, leading to an increase in its static nationalization. The calculations on only those districts in which the parties competed (i.e. eliminating ‘zero’ districts) suggest that the parties’ ‘natural’ static nationalization rates in legislative elections are considerably higher. In other words, when voters have the opportunity to vote for their preferred party, territorial support is more homogenous than the previous analysis suggests.
Municipal election results give further evidence of electoral system effects. In spite of an inherently greater local focus, post-1989 municipal elections generally and counter-intuitively produced higher static nationalization scores than legislative elections. Except for the UDI in 1996 (the single sharply declining line), no party registers below 0.55, and most are between 0.6 and 0.8, while the legislative scores are clustered between 0.45 and 0.7 (the solid lines in the lower left graph in Figure 4). In more recent elections, the RN and UDI’s legislative scores approach levels similar to those in the municipal elections, but the PDC falls from this range. Furthermore, the PPD and PS scores are more than twice as high in the local elections than in the legislative ones during this period.
The electoral system is the source of these differences. From 1992 to 2000, mayors were elected indirectly from municipal councils, who were in turn chosen through proportional representation. The PR system, as we argued, should have helped the parties build higher static nationalization than the two-member system used in legislative elections. Further evidence of the electoral system effects comes by comparing results of a 2004 reform. Beginning in 2004, municipal councils continued to be elected via PR, but a reform established direct mayoral elections. Figure 6 shows that, as expected, plurality mayoral races produced static nationalization scores that were 30–50 percent lower in mayoral than council races in both 2004 and 2008.
Comparison of municipal election static nationalization (2004-2008).
The data suggest that static nationalization scores are strongly influenced by the electoral system. Without the binomial system, scores for the legislature would undoubtedly be higher, increasing the slopes of the overall static nationalization growth curves, and providing even stronger evidence of progressive nationalization (and perhaps ‘freezing’ after 1989 given the constant scores in this period) that mirrors European tendencies of party growth.
6. Effect of institutions on dynamic nationalization
In Figure 7 we display dynamic nationalization levels over time and at different levels of elections to confirm the high level of localism for both parties and coalitions. In this sub-section, we present data that explain this outcome. Our first recourse is to the electoral system, but we also explain that independent parties and candidates foment the high level of localism. Lastly, we use presidential, legislative and municipal electoral results to show presidentialism’s role in generating low levels of dynamic nationalization.
Evolution of Dynamic Nationalization in Chile (1891-2009).
In the early period of Chilean democracy (1891–1924), there were four important parties that competed consistently. Their dynamic nationalization scores were generally over 100, and frequently twice that high, indicating a strong local focus in elections. After the disruption of democracy in the 1930s, dynamic nationalization scores shrank as the average scores for the two long-running parties, the Liberals and Conservatives, fell to 55. 16
Dynamic scores in the post-Pinochet era surprisingly yield contradicting characterizations of parties and coalitions. Statistics for the full 1989–2009 period yield scores of 57.9 for the Concertación and 29.3 for the Alianza. For the Concertación this was the result of shifts in support from –23 to +59 points. Moreover, these endpoints are not masks for otherwise consistent movements; for that same period, the Concertación lost at least 10 points in four districts and gained at least 10 in 14 districts. In the most recent period (2005–2009) the Concertación lost at least 10 points in 15 different districts, but showed a positive swing in nine.
For parties, we begin with a conservative analysis, calculating scores only where parties competed in successive years (i.e. excluding ‘zero’ districts). Even using this, Figure 8 shows some parties with scores over 50, implying that once accounting for temporal and cross-sectional variance there is still a local effect with a standard deviation of about 7 points. This implies that the range of vote change in 40 districts is about 14 points, and even greater for the remaining 20 districts. To take the extreme case of the PPD in 1993–1997, the standard deviation of the swing in the 20 districts where the party competed in both elections was 13.9. In six of these districts, however, the swing was over 20 points. Since support levels are highly variable, personalities or local issues must play a large role in driving voters’ decisions. These numbers imply a large role for candidates or their districts in determining voting outcomes. And these high numbers are considerably higher – often over 100 – when recalculated using all districts (including the zeros). 17 This again points to the impact of the binomial electoral system on the level of localism. Because parties do not consistently compete in the same districts, the swing from year to year yields wide oscillations, since voters cannot express their support for some of the parties
We also find that the low level of dynamic nationalization in Chile is a function of the overlooked support for third parties and independent candidates. While most discussions of Chilean electoral politics focus on the two main coalitions, other competitors win significant vote-shares in many districts. Figure 9 plots these parties’ district-by-district support. While independents win an average of 5–10 percent each year, there is a wide variance: many districts have had representatives of ‘third coalitions’ winning at least 20 percent of the vote, especially prior to 2001. Importantly, these districts with strong third-coalition support do not sustain that support from year to year. As a result, support for the two main coalitions necessarily fluctuates.
Dynamic Nationalization Scores (XTMixed) (Districts in which party competed in successive years).
Vote Share for Parties Outside of Two Main Coalitions (1989-2009).
Independent parties are not just external to the coalitions, but also operate within them, complicating intra-coalition bargaining. For the right, one of the two parties is displaced every election by Alianza-aligned independents in dozens of districts, and a significant number of these candidates win, as shown in Table 1. From the 1989 legislative elections to 2009, non-RN or UDI candidates participated in between 11 and 46 districts as members of the Alianza, winning from two to eight seats per election. The Concertación’s story is similar, though frequently a fourth party, the PRSD, has been able to nominate a candidate for a spot. Between 1989 and 2009 the PRSD and other Concertación-aligned independents competed in between 8 and 29 districts, winning between 3 and 15 seats. Once again, these minor parties and independents did not appear in the same districts from year to year, exacerbating low dynamic nationalization.
Number of districts with Concertación or right-wing candidates from small parties or independents.
We now turn to other evidence of the impact of presidentialism and the electoral laws. First, since presidentialism separates elections for the executive and legislature, it allows the latter to emphasize local over national politics. We therefore expect that presidential elections will produce more nationalized parties than legislative. Furthermore, municipal elections should show greater local effects, since politicians elected to these posts are further removed from national politics and more tied to local issues. The data bear out these predictions. The analysis on the presidential elections yields values of about 30 for the Socialists in the pre-Pinochet period and 10 and 12 for the two coalitions in the post-dictatorship presidential races, numbers that contrast with much higher numbers on Chilean legislative elections but are comparable to the Western European data shown above. 18
The data also support the role of the electoral system in privileging local over national politics for both the pre- and post-dictatorship periods. Prior to 1973, Chile used an open-list PR electoral system that encouraged candidate-centred voting (Carey and Shugart, 1995), while we theorize the post-Pinochet binomial system also supports a personal vote. As we have emphasized, the manner in which the electoral system limits the number of party candidacies has a direct and large impact on dynamic (and static) nationalization. For example, in legislative district 3, the PDC earned 26 percent of the vote in 1989, 0 percent in 1993, 21 percent in 1997, 22 percent in 2001, 23 percent in 2005 and then only 11 percent in 2009. With the exception of 1993, these fluctuations followed the trends for the PDC support for the rest of the country, with the party’s average support at 24, 26, 23, 19, 21 and 14 percent in those same years. However, the dynamic nationalization score is high because the PDC did not run a candidate in district 3 in 1993. This system, then, produces scores that suggest parties are less dynamically nationalized than coalitions in legislative elections.
The municipal elections in the post-Pinochet era allow a last test of the electoral system. Since these elections are based on closed list PR, they should allow national leaders and events to influence the races. At the same time, since municipal elections are local affairs, they should by definition be less nationalized. Comparing the left and middle graph in the row of Figure 7, however, suggests that national issues are more influential for municipal elections than they are for legislative elections. Most starkly, for 2004–2008, the dynamic nationalization scores fell in municipal elections, but averaged over 50 for parties in the legislative elections, suggesting that the legislative electoral system drives the low levels of dynamic nationalization in these elections.
In sum, before the Pinochet dictatorship, Chile’s open-list PR, presidential system produced low levels of dynamic nationalization, especially with respect to legislative elections. In the post-Pinochet period, presidentialism has continued, while the new electoral system continues to promote a personal vote. Further, the electoral system artificially inflates localism scores for parties by limiting the number of competitors within each coalition. Finally, because the binomial system allows independent pacts and independent candidates within established lists, it sometimes siphons votes away from major parties, enhancing localism.
7. Conclusion
While Chile’s pre-dictatorship parties showed static nationalization traits reminiscent of Europe, their dynamic nationalization scores were closer to the Latin American pattern. Since that time, parties have formed two major coalitions, which has lowered their static nationalization, but yielded high scores for the coalitions. On the dynamic scale, our data suggest that local factors are highly influential (i.e. parties are not highly dynamically nationalized). We propose that these outcomes are a result of politically relevant social cleavages that were sharper in Chile than in other pre-1970s Latin America, combined with presidentialism and a permissive open list PR electoral system that created conflicting incentives for personalism in the pre-Pinochet era. In post-authoritarian Chile the binomial electoral system plus presidentialism has lowered parties’ nationalization along both the static and dynamic dimensions. In part, the system creates artificially low scores, since the parties cannot compete in all districts. However, we show that legislative party results for dynamic nationalization are not artificial; coalitions retain high static nationalization in legislative elections but high levels of localism are evident at all levels.
Chile’s parties do indeed appear to fit the label of ‘ni chicha ni limoná’, supporting both conventional wisdom that regards the Chilean parties as ‘European’ while simultaneously supporting literature that characterizes Chile’s parties as personalistic and ‘Latin American’. Support for this dual characterization requires separating nationalization into its two constituent dimensions, static and dynamic nationalization, and analysing Chilean parties and coalitions both over time and at different levels of elections. As analysts puzzle over how it is possible for a country with a party system as ‘European’ to have experienced such a dramatic breakdown of democracy and to be currently experiencing a profound crisis of representation (Siavelis, 2009), our findings of the ‘mixed’ nature of the parties provide some crucial answers to these questions.
The use of the Chilean case, which lends itself to analyzes because of its long and well-documented electoral history, its seemingly contradictory patterns of party growth and the abundance of literature on its parties and party system, has additional implications for the party nationalization literature. To begin with, patterns of party growth and localism in Chile support Morgenstern et al.’s (2009) conclusions that static and dynamic nationalization are empirically unrelated. However, we differ in our reasoning. They believe static nationalization is a result of electoral variables while dynamic nationalization is a function of executive type. We show that electoral factors matter for static nationalization (at least in the context of a political system divided along a clear class cleavage) as well as dynamic nationalization. Future nationalization research should take this into account, and include the often overlooked dynamic nationalization. The findings also support Caramani’s (2004) thesis of progressive (static) nationalization being driven by politically manifested social cleavages, but add an element of conditionality via the mechanical effects of institutions, chiefly presidentialism, the number of electoral districts, list type and district magnitude. Likewise, the article shows how institutions can increase or decrease dynamic nationalization.
Moving forward, a challenge in the study of political parties is to examine the roots, trends and consequences of party nationalization. Scholars must be careful not to conflate the two dimensions of party nationalization because, as Chile shows, this mistake could severely bias conclusions. More research is needed, but the oddities of the Chilean case can properly illustrate the caution that must be taken in simply interpreting nationalization numbers and the consequences of party nationalization for the quality of political representation.
