Recent research on Latin America has gone a long way toward correcting the old stereotype of the region's parties and party systems as excessively pragmatic, clientelistic, personalistic, volatile, uncohesive, and therefore weak. A new conventional wisdom has developed that emphasizes the variety among Latin American countries rather than their common deviation from the norms of the industrialized north. Our understanding now needs to go a step farther, by recognizing that there is almost as much difference within each country as there is across the countries of Latin America. This recognition requires us to be more cautious in generalizing about cross-national differences. Also, scholars must now pay more systematic attention to the substance of party competition, which has been neglected in favor of more objectively measured party-system characteristics such as fragmentation and volatility. This article attempts to improve the new conventional wisdom in both ways, by describing the ideology, polarization, mean left-right positions, fragmentation, and institutionalization of the party systems of 20th-century Latin America on an election-by-election basis.

The conventional wisdom about Latin American parties and party systems developed in four stages. From the turbulence of the 19th century until about 1960, the prevailing stereotype was that Latin America was a region of caudillismo, or domination by one or more strong personalities, often military figures, who generally lacked a strong organizational base of support. However, Chile was always recognized as the exception: the most ‘European' of the Latin American cases, with well-organized, deeply rooted political parties possessing clear left-right ideologies (Johnson, 1958: 19, 66-93). From about 1960 to about 1967, Latin American parties were common objects of study in the United States. This research brought two innovations. First, scholars recognized that the democratic mass politics of the postwar era had given rise to new sorts of parties. Some were communist or socialist parties, which were appropriately considered quite ideological. But ‘Aprista' or ‘national-revolutionary' parties were also emerging, characterized by middle-class leadership, a multi-class social base, and a reformist ideology (Blanksten, 1960; Martz, 1964; DiTella, 1965; Dix, 1966: 292). Second, this wave of research produced detailed case studies of parties as institutions, thereby building up a reserve of knowledge about organizational structure, candidate selection, cohesion, campaigning, factionalism, and elections (Alexander, 1964; Martz, 1966; Williams, 1967; Payne, 1968; Hilliker, 1971). In the late 1960s, however, the study of political parties became a marginal pursuit among Latin Americanists, shouldered aside by military rule and a radicalized environment in which only truly revolutionary change counted and reformism was considered indistinguishable from personalism, clientelism or outright reaction. Some research on parties continued, but it was considered less interesting than research on the military, the church, economic policy, or social classes.

Research on parties continued to be neglected until the mid-1980s, when redemocratization was well under way and parties once again became important political actors. The wave of democratization inspired case studies of parties in Argentina (Gibson, 1996), Brazil (Keck, 1986; Kinzo, 1988; Mainwaring, 1993), Chile (Scully, 1992), Colombia (Hartlyn, 1988), Ecuador (Conaghan, 1988), Mexico, Peru (Graham, 1992), Uruguay (Gillespie, 1991; González, 1991), and Venezuela (Coppedge,1994). The integration of Latin American cases into mainstream debates about presidentialism, legislative behavior, and electoral reform also motivated scholars to collect comparable cross-national data on Latin American party systems for the first time (Taagepera and Shugart, 1989; Remmer, 1991; Mainwaring, 1993; Jones, 1994; Mainwaring and Scully, 1995; Coppedge, 1997b).

In the new conventional wisdom that grew from these studies, admirably synthesized in Mainwaring and Scully's Building Democratic Institutions (1995), Latin American party systems vary principally according to institutionalization and number. The fundamental distinction divides the institutionalized party systems from the ‘inchoate' ones. A secondary division is the more traditional classification into dominant-party, two-party, and multiparty systems. These are the two dimensions that have received the most attention and on which the greatest quantity of data exists. Roughly speaking, Latin American party systems are thought to fall into six distinct types defined by their degrees of institutionalization and fragmentation. The ideal combination is the institutionalized 2-2.5-party system, which is identified with Costa Rica, Colombia, Venezuela, Uruguay before 1971, and in some respects Argentina. Uruguay after 1971 and Chile are classified as institutionalized but fragmented party systems, which are also considered viable, although prone to crises of governance if majority coalitions cannot be built. A third, marginally democratic is the institutionalized dominant-party system, which is identified with Mexico and Paraguay, and a fourth is the inchoate 2-2.5-party system, believed typical of Peru. The worst combination is defined as the inchoate and fragmented party system, which has been diagnosed in Ecuador, Bolivia, and Brazil during its democratic years. The party system of authoritarian Brazil could be classified as an inchoate dominant-party system even though many observers consider authoritarian elections intrinsically incomparable to the democratic systems.

The new conventional wisdom is a vast improvement over the spotty coverage and stereotypes that prevailed 30 years ago. However, there are two respects in which it can be improved. First, it is important to revive the systematic analysis of the substance of party politics. A purely institutional focus limits us to a dissection of party systems that have been drained, gutted, and picked clean of the flesh and blood of politics--ideology, personalities, interests, ideas, platforms, slogans, images, issues--in short, the substance of political competition. Crucial aspects of the democratic process such as alliances and coalitions, policy choices, and polarization, cannot be understood well without considering the ideas, interests, and images of parties, in addition to their number and sizes. The next two sections will evaluate how ideological and polarized Latin American party systems are and describe the extent to which they lean to the left or the right.

Second, improvement of the conventional wisdom requires coming to grips with party-system change. We are all aware that most Latin American party systems change rapidly, especially in contrast to those of the industrialized north. However, the conventional wisdom has not yet fully realized how this variability affects our ability to describe and analyze them. The more volatile a party system is, the less sense it makes to generalize about any of its characteristics--fragmentation, ideological tendency, polarization, even volatility itself. A useful typology by country is out of the question: these party systems defy taxonomy. Averages over a long span of time tend to be poor indicators of the nature of a party system at any one point along the way, and averages over a short span of elections tend to be unrepresentative of other periods. Even generalizations about different periods of a country's history are inadequate because the periods of relative homogeneity are usually frustratingly brief, and because a periodization that is useful for describing one characteristic is rarely useful for describing any others. Most Latin American party systems are changing, and changing often, in several dimensions at once, all on staggered timetables. There is often, therefore, considerable uncertainty about what, if anything, is ‘typical' of the party system in any given country.

Figure 1 illustrates the nature of the problem more precisely using elementary statistics. The two axes represent the two most familiar characteristics of party systems--fragmentation (the effective number of electoral parties) and volatility.(1) Each ellipse in the figure corresponds to the entire 20th-century experience of one country, except for Brazil, which has separate ellipses for its democratic and authoritarian party systems. Each country's mean fragmentation and volatility lies at the exact center of its oval. Each oval traces the 95% confidence intervals corresponding to those means, with the vertical extremes indicating the interval for fragmentation, and the horizontal extremes indicating the interval for volatility. For example, we are 95% confident that the average of the effective number of parties in Uruguay lies between 2.25 and 2.67 (the vertical extremes), and that the average volatility rate in democratic Brazil lies between 29 and 51 (the horizontal extremes).

Figure 1

If one country's mean falls within another country's confidence interval, the two countries' means are not significantly different. Therefore, contrary to the conventional wisdom, we cannot say with confidence that 1) Mexico's party system has been less fragmented than Colombia's, 2) Costa Rica's has been more fragmented than Uruguay's or less fragmented than Bolivia's, 3) Venezuela's has been less fragmented than Argentina's, or 4) Chile's has been less fragmented than Ecuador's. Neither can we confidently say that 1) Venezuela's party system is less volatile than those of Peru, Chile, Bolivia, or Ecuador; 2) democratic Brazil's is more volatile than Ecuador's or Argentina's; or 3) Costa Rica's is more volatile than that of authoritarian Brazil.

In view of the uncertainty introduced by the variability of Latin American party systems, generalizations about countries should be avoided, and if unavoidable, should be made with greater caution, for example by reporting standard deviations or confidence intervals. Ideally, the units of analysis should be individual elections, not countries. The true values of the parameters of interest are known to a much higher degree of certainty when elections are the units because the variance for a single election is close to zero (depending only on the reliability of the indicator itself). Election-by-election data may seem less meaningful than a sweeping generalization about a larger swath of a country's history, but they are in reality more meaningful. Political actors probably care far less about averages than they do about the results of the last election, which defines their political reality for 2 to 5 years at a stretch.

Some conventional wisdom about ideology in Latin American party systems exists, although it has not been studied as systematically or rigorously as fragmentation or volatility. The longstanding perception of Chilean parties as highly ideological persists, as does the perception of parties in Ecuador and Bolivia as highly personalistic, nonideological, and clientelistic (Scully, 1995: 100; Gamarra and Malloy, 1995: 399; Conaghan, 1995: 436). Brazilian parties are widely believed to have fit this description before the coup of 1964, but Scott Mainwaring argues that ‘these old adages must be put to rest' for the post-1985 party system (Mainwaring, 1995: 375-6). Most of the other systems are thought to fall somewhere between these extremes. The Costa Rican party system is considered moderately ideological, as it has a clearly reformist large party (PLN) opposed by an almost equally large but ideologically heterogeneous opposition grouping (PUSC) (Yashar, 1995). In the early 1980s the same could have been said of Peru and Venezuela, but in both countries the traditionally large reformist parties have lost ground to independent candidates--Fujimori in Peru and Caldera in Venezuela (Graham, 1992; Coppedge, 1996). It has always been difficult to characterize the ideology of the large parties in Colombia in Uruguay because they are divided into factions, some of which are ideological and others, merely personalistic (Hartlyn, 1988; Gillespie, 1991). Finally, parties in Argentina and Mexico are considered ideological, but only if a cleavage other than left vs. right is recognized as dominant. In Mexico, the left-right dimension is subordinated to a division between democracy and continued official-party dominance (Domínguez and McCann, 1996). In Argentina the fundamental political divide separates Peronists from everyone else, although the nature of the division is controversial (McGuire, 1995; Ostiguy, 1997).

Before we can assess more accurately how ideological Latin American parties and party systems are, it is necessary to clarify three issues. First, personalism and ideology are not necessarily mutually exclusive qualities. Some of the most rigidly ideological parties in the world have been closely identified with, and tightly controlled by, strong personalities, and parties that are known primarily as vehicles for strong personalities may nevertheless stake out clear ideological positions. Second, clientelism and ideology are not necessarily mutually exclusive. Many successful parties all over the world trade personal favors for political support. Even in supposedly highly ideological Chile, party officials of all tendencies engaged in the same sorts of clientelistic activities (Valenzuela, 1977: 166). Clientelism is merely a means to build and maintain a power base; ideology, where it exists, is what guides what that power is used for. Many parties are to some degree clientelistic, to some degree personalistic, and to some degree ideological: these three qualities vary independently.

Third, for some purposes the ideological sophistication of party leaders is less relevant than the clarity of a party's image among voters. Sartori's definition of an ‘ideological mentality' as ‘a state of dogmatic impermeability both to evidence and to argument' (Sartori, 1969: 403) is a good one if we wish to understand a party's behavior at the elite level--its rigidity or willingness to compromise on policy or coalition partners. But if we wish to understand relations between parties and voters--the rationality of voter choice, the quality of representation, the possibility of holding elected officials accountable--then a different standard applies. Very few voters care about whether a party leader can debate the finer points of Althusser, Maritain, or Hayek. If they wish to vote ideologically, all they need is a sense of the approximate relative positions of the available parties on the ideological spectrum.

‘Sufficiently ideological' parties thus are parties that take clear, widely understood positions on a conventionally interrelated set of issues. Definitions of ‘left' and ‘right' do not always travel or age well: they can vary greatly from region to region, country to country, decade to decade, and even person to person if they are made very precise. The more diverse the regions to be analyzed, and the longer the span of time to be covered, the less specific the criteria for the left-right dimension can be, because dimensions that are relevant in only a few countries must be dropped from the cross-national criteria for comparison. Nevertheless, if the comparison is limited to one region and a manageable span of time, the criteria for the left-right dimension are usually well understood.

My recent classification of Latin American parties is useful for assessing how ideological the parties of the region are.(2) I began by drafting a classification of all the parties (loosely defined as any label reported in election returns) that contested lower-chamber elections in Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, Mexico, Peru, Uruguay, and Venezuela. The classification scheme identified parties with a position on a left-right spectrum (left, center-left, center, center-right, or right) and a religious-vs.-secular dichotomy; or, if that was not possible, it classified parties as ‘personalist,' ‘other' (e.g., regional, ethnic, environmental), or simply ‘unknown'. ‘Personalist' was therefore defined very strictly, so that only parties with no identifiable left-right position would qualify. I then sent this draft and my explicit coding criteria to 80 country specialists, asking for their advice in correcting any misclassifications. Fifty-three of the experts provided feedback, which I then used to make corrections.(3) The final version successfully classifies about 800 of the approximately 1,200 parties in the sample, which won 97 percent of the votes cast. Although a more rigorous classification is possible in principle, this classification is the most comprehensive and systematic one in existence for Latin American parties and, as will be shown below, its reliability is fairly high. Although any case study would provide a firmer basis for rating parties in any one election, there is currently no alternative to this classification for comparing Latin American party systems historically and across countries.

The classification provides two kinds of evidence about party and party-system ideology. First, it reports the percentage of the vote won by strictly personalist parties and candidates in each election. By this criterion, personalist parties are rather small and rare. The personalist vote is greater than 10 percent in only 23 of the 149 elections in the sample. The only country in which the personalist vote has been consistently greater than 10 percent is Ecuador. It is also high, as the new conventional wisdom suggests, in Peru 1990-5 and Bolivia 1989-93. However, it is also greater than 10 percent in the Chilean elections of 1932 and 1949-57. There were also surges of personalism in conventionally ideological Costa Rica in 1958, 1962, and 1974, and in Venezuela in 1968 and 1993. This criterion also rates personalism surprisingly low in Brazil--in fact, lower than in Chile, on average.

Because the criteria for ‘personalism' are so strict, however, it is prudent to consider the reliability of the classifications. Some parties seem nonideological because they are strictly personalist, but others seem nonideological because it is not clear where they stand on the ideological spectrum. In these latter cases the reliability of the classification is low. Some Latin American parties are very hard to classify, and this is reflected in a lack of consensus among the country specialists. A few are large, important parties--the Peronists in Argentina, the Mexican PRI in certain elections, and the Liberals and Conservatives of Colombia during much of their history. But because others are quite small, nearly insignificant parties, it would be misleading to report reliability statistics on individual parties. Instead, I base the analysis here on an aggregated measure of reliability for each election. Specifically, reliability is the proportion of the experts concurring with my final classification of a party, multiplied by the proportion of the vote won by that party, summed over all parties winning votes in that election.(4) In other words, reliability is a weighted average of the degree of concurrence with the classification of all the parties in an election, or the degree of concurrence with the average vote. For example, if three parties contested an election and all experts concurred with my classification of the largest party (A), 4 of 5 experts agreed on the second party (B), and 3 of 5 agreed on the smallest party (C), the reliability would be 86 percent, calculated as follows:

Party	Vote	Agreement	Product
A	.50	1.00	.50
B	.30	0.80	.24
C	.20	0.60	.12
			.86

A low reliability score could result from several factors besides an unclear ideological position on the part of the party.(5) However, it is very unlikely that a classification could have a high reliability score if it did not have a clear ideological position because any vagueness or inconsistency of position would create disagreements among the experts.

The mean reliability for this sample is 85.6, with a standard deviation of 12.6 and a median of 88.7. (These figures exclude Bolivian and Uruguayan elections because too few country specialists provided comments on these cases to calculate reliability.) If we establish a cutoff for reliability at 85 percent--equivalent to 17 out of 20 experts concurring on the classification of the average vote--then 63 percent of the elections were reliably classified and 37 percent were not.

The best standard for identifying sufficiently ideological party systems combines both criteria: reliability of 85 percent or more and a personalist vote of less than 10 percent. Table 1 sorts all the elections according to these twin criteria. By these stiff criteria, 55 percent of the Latin American elections in the sample (52 percent counting Bolivia and Uruguay) were sufficiently ideological; 15-16 percent probably had significant personalism; and 24-9 percent were not clearly ideological without being clearly personalist. These figures provide some support for the old conventional wisdom in that elections that were either personalist or not clearly ideological (or both) have been fairly common in Latin America, perhaps more common than in Western Europe. However, the new conventional wisdom is also vindicated because such elections have not been the norm and because some countries have had sufficiently ideological elections more frequently than others. It is no surprise that Chile ranks highest in this regard and Ecuador ranks lowest.

The new conventional wisdom is not fully vindicated, however. Table 1 shows that almost every country has experienced elections that deviated from its current reputation for being ideological or not. As already noted, Chile, Costa Rica, and Venezuela have had repeated bouts with personalism. The Brazilian party system may be volatile and fragmented, but it has not lacked for ideological clarity in the 1990s. The Argentine party system has a justifiable reputation for being sui generis due to controversy about what the Peronists stand for, but in the pre-Peronist era (1912-30) and during the period of proscribed Peronism (1958-65), this system was sufficiently ideological. The major Colombian parties have a reputation for being virtually indistinguishable in left-right terms, but this did not prevent the experts from concurring with classification of the Liberals in the center and the Conservatives on the center-right during the National Front years (1958-74) and in the 1990s. The ideological swings of Mexican presidents from 1970 to 1988 clouded perceptions of the position of the official PRI, but in the 1960s and the 1990s the relative positions of Mexican parties are clear enough. Peru's reputation for personalism is based on its experience before redemocratization in 1978 and the rise of Fujimori in 1990; but in the 3 intervening elections its parties possessed sufficient ideological coherence. And although there are no reliability figures for Bolivia, I suspect that it would turn out to be sufficiently ideological during the 1979-85 period as well, as the party system was very polarized in left-right terms during those years, as we shall see.

It is also hard to generalize about two other aspects of party ideology in Latin America--how skewed to the left or the right the party system is, and how polarized it is. Mean Left-Right Position (MLRP) measures the how far to the left or the right the average party was in each election, based on the left-right positions of all the parties and their shares of the vote. This indicator assumes that all parties classified left (whether Christian or secular) are approximately twice as far from the center as parties classified center-left; and right parties are twice as far to the right as the parties of the center-right. This assumption permits the calculation of MLRP as

where ‘XR' represents the percentage of the vote won by all the parties in the Christian right bloc, and so on for the other bloc abbreviations. This index would equal 100 if all parties were on the right, -100 if all parties were on the left, 50 or -50 if all parties were center-right or center-left, respectively, or zero if all parties were centrist, personalist, other, or unknown, or if the parties to the left perfectly counterbalanced the parties to the right.(6) In this sample, MLRP ranges from -42 (Peru 1985) to +69 (Brazil 1970), with an overall mean of 5.

Left-right polarization is a different aspect of party-system ideology. Here it is defined as the dispersion of the vote away from the relative center of the party system. The relative center can be farther to the right or the left than the absolute center as defined in the classification criteria, and is operationalized here as MLRP. The index of polarization (IP) makes the same assumptions about the positions of the blocs in a (-1,+1) range.(7) Its formula is

Table 2 categorizes elections by their mean left-right position and degree of polarization. ‘Polarized' party systems are those tending toward a bimodal distribution of the vote on the left-right spectrum. ‘Unimodal' party systems have a more prominent peak distribution near their relative center, often indicating a lack of meaningful competition among parties. ‘Flat' party systems have a profile that is more evenly distributed across the range of left-right options, and are therefore more competitive across ideological tendencies. Each of these types can also ‘lean' farther to the left or the right, or be ‘centered', in accordance with its mean tendency. Table 2 shows that no country's party system has consistently leaned in one direction, and the only party systems that have remained within one range of polarization are those of Costa Rica after 1953, Uruguay, and Venezuela, which are all flat (and not right of center). It may be no coincidence that these are three of the most stable democracies in the region.

Although Table 2 dramatizes the difficulty of generalizing about mean tendencies or polarization by country, it also identifies selected elections from different countries that are comparable in terms of mean tendency and polarization. For example, authoritarian Brazilian elections and several elections in authoritarian Mexico all fit the category of right-leaning unimodal systems, as do a couple of early, less competitive elections in Colombia and Ecuador. In the centered unimodal cell we find Argentina when it was dominated by Irigoyen, Bolivia when it was dominated by Barrientos, and contemporary Peru dominated by Fujimori. Some of the elections in other cells may appear to be strange bedfellows, as they lump together party systems that are institutionalized and inchoate, fragmented and not; but the apparent oddity of the combinations only underscores how independent one characteristic of party systems is from others. It is interesting to note that Chile 1969-73 was not the most polarized or the farthest-left-leaning system, but it was the most left-leaning of the polarized systems and the most polarized of the left-leaning systems. The combination of polarization with a leftist tendency may be hazardous to a regime's health: 3 of the 5 left- or center-left-leaning polarized cases succumbed to military coups before the next election, a far higher proportion than that in any other cell of the table.

There is some truth to the old stereotype that Latin American parties and party systems are poorly institutionalized. Volatility in 20th-century Latin America averages 21.3, compared to 8.6 in Western Europe.(9) However, this average masks great variation within the region that has been recognized for quite some time. In his 1976 classic, Parties and Party Systems, Sartori made a distinction between ‘structured' party systems and ‘fluid' ones (Sartori, 1976: 310-1). Twenty years later, Mainwaring and Scully made a similar distinction between ‘institutionalized' and ‘inchoate' party systems, and produced a summary index of institutionalization (Mainwaring and Scully, 1995: 19). As Table 3 shows, both ratings identify clear differences among the Latin American cases, and the two ratings agree almost perfectly on which systems have been well institutionalized and which have not.(10) Based on their combined judgments, the conventional wisdom is that Uruguay, Costa Rica, Chile, and Venezuela (and possibly Colombia and Argentina) have institutionalized party systems, while Bolivia, Ecuador, and Peru (and probably democratic Brazil) do not. There is no consensus on a few countries--Argentina, Colombia, and Mexico--largely because they are institutionalized in some respects but not others. Mainwaring and Scully (1995: 6-21) wisely examined many different aspects of institutionalization: electoral volatility, the difference between presidential votes and legislative seats won by parties, some evidence on the strength party identification, the strength of linkages between parties and social organizations, the percentage of legislative seats held by parties founded by 1950, the popular legitimacy of parties and elections, and the strength of party organizations. Here, however, I will focus on volatility alone because comparable historical data on other aspects is incomplete.

Table 3 reports confidence intervals by country to clarify the significance of the average differences. The conventional wisdom is basically correct about volatility, with three exceptions. First, there is a big gap within the more institutionalized group, between the extremely low mean volatility in Uruguay, Colombia, Mexico, and authoritarian Brazil and the significantly higher levels in Costa Rica, Chile, and Venezuela. Second, Argentina clearly belongs in the most volatile group, with democratic Brazil.(11) And third, Bolivia, Ecuador, and Peru's volatility rates have been so variable that they are not statistically distinguishable from any but the three least-volatile systems.

Party-system volatility is a good indicator of the outcome of all the things that cause change in a party system, which are manifold: 1) voter defections, 2) generational turnover in the electorate, 3) extension of the suffrage, 4) conjunctural variations in turnout, 5) party mergers and alliances, 6) party splits, 7) election boycotts, and 8) the proscription of certain parties. But it is not the only possible indicator of volatility: the same index can be calculated using the vote shares won by ideological blocs rather than parties. Such an index takes us closer to measuring volatility due solely to voter defections from blocs, which is a narrower but more coherent aspect of institutionalization. To take us closer still, we can remove the components of volatility due to ideological shifts and adjust for differences in the spacing of elections. The resulting statistic is adjusted bloc volatility (ABV):

where Bi,t is the percentage of the vote won by bloc i in election t, Ti is the number of years elapsed since the previous election, and Si,t is the vote share of party i that will shift to a different bloc in election t+1. ABV can be low even when party-system volatility is high if voters have firm loyalties to ideological blocs that transcend loyalties to specific parties: a looser but still useful sort of institutionalization.

One of the distinctive features of the Chilean party system is that it is supposedly structured by precisely the kind of firm identification with left, center, and right blocs that ABV is designed to capture (Valenzuela, 1978; Scully, 1992). But by this measure (last column of Table 3), Chile is no more ideologically structured than Costa Rica or Venezuela. Even more striking is the fact that democratic Brazil clearly belongs in this relatively institutitionalized group. Otherwise, bloc volatility yields the same groups as party-system volatility except that Ecuador is now more clearly extremely volatile.

Party systems have long been classified into types defined by the number and relative sizes of the parties (Duverger, 1954, Almond and Coleman, 1960; Sartori, 1976; McDonald and Ruhl, 1989). This practice has usually led to the classification of Mexico and Paraguay as dominant-party systems; Colombia and Uruguay as historically two-party systems; Costa Rica, Venezuela, and perhaps Argentina as two-and-a-half-party systems; and Bolivia, Chile, and Ecuador as multiparty systems. However, these labels sometimes have to be amended depending on the period being analyzed: authoritarian Brazil had a two-party system, Uruguay became more of a three-party system in 1971, Venezuela's two-and-a-half-party system seems to have been limited to elections between 1973 and 1988, and so on.

The new institutionalism tends to shun simple typologies in favor of continuous measures of fragmentation. The one reported here is Laakso and Taagepera's Effective Number of Parties, based on shares of the vote (ENPV). (See note 1 for its definition.) One can also calculate the effective number of ideological blocs (ENB) using the same formula, but substituting the share of the classified vote won by each bloc for shares won by parties.(12) ENB provides some indication of how meaningful the fragmentation of the party system is. Some fragmented party systems are also ideologically fragmented, but others are not; instead, there is more competition among parties that are very similar in left-right and Christian-secular terms. In these systems, some of the divisions among parties are programmatically superfluous, the products of politiquería (superficial pseudo-politicking).

According to means tests, we can safely say that party systems are more fragmented (in terms of ENPV) in Chile, Ecuador, and democratic Brazil than in the other 8 countries; and that they are less fragmented in Colombia, Uruguay, Mexico, and authoritarian Brazil than in Chile, Ecuador, Argentina, Venezuela, and democratic Brazil. In terms of ENB, the party systems of Ecuador and Chile are significantly more ideologically diverse than those of Bolivia, Peru, Costa Rica, Venezuela, and Argentina. Beyond these simple generalizations averages can be seriously misleading, so it is preferable to compare elections rather than countries. Table 4 cross-classifies party systems in each election by the number of blocs and the number of parties. The dispersion of each of the countries among multiple cells drives home the folly of trying to classify Latin American countries by the number of parties.

This table also groups together cases that are comparable with respect to the diversity of choices offered in elections. The dominant-bloc, dominant-party systems listed in the note at the bottom of the table--e.g., early authoritarian Brazil, Colombia early in the National Front period, and Mexican elections before 1985--presented voters with very limited and relatively meaningless choices, aside from the certainty about which party would win. The lower-right corner of the table defines the opposite situation: extreme multi-bloc, multi-party systems that offered voters extremely diverse choices and few clues about which was likely to emerge as the winner. This was the case in the Brazilian election won by the dark horse Fernando Collor de Mello, Chile at the beginning and end of the Ibáñez interlude, and several Ecuadorian elections. Finally, the three cells in the upper-right corner of the table contain the party systems in which voters had many choices, but not the most meaningful ones. In Argentina 1912-6 and 1922-6, this situation resulted from the fragmentation of the center into a variety of Radical parties; in Venezuela 1958-68, from the division of center-left among URD, Acción Democrática, and its splinters; and in Brazil 1945-50 and 1958-82, from the division of the center-right among the PSD, UDN, and a great many elite-led regional parties and alliances.

This article has been an exercise in description, but it conveys an important message about explanation: party systems as diverse and dynamic as these will not be easily explained. The explanatory factors in the standard toolkit--level of development, class structure, ethnic cleavages, demographic change, and electoral laws--change too slowly or gradually to capture much of the variance described here. Explanations that work in Latin America are more likely to concern factors that are more easily, rapidly, and completely manipulated by governments, party leaders, and other elites. These factors include party splits, mergers, and alliances; campaign tactics; programmatic shifts; perhaps short-term economic performance; and in isolated cases, election boycotts and the proscription of certain parties or candidates.

The diversity and dynamism also pose severe methodological challenges. In order to identify the causes of change, we must have some basis for knowing what the party system would have been like in the absence of any explanatory factor of interest; volatility often makes this extraordinarily difficult. Elaborating typologies or calculating averages will not take us very far. Our only options are either to delve into intensive and rich case studies, which avoid generalization altogether, or to undertake quantitative comparisons of many elections in a large sample over a long period of time, which risk superficiality. The quantitative strategy has two challenges of its own. First, the diversity across countries makes it hard to separate the causal impact of the various conditions that remain relatively constant in each country. And second, the different variance across countries (a type of heteroscedasticity) makes it imperative for analysts to standardize the data so that, for example, a small change in Uruguay is equivalent to a big change in Ecuador.

Although sound explanations will be devilishly hard to substantiate, it is important to try, because whether party systems are weak or strong, left or right, ideological or pragmatic, fragmented or monolithic, they have important consequences. The nature of party systems affects the meaning of elections, the quality of representation, the nature of economic policy choices, and the legitimacy and survival of governments and the democratic regime itself, especially in Latin America.