Abstract
This paper contributes to the debate about the ideological coherence of the Brazilian party system. Using discriminant function analysis of 20 years of surveys of Brazilian legislators, we find that the party system now exhibits relatively little coherence. Though the Worker's Party (PT) is clearly distinct, no clear ideological differences exist between the placement of the system's three other main parties. Moreover, the spatial distance between the PT and the other parties is diminishing over time. Given the importance of a coherent ideological map to any consolidated party system, we question the notion that the Brazilian party system has gradually consolidated. Indeed, our results suggest the opposite: in recent years the Brazilian party system has become relatively more “inchoate.”
Introduction
Since Brazil's redemocratization in the 1980s, debate over the nature of the country's political parties has gone through two phases. In the first phase, scholars settled on the position that Brazil's parties were weak and its party system was “inchoate” (Mainwaring 1999). This view gained traction initially in large part because it resonated with longstanding notions about Brazilian politics, history, and culture. Consider the first sentence of Lamounier's influential book Partidos e Utopias:
There is, in Brazilian political culture, a profound malaise in relation to political parties; more than this, a generalized conviction exists that the country has not yet developed a viable party system, and that it does not seem to exhibit the characteristics that would permit the process of sedimentation of such a structure (1989: 19, emphasis added).
Lamounier's argument is rooted in the hypothesis, found in Weber (1919 [1958]) and Lipset and Rokkan (1967), that parties and party systems only consolidate under certain social-structural conditions.
In particular, Lamounier drew attention to the combination of social-structural conditions that boded poorly for strong parties and an institutional environment that politicians deliberately designed to derail the formation of strong parties. In terms of the cultural context, consider Carvalho (1987) or DaMatta (1979), who both argued that Brazil's republican institutions have long masked a patrimonial and authoritarian culture that does not support the emergence of autonomous mass civil-society institutions. Diniz (1982) and Hagopian (1996) likewise emphasized the persistence of clientelism and personalistic politics, rather than the emergence of deep ideological or other political cleavages during Brazil's transition to democracy. Almeida's recent book (2007) reaffirms this high degree of cultural continuity.
In terms of the institutional context, Souza (1976) emphasized the path-dependent impact of Vargas-era politicians’ efforts to deliberately weaken civil society organizations, by co-opting autonomous agents of popular representation and incorporating them into the state. Similarly, Schwartzman (1982) highlighted the historical importance of authoritarianism, centralization, and the state to explain the weakness of representative institutions throughout Brazilian history.
Thus the reason Mainwaring's view had such appeal was because decades of state manipulation of popular interests, the chaotic experience with mass democracy from 1945 to 1964, and the manipulation and control of mass politics during Brazil's military dictatorship from 1964 to 1985 provided little reason to believe that strong parties would emerge following redemocratization. Real-world events lent support to this expectation: Partysystem fragmentation and electoral volatility were among the highest in the world; party-switching was rampant; the two main candidates for president in 1989 were supported by relatively new and insignificant parties while the main government and opposition parties during the dictatorship failed to mount serious candidacies; voters hardly knew which parties existed.
With this background, it is relatively unsurprising that a narrative emerged in the late 1980s highlighting partisan weakness in contemporary Brazil. Abranches's (1988) seminal article, along with the research of Abrucio (1998), Mainwaring (1999), Ames (2001), and Samuels (2003), among others, bolstered the conclusion that virtually every facet of Brazil's political institutions worked against the emergence of a consolidated party system marked by the existence of strong party organizations with coherent ideological profiles. Reflecting this view, Power concluded that, “Of all the institutional variables in Brazil, it is the party system that has generated the most dissatisfaction and the loudest calls for reform” (2000: 28).
Yet even as these scholars were compiling evidence of partisan weakness, other scholars began to challenge this narrative. Most prominently, Figueiredo and Limongi (e.g., 1995, 2000) argued that parties were strong if one looked at cohesion on legislative roll-call votes. Other scholars pointed to a decline in electoral volatility and in party-switching, and have extended this argument into other relevant areas such as party strength in the electorate and party organizational strength (e.g., Lyne 2005). Hagopian, Gervasoni, and Moraes (2009), for example, point to legislators’ changing self-perceptions about the programmatic and/or ideological content of their election campaigns as evidence that Brazil's non-leftist parties have set down deeper roots in society and have evolved away from their clientelistic and largely non-ideological roots, while Power and Zucco (2009) argue that the party system has acquired some degree of ideological coherence. This view of Brazil's parties takes the position that party weakness is exaggerated, that the party system is consolidated rather than incoherent, and that no political reform is necessary because Brazilian democracy functions at least as well as others in the region, if not better.
This debate has engaged so many scholars because – as Lamounier's quote implies – it is not “just” about Brazil's parties but touches on the question of how observers should interpret the nature of contemporary Brazilian politics in a broad sense. In this paper we side with those who believe that Brazil's parties remain rather weak. Contrary to Figueiredo and Limongi's claims, Carey has recently shown that Brazil's legislative parties remain relatively undisciplined in comparative perspective (2009: 108). More importantly, party organizations still do not perform the core tasks that parties in most other democracies do, such as directly controlling nominations for legislative office and distributing campaign finance. Finally, the party system is relatively incoherent from the voters’ point of view: with the notable exception of the Worker's Party (PT), no party today can claim more than 5 percent of voters as “partisans” (Samuels 2009), meaning that political campaigns remain more personalistic than programmatic. In short, in terms of organizational coherence and connections with voters, Brazil's parties do not resemble anything close to the ideal-type “strong” party, and its party system lacks the sort of coherence that scholars associate with collective responsiveness and accountability.
The Question of Systemic Ideological Coherence
In what follows we engage a key facet of this debate: the question of whether Brazil's party system reflects a reliable, coherent ideological map. As Power and Zucco noted, “Quality political representation requires a party system with a reasonable level of ideological differentiation” (2009: 218). After all, if voters cannot differentiate parties from one another, then democratic representation and accountability will suffer (see also Rosas 2005).
There are several ways scholars can “map” a party system: through the use of expert surveys, with roll-call data, or by comparing the content of party manifestos, for example. Table 1 represents the conventional way scholars array Brazil's parties on the left-right spectrum, using expert opinion (Power and Zucco 2009; see also Wiesehomeier and Benoit 2009).
Approximate Ideological Ordering of Brazilian Parties
Note:
Parties that no longer exist.
As a check on experts’ opinions and in an effort to confirm this general portrait of Brazil's party system, Power and Zucco (2009) used information that legislators themselves provided: the perception of their own party's placement on the left-right spectrum along with their perception of other parties’ placement. Power has surveyed Brazilian federal deputies once per legislature since 1990. To measure legislators’ positions, Power and Zucco (2009) used two questions in five of Power's surveys: 1) legislators’ self-placement on an ideological scale ranging from 1 (left) to 10 (right), and 2) legislators’ placement of all the other parties on the same scale. The answers to these questions allowed the authors to scale parties, controlling for systematic bias in legislators’ self-assessment and their assessment of other parties.
Power and Zucco summarized their findings simply: “The main legislative parties can be arrayed clearly on a classic left-right scale and […] the overall ordering has been relatively stable across time” (2009: 218). Figure 1 replicates Power and Zucco's findings: In terms of the main parties, legislators put the PT on the left, the Brazilian Social Democracy Party (PSDB) to the right of the PT but slightly to the left of the Brazilian Democratic Movement Party (PMDB), and all three of those parties to the left of the Liberal Front Party (PFL – now Democrats or DEM). This conclusion echoes the views of Figueiredo and Limongi (2000) and Leoni (2002), for example, who came to similar conclusions about parties’ placement using roll-call data from the 1990s. Given this, Power and Zucco (2009) conclude that Brazil's party system today qualifies as one in which a reasonable level of ideological differentiation exists.
Comparable Party Position Estimates, 1990-2005
We question this conclusion, using the answers to different questions from Power's surveys. Two points spurred our investigation. First, among the main political parties the picture of ideological coherence is murkier than Power and Zucco (2009) suggest. Power and Zucco even acknowledge that in the three most recent surveys Brazilian legislators placed the PSDB to the right of the PMDB, or failed to distinguish those two parties from each other at all (2009: 228). This raises the question of whether there is relatively less coherence to the party system today than there was even in the early 1990s, during the heyday of the “inchoate parties” thesis.
Second, Power and Zucco's empirical results suggest that by 2005, the party system had coalesced into two blocs, with the Communist Party of Brazil (PCdoB), PT, Brazilian Socialist Party (PSB), Socialist People's Party (PPS), and Democratic Labour Party (PDT) all indistinguishable from one another on the left and the PSDB, PMDB, Brazilian Labour Party (PTB), Liberal Party (PL), PFL, and Progressive Party (PP) – also indistinguishable from one another – occupying the center, with the right side of the political spectrum actually empty. In a telling aside, Power and Zucco recognize that “more parties have become indistinguishable in recent years” (2009: 10). Of course, such a conclusion does not jibe with the neat portrait they paint in Figure 1 and it clashes with the idea that the Brazilian party system has come to exhibit a “reasonable level of ideological differentiation.”
To further explore the question of the evolution of the ideological map of the Brazilian party system, we use different information from the same surveys Power and Zucco (2009) explored. Instead of relying on perceptual data – legislators’ perceptions of parties’ left-right placement – we use attitudinal data drawn from legislators’ responses to a series of other questions. In the next section we describe our method. We then report our results, focusing on the relative positions of the four largest parties in Brazil's legislature over the last 20 years: the PT, PSDB, PMDB and PFL. Our findings can be summarized in the following way: Brazil's party system is characterized as “the PT vs. the rest.” This could be seen as a form of coherence, but high fragmentation means that the party system does not offer Brazilian voters the “reasonable degree of ideological differentiation” that might bolster democratic responsiveness and accountability.
Data and Method
Our empirical analysis utilizes data from the Pesquisa Legislativa IUPERJ-Oxford (PLIO) surveys of Brazilian legislators that Power has administered six times between 1990 and 2009. Following Rosas (2005), we employ discriminant function analysis (DFA) to assess parties’ relative positions in ideological space. This method uses a set of independent variables (in this case, the different responses that legislators provide to questions on Power's surveys) and a categorical dependent variable (the party affiliation of each legislator) to construct a set of new variables (referred to as discriminant functions or dimensions) that can sort the legislators into groups (the political parties to which they belong) more or less successfully.
DFA generates two sets of coefficients for each independent variable on each dimension: a standardized discriminant function coefficient and a canonical structure coefficient (also standardized). The standardized discriminant function coefficients, which are partial coefficients reflecting the association of the independent variable with the discriminant function when controlling for the other independent variables in the model, maximize the distance between the group means on that particular discriminant function. These coefficients allow us to calculate a discriminant score for each observation (i.e., for each legislator) along each dimension, which in turn allows for the calculation of group centroids (i.e., the mean discriminant score for each group of legislators).
The canonical structure coefficients, on the other hand, reflect the uncontrolled correlation between the independent variable and the discriminant scores associated with the discriminant function. As in factor analysis, these canonical structure coefficients are often used to assess the meaning of each dimension (Klecka 1980) – a question we turn to later.
We first identified survey questions that appeared in at least three consecutive waves of the PLIO survey that might distinguish legislators from different parties. A total of 33 questions fit this criterion; these are our independent variables. Appendix 1 provides the full text of these questions. Because not all of these questions were asked in each of the six PLIO surveys, we created three separate samples of questions and ran our tests on each sample. 1 To the extent that our results are consistent across all three samples, we gain confidence in our conclusions.
We also ran DFA on a fourth sample that included data from all six waves of the PLIO survey. Unfortunately, only seven of the 33 relevant questions appear in all six waves of the survey. Given this, results from this model did not discriminate between the four groups (parties) and we did not include it in the results we present.
Sample 1, which includes data from the four most recent waves of the PLIO survey, allows for the greatest number of questions (14) while including data from the 2009 survey, which gives the most up-to-date picture of the party system. Sample 2 excludes the 2009 survey, but it allows for more robust inter-temporal comparisons than does Sample 1 because it includes five waves of the survey and because it includes a greater number of questions (20). Sample 3 maximizes the number of questions by restricting the analysis to the 1997, 2001, and 2005 waves of the survey – the only three surveys in which all 33 relevant questions appeared. The first three rows of Table 2 describe the surveys, number of questions, and number of respondents in each sample.
Discriminant Function Analysis Summary Statistics
Our analysis focuses only on the four largest parties in Brazil's legislature during this period: the PT, PSDB, PMDB, and PFL. Three additional parties – the PDT, the PTB, and the PP (together with its various predecessors) – were initially considered for inclusion in our analysis, but they were dropped due to methodological and substantive concerns. Methodologically, the inclusion of these three parties would have greatly inhibited our ability to produce meaningful inter-temporal comparisons because of small party/survey sample sizes.
Substantively, the reason for excluding the PDT, PTB, and PP from our analysis concerns the size of their legislative contingents. The PT, PSDB, PMDB, and PFL combined to account for 72 percent of the seats in the Legislature in 2001, and for 58 percent in both 2005 and 2009. By comparison, the PDT, PTB, and PP combined to account for only 16 to 20 percent of the seats during this same period (other, even smaller parties comprise the rest). For this reason, the question of whether these three parties – or any of the many other minor parties, for that matter – are ideologically coherent is peripheral to the debate regarding the relative ideological coherence of the Brazilian party system as a whole. If Brazilian voters are unable to distinguish the ideological placement of the four main parties, it is even less likely that they would be able to produce a coherent map of the system that includes all the smaller parties as well. Therefore, we focus on the question of the ideological coherence of the system's main parties.
Table 2 provides the summary statistics for the discriminant function analysis. The “hit ratio,” which reports the percentage of legislators’ party affiliations that the model correctly predicts, and the proportional reduction of error statistic (PRE), which measures the extent to which the classification of respondents produced by the model reduces erroneous predictions, both indicate the relative discriminating power of the independent variables in each model. The “proportion” provides the percentage of the total discriminating power that each dimension explains. The fact that the first dimension accounts for 79 to 88 percent of the discriminating power in our three samples suggests that legislators’ responses to the surveys can be classified largely along a single dimension. The party centroids reflect relative placement along a line. For purposes of discussion we shall refer to lower numbers as being on the left and higher numbers on the right; later, we discuss the “content” of this dimension and justify our use of the party centroids as indicators of the parties’ placement on a left-right ideological continuum.
Party Positions in the Brazilian Congress, 1990-2009
At this point we turn to our findings. Table 3 presents the party centroids and standard errors on the first dimension for each party in each pooled sample. This provides three different portraits of the relative placement of the four main parties in the Brazilian Congress over the entire 1990-2009 period.
Party Centroids and Standard Errors (Pooled Data)
As expected, the position of the PT is consistently far to the left of the other three parties. Yet in contrast to many observers’ impressions, the PMDB is positioned to the “left” of the PSDB in two of the three samples. Moreover, although the PFL appears to be furthest to the right in all three samples, its position is statistically indistinguishable from the positions of both the PSDB and the PMDB in Sample 1. Thus, Table 3 suggests that while the PT is clearly distinct from the other three parties, the relative positions of the PSDB, the PMDB, and the PFL cannot be ascertained with a high degree of certainty, no matter which set of PLIO surveys we use. 2
To ensure that these findings were not unduly influenced by the exclusion of the PDT, PTB, and PP from our analysis, we also ran DFA on pooled data incorporating all seven parties. In each sample, the order of the party positions of the PT, PSDB, PMDB, and PFL remained unchanged.
The inability to clearly distinguish between these three parties could be an artifact of using pooled data; it is possible that party positions have changed over time. To address this possibility we now turn to the discriminant scores for each party in each survey within our three samples. Figure 2 presents the party centroids (bracketed by 95 percent confidence intervals) for each survey in Sample 1. Even though this sample includes only 14 questions, the discriminant scores produce a significant reduction in classification errors, suggesting that this sample sheds light on the relative placement of Brazil's four main parties. The patterns in Figure 2 suggest four things:
Single-Survey Party Centroids, Sample 1, 1997-2009
As in the analysis of the pooled data, the PT is consistently on the left of the spectrum in each survey in Sample 1. Indeed, the difference between the PT's mean discriminant score and those of each of the other three parties is statistically significant in each survey. Moreover, even when the gap between the PT and the nearest party to it is at its smallest (2.20 standard deviations, the difference between the PT and the PMDB in 1997), this difference between “the PT and the rest” is much larger than the largest gap found between any of the other three parties in any survey year (0.94 standard deviations, the gap between the PMDB and the PFL in 1997).
As other scholars have suggested using different indicators (e.g., Power and Zucco 2009), the PT displays a striking move towards the center following Lula's first presidential victory. Between 2001 and 2005 the mean position of the PT legislators shifted 0.96 standard deviations to the right – the largest movement observed by any party from one survey to the next.
At the other end of the political spectrum, the PFL shows some signs of a fairly steady movement towards the center. Between 1997 and 2009, the party's mean position shifted 0.34 standard deviations to the left.
In contrast to the implications of several scholars’ research (e.g., Hagopian, Gervasoni, and Moraes 2009; Power and Zucco 2009) the positions of the PFL, PMDB, and PSDB are fundamentally indistinguishable from one another. Indeed, of the 12 party/survey dyads formed by these three parties (e.g., PSDB-PMDB 1997, PSDB-PFL 2001, or PMDB-PFL 2009), the mean discriminant scores are significantly different in only a single instance – the PMDB is significantly to the left of the PFL in 1997. The PMDB is positioned to the left of the PSDB in each of the first three surveys (1997-2005), but the differences between the party centroids for these two parties are miniscule (less than 0.12 standard deviations) in every survey after 1997. Moreover, the position of the PSDB cannot be distinguished from the positions of the PMDB or the PFL in any of the four surveys. Finally, even setting aside questions of statistical significance, the simple ordering of these three parties’ positions along the left-right dimension raises doubts about claims that there is a coherent “ordering” of parties across Brazil's political spectrum. As Table 4 shows, the order in which these three parties are positioned has been highly inconsistent, and in none of the four surveys does it follow the conventional wisdom, which places the PFL on the right, the PSDB on the left, and the PMDB in the center.
Relative Positions of the PT, PSDB, PMDB, and PFL, Sample 1
Based on the results from Sample 1, it appears that the Brazilian party system is consolidating into a system best described as “the PT vs. the rest” – and that the difference between the PT and “the rest” is diminishing. Of course, it is reasonable to wonder whether these conclusions might be unduly influenced by the relatively small number of independent variables included in this particular sample. In order to test the robustness of these findings, Figure 3 presents results obtained from the analysis of Sample 2, which includes more questions and more waves of the PLIO survey than does Sample 1.
Single-Survey Party Centroids, Sample 2, 1990-2005
Figure 3 further supports the four initial conclusions stated above. The PT is again located far (1.80 to 4.58 standard deviations) to the left of the other three parties, and it once again exhibits movement towards the center of the political spectrum following Lula's election (moving 1.18 standard deviations to the right between 2001 and 2005). The relative positions of the PSDB and PMDB are statistically indistinguishable in all five surveys. The PFL, whose movement towards the left appears less gradual in this sample than in Sample 1, was located significantly to the right of the PMDB in each of the first three surveys, and it was significantly to the right of the PSDB in 1990, 1997, and 2001, but it is indistinguishable from either of these parties in 2005 – again suggesting that the party system now exhibits less ideological consistency than in previous years.
Finally, we turn to Sample 3, which includes the full set of 33 questions. We first note that the discriminant scores produced by the model for Sample 3 correctly predict the partisan attachment of 80 percent of the respondents, which represents a substantial 70 percent reduction in classification errors and suggests that, as a group, the PLIO questions included in this model have a large degree of discriminating power. To what extent do the single-survey party centroids from Sample 3 – shown in Figure 4 – support our analysis of Samples 1 and 2?
Single-Survey Party Centroids, Sample 3, 1997-2005
First, as before, the PT is far to the left of the other three parties. The difference between the PT and each of the other three parties is again statistically significant in all three surveys in this sample, and these differences are much larger (ranging from 3.62 to 6.17 standard deviations) than are the gaps that separate the other three parties (which range from 0.23 to 1.88 standard deviations).
Although this sample limits our ability to draw any firm conclusions about party movement because it contains only three surveys, the 1.09 standard deviation difference between the PT's position in 2001 and its position in 2005 again supports our conclusion that the PT appears to have shifted to the center during Lula's presidency. As for the PFL, although it appears to be inching towards the center, the differences in this party's position from one survey to the next are neither statistically nor substantively significant.
Finally, the most interesting difference between Figure 4 and the previous two figures concerns the relative placement of the PFL, the PMDB, and the PSDB. Whereas the ordering of these three parties is highly inconsistent in the models that correspond to Samples 1 and 2, it is uniform when we use Sample 3 – in all three surveys (1997, 2001, and 2005), the PSDB is to the right of the PMDB and to the left of the PFL. However, the distances that separate these three parties are not great. Indeed, the PSDB and the PMDB are statistically indistinguishable from one another in 2001 and 2005, while the PFL and PSDB are statistically indistinguishable from one another in 2005. Given this, the results of our analysis of Sample 3 do not contradict our conclusion that the Brazilian party system appears to be slowly consolidating into a system of “the PT vs. the rest.” 3
To ensure that the results reported for Sample 3 are not driven by partisan differences regarding legislators’ evaluations of the Collor and Cardoso presidencies or their evaluations of the privatizations enacted during Cardoso's presidency, we ran a series of restricted models excluding each of these three questions individually and in tandem. The hit ratios and the patterns of canonical structure coefficients in these restricted models differ very little from those of the full model. As for our estimated party placements, although the single-survey party positions do move about slightly from one model to the next, the overall conclusions remain the same: the PT is significantly and substantially to the left of the other three parties and the PFL is significantly to the right of the PSDB and PMDB in 1997 and 2001 in all of the restricted models. Although the relative positions of the PSDB and PMDB party centroids in 2001 and 2005 vary depending on which combination of variables are excluded, these two parties remain statistically indistinguishable in these two survey years in each of the restricted model.
In sum, these three different samples paint three relatively similar pictures of the Brazilian party system. First, given that its position is significantly different from all three other parties in every survey in each of our three data samples, we can confirm the conventional wisdom that the average PT legislator is positioned ideologically to the left of legislators from the other three main parties. However, PT legislators have on average moved toward the center of the political spectrum since Lula assumed the presidency.
In addition, our analysis of the positions of the PSDB, PMDB, and PFL contradicts Hagopian, Gervasoni, and Moraes, who argue that “we should not be using the term ‘catchall’ any longer as synonymous with Brazilian parties of the center and right” (2009: 381). If Hagopian, Gervasoni, and Moraes were correct, legislators from the PSDB, PMDB, and PFL would tend to provide distinct answers to Power's survey questions, and consistently so. However, the difference between the party centroids of the PSDB and the PMDB is statistically significant in only one of 12 dyads (the 1997 survey in Sample 3). These two parties are indistinguishable. Moreover, although the PSDB is significantly to the left of the PFL in six of 12 dyads and the PMDB is significantly to the left of the PFL in eight of 12 dyads, these three parties are essentially indistinguishable from one another in 2005 and 2009. 4 That is, our findings – which use more recent data than Hagopian, Gervasoni, and Moraes (2009) – suggest that membership in the PSDB, PMDB, and PFL is now interchangeable. The overall picture that emerges is of a two-bloc system, with the PT on the center-left and the PSDB/PMDB/PFL in the center. In short, legislators’ own attitudes suggest that the Brazilian party system has consolidated into “the PT and the rest.”
Of the 12 dyads formed by these three parties in 2005 and 2009, only the difference between the PMDB and the PFL in Sample 3 is statistically significant.
Identifying the Difference
What does the first discriminant function “mean?” Earlier in this paper we argued that one of the virtues of using attitudinal rather than perceptual data to analyze party placements on a left-right scale is that such data allow us to examine the substance of what we have been calling the left-right spectrum. That is, in contrast to (for example) Power and Zucco's (2009) use of self-reported ideological placement plus assessment of other parties’ placements, the canonical structure coefficients generated by our three models can be used to assess the substantive meaning of the differences between parties’ placement on the left-right spectrum. In this section we explore the issues that classify legislators into parties.
Table 5 clusters the questions into five categories and presents the canonical structure coefficients for each of our three models. Researchers who examine canonical structure coefficients to identify the independent variables that are most strongly correlated with the discriminant function typically employ an arbitrary cutoff point to determine which variables merit the greatest amount of attention. Rather than employing a single cutoff point for all three models, we employ a cutoff ratio; the coefficients that appear in italics in Table 6 are those with an absolute value that is at least 40 percent greater than the average of the absolute value of the canonical structure coefficients for that model. 5
Canonical Structure Coefficients
What Type of Economic System is Most Adequate for Brazil?
In our case, the use of a single cutoff point is inadvisable because there is an inverse relationship between the average of the absolute value of the canonical structure coefficients for each model and the number of discriminating variables included in the model. That is, the more questions included in the sample, the smaller the coefficients.
The first two variables that exceed our cutoff ratio both come from a group of questions about the relationships between the state and the private sector and between civilian and military authorities, and thus both address policy issues that are strongly associated with conventional notions of what constitutes the left-right political spectrum in Brazil. First, when asked in the 1997-2009 waves of the PLIO survey what type of economic system is most adequate for Brazil – 1) a predominantly market economy, 2) an equal distribution between public and private ownership, 3) the prevalence of state ownership without eliminating private business, or 4) the complete absence of private capital in the principal economic sectors – legislators provided the responses displayed in Table 6.
The party means on this variable are highly consistent across time (the variation between the highest and lowest single-survey means does not exceed 11 percent for any of the four parties, and none of the parties display any consistent directional trends); therefore, we can analyze the pooled data for this variable. As one would expect, members of the PT were the least likely to identify a predominantly market economy as the ideal for Brazil, and they were the most likely to favor both the prevalence of state ownership and the outright elimination of private capital. Members of the PSDB are more likely to favor a predominantly market economy than are members of the PMDB, but their position is indistinguishable from the position of the PFL. 6
Results from t-tests with unequal variances run on the mean score on this variable confirm that the difference between the PSDB and the PMDB is statistically significant (p-value 0.0055), while the difference between the PSDB and the PFL is not (p-value 0.21).
The second question asked whether legislators favor giving the Armed Forces the right to intervene to guarantee internal order. As Figure 5 illustrates, a clear “PT versus the rest” division emerges. Members of the PT overwhelmingly respond negatively to this question; in all, only 5 percent of petistas (and none of them prior to 2005) supported giving the Armed Forces this right. Support for allowing the Armed Forces the right to intervene in internal affairs is significantly greater amongst members of the PSDB (43 percent), PMDB (54 percent), and PFL (70 percent). 7 One might expect this ordering of the parties given the PSDB's origins as a center-left faction of the PMDB and the PFL's origin in the former governing party under the military dictatorship, ARENA. However, as Figure 6 illustrates, this ordering has dissipated over time. When we exclude the 1990 and 1993 surveys, the partisan order appears to remain the same – PT (6 percent), PSDB (51 percent), PMDB (52 percent), PFL (61 percent) – yet the PSDB, PMDB, and PFL become statistically indistinguishable from each other. 8 Indeed, Figure 6 reveals that as of 2009, there is absolutely no difference in terms of sentiments about civil-military relations between the former government party and the former left-wing faction of the opposition party.
Support for Allowing the Armed Forces the Right to Intervene in Domestic Affairs
Percentage of Legislators who Attribute their Electoral Success to their Party's Organization and/or its Reputation
Results from t-tests with unequal variances confirm that the differences between adjacent parties on this pooled data are all statistically significant.
Here, the p-values from t-tests with unequal variances on the differences between adjacent parties are 0.90 for the PSDB-PMDB difference and 0.20 for the PMDB-PFL difference.
Our second group of questions focuses on evaluations of previous governments and government policies, which should again correlate with parties’ general placement on the left-right spectrum. Here, we found three questions strongly correlated with the first discriminant function. First, when asked (in 1997-2005) to evaluate the privatization of government enterprises on a 10-point scale, members of the PT predictably provided strongly negative evaluations: a mean of 3.3, considerably lower than those for the PMDB (5.9), PFL (6.9), and PSDB (7.1). (The PSDB's legislators perhaps offer the most positive evaluation of privatization because the bulk of such reforms were undertaken under PSDB President Fernando Henrique Cardoso.)
The other two questions concerned evaluations of the Collor and Cardoso presidencies. Collor's presidency was rated most negatively by the PT (2.0), followed by the PSDB (3.2), PMDB (3.8), and PFL (4.1). Cardoso received higher marks across the board, but there is still a significant gap between the evaluations offered by members of the PT (4.4) and the ratings given by legislators from the other three parties (PMDB 6.8, PFL 7.5, PSDB 8.1). None of these three variables present indications of any significant change over time.
The third group of questions focuses on legislators’ explanations for their own electoral success – a set of questions that is closely related to the sorts of issues that concerned Hagopian, Gervasoni, and Moraes (2009). Here, responses to three questions exceeded our cutoff ratio. First, when asked which factor was most important to their election, their party organization and/or its reputation or their own individual competence and political activity, PT legislators were far more likely (64.5 percent) to credit their party affiliation for their electoral success than were members of the PMDB (21.8 percent), PFL (10.2 percent), or PSDB (8.3 percent). However, as Figure 6 shows, PT legislators’ opinions on this matter have changed over the past two decades. With the exception of an increase in the 2001 survey, the percentage of PT legislators who attribute their election to the party's organization or its reputation has experienced a steady and steep decline, falling from 100 percent in 1990 to 40 percent in 2009. In contrast, the responses for the other three parties are fairly low and stable. These findings, which come straight from the mouths of legislators themselves, directly contradict research using either perceptual data or other information (e.g., Hagopian, Gervasoni, and Moraes 2009; Lyne 2005) that suggests party labels and/or party organizations have grown relatively more important to individual legislators since redemocratization.
A second, similar question asking legislators to rate (from 1 to 10) how important their party organization was to their electoral success elicited a similar response; the mean rating offered by members of the PT (7.7) is significantly higher than that offered by members of the other three parties (PMDB 6.0, PSDB 4.9, PFL 4.5). Again, however, as Figure 7 reveals, there is evidence of a steady (though less precipitous) decline in the degree to which PT legislators credit the party organization for their success; the mean rating offered by PT members fell from 8.8 in 1990 to 7.2 in 2005 (this question was not asked in 2009).
Rating of the Importance of Party Organization to Legislators’ Electoral Success
Finally, a question asking legislators to rate (again from 1 to 10) the importance of the contribution made by family traditions and/or historical loyalties to their electoral success also finds a substantial difference between the PT (3.3) and the other parties (PSDB 5.4, PFL 5.9, PMDB 6.4). If arguments about “party consolidation” were correct – if Brazil's main political parties have become more ideologically coherent – we should see evidence of an across-the-board decline in the importance of such personalistic political ties, even if the relative importance of family and personal connections continued to vary across parties. However, as Figure 8 shows, personal connections have grown more important for PT legislators, and they have remained highly important for legislators in the other parties. To the extent that legislators in “consolidated” party organizations should believe that their party label and organization are important while personal factors are relatively less important, the findings from these three questions suggest that PMDB, PFL, and PSDB legislators do not believe that they belong to “programmatic” parties, and they also suggest that the programmatic nature of the PT may have declined over the last decade.
Rating the Importance of Family Traditions and/or Historical Loyalties to Legislators’ Electoral Success
None of the questions that exceed our cutoff ratio come from our fourth group of questions, which concern the rules of political competition and the exercise of political power. However, two questions about the legislators’ relationship with their party and constituency do exceed the cutoff. First, asked whether they should always vote in line with their party's indications or with their individual beliefs, members of the PT (92 percent) were much more likely to support their party than were members of the PSDB (63 percent), PMDB (54 percent), or PFL (45 percent). If arguments about the consolidation of the Brazilian party system were correct, we should see an upward trend on this question over time for all parties. However, as Figure 9 shows, this is only the case for PSDB and PFL legislators, who are now somewhat more willing to say they “should” vote with the party leadership. In contrast, support for party leadership on roll-calls in the PMDB in 2009 is exactly where it was in 1990, and support within the PT for the notion that legislators should vote according to the indications of their party's leadership has actually declined slightly since 2001.
Should Legislators Vote According to the Indications of their Party Leadership?
Moreover, support for the notion that legislators now are more likely to bend to the will of party leadership evaporates when we consider a related question, which asked legislators not what they should do but how they actually behave when there is a conflict between their region's needs and their party's position – do they vote with their party most of the time, or do they base their vote on their region's needs? This question generated one of the greatest gaps between the PT and the other parties. Whereas 81 percent of the members of the PT report voting for the party's interests most of the time when these interests conflict with the needs of their region, only 26 percent of PSDB legislators, 19 percent of PMDB legislators, and 16 percent of PFL legislators choose their party over their region. Moreover, as Figure 10 reveals, contrary to the trend reported by Hagopian, Gervasoni, and Moraes (2009: 371), our data show that from 1997 through 2009 legislators from all four major parties grew less party-oriented and more focused on local interests.
Percent of Legislators Choosing Party Interests when they Conflict with Local Needs
In the Brazilian context, the set of questions explored above are all conceptually related. Questions about the extent to which the state should intervene in the economy and about the role of the Armed Forces have long been prominent topics of political debate throughout Latin America. The distinct positions taken by various political actors were employed as indicators of their placement on the left-right ideological continuum. In particular, legislators’ responses to the survey question regarding the type of economic system that would be most appropriate for Brazil and their evaluations of the process of privatization initiated by President Cardoso both relate to the economic-distributive divide identified by Kitschelt et al. (2010) as the most important dimension of programmatic party competition in contemporary Latin America. This provides good reason to believe that the answers that legislators from a given party provide represent a decent approximation of the ideological placement of that party relative to other parties.
Furthermore, questions regarding legislators’ explanations for their own electoral success and those that explore legislators’ adherence to the indications made by their party's leadership serve as indirect indicators not of ideology per se but rather of the legislators’ sense of the ideological coherence of their own party; legislators who assign relatively little electoral value to their party “brand” and who freely admit relatively little allegiance to their party leadership's policy positions when these conflict with personal or regional interests are essentially defining their own party as a catchall party. At a minimum, relative to Power and Zucco's (2009) use of legislators’ self-placement on a left-right scale and their placement of other parties on the same scale, our analysis provides some degree of policy content to legislators’ placement on the left-right dimension. Taken together, our analysis of the ten questions that drive the first discriminant function supports the conclusion that, with regards to issues that are often identified as key components of the ideological division between the right and the left, the Brazilian party system is most accurately described as one that pits the PT versus the inchoate rest.
Conclusion
Democracy requires political parties. Indeed, notions of interest representation and electoral accountability require that voters be able to distinguish parties from one another. Have Brazilian parties grown more programmatic with time, so that the Brazilian party system has consolidated into one in which voters can distinguish the major parties from each other? Although our answers are not definitive, given data limitations, our findings cast doubt on other scholars’ results: According to legislators’ responses to a battery of questions over nearly 20 years, the answer to this question is “no.” Given this, we tend to agree with Santos and Vilarouca, who suggest that although the Brazilian party system is not as chaotic as its detractors sometimes imply, certain distortions continue to “hamper the system's intelligibility for voters” (2008: 80). In particular, the party system's ideological map has grown relatively more incoherent after Lula took office due to the mishmash of parties in Lula's cabinet coalitions as well as the continued absence of any effort at collective image-building on the part of any of Brazil's center-right parties.
The comparative literature on political parties often speaks of three “faces” of political parties: in the Legislature, as organizations, and in the electorate. Brazil's legislative parties remain fairly undisciplined in comparative perspective (Carey 2009). As for party organizations, Brazil's national parties have never and still do not possess the two most important resources any party can possess: control over nominations, and control over campaign finance. Finally, the idea of a coherent ideological map in today's Brazil also does not jibe with what we know about voters’ beliefs. Despite 20 years of elections and intense competition, only the PT currently has attracted more than 5 percent of the electorate in terms of party “identifiers.” Party sedimentation in the electorate remains extremely weak (Samuels 2009), even among the supposedly “programmatic” non-PT leftist parties. Few voters cast partisan votes because few voters know where any party stands on any of the issues – or even understand what “left” and “right” mean. Over the last 20 years, parties have done little to clarify things for voters; indeed, one could say that they have endeavored to confuse Brazil's citizens. Given voter ignorance and the importance of personalism, it is hardly surprising that politicians on the center-right can adopt indistinguishable positions.
In the end, our findings suggest that seeing a clear ideological map in the Brazilian party system is wishful thinking. Since 1990 the only consistent distinction between the main parties in the system has been between the PT and the rest. And over time, even this difference is diminishing as both the PT and the DEM (formerly PFL) move towards the center of the spectrum. To return to Lamounier's quote at the start of the paper: does Brazilian political culture exhibit the traits that permit the sedimentation of a structured party system? At both the elite and mass levels, the answer remains “no.”
Footnotes
Variable Names and Question Wording
Do you fully agree, partially agree, partially disagree, or fully disagree with the following statements:
Sample Statistics
Party/Sample Sample Sizes
| Party | Sample 1 | Sample 2 | Sample 3 |
|---|---|---|---|
| PT | 63 | 57 | 36 |
| PSDB | 66 | 73 | 40 |
| PMDB | 68 | 93 | 36 |
| PFL | 78 | 101 | 57 |
| Total | 275 | 324 | 169 |