From Antipetismo to Generalized Antipartisanship: The Impact of Rejection of Political Parties on the 2018* Vote for Bolsonaro

This article analyzes the effects of antipartisanship in explaining the Bolsonaro vote in the 2018 presidential elections. The expansion of the anti-Workers’ Party sentiment known in Portuguese as ‘antipetismo’ into an antipartisanship that accommodates a wider range of targets and a growing association between antipartisanship and political intolerance provide the context for and underscore the relevance of our tests. Our expectation is that the vote for Bolsonaro was strong related to both antipetismo and negative attitudes towards mainstream political parties, especially when such attitudes were intense. The empirical basis of the present work was data from the Barometer of the Americas (LAPOP), particularly from the 2018/2019 wave. Empirical tests indicate that antipartisanship in its varied forms is a relevant phenomenon and had a major role in defining the electoral choices of 2018, particularly in the vote for Bolsonaro.

n issue of concern for researchers in the area of electoral behavior in Brazil is the vote for Bolsonaro in 2018. How can one explain the victory of a candidate with no public speaking skills, little name recognition, an incendiary discourse that made no concessions and a political party with few campaign funds and little media coverage? Bolsonaro's unexpected victory is usually explained by his appearance at an exceptional historical moment preceded by the 2013 protests, the impeachment of President Dilma Rousseff, the prosecution and imprisonment of the political leaders of the main political parties and an acute and prolonged economic crisis.
In this context, an increase in negative attitudes towards political institutions and elites is to be expected. This article focuses on these attitudes, specifically in relation to political parties. The 2018 election was the first since 1989 in relation to which it makes sense to inquire about the influence of the rejection of the main Brazilian political parties on the vote. Hence the question we intend to investigate: did antipartisan attitudes contribute to the vote for Bolsonaro? If so, was this merely a case of antipetismo or a wholesale rejection of the major political parties?
The main theoretical contribution of the article lies in our linking the literature on antipartisanship and studies on political tolerance, which leads to the development of a two-dimensional typology of antipartisanship. This typology takes into account not only the extent but also the intensity of voters' negative attitudes towards parties. The main empirical contributions of the article are: 01. presentation of evidence of the (unprecedented) impact of antipartisanship in the vote for Bolsonaro in 2018, well beyond that of antipetismo alone; 02. and demonstration that part of that antipartisan vote was also anti-system sentiment congruent with the political context and the option of an antiestablishment electoral alternative.
The article is divided into four sections. In the first, we present the theoretical debate on antipartisanship and voting and include contributions from the political tolerance agenda. The second section is dedicated to data and methodology. In the third section, we analyze longitudinally the picture of negative party feelings in the national context and then we present and discuss A Mario Fuks, Ednaldo Ribeiro, Julian Borba (2021) 15 (1) e0005 -3/28 our results from testing different types of antipartisanship and the 2018 elections.
Finally, we return to the main findings of the study and suggest some directions for future research.

Antipartisanship and Brazil
Since at least the mid-1990s the concept of antipartisanship has been studied as a component of political dynamics (POGUNTKE, 1996;POGUNTKE and SCARROW, 1996). As a result, in the last twenty years, studies on electoral behavior have consolidated the view that parties are not only the axes of attraction for voters but also the objects of their rejection.
The first approach considers the target of antipartisanship to be either the main parties that make up the current party system or the political parties per se (POGUNTKE, 1996). As a consequence, antipartisanship, especially that which rejects party associations in any circumstance, can contribute to the emergence of neopopulism, to a decline in support for a democratic regime and to patterns of antiestablishment behavior (MUDDE, 1996;SCHEDLER, 1996), especially in contexts characterized by the historical fragility of political parties, as is the case in Latin American countries (YERBA, 1998).
The second approach falls within the scope of studies of party identification and focuses on aversion to the party opposed to the party that the individual identifies with. In other words, identification with a party is complemented by the rejection of the party(ies) against which it competes (McGREGOR et al., 2015;MASON, 2018). A recent study by Abramowitz and Webster (2016) points out that negative partisanship is perhaps the most consistent political attitude of the American electorate today. An attitude held regardless of which party is identified or the strength of its identification, "American voters in the 21st century are much more likely to hold strongly negative views of the opposing party than in the past" (ABRAMOWITZ and WEBSTER, 2016, p. 21).
According to this view, the behavioral expression of negative partisanship is to not vote for the party that is the target of disaffection. Unsurprisingly, the principle method for measuring this phenomenon is by questioning interviewees about the party for which they would never vote (GARRY, 2007;MEDEIROS and NOEL, 2014;ROSE and MISHLER, 1998).
In Brazil, the topic of negative partisanship 1 first went under the microscope in the studies carried out by Carreirão and Kinzo (2004) and Barbetta (2004), Carreirão (2008;2007a;2007b), Samuels (2008Samuels ( , 2006, Ribeiro, Carreirão and Borba (2016;2011) andNeves (2012). These studies describe the negative feelings of Brazilians towards political parties as being of no lesser importance than their positive feelings, especially in the structuring of the vote.
When framing Brazilian negative partisanship within the scope of the international literature on antipartisanship, Samuels and Zucco (2018) Reis (1978) on the nature of partisanship in the context of the authoritarian regime; the study by Pereira (2014), which deals with the stability and effectiveness of party preference in the country. See also the study by Baker et al. (2016) on the dynamics of party identification in the recent period, and the study by Baker and Rennó (2019) which addresses the effects of different measures of partisanship on the levels of party identification in Brazil. A recent overview of the literature on partisanship in Brazil can be found in Fuks and Borba (2020, in press). main political parties. And, contrary to the argument of Samuels and Zucco, widespread antipartisanship has, at least in the case of 2018, "very precise implications" (SAMUELS and ZUCCO, 2018, p. 144) in terms of electoral choice.
We also argue that in addition to its scope, we must consider the intensity of antipartisanship. This intensity is higher when rejection of political parties is accompanied by political intolerance (POGUNTKE and SCARROW, 1996); that is, when the negative attitude is not restricted to disaffection, but also includes the desire to prevent parties from exercising their political activities.
Traditional political tolerance studies are restricted to radical groups. Thus, considering the theoretical debate and the political context presented above, we believe that it is necessary to expand the discussion on theories and typologies capable of explaining the various expressions of antipartisanship in Brazil. Studies that take a more culturalist line divide antipartisans between those who reject existing parties and those who adopt a general antipartisan stance, according to which parties are evil in themselves. In the tradition of studies on party identification, the pattern is to oppose two types of partisanship: positive and negative. In other words, in this second approach, there is no distinction between types of antipartisanship. Even models that intend to go beyond the binary typology do not escape this duality. In this tradition, Rose and Mishler (1998) present a more general typology of partisanship, but the negative partisan still boils down to one type: the voter who has no party identity and who would not, under any circumstances, vote for a particular party. The difference between the negative and positive partisan is that supporters identify themselves with a party whether or not they have negative identification with another party.
In order to fill the gaps, Fuks, Ribeiro and Borba (2018) constructed a typology that is based on the idea that antipartisanship is a bidimensional phenomenon and that, therefore, is only properly understood if we capture its variations in terms of disaffection to parties and of political tolerance.
The two variants of our typology, following the study mentioned above, distinguish voters according to the scope and intensity of their attitudes towards political parties, as can be seen in Table 01 and 02.
In both cases, the typology divides up antipartisans according to intensity (tolerant or intolerant). The difference lies in the target and the extent of their rejection of the parties. In the first, disaffection and intolerance have two possible targets: the PT (restricted) and the three parties together (generalized).
Thus, we have four profiles of antipartisan voters: the tolerant antipetista, the intolerant antipetista, the tolerant generalized antipartisan and the intolerant generalized antipartisan. The second variant of the typology does not target a specific party for disaffection. In this case, the factor that determines its scope is the number of parties: whether the target is one or two parties (restricted) or covers the three parties (generalized). So, again, we have four profiles of antipartisan   Our expectation is that antipetismo and generalized antipartisanship, especially the intolerant types, are more likely to vote for the 'anti-establishment' candidate Jair Bolsonaro than non-antipartisans (reference group). From this general proposition, we derive our two hypotheses:

H1:
Antipetismo had an effect on the Bolsonaro vote in the 2018 elections, and that effect is even greater when accompanied by political intolerance.

H2:
The rejection of all parties (generalized antipartisanship) had an effect on the vote for Bolsonaro, and this effect is even greater when accompanied by political intolerance.
Our hypotheses are based on the idea that, in an electoral context with a high rejection of the main political parties and with a candidacy that presents itself as an alternative to the conventional parties, both antipetismo and more general antipartisanship are important predictors of vote. In addition, this same context stimulates the more general and more intense types of antipartisanship, both of which have been given scant treatment in the Brazilian literature and both of which are important for helping explain the vote for Bolsonaro.

Data and methodology
By combining analysis of negative party feelings and political tolerance our study seeks to articulate, based on a typology, the scope and intensity of the rejection of parties by Brazilian voters.
Our What differentiates the models, therefore, is the focus: 01. in the first model, the focus is on antipetismo, compared to the rejection of all parties; 02. in the second model, the focus falls on antipartisanship regardless of specific parties. All other independent variables are identical.

Antipartisanship and vote for President in 2018
Before presenting the results of the tests that weighed the importance of the various types of antipartisanship in explaining the 2018 vote, we present a longitudinal view of the negative sentiments towards the parties.
Therefore, in addition to the 2019 data, we include here equivalent information  2018,2014,2010,2006,2002) and LAPOP (2019). Note: Eseb scales from 0 to 10; Lapop scale 01 to 10. The percentages refer to the initial points of the scales (0 or 01). The error margins (in %) for each survey are, respectively: 02; 3.3; 2.2; 02; 2.2; and 2.5. Graph 01, below, shows the distribution of respondents only among the proposed disaffection profiles, therefore, it does not consider the 42% of the sample  Reinforcing evidence from previous studies (RIBEIRO, CARREIRÃO and BORBA, 2016;SAMUELS and ZUCCO, 2018), it is clear that rejection of the PT is a consolidated and widespread phenomenon, but that other parties are also targets for negative attitudes. In particular, a finding from the data is the existence of a more recent phenomenon, one that is just as noticeable as antipetismo: disaffection and generalized intolerance towards the entire set of parties. In other words, antipetismo coexists today alongside broader forms of antipartisanship. We established the PT as a reference party for the construction of the first variant of our typology. As our goal is to investigate antipartisanship, Graph 03 shows the percentage distribution of respondents in these groups, excluding nonantipartisans. Thus, considering only those who manifest party disaffection, there is a predominance (39.5%) of intolerant antipetismo, i.e., of those who, in addition to rejecting the PT, also oppose the PT's right to launch candidacies in presidential elections. It is noteworthy, however, that one fifth of the sample defends the withdrawal of this political right for the three parties (generalized antipartisan), clearly configuring an anti-system posture. We will now see if the different types of antipartisanship had an effect on the Bolsonaro vote in 2018 6 . In model 01 (Table 05) Our data also confirm the expectation regarding the voting trends of certain groups in society. In comparison with other groups, whites 9 are 46% more likely to ______________________________________________________________________________________________ 6 An alternative model was conducted by changing the cutoff points to create the initial measures of disaffection and intolerance. The extreme cut, originally at point 01, was replaced by 03. The typology, considering these new cuts, produced similar effects, as can be seen in Table 02 of the annex, indicating considerable robustness of the procedure. Similar tests considering the tertiles of the measures of disaffection and tolerance are not feasible, since they are different between the three parties covered by the data in the two dimensions (disaffection and tolerance). The first tertiles range from 01 (in the case of PT in the measure of disaffection) to 03 (in the case of the PMDB in the measure of tolerance), i.e., all were covered by the cut used for this robustness test. 7 Alternative multinomial models, considering the three main applications, can be consulted in Tables  03 and 04 of the annex. 8 Question writing: Do you consider that the country's economic situation is better, equal or worse than 12 months ago? (worst = 0, equal = 01 and best = 02) (LAPOP, 2019). 9 Originally, the Lapop question comprised six categories (white, indigenous, black, mulatto, yellow and other). We recode the measurement to isolate whites and non-whites (white = 01 and nonwhites = 0).  and taking unequivocal and vehement positions on issues such as sex education in schools, same-sex marriage, public security and threats from communism, and consistently affirmed the traditional family and religion as pillars of Brazilian society. The impact of values and religion on voting is confirmed in our model.
Protestants 10 are 77% more likely to vote for Bolsonaro than are Catholics. Even more striking is the increase in the probability of voting for Bolsonaro among those who express a desire to have a firearm 11 -145%. Consistently, attitudes favorable to same-sex marriage decreased the probability of voting for Bolsonaro by 5%.
Ideology provides more evidence that attitudes played a decisive role in the 2018 election. It was unmistakably influential: voters who identified themselves as 'right-wing' 12 were 185% more likely to vote for Bolsonaro. This link between ideology and voting in 2018 was strengthened by the campaigns, that repeatedly framed Bolsonaro as the most extreme rightwing candidate on the ballot. Centrist voters were more divided between candidates, as there was no statistically significant correlation between a centrist position and support for Bolsonaro.
Certainly, a broader study of the determinants of voting in 2018 would require a more detailed analysis of the role of all these factors. However, our concern here is to test the effect of antipartisan sentiments on the vote for Bolsonaro.
The first model shows that antipetismo was an excellent predictor of voting in 2018, especially in relation to the vote for Bolsonaro. Compared to 'nonantipartisan' voters, tolerant and intolerant antipetistas were more likely to vote for Bolsonaro, 181% and 352% respectively. It is important to note that the intensity of negative feeling towards a party matters, as apparent in respect to the differences between tolerant and intolerant voters. The same is true among generalized antipartisans although the difference in relation to the reference category is smaller.
A tolerant generalized antipartisan has his or her probability of voting for Bolsonaro ______________________________________________________________________________________________ 10 The categories listed by Lapop were: Catholic, Protestant, non-Christian Oriental religions, Evangelical, traditional, Spiritist, agnostic/atheist, none and other. We reduced for Catholics, Protestants (aggregates with evangelicals), others (aggregated with those who believe in a Higher Being but do not belong to any official religion) and atheists. 11 The question asked by Lapop was: If you could, would you have a firearm for your protection? (yes = 01, no = 0). 12 The question asked respondents to position themselves on a 10-point scale, where 01 was equivalent to 'left' and 10 to 'right'. We recode this original measure by gathering points 01 to 03 as 'left', 04 to 07 as 'center' and 08 to 10 as 'right'.
increased by 121%, while the intolerant voters are 212% more likely to vote for Bolsonaro.
Additionally, we also estimated the predicted probabilities of each of these profiles, keeping all other variables stable (Table 06) Table 06 reinforce the analysis presented above, highlighting the considerable differences between intolerance groups, especially among PT members, but also among the generalized ones. These results only confirm the strong and consistent electoral impact of voters' attitudes towards the PT, whether positive or, in this case, negative. Petismo The results of the second model (Table 07), based on the typology that takes into account only the number of parties (without party distinction), test our second hypothesis. The results indicate, again, that antipartisanship had an influence on the vote. Compared to 'non-antipartisan' voters, restricted and generalized antipartisan voters are more likely to vote for Bolsonaro, but only when they are intolerant. While an intolerant restricted antipartisan has a 91% greater chance of voting for Bolsonaro, for a intolerant generalized voter the increase is an impressive 171%.
Therefore, the effect on voting in this second typology is conditioned by the intensity of the negative attitude towards the parties (intolerance), confirming our hypothesis about the vote for Bolsonaro as partially a result of a generalized disaffection towards political parties in Brazil, in addition to antipetismo. Although, in this second model, the effects of more comprehensive antipartisanship depend on the presence of intolerance, this result reinforces the relevance of our twodimensional typology and validates the emphasis on intolerance in our second hypothesis.
As we have done previously, we estimate the predicted probabilities of these profiles without party distinction. The hypothetical scenario exhibits the same demographic and attitudinal characteristics, but the distinction is now made between tolerant restricted antipartisan, intolerant restricted antipartisan, tolerant generalized antipartisan and intolerant generalized antipartisan. The results (Table 08) only reinforce our previous findings that voters who express both party disaffection and political intolerance were more likely to vote Bolsonaro.
The results presented above show the electoral impacts of a context in which politicians, political institutions, parties and everything related to politics loses the little appreciation it once enjoyed in society. With the offer of a candidate who represented such sentiments, dissatisfied voters found an electoral alternative to express their negative party identification.  This scenario allows us to understand the expansion and intensification of the antipartisan phenomenon in Brazil. In this study, we found empirical evidence of a much broader phenomenon than that presented in the main work on the topic (SAMUELS and ZUCCO, 2018), according to which antipartisanship in Brazil is synonymous with antipetismo. The evidence presented here indicates that antipartisanship goes far beyond antipetismo and that its widespread manifestation, in 2018, has the same electoral consequences as antipetismo.
In addition, the inclusion of a dimension that has not yet been explored in studies on antipartisanship, has allowed us to verify the intensity of negative attitudes towards parties. As seen above, party intolerance, whether directed at one party or all of them, provided an important attitudinal basis for the election of Bolsonaro.

Final considerations
As The results that we present throughout this article show that antipartisanship, in its various manifestations, is a relevant phenomenon that had a major role in defining the electoral choices of 2018. Focusing simultaneously on the extent and intensity of these negative feelings, through an unprecedented combination of agendas on partisanship and political tolerance, we demonstrate through two variants of our typology that the breadth and strength of antipartisanship today is considerable and goes beyond one-dimensional rejection of a specific party.
In addition, it remains to be seen whether, in Brazil, antipartisanship has an impact only on electoral behavior or whether it is related to a broader and more articulated set of political behaviors and attitudes.