The two Brazils in the 2022 election

Image: Messala Ciulla


The dangerous (co)relationships between Bolsonarism, COVID-19 and Emergency Aid

Determinations of the relationship between COVID-19 and Bolsonarism

Throughout 2020 and 2021, a set of surveys were carried out and disseminated about the relationship between voting for Bolsonaro and contamination by Covid. The surveys revealed a positive relationship between Bolsonarism and contagion and were widely publicized in magazines. scientific, in the big media and Blogs critics of the government. Despite the differences in the size and representativeness of the samples, as well as in the methodologies of the different researches, the size and significance of the correlations left little room for doubt about the existence of the aforementioned relationship.

Correlation analysis, however, is not causality analysis and the theoretical connection that a significant number of researchers (and the vast majority of journalists) proposed between these two variables seemed premature. As a general rule, the hypothesis suggested was that voters and supporters of President Bolsonaro would underestimate the danger of the ongoing pandemic, which would result in the relative relaxation of social distancing and other safeguarding practices capable of depressing contagion.

The hypothesis that structured our research was discreetly different. It seemed obvious to us that “denialism” – so widespread among Bolsonaro supporters and followers – played a relevant role in promoting contamination. However, we doubted that this element, by itself, could explain the high disparity in the percentage of contamination among the thousands of Brazilian municipalities.[I]. It seemed to us that, in addition to denialism, variables of a more “structural” nature should be behind such high incidence differentials. More: we believed that the structural variables co-responsible for the increase in the incidence rate of COVID-19 could also contribute to understanding the socioeconomic and cultural profile of the typical Bolsonaro voter.

In an attempt to test this hypothesis, we created a database with 103 variables for the 5.569 Brazilian municipalities plus the Federal District, generating 5.570 municipalities. Among these variables, 60 are raw data, with geographic, demographic, economic, political and sociocultural information from the most diverse official sources, in particular the Brazilian Institute of Geography and Statistics (IBGE), the Superior Electoral Court (TSE), the Ministry of Citizenship (MC), the Ministry of Health (MS-DATASUS), the Annual Social Information Report of the Ministry of Labor and Employment (RAIS-MTE), among other bodies.

In turn, the raw data were appropriated and transformed into 69 municipal-based indicators. As a general rule, these indicators were relativized (or normalized) by the population residing in the municipality; or they are GDP growth rates, different GVAs, formal occupation, etc. The indicators were grouped into five classes: (1) political alignment of the population (eg, percentage share of votes for Bolsonaro or Haddad in the first and second rounds of the 2018 elections); (2) economic structure of the territory (e.g. GDP per capita in 2018; share of formal industrial employment in total formal employment in 2019; share of formally employed in the total population of the municipality in 2019; ratio between formally employed and total employed in 2010 ; etc.); (3) local sociocultural structure (eg: participation of the black, evangelical, illiterate, higher education population in the total population in 2010); (4) impact of the pandemic on global health (eg, percentage of infected people and deaths from COVID-19 in the total population in 2020); and (5) assistance coverage (e.g.: percentage of people benefiting from Emergency Aid – henceforth, AE – in the total population; ratio between the total amount of AE received in a municipality between May 2020 and April 2021 and the municipal GDP in 2018).

Differences in data dates and indicators relate to data availability. Sociocultural data, for example, are based on the most recent Demographic Census, which dates from 2010; the municipal GDP is calculated with a lag of three years in relation to the national GDP. As the variables are of a structural nature, we believe that the changes that occurred over the period were not very significant. The database built by us is available here for the use and/or investigation of any interested parties.


Basic results: two Brazils in contention

The results we found not only confirmed our initial hypotheses, but also brought to light new dimensions of the political-electoral “field” of Bolsonarism and about the economic, political and health consequences of the Emergency Aid. In fact, some of the results achieved are so counterintuitive that they apparently raised doubts about the consistency of the tests we performed.[ii].

Table 1: Selected Correlations between Structural Socioeconomic Indicators,
Vote for Jair Bolsonaro and Fernando Haddad, Incidence of COVID-19 and Emergency Aid

FDB: IBGE, RAIS-MTE, Datasus, Min. Citizenship, TSE

The 29 indicator variables in each of the lines in Table 1 above are ordered according to their correlations with the vote for Bolsonaro in the first round of the 2018 elections. The column “Rank1” corresponds to the ranking of correlations: the lower the value of Rank1 (first, second, etc.) the higher the correlation, and vice versa. Not all tested variables are represented in Chart 1. We extracted variables with high auto-correlation (I voted for Bolsonaro in the first and second rounds, for example), as well as variables with correlation close to zero and/or low reliability (significance ). The significance of all presented correlations is less than 0,01%. That said, in the upper lines are those variable-indicators that have a strongly positive and significant relationship with the vote for Bolsonaro. As we “go down” towards the bottom of Chart 1, the correlations fall, becoming strongly negative.

As expected, the relationship announced in several polls between voting for Bolsonaro and the incidence of COVID-19 is confirmed. If we abstract the two tautological correlations (correlation of the rate of contamination with itself and with deaths from COVID-19), taking only the strictly independent variables, the vote for Bolsonaro in the two rounds emerges as the variable with the greatest explanatory power of the differentials in the rate of municipal contamination: the higher the percentage of votes for Bolsonaro in 2018, the greater the percentage of the population contaminated (correl 0,416, sig 0,0000). From our point of view, the fact that this correlation emerges in the first place demonstrates the existence of an autonomous political-ideological component – ​​expressed in denialism – that increases contagion due to resistance to social distancing, vaccination and the hygiene care necessary for the control and depression of virus diffusion factor. In summary: our tests fully corroborate the previously mentioned studies.

On the other hand, our tests also prove that the relationship between COVID-19 and Bolsonarismo transcends – and a lot! – the strictly ideological dimension. Let's take the third and fourth lines of Table 1. What they tell us is that the greater the percentage of the population employed (formally or informally, Census data) in 2010 and the greater the percentage of the population formally employed in 2019 (data from RAIS-MTE ) in the total population of municipalities: (1) the greater the number of votes for Bolsonaro; (2) higher incidence of COVID-19; (3) lower percentage of citizens who received Emergency Aid; (4) lower relative expression of the total value of the AE compared to the municipal GDP; and (5) a lower percentage of votes for Haddad in the second round of presidential elections.

And most important of all: this series of relationships – more Bolsonaro, more Covid, less AE and less Haddad – persists and monotonically reaffirms itself for all 15 variables that, in some way, translate social inclusion. Thus, municipalities with better educational indices, with a greater number of formal employees in the total population, with a higher GDP per capita, with a greater number of families receiving more than ten minimum wages tended to vote for Bolsonaro and to present contamination rates above from the average.

Among these inclusion variables, one in particular deserves attention: the more industrialized municipalities, with a greater number of employees in the Manufacturing Industry – that is to say, where the working class is relatively more expressive – and with a greater participation of Industry in the municipal GDP followed the same rule: they voted preferentially for Bolsonaro, had higher rates of contamination, received a smaller percentage of AEs and, as a general rule, did not vote for Haddad in the second round. It is also worth noting the variable “Code of the Federation Unit (UF)”. The IBGE code system is such that municipalities in the North Region start with the number 1, in the Northeast Region with the number 2, in the Southeast Region with the number 3, in the South Region with the number 4 and in the Midwest Region with the number 5. The “Sul Maravilha” and the Cerrado do Agribusiness were Bolsonaro’s electoral bases and, simultaneously, they were the regions with the highest incidence of COVID-19; while the North and Northeast Regions (with a smaller code) were (as a general rule open to exceptions) Haddad's main electoral strongholds and the territories with the highest percentage of people benefiting from the AE in 2020.

Finally, there is a variable that helps to assemble a general picture of great explanatory power: the vote for Bolsonaro presents a positive correlation with the degree of urbanization (urban population / total population). In Table 1, this correlation is shown in the sixteenth line by its opposite (degree of rurality = rural population / total population), and a negative correlation (-0,288) is shown between voting for Bolsonaro and the degree of rurality. But this result is the “mirror version” of a positive correlation between this same vote and the degree of urbanization (0,288). Now, the more urban and industrial a municipality is and the greater the percentage of the population engaged in formal activities (employment with a formal contract and civil servants), the greater the difficulties in maintaining social distancing. By contrast, the more rural, agricultural and informal the economy of a municipality is, the easier it is to maintain a minimum distance. Especially due to the high coverage of Emergency Aid. And, not for free, the variables-indicators of AE coverage (columns 3 and 4) also present a positive correlation with “degree of rurality” (correl 0,217 and 0,391, respectively) and with informal employment (variable 19, correl 0,396 and 0,557 ).

Let us now look more closely at the bottom of Chart 1 (from row 16 downwards). There, the exchange of signs of the correlations between the variables in columns 1 and 2 (vote for Bolsonaro and incidence of COVID), which become negative, and between columns 3, 4 and 5 (emergency aid coverage and vote in Haddad), which become positive. The main determinant of this change is that the variables listed at the bottom of Chart 1 are variables indicative of relative social exclusion; and, for this very reason, they refer to that portion of the population that significantly entered the Federal Budget during the PT governments. Objectively, what Table 1 informs us is that the % of votes for Bolsonaro will be much lower (and the % of votes for Haddad will be much higher) in those municipalities where it is greater: (1) the % of people domiciled in the countryside (as opposed to mere rural landowners with urban domicile); (2) the % of the population under 35 years of age (segment where unemployment and informality are higher); (3) the % of blacks and browns (who only start to benefit from a specifically ethnic-racial policy of inclusion based on quota policies); (4) the % of illiterates (portion of the population with greater difficulty in entering the formal labor market); (5) the percentage of poor people (with per capita income below the minimum wage, who benefited from programs such as Bolsa Família); (6) the % of informal workers; and (7) income inequality (measured by the Gini Index).

Now, this picture illuminates a reality already intuited, but which – as far as we know – had not yet been rigorously demonstrated: the fact that the 2018 electoral campaign was marked by a confrontation of radically antagonistic projects in terms of the role of the State in the process. of social inclusion and facing/overcoming inequalities. As much as the conservative media tried to mask the dispute as a confrontation between “friends and enemies of corruption” (rescuing and re-heating the battered udenist discourse, which supported the coups of 1954 and 1964) and liberals of all stripes (especially those who militate against taxes and in favor of tax subsidies for the promotion of investments) envisaged a dispute between “populist politicians X meritocratic business community”, what was effectively at stake was the right (or not) of popular governments to use a portion of the budget for support the minimum income, bring water and light to the sertão, facilitate the creation and relieve micro-enterprises (including individual ones), support the socialization and access to higher education of descendants of slaves, support the productive social insertion of small rural producers, indigenous peoples and quilombolas, among other neglected sectors for centuries. Even more important: the data collected and systematized for 5.570 municipalities in the country reveal that, in general (and despite exceptions) the people voted according to their strategic interests. Even in the face of a situation of absolute exceptionality, determined by the impeachment of Dilma, by the process of lawfare which ended up leading to Lula's arrest, due to the imposition of silence on the ex-President (prohibited from giving interviews by the STF), even so, popular conscience manifested itself, not only in the expressive percentage of votes given to Haddad in the second round (45 %) but, above all, due to the fact that the sectors that supported Haddad were precisely those that the PT governments effectively supported and contemplated.


Surprising results: did the pandemic in Brazil spare the poorest?

There is a vast literature on the regressive selectivity of pandemics, which hit the bottom of the social pyramid more sharply. This relationship is recognized even in publications by technical bodies of the UN system, famous for the predominantly conservative inflection of their analyzes (see, for example). However, a superficial reading of the results we found could lead to the conclusion that the perverse selectivity of the pandemic had not manifested itself in Brazil. But that's not what it's about. Then let's see.

When operating with a municipalized database, it is necessary to understand that the “sample individual” is not a real individual, but a collective composed, itself, of highly differentiated groups and individuals. An example may help in understanding this point. Let's imagine two neighboring municipalities: A and B. The first (A) has an illiteracy rate of 2% of the total population, while B has 20% of those residing in this category. The neophyte in spatial statistical analysis could conclude that “the problem of illiteracy” is greater in the second municipality than in the first. However, this is a hasty conclusion and disappears as soon as he is informed that: (1) the first municipality has 500 inhabitants and the second with only 8 inhabitants. The 2% of illiterate people in the first municipality (10 inhabitants) make up a population greater than all the inhabitants of the second. Let us now imagine that almost all of the illiterate people in municipality A live in a quilombola community that, for years, has been requesting the installation of elementary schools and education for young people and adults in its interior. Where, in what territory, is it urgent to invest in the fight against illiteracy?

Bringing the issue to our field: when we say that the % of the vote for Bolsonaro was higher in municipalities characterized by high industrial GVA and significant participation of workers among the occupied, we are not saying that the working class has become Bolsonaro's electoral base. This movement may even have occurred. Or not. The accelerated de-industrialization in the country, the low growth rate and the fall in wages in some industrial sectors may even have generated discontent among working-class sectors that migrated to a conservative vote. It's possible. But this is not what the data reveals. Because we are not operating with social strata, but with municipal averages. And every community is marked by specific patterns of stratification. “Rich counties” are not counties where everyone is rich.

In fact, it is very likely that the high rates of contamination by COVID in industrial cities with a high GDP per capita have mostly affected the social strata located at the base of the pyramid, such as the working class. After all, this is a segment that: (1) cannot opt ​​for “work at home”; (2) in the long term, it earns higher returns than those provided by the AE; and (3) in the vast majority of cases, commuting between work and home in public transport vehicles that are crowded during peak hours. Thus, when we operate with data and indicators systematized and grouped by criteria other than those chosen by us (municipalized variables), the socially selective nature of COVID-19 is easily evidenced.

For example: in 2020, the morbidity rate among hospitalized black (42,78%) and brown (39,22%) patients was significantly higher than the morbidity rate among Caucasians (36,55%) and Asians (36,48% ). While the hospitalization rate of black and brown patients (4,54% and 34,62% respectively) remained below the relative participation of these ethnic groups in the total population (7,61% and 43,13%, respectively).

However, none of these relativizations nullifies or denies two important facts: 1) Emergency Aid was channeled primarily to the social segment and to the municipalities that, in 2018, gave the highest percentage of votes to Haddad, that is to say, to those citizens who they were entitled to it; 2) the AE contributes to the social distancing of citizens from less favored social strata. It is not a matter of denying the fact that a portion of the beneficiaries of the AE were not entitled to it. It is only a question of recognizing that this was not the dominant case. In smaller municipalities with a less diversified economy – that is to say: in municipalities where Haddad achieved the highest number of votes in 2018 – AE corresponded to up to 30% of GDP and reached more than 60% of residents.

We don't think there are any doubts about the impact of this national coverage on the results of the 2020 municipal elections, in which the left's performance fell short of expectations and the performance of allied parties and supporters of the Bolsonaro Government went beyond expectations. And the most interesting thing is that this “political appropriation” of the AE does not seem to have been planned. The AE was imposed on Jair Bolsonaro and Paulo Guedes by Congress, based on a strong mobilization of opposition parties. And it only had significant electoral impacts because, even after its imposition, the Government did not pay due attention to it, using the Single Registry (created in 2001, in the FHC government, but disseminated and consolidated in PT governments) as the main eligibility instrument of coverage. It is worth saying: the AE had electoral effects precisely because the Government did not manipulate it electorally. One of the least republican governments in the history of the Republic, by underestimating the political strength of the AE, achieved what seemed impossible: attracting part of the PT electorate, which had been conquered by the conversion of the Federal Budget into an instrument of social inclusion.


Lessons for 2022

Since Lula got out of prison, the situation has accelerated a lot. Vaza-Jato made it clear that Lula's trials and convictions were nothing more than lawfare. The entire Judiciary was contaminated by the operation and has been seeking to redeem itself with the (late) recognition of Moro's partiality and the archiving of all the processes that were still ongoing against the former President. Meanwhile, Lula is received as Head of State in several countries around the world and the set of candidacies that want to be a “third way” cannot, even added together, reach the intention to vote for Bolsonaro and, much less, for Lula. At the same time, however, rejection of Bolsonaro continues to grow and is already the highest among all likely participants in the 2022 election.

From our point of view, definitely not. And this for three reasons. Firstly, because the system that fueled and carried out the 2016 coup-impeachment and Lula's arrest the following year is complex, strong and very well articulated. It involves media, Judiciary, Parliament, Armed Forces, big capital (Brazilian and multinational), Embassies and foreign Intelligence Agencies and much more. We cannot underestimate the risks of Brazil's fragile democracy. Secondly, because even in fits and starts, Bolsonaro has managed to stay in power and circumvent the opening of an impeachment process.

And, whether due to miscalculation, ignorance or omission, his government implemented a minimum income program that prevented the economy from collapsing in 2020 and ensured the electoral success of the right that year. Finally, it is necessary to understand that any blow against democracy does not need to be given before the elections or the inauguration of the new President. The blows against Getúlio, Jango and Dilma took place during the management process. Sometimes it's easier to let the opponent "win" the game, and strike later.

And here it is worth remembering that the Brazilian economic crisis is serious and structural. It is based on deindustrialization. Which have been imposing for years. Even in the 13 years of PT management. And the draft of the PT's economic program available so far is not very clear on how this issue will be faced. We would not be surprised if part of the intelligentsia anti-PT was, at this moment, meditating on whether it would not be better to let “Lula take, but not win”. Like Dilma in 2014-2016.

Perhaps the most auspicious aspect of the 2022 electoral clash is to be found in the fact that – instead of adopting the 2020 budget management standard – the Bolsonaro government seems determined to administer it “electorally” in 2022, delaying precatorios, reducing the amounts destined for Education, Health and the new minimum income program, in favor of the release of the resources foreseen in the “Amendments of the Rapporteur”, negotiated by parliamentarians from the Bolsonarist base for the purchase of votes in different electoral corrals. This shot has a high chance of backfiring. Apparently, neither the left nor the right fully understood how much the results of the 2020 election (as well as the results of the presidential races between 2002 and 2014) were influenced by the Republican management of the Budget and the inclusion of the poor in it. As in intersubjective human relationships, often the most effective way to conquer the Other is to give up the games of seduction and operate in the field of transparency and honesty.



By way of conclusion, we would like to draw attention to the fact that voting intention is not voting. In a way, this is what the correlations between the variable-indicator in the second line of Chart 1 – “% of valid votes in the first round” – and the 5 variable-indicators in the columns of the same Chart tell us about this. The correlation with “vote for Bolsonaro” is 0,909 positive, while “votes for Haddad” is -0,989 and “% of AE Value in GDP” is -0,787. Once again, part of these correlations reflect exogenous determinations, that is to say, they express cross-correlations. The percentage of valid votes is higher the more urban the municipal population is (0,447), the lower the illiteracy rate in the municipality (-0,829) and the higher the percentage of people with higher education (0,614).

Now, we have already analyzed the overlapping of the cleavage “inclusion X exclusion” with the cleavage “Bolsonaro voters X Haddad voters”. But this is not all. There is a specifically political-ideological dimension to these correlations. In fact, Bolsonaro’s voter in 2018 was a more “militant” voter, with a degree of belief in the need to “change the country through the election of the Myth” that exacerbated and maximized his decision to participate in the voting process. In contrast, Haddad's voters had to overcome a wide range of doubts and uncertainties arising from the media's bombardment of the apologetic dissemination of Lava-Jato and the proliferation of criticisms of PT corruption.

It seems to us that this difference between degrees of “conviction” has not yet been overcome. She was present in the 2020 elections and should return in the 2022 elections. In the case of the 2020 elections, many were surprised by the distance between the polls of voting intentions (up to the exit of the ballot box) and the electoral results. At least part of this discrepancy is associated with conviction and the militant vote: the PT and the leftist parties in this country still carry on their shoulders the weight of negative media campaigns that work very well with common sense: if Justice investigates and arrests and does not investigate or arrest Z, then A is guilty and Z is innocent.

For this very reason, it is very important to understand that the 2022 elections will not be defined by intentions, but by winning over a militant electorate, capable of proudly exposing its political-ideological option. As far as we can see, there is still a gap between Bolsonaro voters, who proudly expose their option for preserving the status quo through the use of the national flag and the T-shirts of the Brazilian team as (supposedly universal and anodyne) symbols of conservatism, and the voter of Haddad (in 2018) and Lula (in 2022), who no longer wears his red T-shirts and his stars with the ease of the eighties, nineties of the last century and the first decade of the current century.

This is a point that should be the object of reflection on the part of political strategists: victories are won with adherence and enthusiasm. Voting intentions are necessary but not sufficient conditions.

*Carlos Águedo Paiva he holds a doctorate in economics from Unicamp.

*Allan Lemos Rocha is a statistician and is studying for a master's degree in Urban and Regional Planning at UFRGS.



[I] When we ended the data collection for the construction of contamination and morbidity indicators by COVID-19 in mid-2021 (with a view to further statistical and theoretical analysis), the disparity in the rate of contagion was still very high. In the 100 Brazilian municipalities with the lowest contagion rate, it corresponded to less than 1% of the population, while in the 100 municipalities with the highest incidence, it was close to 20%.

[ii] The authors produced an article with the global analysis of the results, which was offered for publication in the most important Brazilian journals in the area of ​​“Saúde e Sociedade”. To our surprise, the paper it has been turned down on the surprising grounds that the work would be irrelevant. Given that the themes “pandemic”, “Bolsonaro's election”, “political and health developments of the Emergency Aid” and the connections between them are of unquestionable relevance, we interpret the refusal as a fear that, whether it be the Database or the statistical tests and the results found contained errors. However, the risk of errors in these bases are minimal: the Bank produced by us was made available, the sources used for its assembly are public and official and can be audited, and the tests are simple correlation exercises that can be replicated by any social scientist. . In the end, we came to the conclusion that the problem lies in the lack of knowledge of characteristics of Spatial Statistics, which complicate the interpretation of the results obtained. For this very reason, we created a specific section (the third) in this summarized version of the previously produced work aimed at clarifying how statistical results should be interpreted in regionalized analyses.

See this link for all articles