Introduction
The global political landscape has witnessed a surge in anti-system politics, marked by the rise of populist leaders who challenge established norms and institutions. Jair Bolsonaro, Brazil’s former president, exemplifies this movement, which has remained resilient even after his electoral defeat by Lula in 2022, as evidenced by the strong performance of his allies in gubernatorial and legislative elections. Bolsonaro’s political ideology and rhetoric align with the anti-system wave that has swept through the United States, Europe, the Philippines, and India.
Firms, I argue, are not neutral transmitters of market conditions, but active producers of economic inequality and, by extension, political attitudes. The theoretical focus is not limited to Bolsonaro’s support, but instead considers how experiences of declining job quality and perceived unfairness in the workplace can erode democratic commitment. Bolsonaro serves as an empirically tractable case for exploring this broader relationship: during the period under study, he was the principal vehicle for anti-system sentiment in Brazil, and his support base provides a window into the mechanisms by which firm-structured inequality translates into political estrangement.
In recent years, Brazil’s political landscape has significantly shifted with the rise of Bolsonaro and a political movement characterized by anti-system and authoritarian tones and a loose far-right ideology. This movement’s rapid growth is paradoxical in a country that had long been governed by the center-left Workers’ Party (Partido dos Trabalhadores – PT), which made significant progress in reducing poverty and inequality.
A conventional explanation for this paradox is that Bolsonarismo represents a backlash from the upper classes against the progress made by the poor under the PT. Instead, this article provides empirical evidence supporting the argument that firm-based origins of wage and job quality inequality in Brazil played an important role in shaping the political sentiments that paved the way for Bolsonaro’s rise. While Bolsonaro is the empirical anchor of this study, the broader theoretical interest lies in understanding how experiences of economic decline and perceived unfairness in the workplace may fuel democratic disaffection more generally.
This article demonstrates that the role firms play in shaping inequality can lead to anti-system politics and populist right support. Bolsonaro’s prominence during the period of study – when he was the dominant, if not exclusive, channel for anti-system sentiment in Brazil – makes his base an analytically useful case through which to observe these broader dynamics.
While Bolsonaro provides a central case for empirical leverage, the article’s core concern lies in the broader relationship between firm-based experiences and the development of anti-system political attitudes. Rather than offering a narrow account of vote choice, the argument advances a political economy framework for understanding how workplace dynamics – particularly declining job quality and perceived unfairness – can erode democratic commitment. In line with recent findings in labor economics, I emphasize that firms are not merely passive arenas for broader market trends but are themselves generators of inequality. Crucially, this inequality operates along two axes: between firms, where otherwise similar workers receive systematically different compensation and treatment depending on the employment strategies of their firms; and within firms, where disparities in pay, stability, or advancement opportunities among colleagues can heighten perceptions of injustice. Both dimensions are politically salient and may contribute to anti-system attitudes, especially when opportunities for mobility across firms are limited by regulatory, geographic, or social constraints.
To explore this hypothesis, I combine individual-level administrative data with observational and experimental evidence from an original online survey. By utilizing these diverse methodologies, I establish a comprehensive understanding of the relationship between job quality, political attitudes, and support for Bolsonaro, highlighting the firm-based origins of anti-system politics in Brazil.
First, using individual-level matched firm–worker data and a two-way fixed effects model, I examine two subsets of workers who are comparable in terms of all observables available in formal employment data recorded in social security records. The analysis focuses on comparable workers who actively expressed support for Bolsonaro and those who did not. This analysis shows a significant association between Bolsonaro supporters and a relative intertemporal decline in their salaries compared to their occupation, industry, and the median wages offered by their employing firms.
Second, I present findings from a large original online survey conducted among full-time workers in Brazil. The survey reveals a strong connection between workers’ negative evaluations of their job quality and employment situation in their firms across various dimensions and their propensity to express anti-system political attitudes. These observational results are supplemented with a conjoint experiment indicating that respondents perceive workers in low-road employment conditions as more likely to manifest anti-system political attitudes and behaviors.
Specifically, this experiment explores the impact of low-road firm-level contractual conditions and affirmative action policies, similar to those implemented by the PT at the national level. The analysis demonstrates the significance of non-wage components of job quality at the firm level as well as the role of affirmative action policies in shaping anti-system attitudes and behaviors.
Finally, I present an information experiment that evaluates the impact of wage-related information on democratic satisfaction. By providing respondents with disaggregated earnings data by race, indicating that they earn less than their peers, I show that this information negatively affects their satisfaction with democracy when they see they earn less than comparable workers of different races in terms of occupation, skills, industry, and gender.
By integrating these empirical findings, this study contributes to our understanding of the economic roots of anti-system politics by tracing how firm-level variation in job quality shapes political attitudes. It shows that the experience of decline or perceived injustice within the workplace – whether relative to peers in other firms or to colleagues within the same organization – can erode democratic commitment among workers. In the Brazilian case, the rise of anti-system sentiment was not merely a reaction to macroeconomic crises or cultural polarization but was also grounded in micro-level experiences of status loss and inequality in the world of work. The findings highlight how public policies intended to reduce inequality – such as wage compression, minimum wage hikes, or affirmative action – can have politically destabilizing effects when implemented in a labor market marked by deep heterogeneity across firms. As Hagopian (Reference Hagopian2018) argues, the stratified nature of Brazil’s welfare and employment systems created symbolic and material boundaries among workers, intensifying resentment when redistributive policies blurred those distinctions. This article advances a framework for understanding how such workplace-driven grievances – especially in formally segmented labor markets – can become politically consequential.
Theoretical Framework
Firms are central to the organization of social life, establishing hierarchies and categorical boundaries that construct and reproduce inequalities. As Tilly (Reference Tilly1999) argues, workplaces are primary sites where inequalities are both created and maintained. Consequently, changes in production impact relationships of exploitation, opportunities, and claims over resources among different groups.
This article presents a firm-level theory arguing that firms play a crucial role in shaping anti-system political attitudes. While material explanations for changing political preferences often highlight employment shocks, income changes, and technological advancements, they overlook significant changes within firms. Unlike macro-level economic explanations, this theory focuses on firm-level decisions and strategies. It argues that diverging strategies within firms create a disconnect between workers’ expectations and their economic realities, fostering anti-system sentiments.
Recent decades have seen significant changes in production methods, with workplaces becoming increasingly fragmented (Weil Reference Weil2014) and traditional contracts breaking down (Rahman and Thelen Reference Rahman and Thelen2018). Despite a focus on skills and worker characteristics, evidence shows significant wage and benefit disparities among similar workers across different firms (Song et al. Reference Song, Price, Guvenen, Bloom and vonWachter2019; Kristal et al. Reference Kristal, Cohen and Navot2020). This indicates that ‘where you work’ is now more critical than ever in determining employment outcomes (Barth et al. Reference Barth, Bryson, Davis and Freeman2016).
High-road and low-road firms, those that provide good levels of job quality, internal career opportunities, benefits, and mobility, versus those that do not, contribute significantly to income inequality, benefits, job security, mobility, and opportunities across comparable workers (Guiso et al. Reference Guiso, Pistaferri and Schivardi2005; Babecký et al. Reference Babecký, Caju, Kosma, Lawless, Messina and Rõõm2012; Song et al. Reference Song, Price, Guvenen, Bloom and vonWachter2019; Kristal et al. Reference Kristal, Cohen and Navot2020). These differences are not rooted in industry fundamentals but in firm-specific strategies and policies. This divergence, termed the ‘great separation’, reflects a new industrial divide driven by managerial decisions rather than trade, industry, or technology.
The theory suggests that workers in low-road firms experience a disjunction between expectations and employment reality, fueling frustration and anti-system attitudes. Traditional economic explanations for populism focus on macro-level shocks like trade and automation, which often fail to connect individual economic deprivation with populist support. Instead, this theory locates the immediate causes of anti-system politics at the workplace level.
While standard models of labor market adjustment might predict that workers facing lower pay or worse conditions would move to higher-paying firms, recent evidence in labor economics reveals persistent between-firm wage inequality, even after controlling for occupation, education, and sector (for example, Barth et al. Reference Barth, Bryson, Davis and Freeman2016; Song et al. Reference Song, Price, Guvenen, Bloom and vonWachter2019). This persistence reflects labor market frictions – including statutory protections, regional immobility, firm-specific human capital, and information asymmetries – that constrain workers’ ability to move across employers.
Within this context, I argue that political frustration arises not merely from wage disparities across similar workers, but from the combination of inequality and perceived immobility. Workers in low-road firms are not just paid less – they are also less likely to see viable paths for advancement. These combined experiences – of relative deprivation and blocked opportunity – can erode democratic commitment, especially when workers attribute their stagnation to institutional or political failures. While classic labor market models would predict convergence through worker mobility or wage competition, a growing body of research in labor economics has documented persistent frictions and rigidities – ranging from geographic immobility and family constraints to contractual path dependence and firm-level monopsony – that dampen adjustment even in competitive sectors. This article builds on that literature to treat limited mobility not as an additional explanatory factor but as a scope condition that gives firm-level inequality its political salience. In doing so, it aligns with theories of status anxiety and political alienation, but locates these phenomena in the concrete organizational structure of the workplace rather than in abstract expectations or purely cultural frameworks.
In Brazil, this dynamic is particularly relevant. Before Bolsonaro’s election, the country faced a severe economic crisis marked by stagnant growth, escalating public debt, and a series of high-profile corruption scandals. These structural disruptions help account for the widespread public dissatisfaction with the political status quo. However, such conditions alone do not fully explain the authoritarian turn in political attitudes that defined the rise of Bolsonarismo. Disillusionment with incumbents or mainstream parties may be a necessary condition for political change, but they are not sufficient to account for the embrace of anti-system rhetoric and democratic ambivalence.
The rise and continuity of Bolsonaro’s radical-right populist movement remain insufficiently explained in the political science literature. Standard narratives emphasizing class conflict fail to capture the diversity of his support base. Similarly, accounts that foreground corruption scandals or Evangelical mobilization do not fully illuminate the breadth, ideological ambiguity, and persistence of Bolsonaro’s appeal – nor the deeper political shifts it reflects.
The available evidence indicates a clear disparity in wages between firms in Brazil, irrespective of worker, industry, or product market characteristics. This aligns with the global phenomenon of the great separation. Helpman et al. (Reference Helpman, Itskhoki, Muendler and Redding2017) show that between-firm differences in Brazilian workers’ wages across detailed occupations and sectors increased significantly after trade liberalization in the 1990s, explaining a large portion of total wage inequality.
During the period of economic stability and growth in the early 2000s, income inequality in Brazil declined. Policies such as increasing the minimum wage and expanding social programs helped improve living standards for many. However, recent research shows that firm effects accounted for a significant portion of changes in inequality, driven by pay policies rather than changes in firm fundamentals (Alvarez et al. Reference Alvarez, Benguria, Engbom and Moser2018). While these policies had positive impacts, they did not fully address broader firm-level productivity issues, leaving workers vulnerable to economic fluctuations.
The dynamic labor market in Brazil, characterized by significant informal employment and low geographical mobility, further exacerbates these disparities. Firms with different productivity levels and pay policies contribute to a fragmented labor market where workers’ experiences and outcomes vary widely. This fragmentation can lead to feelings of frustration and disillusionment among workers, fueling anti-system attitudes.
Drawing on previous research, I argue that to understand the paradox where workers with similar observable material characteristics show different political behaviors, we need to focus on workplaces and firm employment strategies, as well as different components of job quality. Following the literature that sees relative economic decline as a precursor to populist voting, I contend that firm strategies shape workers’ subjective perceptions of their well-being, life satisfaction, and frustration with the political system.
In line with the existing literature, I do not develop a new theory regarding relative inequality or workers’ expectations. The saturated and often contradictory literature on populism converges on the empirical regularity that relative position and mismatches between labor market outcomes matter for explaining workers’ dissatisfaction.
The literature indicates a connection between relative material decline, expectations, and support for populists (Burgoon et al. Reference Burgoon, van Noort, Rooduijn and Underhill2018; Kurer Reference Kurer2020; Häusermann et al., Reference Häusermann, Kurer and Zollinger2023). I argue that anti-system attitudes are related to the experiences workers face in their firms, specifically their lived experiences in workplaces and their subjective perceptions of their employment situation relative to their expectations.
Context
Before Jair Bolsonaro was elected President of Brazil, the country faced a severe economic crisis. Brazil’s economy had experienced a period of relative stability during the early 2000s, fueled by rising commodity prices and government initiatives under the Partido dos Trabalhadores. However, the global financial downturn in 2008 exposed vulnerabilities in Brazil’s economy, leading to prolonged economic decline. The country struggled with fiscal imbalances, corruption scandals, and a loss of investor confidence. A sharp decline in commodity prices further hindered economic recovery. As Avritzer and Rennó (Reference Avritzer and Rennó2021) show in the context of the COVID-19 pandemic, democratic legitimacy in Brazil has been increasingly undermined by a perceived inability of institutions to respond effectively to citizen needs.
This context may help explain the electorate’s appetite for outsider figures and rejection of established parties, even if the chosen candidate – Bolsonaro – had long held office and cycled through various minor parties. Yet what remains less understood is not merely Bolsonaro’s personal success, but the resonance of a broader anti-system orientation – marked by authoritarian appeals and institutional distrust – across a wide swath of the electorate. This orientation has proven remarkably durable, continuing to shape Brazilian politics even after Bolsonaro’s defeat.
The movement that took Bolsonaro to power in 2019 predates his candidacy by several years, originating from the early 2015 protests against Dilma Rousseff’s government. At that time, diverse and ideologically heterogeneous groups like the economically liberal Free Brazil Movement (Movimento Brasil Livre), the anti-corruption movement Come to Streets (Vem Pra Rua), Revolted Online (led by an Evangelical leader), and several pro-military groups coordinated protests despite lacking a unifying agenda or political program other than rejecting mainstream parties on both the left and right (Rocha et al. Reference Rocha, Solano, Medeiros, Rocha, Solano and Medeiros2021). Eventually, after various attempts to shape candidacies for different offices, this loose movement laid the groundwork for Bolsonaro’s electoral campaign.
The rise and continuity of this anti-system political movement are not yet fully explained in the political science literature. Standard narratives like class conflict do not pass empirical tests, as Bolsonaro’s electorate is highly diverse. Other explanations, such as corruption scandals or the role of Evangelicals, are insufficient to explain its broad reach, ideological heterogeneity, and temporal persistence. In this article, I address this gap by analyzing how firms shape inequality in Brazil and how this affects workers’ feelings of frustration with the political system and the development of anti-system political attitudes.
The Puzzle of the Bolsonaro Voter
The existing literature provides limited insights into Bolsonaro’s electoral base (Hunter and Power Reference Hunter and Power2019; Rocha et al. Reference Rocha, Solano, Medeiros, Rocha, Solano and Medeiros2021; Chaguri and do Amaral Reference Chaguri and do Amaral2023). While it is not uncommon for radical-right populists to attract a heterogeneous and ideologically diffuse electorate, the Brazilian case underscores this pattern with particular intensity. Rather than coalescing around a clear programmatic core or a stable social constituency, the movement that eventually propelled Bolsonaro emerged from a fragmented grassroots field marked by diffuse grievances and limited organizational direction. This lack of consolidation – especially among working-class voters – parallels the experience of populist waves in Southern Europe, and stands in contrast to the more ideologically coherent radical-right formations seen in Northern Europe.
What sets this case apart is that, contrary to the typical trajectory of a movement evolving into an institutionalized network of local political organizations, eventually transforming into a more conventional party (as seen with Lega Nord or Ciudadanos), Brazil’s unique and turbulent political climate in 2019 propelled this movement directly to the Planalto, the seat of executive power (Rocha et al. Reference Rocha, Solano, Medeiros, Rocha, Solano and Medeiros2021).
Despite rhetoric about class conflict, neoliberalism, and late capitalism prevalent in the qualitative academic literature in Brazil (Richmond Reference Richmond2020; Søndergaard Reference Søndergaard2023), there is no clear evidence that explanations based on an upper-class backlash against the poor have much traction regarding the composition of Bolsonaro’s electoral base.
Demographically, Bolsonaro’s electorate is challenging to characterize in terms of social stratification (Chaguri and do Amaral Reference Chaguri and do Amaral2023). Education, income, being a formal worker, or being an impoverished formal worker are not predictors of Bolsonaro support. Instead, holding authoritarian values (believing that a military coup is a solution to political problems) is a strong predictor of this support (Rennó Reference Rennó2020; Chaguri and do Amaral Reference Chaguri and do Amaral2023).
This pattern is consistent with findings from Latin America more broadly. For instance, Cohen and Smith (Reference Cohen and Erica Smith2016) show that authoritarian predispositions – measured through non-political traits such as parenting values – can significantly predict support for right-wing authoritarian candidates. These traits may not be explicitly political in origin but are activated under specific conditions, such as perceived threats to order or status.
Beyond consensus on some conservative issues, Rennó (Reference Rennó2020) and Chaguri and do Amaral (Reference Chaguri and do Amaral2023) found that a key predictor of voting for Bolsonaro is expressing authoritarian attitudes and sentiments against democracy. Ultimately, behind Bolsonaro is an ideologically multifaceted, socially heterogeneous electoral base, unified primarily by anti-system sentiment: initially anti-PT, but also rejecting all major parties and questioning democracy.
This aligns with Cohen et al. (Reference Cohen, Smith, Moseley and Layton2023), who show that even after Bolsonaro’s electoral victory, his supporters continued to express high levels of support for institutional ruptures. Their work confirms that Bolsonarismo is not merely reactive or episodic but reflects a more persistent orientation toward democratic backsliding.
In my own original survey of full-time workers in Brazil, race is the only demographic factor that consistently correlates with Bolsonaro’s vote. Although race overlaps with class composition in Brazil, it remains a strong predictor when pooled with other variables. This suggests that the ‘classist’ narrative favored by structuralist or Marxist scholars is not a strong factor in explaining Bolsonaro’s rise.
The evidence indicates that Bolsonaro aligns with the type of radical-right populism analyzed by Akkerman et al. (Reference Akkerman, Mudde and Zaslove2014) as an ‘ideology’ in its own right. It resembles populism seen in Southern Europe and the United States, defying strict categorization within a programmatic agenda while incorporating elements that appeal to an ideological and sociocultural right emphasizing race, religion, and conservative values. However, in practice, the composition of Bolsonaro’s support base is far more heterogeneous than his most provocative or ideologically charged rhetoric might imply.
An alternative explanation is based on a cultural backlash partly promoted by the strong Evangelical movement. Arguments centered on the role of conservative ideology and Evangelical religion map Bolsonaro to the ‘culturalist’ explanations for the rise of populism (Inglehart and Norris Reference Inglehart and Norris2017; Hunter and Power Reference Hunter and Power2019). While this connection is strong in the data (Amaral Reference Amaral2020), it still fails to explain the diversity and breadth of Bolsonaro’s base.
It is important to note that Evangelicals have long played a strategic role in Brazilian politics, coalescing with governments and parties across the ideological spectrum (Smith Reference Smith2019). While the match might seem natural with someone like Bolsonaro, he is not their exclusive ally in national politics. Moreover, some of his positions have been rejected by notable pastors, specifically one of the founders of the movement that led Dilma’s impeachment because they opposed the free market and neoliberal parts of the ‘new right’ agendas (Rocha et al. Reference Rocha, Solano, Medeiros, Rocha, Solano and Medeiros2021).
Evangelicals field candidates in almost all districts and are typically part of government alliances. But Bolsonaro’s movement is far more than an issue-based Evangelical party. Its rarity lies in bringing together extremely libertarian and extremely conservative factions of the movement that emerged in opposition to Dilma (Rocha et al. Reference Rocha, Solano, Medeiros, Rocha, Solano and Medeiros2021). Just as with other populists with a heterogeneous base like Trump, electorates with conservative values are an important part of their core coalition but do not explain the large electoral shares they obtain.
In terms of programmatic issues, Bolsonaro, like other radical-right populists, ran under a loose program mixing identity politics, authoritarianism, and neoliberal or libertarian economic rhetoric. These issues are diverse and sometimes inconsistent. For example, Bolsonaro supporters are more likely to oppose the legalization of abortion but not more likely to oppose gay marriage or support the imprisonment of women who undergo abortion (Rennó Reference Rennó2020). Among protestors at the March 2015 protests, one of the largest that led to Bolsonaro’s rise, 61 per cent opposed the statement that ‘gay relationships are not natural’ and 50 per cent supported marijuana use (Rocha et al. Reference Rocha, Solano, Medeiros, Rocha, Solano and Medeiros2021).
Given that holding anti-system and authoritarian attitudes strongly predicts Bolsonaro support, we need to explain how these attitudes develop. Data show that these attitudes move independently of cultural or issue-based explanations. While holding conservative values predicts voting for Bolsonaro, cultural backlash explanations do not sufficiently describe his social base. An alternative political economy explanation focusing on firm-based workers’ experiences might provide another path through which anti-system attitudes develop.
This article aims not to adjudicate among potential explanations for a key actor in the worldwide populist wave, but rather to nuance economic explanations of anti-system political attitudes. It highlights subtle dimensions in voters’ economic lives that provide a more complete understanding of how they develop anti-system views, reject democratic values, and support populists. Specifically, I argue that firm-based origins of inequality in wages, benefits, job security, and overall job quality significantly influence the political attitudes of workers, independent of their individual characteristics.
Firms and Inequality in Brazil
Evidence from matched firm–worker data indicates a clear disparity in wages between firms in Brazil, irrespective of worker, industry, or product market characteristics. This goes in line with the aforementioned great separation, which describes the division between employees of high-road and low-road companies. In my argument, such high inequality among otherwise comparable workers, due to the differences among the firms that employ them, is highly consequential politically.
Brazil is not exceptional in terms of this great separation between high- and low-road firms (Alvarez et al. Reference Alvarez, Benguria, Engbom and Moser2018; Godechot Reference Godechot2018). Helpman et al. (Reference Helpman, Itskhoki, Muendler and Redding2017) show that between-firm differences in Brazilian workers’ wages across detailed occupations and sectors increased by 115 per cent after trade liberalization in the 1990s, explaining 55 per cent of total wage inequality. Recent trade theories predict that some firms in the same product market benefit from exporting opportunities, while others do not, thus increasing inequality among comparable workers with limited mobility across firms (Dix-Carneiro and Kovak Reference Dix-Carneiro and Kovak2017; Kim and Osgood Reference Kim and Osgood2019). Additionally, low mobility across local labor markets in Brazil increases firms’ monopsony power (Dix-Carneiro and Kovak Reference Dix-Carneiro and Kovak2017).
The PT period in government saw a rapid decline in overall inequality in Brazil, particularly during Lula’s years (Roberts Reference Roberts2012; Souza and Medeiros Reference Souza and Medeiros2013). However, recent research contradicts some key tenets of conventional wisdom regarding the association between PT governance and lower inequality. Alvarez et al. (Reference Alvarez, Benguria, Engbom and Moser2018) found that 40 per cent of changes in inequality are attributable to firm effects. According to the authors, the decline in inequality during the PT period may be due to a compression of returns to firm-level pay premia. On the workers’ side, while educational attainment increased, the decline in high school and college education premia offset this increase. In other words, more educated workers are earning the same as less educated workers did a decade ago. On the firms’ side, growth and trade expansion would have predicted a higher degree of convergence in firm productivity, but instead, productivity levels across firms became more dispersed. Firm-level pay differences explain the initially high levels of earnings inequality, and the decrease in inequality is due to the weakening of the pass-through effect between firm productivity increases and workers’ earnings.
In other words, inequality decreased without changes in Brazilian firms’ fundamentals and was mainly driven by pay policies. According to Alvarez et al. (Reference Alvarez, Benguria, Engbom and Moser2018), two key policies explain this: first, the increase in minimum wages and second, the higher centralization of collective bargaining under PT governments.
First, for economically active workers whose real incomes increased during the PT period, the driver was mainly changes in the within-firm distribution of earnings, not changes in firm fundamentals. This might have made these workers temporarily better off in absolute terms during periods of growth, but did little to protect them during the recession that began at the end of the PT period. Second, for workers in high-paying firms, the compression achieved through higher centralization of wage bargaining meant that their incomes were relatively closer to comparable workers in less productive firms. Third, the decline in educational premium could have created a mismatch between expected and realized incomes for workers in the early stages of their careers.
As Feierherd (Reference Feierherd2022) has shown, the actual implementation of labor legislation and the local regulatory environment under the PT government was contingent on local politics. Beyond more centralized wage bargaining and increases in minimum wages, there were no significant reforms to reduce the flexibility in determining employment conditions that firms obtained during Cardoso’s government. The PT’s agenda, which focused on wage and income policies rather than addressing labor market segmentation (Feierherd Reference Feierherd2022), made Brazilian workers’ relative situation less resilient to recessions.
To the extent that PT policies had a clear target, such as university quotas, they affected those more likely to have relatively higher incomes at the beginning of their period. Not surprisingly, among all the issues assessed by Rennó (Reference Rennó2020) as predictors of Bolsonaro’s vote, one of the strongest effects (comparable to thinking that ‘elections are not honest’) is the enactment of racial quotas in universities, which has clear labor market consequences.
The decline in earnings inequality during the PT years was real and rapid, driven by deliberate public policies that favored redistribution through the labor market. Decomposition analyses show that from 1995 to 2012, a 20 per cent decline in earnings inequality outpaced the 12 per cent drop in income inequality, primarily due to minimum wage hikes, labor formalization, and wage compression among formal workers (Ferreira et al. Reference Ferreira, Sergio and Julián2022). These changes were accelerated after 2001 when Brazil’s formal wage floor lifted unskilled workers’ earnings, especially in a context of stronger labor demand, contributing to sharp reductions in race-, gender-, and urban–rural wage gaps. Importantly, formal workers, those recorded in the RAIS dataset (that is, Annual Social Information Report), experienced not only real wage increases but also stronger coverage by INSS (Instituto Nacional do Seguro Social) social security, targeted transfers such as Benefício de Prestação Continuada (non-contributory pensions and disability benefits), and higher protections tied to their formal status (Hagopian Reference Hagopian2018). However, these benefits were not uniformly welcomed. As Hagopian (Reference Hagopian2018, 381–382) notes, because ‘workers of color constitute a majority of those earning the minimum wage’, the narrowing of racialized wage gaps and the enactment of affirmative action programs – such as university quotas – produced visible redistributive shifts. These dynamics made workers in higher-paying or higher-prestige positions within the formal labor market feel their status was being eroded relative to less skilled workers.
These perceptions were exacerbated by what Hagopian (Reference Hagopian2018) terms Brazil’s ‘truncated welfare state’, in which access to social rights remained stratified and formally conditioned. Rigid labor market laws, rooted in the legacy of the Vargas-era system, cemented boundaries between insiders and outsiders. For formally registered workers, often white, male, and better educated, job-linked social protections and labor standards reinforced a sense of exclusivity. In turn, they opposed policies that threatened to dilute their relative privileges. As Hagopian notes, many such workers ‘fought to prevent the dilution of their benefits to compensate the worse off’ (2018, 378). These feelings of loss were not merely material but also symbolic: the PT’s redistributive agenda – through minimum wage policy, racial quotas, and wage compression – flattened longstanding labor hierarchies. For workers who perceived themselves as meritocratic achievers within the formal sector, these changes fostered a sense of relative decline and injustice. This is a key mechanism linking PT-era labor policies to the rise of anti-system sentiment among precisely those formal workers who, while better off in absolute terms, came to feel alienated by a model of redistribution that blurred the boundaries of status and reward.
Data and Results
To test the theory, I combined extensive administrative longitudinal data from Brazil’s social security administration (the RAIS dataset) with data on individuals’ support for Bolsonaro’s attempt to create his own party. Additionally, I used an online survey targeting a nationally representative sample of full-time workers, measuring various dimensions of job quality and political attitudes, a conjoint experiment varying firm-level employment characteristics, and an information experiment varying information on wages.
Although testing this argument was challenging due to many unobservable factors influencing both firm selection and political attitudes, the combination of these empirical approaches – including rich and original data – supported the role that firms play in shaping workers’ political attitudes through their employment strategies. The first design showed that Bolsonaro supporters experienced a relative decline in their real wages in the five years before Bolsonaro’s election compared to very similar workers who did not express support for Bolsonaro’s party.
The online survey indicated that workers dissatisfied with different aspects of job quality were more likely to express anti-system political attitudes. The two experiments in the survey provided additional insights. First, various firm-level characteristics (such as contract stability and ethnic diversity quotas) influenced the likelihood of associating workers in those firms with anti-system political attitudes. Second, respondents informed that they earned less than comparable workers of different races were significantly more likely to believe their situation was unfair and that Brazilian democracy was not functioning well.
Employment Trajectories and Bolsonaro Sponsors
In my first empirical approach, I used individual-level administrative data to evaluate the determinants of expressed support for Bolsonaro among Brazilian workers who were comparable in terms of all observable characteristics in their formal employment records. At the end of 2019, Bolsonaro found himself without a partisan label to compete in the next presidential election after a dispute with the leaders of the Social Liberal Party, the party he had used as his platform in the previous election. He announced the creation of a new party, Alliance for Brazil (Aliança pelo Brasil), and its followers quickly came to be identified as ‘Aliados’. The key task for Bolsonaro’s movement was to gather enough signatures to allow this new party to compete in the upcoming national elections.
Knowing that their strongest base came from the online movement that preceded his election, Aliados pressured Brazil’s Superior Electoral Court (Tribunal Superior Eleitoral) to enact a previous ruling allowing for digital citizen sponsorships. However, delays and disagreements on implementation slowed this process. Consequently, they turned to their considerable social media reach to organize in-person events in highly transited places (for example, public beaches) to gather support. At these gatherings, citizens could go to a table where a representative of the movement would register their signature expressing support for the creation of the Aliança.
An overview of Aliados’ official communication channels indicates that once again Bolsonaro was driving a heterogeneous grassroots movement. By the end of 2021, however, it was clear that they would not be able to obtain sufficient signatures by the Tribunal’s deadline, despite their efforts. Bolsonaro therefore turned to an alliance with an existing party, the Liberal Party (Partido Liberal).
Data
First, the data on supporters of the creation of Bolsonaro’s party were provided under a freedom of information request I presented to the Tribunal Superior Eleitoral in 2021. While the Tribunal subsequently restricted access to these data following a change in privacy regulations, Bolsonaro abandoned the idea of creating his own party shortly after this decision and terminated signature collection.
These endorsements are an extremely useful data source, despite the Aliados’ failure to collect a sufficient number within the prescribed deadline. If we consider them an honest expression (or a revealed preference) for a second Bolsonaro term, then they are a valuable indicator of radical-right support, even more so than a vote for Bolsonaro, which voters could cast for strategic reasons.
While the Aliança endorsement data captures a subset of strong Bolsonaro supporters, I do not claim that these individuals exhaust the universe of anti-system voters. Rather, I treat their support for the creation of Bolsonaro’s new party as a meaningful, if imperfect, indicator of intense political alignment with his movement. The act of endorsing was relatively low-cost – requiring a simple online submission of name and ID – but not entirely without consequences. While some reputational risks may have existed, the data were not publicly searchable and remained largely obscure, accessible only via batch-based disclosures on the Tribunal Superior Eleitoral’s website, and ultimately retrievable in full only through a formal freedom of information request. This suggests that, although the action was more visible than a private opinion, it did not require the kind of social exposure or organizational commitment typical of formal partisan affiliation.
Nonetheless, I acknowledge that this group may not be sharply distinct from other Bolsonaro supporters, and that some matched individuals who did not endorse the party might still have held similar views. To address this, I supplemented the administrative data with an original representative survey of full-time workers that included observational and experimental components, allowing me to test the same theoretical mechanisms within a broader and more generalizable population.
Second, I drew on the RAIS for employment characteristics. The RAIS is a detailed, matched, firm–worker dataset comprising the contractual characteristics, demographic information, occupation, and wages of every formal worker in Brazil, which includes hundreds of millions of longitudinal employment history data points from each worker. These data were accessed through an agreement with Brazil’s Ministry of Labor.
To assess the reach of this empirical strategy, it is important to consider the relevance of formal sector voters in terms of vote intention data disaggregated by labor market status. In the last Datafolha survey before the second-round election (n = 4,580), Bolsonaro led among voters with formal employment contracts, receiving 48 per cent of their support compared to Lula’s 42 per cent. Among informal workers, however, Bolsonaro’s support dropped significantly to 37 per cent, while Lula’s rose to 51 per cent. This 11-point drop among informal workers represents a sharper shift than that observed across most income brackets above 2 minimum wages (1,996 reais in 2019), where variation tends to be modest (for example, Bolsonaros’s support ranged narrowly from 52 to 58 per cent). This pattern underscored that registration status, more than income per se, captured a key divide in Bolsonarista support. Since over 61 per cent (IBGE 2025) of the economically active population is formally employed in the private sector, examining Bolsonaro’s appeal among these workers offered insight into how political preferences are shaped within a key segment of the electorate – one that plays a central role in Brazil’s economy and labor relations.
My initial database of sponsors comprised 69,355 individuals. I was able to match 18,117 to the RAIS data using Levenshtein distance matching on full names and by manually validating the results. This approach excluded common names, and of course Aliados supporters outside the formal workers’ population (non-economically active citizens, informal workers, and business owners).
Research design
To understand how Aliados’ employment trajectories compare to other similar workers, I exploited the extensive RAIS dataset to obtain a comparable group of workers who were very similar in their observable characteristics but who did not sign support for the creation of Aliança pelo Brasil. Specifically, I performed one-to-many matching (each Aliado matched to five equivalent workers) by exact matching on the occupation at the highest level of disaggregation (six-digit occupational codes that follow International Labour Organization standards), municipality, educational level, gender, and distance matching on age. This controlled for the main observable sources of wage heterogeneity to a very detailed level, as occupations were matched on six-digit codes (for example, barista, sushi chef, janitor, receptionist, pinewood worker, welder). This strategy accounted for the main observable sources of wage heterogeneity and also helped attenuate concerns regarding spatial variation in purchasing power, since municipality-level matching controlled for regional cost-of-living differences in a statistical sense.Footnote 1
To consider a reasonably long and politically relevant period, I compared the employment situation of each group in 2014 and 2019. I chose 2014 because it marks President Dilma Rousseff’s successful re-election. This five-year period is important because the country entered a series of economic and political crises that culminated in Dilma’s impeachment and President Temer’s unpopular government. It is therefore an important period to evaluate whether there are observable differences between the two groups that could be tied to their employment trajectories.
To do this, I used a two-way fixed effects model that allowed me to compare the change in relevant observable employment characteristics, while controlling for individually constant characteristics. My quantity of interest was the change in different measures of compensation and relative employment position in the post-period between the group that expressed support for Bolsonaro and those in the comparison group that did not.
Results
Simple descriptive statistics (see Figures 1 and 2 and Table 1 in the Appendix) showed that despite being equal in key observable characteristics, the trajectory of Bolsonaristas relative to their reference groups (assessed by educational achievement and occupational groups) and in terms of the median wages paid by their firms has been one of relative decline. This decline, however, was not seen among workers who did not express support for Bolsonaro.
Figure 1. Five-year change in average wage difference with respect to the individual’s educational group.
Note: 95 per cent confidence intervals.
Figure 2. Five-year change in average wage difference with respect to the individual’s six-digit occupational group.
Note: 95 per cent confidence intervals.
This result was tested in the two-way fixed effects design presented in Table 1. In this model, the coefficient on the interaction between being an Aliado and the time period in which citizens could express their support for Bolsonaro’s party identified the change in wages among Aliados relative to neutral workers, after controlling for individually fixed characteristics and across two groups that (thanks to the initial matching procedure) were already very similar.
Table 1. Ordinary least squares – robust standard errors
Note: **p < 0.05; ***p < 0.01.
Interestingly, if we look at the absolute values, the profile of workers that supported Bolsonaro seems to be better compensated than others in their same occupational groups. Yet they also seemed to experience a rapid decline that was not seen among the non-Bolsonaristas.
Being a Bolsonaro supporter was associated with the average yearly wage of a worker having declined by 102 reais, and the educational and occupational ‘premia’ above the reference groups being 95 and 161 reais respectively. To understand the magnitude of these differences, consider that the decline in occupational premium represents almost 15 per cent of the average premium in 2014.
While it was not possible to attribute this decline to wage-compression policies induced by PT’s policies, the wage differences do indicate that diminishing wage premia and inequality within occupational groups seem to correlate with a higher likelihood of developing anti-system political attitudes among those experiencing the relative decline. In this period, too many factors were affecting the Brazilian economy, from a large-scale recession, fiscal tightening, and labor reforms, to be able to attribute these changes to any particular policy. What is interesting, though, is that declining inequality across very similar groups of workers might be spawning anti-system reactions. This aligns with the notion introduced in the theory that firm-based sources of inequality, rather than large differences across the overall income distribution, might be politically consequential.
Experimental Evidence from an Online Survey
While the aforementioned results provided suggestive evidence for my hypothesis, they may indicate the presence of a very specific phenomenon among the most politically mobilized and frustrated workers. Additionally, though the approach combining matching and worker-fixed effects minimized heterogeneity, it did not solve the key endogeneity problem when trying to make causal inferences about politics based on firm–worker data. Specifically, time-varying unobservable characteristics of workers affected both their selection into jobs or firms and their political attitudes. For example, depressed or sick workers might underperform at work, and their frustration or anger might also make them more likely to support Bolsonaro, without a causal relationship between job characteristics and political choice.
In other words, the design above faced two problems: first, generalizability or ‘external’ validity, as it was based on a very specific population subset; second, while illustrative of an interesting pattern, it fell short of being able to causally identify the effect of changing employment characteristics on anti-system political attitudes.
To address these concerns, I implemented an online survey targeted exclusively at full-time workers (n = 1,373). The survey included both a battery of job quality questions (see Table 2) and two survey experiments. Together, these pieces provided strong evidence that there was a relationship between firm-level working characteristics and the development of anti-system political attitudes. The design replicated an instrument that I had already fielded in Italy, with the questionnaire adapted in translation for cultural and context adequacy.Footnote 2 The survey was fielded online using quota sampling, with nationally representative quotas for gender and age. In the following sections, I present my observational results before doing the same for each embedded experiment.
Observational results
First, I implemented a reduced battery of employment questions based on the European Working Conditions Survey and a standard battery of anti-system political attitudes (see Tables 2 and 3) (Muller et al. Reference Muller, Jukam and Seligson1982). This section aims to answer whether job quality, measured through detailed items and disaggregated into its different dimensions, has a relationship with holding anti-system political attitudes, independently of the most obvious predictors that relate both political attitudes and why some workers are employed in certain firms (such as educational levels or industry).
Table 2. Battery of employment-related questions
Table 3. Battery of political attitudes
The components in the employment battery measure different aspects of job quality, which the literature assumes will be lower in low-road firms. These included compensation and benefits, relationships and experiences of autonomy and fairness within the firm, job security, work–life balance, and availability of training opportunities in their workplace. Each dimension was reduced and scaled in the same direction. The political anti-systemness of each respondent was reduced using the same procedure. All measures are standardized for ease of interpretation.
The results (see Tables 4, 5 and 6) showed an inverse relationship between job quality and anti-system attitudes across all dimensions of job quality. More interestingly, even when pooling all the (highly collinear) dimensions together, my measures of autonomy, engagement, and fairness in how workers perceive they are treated by their firm had an even larger effect on anti-systemness than compensation and benefits. Additionally, the measure of work–life balance also had a large effect in this more saturated model.
In other words, even after accounting for detailed industry, firm, and worker characteristics among a representative sample of full-time Brazilian workers, those who perceived poorer job quality were considerably more likely to manifest higher levels of anti-system political attitudes.
Table 4. Ordinary least squares estimates of standardized principal components of political items on principal components of employment items
Note: *p < 0.1; **p < 0.05; ***p < 0.01.
Table 5. Ordinary least squares estimates of standardized principal components of political items on principal components of employment items
Note: *p < 0.1; **p < 0.05; ***p < 0.01.
Table 6. Ordinary least squares estimates of standardized principal components of political items on principal components of employment items
Note: **p < 0.05; ***p < 0.01.
Information experiment
The first experiment provided respondents with information regarding the salaries of hypothetical workers who were similar to them in observable characteristics and employed in a firm representative of their industry. The treatment group saw salaries that were between 8 per cent and 16 per cent higher than their reported salary, with the percentage difference varying across race groups, following a realistic distribution of wages in the RAIS data. This way, respondents were informed of two things at the same time: that they earned less than others in their occupational group and how much less they earned compared to similar workers of other races.Footnote 3
In line with the theory and with the results from the section ‘Data and Results’, the expectation was that a worker who observed that their salary was lower than their peers would react with frustration, feelings of unfairness, and eventually develop anti-system political attitudes. The outcomes were measured in a large battery of anti-system attitudes, with perceptions of democracy (being the most salient item in the principal component analysis) positioned right after the treatment uptake questions. The outcomes were measured on five-point scales.
The results in columns 1 and 2 in Table 7 show that the treatment uptake was significant, and that respondents exposed to information showing their salaries were low were less likely to think the salary they were paid for their work was fair. More importantly, we see a similarly sized and significant effect on perceptions of democracy, with workers exposed to the information treatment being 0.13 points (on a five-point scale) less likely to think that democracy works well in Brazil.
Table 7. Average Treatment Effect of receiving relatively lower salary information among comparable workers
Note: **p < 0.05; ***p < 0.01.
Conjoint experiment
In the conjoint experiment, respondents were presented with two tasks in which they saw the characteristics of the firms where two hypothetical workers, who were similar to the respondent in educational level, gender, and net salary, were employed. They were then asked which of the two workers they thought said a phrase manifesting a political attitude or engaged in a certain behavior. The characteristics of the firms (the vertical axis in Figures 3 and 4) varied firm ownership, Chief Executive Officer compensation relative to workers, contract length, benefits, the use of diversity quotas for promotions, working schedules, and informality.
Figure 3. Average marginal component effects on attitudinal items.
Note: 95 per cent confidence intervals.
Figure 4. Average marginal component effects on behavioral items.
Note: 95 per cent confidence intervals.
The results showed that respondents thought comparable workers were considerably more likely to think that there was no real democracy when they were hired informally (average marginal component effect (AMCE) = 0.11), when working times were not flexible (AMCE = 0.11), and when they lacked benefits (AMCE = 0.9). In general, low-road characteristics moved in the expected direction, though the strongest effects on anti-system attitudes and behaviors were on informality, benefits, working times, and diversity quotas.
Interestingly, using diversity quotas as a criterion for promotions (a policy strongly associated with PT’s agenda) had large effects on respondents thinking that they were treated unfairly, that differences between the elite and the people were larger than differences among the people (a key component of populist ideology in the Akkerman et al. Reference Akkerman, Mudde and Zaslove2014 battery), as well as taking part in protests and being part of groups that wanted to overthrow the government.
Conclusion
This article challenges the prevailing narratives that attribute Bolsonaro’s electoral base to class conflict or an upper-class backlash against the poor. Instead, it argues that anti-systemness, along with authoritarian attitudes, is a defining feature of Bolsonaro’s supporters. I posit that the development of these attitudes is related to workers’ firm-based experiences, both in terms of their relative wages and the differentials in several components of job quality. The relative decline in workers’ well-being and job quality due to firm-based strategies contributes to their anti-system views and frustration with the political system. Importantly, declining income inequality in the general population (as measured through the Gini index) might not counteract and could even reinforce this effect.
To evaluate this argument, the study examined the employment trajectories and characteristics of Bolsonaro supporters in Brazil during the period in which he attempted to create his own political party. The data, obtained from supporters’ signatures collected for the party’s creation, provided a unique opportunity to analyze the employment experiences of this group and to compare them to a similar group of non-supporters. The findings indicated that Bolsonaro supporters experienced a relative decline in their employment situation, wages, and occupational premiums compared to non-supporters. The decline in occupational premiums, in particular, indicated a significant change that may have contributed to their anti-system political attitudes.
The article also presented evidence from an online survey that showed that anti-systemness declines with increases in job quality across several dimensions. This suggests that workers’ perceptions of how they are treated within the firm and their overall work experiences play a significant role in shaping their anti-system political attitudes. Furthermore, the study conducted an information experiment in which respondents were provided with information about the salaries of comparable workers in their industry. Those told their salaries were lower than their peers were more likely to perceive their own salary as unfair and to have less confidence in the functioning of democracy in Brazil. Finally, a conjoint experiment varying firm-level characteristics provided more insight into the mechanisms, experimentally finding that firm-level employment conditions were associated by respondents with political attitudes and behaviors.
Together, these results provide valuable insights into the relationship between employment trajectories and the development of anti-system political attitudes in Brazil. Understanding the development of anti-system attitudes among workers is crucial in unraveling the puzzle of Bolsonaro’s support. Firm employment strategies and components of job quality, rather than observable material characteristics, shape workers’ subjective perceptions and their dissatisfaction with the political system. The article argues for the need to examine the intricate dynamics within workplaces to gain a more comprehensive understanding of how anti-system views are formed.
This article contributes to the growing literature on the political consequences of economic inequality by highlighting the importance of firm-level dynamics in shaping individuals’ political views. Moreover, it demonstrates the potential consequences of firm-based sources of inequality and relative decline within occupational groups for democratic politics. While the specific causes of the decline cannot be attributed to any particular policy, the findings suggest that diminishing wage premiums and increasing inequality within similar groups of workers may contribute to the development of anti-system views. Addressing issues related to compensation, autonomy, fairness, and work–life balance in the workplace may therefore contribute to reducing anti-systemness among workers.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0007123425100859
Data availability statement
Replication data and code for the survey section can be found in the Harvard Dataverse: https://doi.org/10.7910/DVN/K9EZ32. The same repository contains the code for replicating the steps of merging the data, creating the variables and running the models that used the RAIS and Apoios data. Instructions for requesting these datasets from Brazil’s authorities can be found in the Supplementary Materials.
Acknowledgements
For helpful comments on various drafts of this article, I thank Ben Schneider, Peter Hall, Kathleen Thelen, and Volha Charnysh. I am especially grateful to Danny Hidalgo and Guillermo Toral for their work in maintaining the RAIS data repository at Massachusetts Institute of Technology, and to Danny for early intellectual guidance on methodological design. I also thank Suzanne Berger for her inspiration. For excellent research assistance, I thank Lisi Ferrero.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing interests
None to disclose.
Ethical standards
The research was conducted in accordance with the protocols approved by the Massachusetts Institute of Technology IRB.
Open access
