English (pdf)
Article in xml format
Article references
Automatic translation
Send this article by e-mail
Cited by SciELO 
Cited by Google 
Similars in
SciELO 
Similars in Google 
Permalink
Worldwide democracy has been in decay over the last 16 years (Repucci & Slipowitz, 2022). This has led to a growth in protests and Brazil’s recent history provides an interesting example of this phenomenon: from 2013 to 2023, approximately 15 nationwide protests occurred, with the majority and the most populous ones concentrated after 2016 (Brazil’s Protests List, 2022). This has been associated with a growing polarised context (Couto & Modesto, 2021; Gloria-Filho & Modesto, 2022; Ortellado et al., 2022; Prazeres, 2022), but polarisation issues alone are not enough since other issues can come into play, such as the erosion of institutions that uphold democratic governance and values (Repucci & Slipowitz, 2022). In early 2023, anti-democratic riots in Brazil against Luís Inácio Lula da Silva’s election are an example of the undeniable effect of polarisation (Nicas & Spigariol, 2023); Lula, a popular left-wing politician, won the election against Jair Bolsonaro, the former right-wing president, making these riots resemble what happened on January 6 in the U.S (Cameron, 2023). Notwithstanding, these far-right protests found little response from those who are politically left, which brings into question why only a few people act, even under unfair conditions.
One reason we might have missed a left-wing uprising is related to protests being only one form of collective action. Collective action is any act that represents a group and aims to improve this group’s conditions (van Zomeren et al., 2008) or maintain its social status (Mikołajczak & Becker, 2019). For example, the hashtag “semAnistiaPraGolpista” (no amnesty for anti-democratic rioters) was second on Twitter trending topics on January 9th, 2023-one day after the Brasilia (i.e., Brazil’s federal capital) riots took place (GetDayTrends, 2023). Another reason for there not being a proportional response from the left-wing population may be related to what we propose as inaction factors. This study aimed to test a model of collective action that considers these factors.
In this paper, we focus on the Integrated Model of Collective Action (Jost et al., 2017)-and its related models, such as the Social Identity Model of Collective Action (van Zomeren et al., 2008). They propose that people will protest if they identify themselves with a group, perceive this group to have suffered an injustice, and believe that their group can make changes through collective action (Bos, 2020; Osborne et al., 2019; van Zomeren et al., 2008). As our study concerns global political views, we focused on a generalised model for collective action. Nonetheless, we also explored system-challenging and system-supporting tendencies, as proposed by Jost et al. (2017), by grouping participants by their motivations to act in favour or against the system and tested it in a multigroup model.
Therefore, the first variable we consider for our proposed model is relative deprivation. Social psychology models prioritise this variable because social comparison processes and individual differences in fairness change the perception of injustice and affect any behavioural outcomes more directly than objective injustice (Klandermans et al., 2008; van Zomeren et al., 2008). Relative deprivation will predict collective action, but only when this relation is mediated by emotions (van Zomeren et al., 2008). Research focus has turned to anger and other negative affective states because they elicit fast reactions and are associated with relative deprivation (Berkowitz, 1989; Jost et al., 2017).
Given the effect of those variables in predicting a rise in collective action, Jost et al. (2017) propose that the system justification theory can help explain collective action (vs. inaction) by expanding on how groups protest for different system-related tendencies. A system is any predominant so-cial, political, or economic arrangement and system justification is a social, cognitive, and motivational rejection of system alternatives by considering the system as fair, legitimate, and justifiable even if that is against individual or group interests (Jost et al., 2004, 2017; Kay & Jost, 2003). Nowadays, overall systems are commonly related to conservative ideologies, which makes system justification an inaction factor for groups motivated to challenge the system and an action factor for groups motivated to support it (Jost et al., 2008). Returning to the early 2023 Brazilian case, system justification may have had a pivotal role in causing people who considered Lula’s election legitimate (i.e., justifiable by the election system) not to act against the anti-democratic protests. Just as it caused people who considered it fraudulent (i.e., justifiable as a coup against the previous system) to act.
We also propose that learned helplessness will reduce collective action. Learned helplessness is a condition in which a person suffers from a sense of powerlessness, arising from a traumatic event or persistent failure to succeed (Maier & Seligman, 1976; Miller & Norman, 1979). People will learn to be helpless if “they explain their inability to control important events by blaming internal (‘It’s me!’), global (It’ll affect everything I do!’), and stable (It’ll last forever!’) causes” (McKean, 1994, pp. 177-178). As a result, this belief leads to behavioural (e.g., withdrawal, passivity, and procrastination), cognitive (e.g., frustration and low self-esteem), and emotional deficits (e.g., fear, dysphoria, and depression). Again, considering Brazil’s recent uprising, people who hold a politically left orientation had gone through four years of powerlessness from several attacks by Bolsonaro’s government against human rights (Werneck & Guevara Rosa, 2021), science, health, environment (Rodrigues, 2022), and other issues. In fear of a coup by Bolsonaro and his supporters, Brazilians who did not support him may have felt helpless and discouraged to act.
Given that societal and political issues usually develop throughout several years, some groups would experience relative deprivation more constantly than others, which we expect would lead to expectations of uncontrollability or helplessness, thus preventing collective action through its resulting deficits. Learned helplessness is also positively correlated with hopelessness (Quinless & Nelson, 1988), which reduces collective action in individual contexts (Stroebe et al., 2019).
Figure 1 presents our proposed model and our hypotheses-System Justification reduces Collective Action by reducing Relative Deprivation and Anger (H1), and Relative Deprivation decreases Collective Action by generating Learned Helplessness (H2). Finally, an exploratory model was estimated considering participants’ motivation towards the system to better understand the differences in the proposed paths.
Method
This study material, data, and code are fully available at Open Science Framework (OSF; https://doi.org/10.17605/OSF.IO/2UKDJ). Supplemental Information materials are also available through OSF (https://osf.io/xwqvu).
Participants
An online survey was conducted using Unipark’s web-based Survey Software and shared by means of Facebook Ads between July and September 2021. Participants were asked for their consent on voluntary participation and data usage before accessing the survey. The study followed its ethical principles and standards, as well as the American Psychological Association standards. 1,092 participants completed the survey; we removed 131 participants due to their age (younger than 18 years), wrong postal code information, and failure to answer the attention check items. The final sample consisted of 961 participants aged from 18 to 79 years (M = 43; SD = 14.55); 52.76% identified as men and 45.37% identified as women; mostly white (84.81%). Participants were from 25 of the Brazilian states and showed an even distribution among the political spectrum (M = 49.19, SD = 28.91, and range from 0 to 100). More characteristics of our participants are presented in the Supplemental Information.
Instruments
Participants responded to the study measures using a slider varying from 0 to 100 with different anchors for each instrument. This allowed us to treat the data as continuous (Norman, 2010; Zumbo & Zimmerman, 1993). We used reduced versions of some instruments in favour of maintaining the estimation of our measurement model, but also for reducing model complexity and warranting stable estimates and statistical power within our sample (Landis et al., 2000). We provide details of the refinements and the analysis considering all items in Supplemental Information. Participants responded to the collective action measures in the order below, then responded to all other measures in a random order and ended with the sociodemographic items. The Codebook (available at Open Science Framework; https://osf.io/abg4v) contains descriptions of each item.
Collective Action. Participants answered six items retrieved from Jost et al. (2012). They were responded as (i) their intent to act on the described behaviours (not at all to absolutely); and (ii) how often they have acted on the described behaviours (never to always). Both scales showed adequate reliability and fit indices (intention: CFI (Comparative Fit Index) = .98, TLI (Tucker-Lewis Index) = .97, RMSEA (Root Mean Square Error of Approximation) = .06, a = .82, w = .83; past behaviour: CFI = .99, TLI = .98, RMSEA = .05, a = .81, w = .82).
Collective Action Motivations. To assess individual motivations for collective action, we developed items describing system-supporting (e.g., “How motivated to take a political stand would you feel to defend your country’s political values?”) and system-challenging actions (e.g., “How motivated to take a political stand would you feel to oppose your country’s customs that hold people responsible for the dificulties they experience?”), to which participants responded on a scale from not at all motivated to totally motivated. We assessed this scale’s validity through exploratory and confirmatory factor analysis, whose results are available in the Supplemental Information. Four items measured system-supporting motivations (a = .78, w = .80) and four measured system-challenging motivations (a = .80, w = .80). The scale presented adequate fit indices, CFI = .94, TLI = .89, except for RMSEA (.12).
System Justification. This version was based on the General System Justification Scale by Kay and Jost (2003). It consisted of six items designed to assess situational aspects of the participants’ perceptions of system justice, legitimacy, and justification (e.g., “Most policies in Brazil seek the greater good”). Participants answered on an agreement scale (from strongly disagree to strongly agree). The instrument showed adequate fit (CFI = .98, TLI = .96) and good reliability indices (a = .90, w = .90), except for RMSEA (.09).
Relative Deprivation. To measure relative deprivation (e.g., “The government respects people like me”), participants indicated their agreement (from strongly disagree to strongly agree) with eight items from the Social Justice Perception Scale (Klandermans et al., 2008). There is a Brazilian version of this scale by Moreira et al. (2018). It is unidimensional and shows adequate fit (CFI = .95, TLI = .93) and reliability indices (a and w of .97), except for RMSEA (.16). Scale items were reverse-coded to assess relative deprivation.
Anger. To measure anger states related to relative deprivation, we used four items from the Positive and Negative Affects Scale (Carvalho et al., 2013; Watson et al., 1988). Participants were asked to indicate, on a scale from not at all (0) to a lot (100), how much, in general, they have felt hostile, irritable, nervous, and upset. The scale displayed adequate fit (CFI = .99, TLI = .97) and good reliability indices (a and w of .87), except for RMSEA (.11).
Learned Helplessness. The Learned Helplessness Scale (Quinless & Nelson, 1988) was adapted for Brazil by Couto and Pilati (2023). It consisted of six items (e.g., “When I do not succeed at a task, I do not attempt any similar tasks because I feel that I would fail at them also”) and participants responded on a scale from strongly disagree to strongly agree. The measure showed adequate fit (CFI = .97, TLI = .95, RMSEA = .07) and reliability indices (a and w of .78).
Procedure and data analysis
Data were analysed using R (v. 4.2.2) with the lavaan (v. 0.6.12, Rosseel, 2012) and semTools (v. 0.5.6, Jorgensen et al., 2022) packages. We employed a structural equation model (SEM) to assess our hypothesis. We conducted Harman’s single-factor test to account for common method variance; poor fit measures and low explained variance suggest that no relevant method biases affect the models (Podsakoff et al., 2003). A multigroup SEM tested group motivation differences in collective action. The multigroup model tested the same structure (i.e., path model, comparison of indirect effects, and measurement model) in each group. We conducted model estimation with the Maximum Likelihood estimator. To evaluate model fit we considered the ratio of chi-square by degrees of freedom (c2/df) below 5, CFI and TLI above .90, and RMSEA below .08 with upper confidence interval (IC) below .10 (Brown, 2006). We also report standardised regression coefficients and effects, 95% confidence intervals, and p-value considering an alpha of .05.
For our multigroup model, we created four groups considering participants’ Bartlett scores on the Collective Action Motivations Scale: (i) System Challenging Motivations (n = 265), which consisted of participants’ scores above the 60 percentile in the System Challenging factor; (ii) System Supporting Motivations (n = 266), consisting of participants’ scores above the 60 percentile in the System Supporting factor; (iii) System Supporting and Challenging Motivations (n = 119), which consisted of participants’ scores above the 60 percentile in both factors-although this group may seem counterintuitive, it represents people who are motivated to support the system in certain issues but challenge it in other; and (iv) No System Related Motivations (n = 311), consisting of participants’ scores below the 60 percentile in both factors. The criterion used attempted to be both conservative in distinguishing participants with high levels of motivation towards the system and in providing homogeneous groups. We opted for using a categorisation process despite its problems (MacCallum et al., 2002) as a parsimonious option to explore how system-related tendencies would affect our model.
Results
Participants answered our criterion variables through the same set of items, which led intention and past behaviour measures to be strongly and positively associated (r = .84, 95% CI (.82, .86), p < .001), suggesting participants did not distinguish them. To avoid collinearity problems, we included only the collective action intentions measure in the model-which was closer to the original measure. We also tested a single-factor model to assess common method biases; this model showed poor fit, c2 (405) = 7550.62, c2/df = 18.64, CFI = .67, TLI = .64, and RMSEA = .14, 90% CI (.13, .14), and was significantly worse than our General Model considering the Chi-square difference test (Dc2 (7) = 5697.02, p < .001). Therefore, we do not expect significant method biases in our data. A correlation matrix of the latent variables is available in Supplemental Information.
The data showed a good fit to the General Model (Figure 2), c2 (398) = 1853.60, c2/df = 4.66, CFI = .93, TLI = .93, and RMSEA = .06, 90% CI (.06, .07). We expected that System Justification would reduce Collective Action intentions through Relative Deprivation and Anger (H1; Figure 2, Paths ‘a’, ‘b’, and ‘c’). A mediation effect was supported; the indirect effect was negative and significant (babc = -.10, 95% CI (-.13, -.06), p < .001). Although expected based on Jost et al. (2017), the direct effect of System Justification on Collective Action intentions was not significant (Figure 2, Path ‘f’; bf = -.04, 95% CI (-.11, .04), p = .35). This suggests that, in general, the more Brazilians justify the system, the less they intend to act collectively when mediated by Relative Deprivation and Anger.
Note. We removed the measurement model to improve visualisation. Dashed lines are nonsignificant paths. * p < .05. ** p < .01. *** p < .001.
Figure 2 General model.
The second hypothesis was that Relative Deprivation would decrease Collective Action intentions through Learned Helplessness (H2; Figure 2, Paths ‘d’ and ‘e’). A full mediation effect was supported. The indirect effect of Relative Deprivation on Collective Action intentions was negative and significant (bde = -.06, 95% CI (-.10, -.03), p < .001), suggesting that Relative Deprivation would decrease Collective Action intentions through Learned Helplessness. We also found that Relative Deprivation predicts higher Learned Helplessness states (Figure 2, Path ‘d’; bd = .35, 95% CI (.28, .41), p < .001); and Learned Helplessness reduces Collective Action intention (Figure 2, Path ‘e’; be = -.18, 95% CI (-.27, -.09), p < .001).
Despite our findings not all people feel the same about the systems currently in place. Participants’ motivation towards the system is expected to affect the relationship between variables. We conducted invariance tests to assess whether there were any relevant group differences to be explored, on Table 1 we present fit indices for the multigroup model with configural and metric invariance tests. Although the results suggest that any identifiable group difference could be due to measuring factors, we believe this variance is due to the political characteristics of our groups and measures, and thus an inherent data difference.
Table 1 Invariance tests for a multigroup model of Collective Action Motivations
Note. Each model was compared to the configural model since no invariance was achieved.
The multigroup model considering Brazilian’s motivations towards the system (Figure 3) showed poorer fit indices when compared to the General Model (c2 (1592) = 3289.92, c2/df = 2.07, CFI = .89, TLI = .88, RMSEA = .07, 90% CI (.06, .07)), but represented an improvement to the General Model considering the Chi-square difference test (Dc2 (1194) = 1436.31, p < .001). It also revealed relevant group differences in the association between variables.
Despite several relevant implications, we will limit our analysis of the Multigroup Model to our initial hypotheses. We found group differences regarding the mediated effect of System Justification on Collective Action through Relative Deprivation and Anger (Figure 3, Paths ‘a’, ‘b’, and ‘c’). In the No System Related Motives (b abc = -.09, 95% CI (-.13, -.05), p < .001) and System Supporting and Challenging (b abc = -.10, 95% CI (-.19, -.01), p = .03) groups-where people had inconsistent group motivation-we found significant negative effects. For the System Challenging (babc = -.01, 95% CI (-.03, .005), p = .18) and System Supporting (babc = -.004, 95% CI (-.03, .02), p = .80) groups-where people had consistent group motivation- we found nonsignificant effects. Furthermore, there is a negative and significant direct effect of System Justification on Collective Action intention in the System Challenging group (Figure 3a, Path ‘f’; bf = -.23, 95% CI (-.38, -.07), p = .004). The results suggest that the inaction effect of System Justification is stronger in people who hold weak political motivations or hold system-challenging political motivations-not being enough to deplete Relative Deprivation or anger towards the system in the latter case.
As for the Learned Helplessness mediation hypothesis between Relative Deprivation and Collective Action, we found non-significant indirect effects in every group. We also found significant negative direct effects of Learned Help-lessness on Collective Action for the No System Related Motives (Figure 3d, Path ‘e’; be = -.16, 9 5% C I (-.31, -.01), p = .04) and the System Supporting and Challenging (Figure 3c, Path ‘e’; be = -.42, 95% CI (-.67, -.18), p = .001) groups. The results imply that people who hold weak political motivations are more susceptible to inaction due to learned helplessness states.
Discussion
Social psychology, among other fields, has been keen on understanding political mobilisation and most of these theories consider relative deprivation a central aspect of collective action (Bos, 2020). But considering the increase in protests worldwide, and Brazil’s polarised political context, we sought to understand why some people do not protest.
Our General Model (Figure 2) suggests that System Justification acts as an inaction factor by reducing Relative Deprivation and group-based emotions (such as Anger) in general contexts, but not directly-corroborating the first hypothesis. We only found the mediated effect of System Justification on Collective Action to be relevant in the General Model and in inconsistent group motivations of the Multigroup Model (i.e., No System Related Motives and System Supporting and Challenging groups). However, there was also a direct effect of System Justification in reducing Collective Action in the System Challenging group (Figure 3).
As for the second hypothesis, we also found support for the derogation of Collective Action intentions by Relative Deprivation and the mediation of learned helplessness in the General Model. Indeed, Relative Deprivation seems to lead to increased states of helplessness, which reduces intentions of Collective Action. Because of the inability of in-dividuals to control their political and economic situation -which depends on societal changes-they may start to believe that nothing they do can change their reality. This may lead to passivity, hopelessness, and giving up (Maier & Seligman, 1976; Miller & Norman, 1979)-or inaction in collective actions (Stroebe et al., 2019). These effects occurred in the General Model and in inconsistent group motivations of the Multigroup Model (i.e., No System Related Motives and System Supporting and Challenging groups).
We also found a moderate and relevant correlation between Anger and Learned Helplessness both in our General and Multigroup Models. Although initially unexpected-anger acts as a mediator increasing Collective Action and Learned Helplessness acts as a mediator decreasing Collective Action-, this relation can be explained by expressive suppression. Solak et al. (2021) found that people who tend to suppress their emotions are less likely to engage in collective action; also, emotion suppression is related to reduced interpersonal and social well-being, such as increased risk of depression (Chervonsky & Hunt, 2017), one of learned helplessness’ consequences.
Considering our Multigroup Model for Collective Action, we must consider that several studies suggested that feelings other than anger might be better at predicting collective action (e.g., sympathy, empathy, and guilt; Osborne et al., 2019). Anger is considered a relevant emotion in predicting collective action because of its capability to elicit fast reactions (Berkowitz, 1989). Anger is also typically associated with unplanned reactions, which is rarely the case in disruptive collective action-such as protests.
The Multigroup Model for Collective Action Motivations raises two other points. Even though only one third of our sample did not identify with a left or right-wing political orientation, the General Model showed closer similarity with the No System-Related Motives. This similarity could suggest that most Brazilians hold weak political motivation-even in a polarised political context (Gloria-Filho & Modesto, 2022). In addition, the presence of an association between learned helplessness and collective action only in groups with inconsistent group motivations provides support for the relevance of social identity in collective action. One’s identification with a group is a prerequisite to collective action, as politicised identities have a greater impact on collective action than non-politicised ones (van Zomeren et al., 2008). Inaction and hopelessness also strongly affect individuals who do not feel motivated and thus do not identify with any social group (Stroebe et al., 2019).
Another relevant variable for the Social Identity Model of Collective Action is perceived group efficacy (Jost et al., 2017; van Zomeren et al., 2008). We chose not to test for group efficacy in this paper, but we expect it would predict learned helplessness since the perception of one’s group inability to change their situation could lead to learning processes related to uncontrollability and helplessness. Yet, this relationship should be tested in future studies.
One limitation of the study was its cross-sectional design. Theoretically, learned helplessness is a process. Its association with relative deprivation predicted that people would learn to be helpless through experiencing relative deprivation. We found a positive association between these variables, but the Learned Helplessness Scale measures only a state of helplessness, and information about the process is lost. A longitudinal study could help to disentangle this association and provide information regarding which deficits produced by helplessness strongly affect collective action.
A longitudinal method could also help identify an association between learned helplessness and group identification. Future research might investigate whether learned helplessness may lead to lower group identification through social isolation (Shaghaghy et al., 2011), which would explain its presence only in non-politically motivated groups in the Multigroup Model. Future research should also assess the causal relationship between variables; some of them -used in the present models-already went through causal relationship testing, i.e., experimental designs in System Justification, Relative Deprivation, Anger, and Collective Action (Jost et al., 2012). The causal effect of Learned Helplessness and its association with other variables must still be tested.
This study provided a test of system justification in the Brazilian context. It proposed a relevant variable for understanding inaction processes underlying collective action. In a nationwide survey, with participants from all political orientations, our findings underscore the risk of relative deprivation as it may lead to helplessness, which is associated with social isolation, and may contribute to a decay in democracy by derogating collective action.
Brazil’s democracy has had a substantive decay since 2020 (Sanches, 2021). The anti-democratic riots in early 2023 are just the last wave of consecutive attacks on democracy with little or no consequence to their actors (Stargardter, 2022). While some of the politically engaged Brazilians acted against these riots, our research points to the effects of system justification in the inaction of the politically engaged; and the effects of learned helplessness in the inaction of the politically disengaged. Political research in Brazil points to most of the population as de-politicised and politically moderated (Ortellado et al., 2022); therefore, learned helplessness may be especially relevant in understanding societal changes1.
