Finding (dis-)advantaged system justifiers: A bottom-up approach to explore system justification theory

System Justification Theory (SJT) postulates that individuals are motivated to justify the status quo, including the paradoxical prediction that those who are disadvantaged (e.g., lower social status) by the prevailing system will justify the system more than those who are advantaged by the prevailing system. We test if this assumption holds in an entire sample, or only among subsets of a sample using a bottom-up approach, Correlational Class Analysis. Using a representative sample from four European countries ( N total = 5157) we found six subpopulations. The first subpopulation ( Justifiers, n total = 1256;24%)wasconsistentwithSJT,thatis,systemjustificationandsocialstatus were negatively correlated. The second subpopulation ( Rejectors, n total = 1688; 33%), however, was characterized by a positive correlation between social status and system justification, which contrasts with the prediction of system justification theory. The other four subpopulations ( N total = 2211; 43%) were characterized by an ambivalent pattern.Thatis,atleastonesocialstatusindicator,butnotall,supportedtheprediction that disadvantaged individuals justified the system more than advantaged individuals. Theseheterogeneouspatternswouldbeundetectedusingtraditionalapproaches.Further, our results show that inequality salience is lower, trust in political institutions is higher, and support for political violence is higher among Justifiers compared to Rejec-tors . We discuss how understanding the interrelations between multiple indicators of social status and how they differ between subpopulations can help us to obtain a more comprehensive picture regarding under which circumstances and for whom system justification theory applies.

examining SJT's assumptions and predictions is sometimes contradictory (e.g., Brandt, 2013Brandt, , 2020;;Owuamalam et al., 2018;Sotola & Credé, 2022;Trump & White, 2018).Typically, contradicting evidence is counted as evidence against the theory, whereas supporting evidence is counted as evidence in favour of the theory.Whether the theory is supported in aggregate depends on whether contradicting or supporting evidence is more convincing (e.g., rigorous method, consistent results, larger effect sizes).Here, instead of testing (again) if SJT is a viable theory in aggregate, we take a different approach and consider for whom the theory most strongly applies.
We aim to contribute to the literature by assessing whether it is possible to identify people who support and contradict SJT within the same sample, using a bottom-up approach.Specifically, we use a Correlational Class Analysis (CCA; Boutyline, 2017) to identify subpopulations which differ in how their objective (i.e., education) and subjective social status (i.e., relative deprivation, subjective income) are associated with system-justifying attitudes.This analysis reveals that the predictions of SJT hold for some subpopulations, but not for other subpopulations.Therefore, our findings neither unequivocally support nor reject SJT.
Instead, we show for whom the theory's predictions hold.

Assumptions of SJT
System justification theory postulates that 'people are motivated to defend, justify and bolster aspects of the status quo, including existing social economic and political systems, institutions and arrangements' (Jost et al., 2015, p. 321).In support of this assumption, scholars have found, for example, that people engage in biased information processing in support of the system (Hennes et al., 2012) and that criticism of the system elicits defence mechanisms ( (Jost & Hunyady, 2005); but see Sotola & Credé, 2022).Further, justifying prevailing systems has many implications for political attitudes and behaviours.System justification attitudes are associated with high trust in institutions (Cichocka & Jost, 2014;Hunyady, 2018;Intawan & Nicholson, 2018;Jost et al., 2003), support for established political parties and candidates (Langer et al., 2022;Satherley et al., 2022), less support of system-challenging movements (e.g., Occupy Wall Street, Jost, 2019), and more support for conservative and/or right-wing ideologies (e.g., Jost et al., 2004).
Here, we focus on reviewing supporting and contradicting evidence for two key claims of SJT: (1) disadvantaged groups report equal or more system-justifying attitudes than advantaged groups, and (2) system-justifying attitudes are higher when inequalities are salient.

1.2
Key Claim 1: Disadvantaged groups tend to justify the system SJT postulates that the system motivation holds-at least to some degree-for everyone, independent of their own position in society.
For advantaged groups, this system-justifying motivation aligns with self and group justification motives because supporting the system maintains their high status.For disadvantaged groups, however, it con-flicts with self and group justification motives.This mismatch between motives for disadvantaged groups is thought to induce cognitive dissonance, an aversive feeling of uncertainty that people seek to avoid (Proulx et al., 2012).Disadvantaged groups can respond to that cognitive dissonance by bolstering their support for the social system and the status quo (Jost et al., 2003).This heightened cognitive dissonance experienced by disadvantaged people and groups implies that they are equally or more strongly motivated to defend the status quo and see the status quo as legitimate compared to advantaged groups (Jost, 2011;Jost et al., 2004;Rudman et al., 2002).Indeed, cross-sectional surveys (Jost et al., 2003;Sengupta et al., 2015), interviews (Durrheim et al., 2014;Godfrey & Wolf, 2016) and experimental approaches (van der Toorn et al., 2015) have found support for the claim that disadvantaged, low-status and low-power groups express more system-justifying attitudes than advantaged, high-status, and high-power groups: 'people are very good at making a virtue out of necessity' (Jost, 2020, p. 3).This prediction is controversial.For example, some scholars have critiqued the cognitive dissonance assumption in SJT, suggesting that system justification serves as an identity management strategy and is most prevalent in people who strongly identify with the devalued group (Owuamalam et al., 2016).Empirically, others found little evidence that disadvantaged groups in general show more system-justifying attitudes in analyses of large and international datasets (Brandt, 2013;Brandt et al., 2020;Caricati, 2017;Davidai, 2018).For example, Brandt (2013) did not find more system-justifying attitudes among disadvantaged groups, including people with lower income or education, and women.In contrast, on average, high-status groups were found to be more likely to support, trust and legitimize the prevailing system (Hetherington, 1998;Jost & Burgess, 2000).Buchel et al. (2021), using the ladder measure of subjective social status, found that objectively disadvantaged individuals were more motivated to defend the system, while subjectively disadvantaged individuals defended the system only in unequal contexts.Other studies also found that disadvantaged groups are less supportive of unequal social conditions (Guimond et al., 2003;Lee et al., 2011) and perceive the current society as less ideal (Zimmerman & Reyna, 2013).In sum, there is contradicting evidence and it remains unclear if objective and subjective social status predicts system-justifying attitudes positively, negatively or not at all.p. 910, Hypothesis 18), empirical tests of the assumption are rare.
Moreover, the existing empirical findings typically do not support the SJT assumption.For example, Trump and White (2018) experimentally manipulated the level of inequality (here the Gini coefficient) and measured overall perceived fairness of the system, trust in institutions and fairness of the economic system.Inequality did not significantly predict economic system justification or institutional trust.Moreover, for general system justification, the opposite effect emerged (i.e., system justification was lower in the high inequality condition).Additionally, their results revealed a mixed pattern for the interaction between social status and inequality.For general system justification a small effect supporting SJT claims was found, while there was no effect for trust in institutions and a small effect for economic system justification contradicting SJT claims.In sum, the authors conclude that across a well-powered representative sample and two replications there is no evidence supporting the claim that inequality increases system justification.
Results from research using large representative data sets (i.e., American National Election Studies, General Social Surveys and World Value Survey, Brandt, 2013;International Social Survey Program, Caricati, 2017) also mostly contradict SJT assumptions.For example, Caricati (2017) used the human freedom index and Gini coefficient as measures of inequality and found that across 36 countries advantaged individuals justify the status more, especially in equal contexts, compared to disadvantaged individuals.Brandt (2013) found mixed results testing SJT assumptions with Gini and Gender Inequality coefficients as measures of social inequalities.Results for the Gini coefficient were not significant or opposed to the SJT assumptions.The findings for the Gender Inequality index revealed some weak support for the SJT claims, but only for social class, and not for any other status indicator like education or income.
In sum, most existing findings do not support the assumption; however, less research examined inequality and previous research exclusively focussed on country-level indicators of economic inequality (e.g., Gini coefficient) or manipulated inequality information.Country-level indicators of inequality might be too broad to capture the potential impact of inequality on system justification.Research focussing on political attitudes and voting patterns indicates that to understand how economic inequality fosters political attitudes, it is important not just to focus on macro-levels of inequality (e.g., Gini coefficient), but also to consider meso-and micro-levels of how individuals experience the places they live in (e.g., McKay et al., 2021;Rodríguez-Pose, 2018).Considering that the proposed underlying mechanism for system justification is reducing cognitive dissonance, salience of inequality in people's surroundings and their personal experience of inequality might be better indicators compared to 'abstract' country-level indicators.
Therefore, we focus on perceptions of inequality in everyday life.
That is, we measure people's self-reported inequality salience in their everyday environment and perceived fairness of resource allocation in society.It may be the case that country-level indicators of inequality are an indirect proxy of people's own inequality perceptions.However, like all indirect measures, the extent to which the country-level indi-cators map onto people's own impressions is likely to be imperfect.
Thus, in this study we apply a more direct measure of subjective perceptions of inequality which can potentially shed new insights on the link between inequality and system justification.We test the prediction that perceptions of inequality will moderate the association between social status and system justification.When inequality perceptions are high, people with lower social status will be more likely to justify the system, whereas the link between status and system justification will be weaker or reversed when inequality perceptions are low.

OVERVIEW
Our review shows that existing work both supports and contradicts SJT's key claims.This is a relatively typical pattern in much of social psychology.Some scholars posit an effect; others find contradicting evidence.The situation is resolved when one side collects enough evidence to show that their side is typically correct (e.g., a meta-analytic estimate significant and in the predicted direction) or an overlooked moderator is identified that explains the different results between studies (e.g., cultural context).The problem with this approach, however, is that it is often not clear in advance what moderator variables might be relevant.Thus, it is never clear whether moderation would have been found in previous studies, or whether there are additional relevant moderators.
We take a different approach.Rather than considering which perspective receives the most support, we instead aim to identify the subpopulations for whom the SJT assumptions hold and the subpopulations for whom the SJT assumptions do not hold.That is, rather than identifying which perspective is right and which perspective is wrong, we aim to identify the people for whom each perspective best applies.
Therefore, the overarching goal of our article is to explore whether the two key claims of SJT that we reviewed hold for the whole sample or only for subpopulations.The claims we test are: Claim 1. Social status negatively predicts system-justifying attitudes.
Claim 2. The association between social status and system-justification is stronger when inequality is salient.
In order to achieve this aim, we take two approaches.First, we take a traditional approach that tests the claims in the whole sample using multiple regression analysis.Second, we take a novel, bottomup approach.Our analysis progresses in a stepwise fashion.First, we inductively identify subpopulations using correlational class analysis (Boutyline, 2017).This technique identifies subgroups who share a similar correlational structure in their responses.The aim of this analysis is to test Claim 1, that is to identify subpopulations who differ in how well SJT Claim 1 applies.To do this, we include all the available items that measure system justification or social status in the CCA.This identifies groups of people who think about the economic system and its relation to their social status in a similar way.In this case, the CCA could identify subpopulations with positive or negative associations between social Next, we test Claim 2 by predicting subpopulation by inequality perceptions.That is, are subpopulations which support SJT Claim 1 more likely to emerge when inequality is salient?We further explore the composition of the subpopulations (e.g., demographics).These subsequent analyses help us to gain a deeper understanding of why SJT is supported among some people, but not among others.
Finally, we are interested whether the subpopulations differ in how system justification is associated with system-legitimizing attitudes like satisfaction with democracy, trust in institutions and support for violence against the government.For political outcomes we deliberately focussed on attitudes that directly refer to important societal institutions.It is the legitimacy of these institutions which will ultimately determine the level of societal unrest (Farrell & Knight, 2003;Putnam, 1993).The analysis serves to explore if how people think (i.e., subpopulation) is meaningful beyond what people think (i.e., mean system-justifying attitude) for the legitimizing of these institutions.

Sample and procedure
The Data and materials necessary to replicate our findings are available on OSF (https://doi.org/10.17605/OSF.IO/GVCFX).

Instruments to detect subpopulations
Economic System Justification.Economic System Justification was measured using three items (i.e., item 1, 7 and 12 of the original scale) 1 developed by Jost and Thompson (2000).An example item reads 'Economic positions are legitimate reflections of people's achievements' .All items were measured on a scale from 1 (Strongly disagree) 1 Due to limited study capacity, only these three items were measured in the study.
to 5 (Strongly agree).Internal consistency was satisfactory, Cronbach's α = .76(France = .77;Poland = .64;Spain = .81;UK = .79)Social status variables.Education was used as an indicator of objective social status.Education was coded into two categories: individuals without university degrees (coded −0.5) and individuals with a university degree (coded 0.5).This decision was made because education effects are mainly driven by differences between individuals with university degrees and all others (Easterbrook et al., 2016).
Two indicators of subjective social status were used: subjective income and relative deprivation.For subjective income, respondents indicated on a 5-point scale to what extent they felt that they were able to live a comfortable life with their current household income (1: very easy-5: very difficult).We reverse-coded this item so that higher scores reflect living more comfortably.

Instruments to describe subpopulations
Subjective Inequality in Everyday Life.Subjective Inequality in Everyday Life (SIEL) was measured using four items (i.e., item 4, 7, 8 and 9 of the original scale) developed by García-Castro et al. ( 2019).An example item reads 'Among the people I surround myself with, some can afford to buy a lot more and better things than others' .All items were measured on a scale from 1 (Strongly disagree) to 5 (Strongly agree).Internal consistency was satisfactory, Cronbach's α = .84(France = .87;Poland = .83;Spain = .83;UK = .80).
Perceived Inequality of Resources.Perceived inequality of resources (PIR) was measured using two items adapted from Sánchez-Rodríguez et al. ( 2019).An example item reads 'The resources in [Country] are distributed unequally' .All items were measured on a scale from 1 (Strongly disagree) to 5 (Strongly agree).The two items correlated with r = .73(France = .82;Poland = .70;Spain = .66;UK = .73).
Identification.Identification with education and income was measured using two items: 'I identify with people with a similar level of education to my own' and 'I identify with people with a similar level of income to my own' .Items were measured on a scale from 1 (Not at all) to 6 (Extremely).The two items correlated with r = .73(France = .68;Poland = .77;Spain = .73;UK = .71).We included these items to explore if levels of identification differ across possible subgroups of participants.
Status Anxiety.Status anxiety was measured using five items (i.e., item 1 to 5 of the original scale) from Melita et al. (2020).An example item reads 'I feel anxious that I will be stuck in my position for life' .All items were measured on a scale from 1 (Strongly disagree) to 5 (Strongly agree).Internal consistency was satisfactory, that is, Cronbach's α = .91(France = .91;Poland = .88;Spain = .90;UK = .93).We included this measure to explore

Political outcomes
Trust in institutions.Trust in institutions was measured using three Support of violence against the government.Support of violence against the government was measured using three items that were developed within the broader research project.An example item reads 'Violent action against the government is needed to bring about real change' .All items were measured on a scale from 1 (Strongly disagree) to 5 (Strongly agree).Internal consistency was satisfactory, Cronbach's α = .85(France = .85;Poland = .91;Spain = .79;UK = .78).

Control variables
As control variables in the regression models, we included participants' age (grand-mean centred) and gender (coded male = 0.5, female = −0.5).For complete information about all items included in the original survey see codebook on OSF.

Analytical strategy
We conduct our analyses in two steps.First, we use the traditional approach to test Claim 1 that disadvantaged groups (i.e., low education, low income and high deprivation) tend to justify the existing system more than advantaged groups in the full sample.We regressed education, income and relative deprivation on system justification, including fixed effects for countries.
In the second step, we use the bottom-up approach by conducting a CCA using the full sample with the corclass R package (Boutyline, 2016).
We included the economic system justification items and social status items (eight items in total, see section 'instruments to detect subpopulations').Using CCA allows us to test our assumptions inductively by sorting individuals into subpopulations based on similarities in their correlation pattern, instead of their absolute scores.This means individuals within one subpopulation can differ in their absolute score of the included variables (e.g., individuals with high and low perceived relative deprivation could be sorted into the same subpopulation) but will be similar regarding the interrelations among system-justifying atti-tudes and social status (e.g., all individuals in the same subpopulation will have a stronger negative association between system justification and social status).This approach differs from other frequently used clustering methods like hierarchical cluster analysis in which clusters are determined based on similarity of scores on a set of items.

RESULTS
Claim 1. Social status negatively predicts system-justifying attitudes

Traditional approach
We conducted multiple regression analysis, predicting systemjustifying attitudes by social status indicators (i.e., income, education, relative deprivation).Income and inequality perception were mean centred in each country and education was coded as −0.5 (lower educated, that is without university degree) and 0.5 (higher-educated, that is bachelor's degree and higher).We included fixed effects for country and controlled for gender and age.Results are displayed in Table 1 (separate country analyses are reported in SOM Tables S1-S4).
The results show that individuals with higher income hold more system-justifying attitudes in the full sample and in all four countries separately.Further, there was no significant relationship between relative deprivation and system justification.This is inconsistent with the claim that disadvantaged groups justify the system (more).However, in the full sample individuals with higher education had less systemjustifying attitudes.This finding held in the Spanish and British sample, but not in the French and Polish sample (see SOM).This is (partly) consistent with the SJT claim.In combination, the findings show that the relationship between social status and system justification might differ for subjective and objective measures and between countries.All findings hold after controlling for gender and age (see Supplementary  Online Material).The findings show that in the full sample, people with more conservative attitudes, who are older, or who are male hold more system-justifying attitudes.Additionally, gender was only a significant predictor in Spain and age was not significant in the United Kingdom.
In sum, when following a traditional approach, our initial findings do not fully support the claim that disadvantaged groups hold more system-justifying attitudes when considering income or relative deprivation, but our findings do support the claim when status is measured using level of education (these associations were not significant in all countries, but were in the predicted direction).These mixed results are emblematic of the status-system justification literature.One way to advance this discussion is to follow a more bottom-up approach and try to identify subpopulations for which the key predictions of SJT (may or may not) hold.

Bottom-up approach: Correlational class analysis
In the following, we used CCA to first find subpopulations who share interrelations between social status indicators and system-justifying attitudes.The CCA identified six subpopulations in the full sample (for country description see SOM Table S5-S18).Below, we will describe each subpopulation in turn.2Subpopulation 1: Justifiers.In the first subpopulation, the system justification items were negatively associated with income (∼r = -0.18,p < .001)and education (∼r = -0.49,p < .001),and positively with relative deprivation (∼r = 0.62, p < .001)(see Figure 1 for a network plot; see SOM Table S19).That is, people who are disadvantaged in one of the three domains (i.e., lower income, lower education and feeling deprived) were more likely to justify the status quo, and vice versa.That is, this subpopulation shows a correlational pattern which fits with the claims of SJT.In total, 24% of the full sample belonged to the Justifiers population (French sample: 21%; Polish sample: 29%; Spanish sample: 24%; UK sample: 22%).
Subpopulation 2: Rejectors.In the second subpopulation, the system justification items were positively associated with income (r = 0.33, p < .001)and education (r = 0.48, p < .001),and negatively with relative deprivation (∼r = -0.27,p < .001)(see Figure 1 for a network plot; see SOM Table S20 for correlation matrices).That is, people who were advantaged in three domains (i.e., higher education, higher income and affluent) were more likely to justify the status quo, F I G U R E 1 Subpopulations depicted as networks.ESJ, economic system justification; RD, relative Deprivation; edge thickness represents correlational strengths (blue = positive, yellow = negative).Blue nodes represent social status variables, red nodes represent economic system justification items.and vice versa.That is, this subpopulation shows a correlation pattern which contrasts with the claims of SJT.In total, 33% of the full sample belonged to the Rejectors subpopulation (French sample: 35%; Polish sample: 34%; Spanish sample: 31%; UK sample: 31%).
They were positively associated with income (r = 0.09, p ≤ .006)and negatively with relative deprivation (∼r = 0.27, p < .001)rejecting SJT claims (see Figure 1 for a network plot; see SOM Table S22 for correlation matrices).In total, 19% of the full sample belonged to the Education-based Justifiers subpopulation (French sample: 20%; Polish sample: 16%; Spanish sample: 22%; UK sample: 19%).We named this population Education-based Justifiers because education had the 3 The second relative deprivation item and third system justification item did not correlate significantly, r = 0.024, p = .564.strongest relation with system justification.However, note that for income and relative deprivation, the results contradict SJT.
Subpopulation 5: Education-based Rejectors.In the fifth subpopulation, the system justification items were strongly positively associated with education (r = 0.66, p < .001),rejecting SJT claims.It was negatively associated with income (r = -0.14, p item 2 & item 3 ≤ 0.006, p item 1 = 0.072) and positively with relative deprivation (∼r = 0.20, p < .001), 4 supporting SJT claims (see Figure 1 for a network plot; see SOM Table S23 for correlation matrices).That is, people with lower income and higher education, and who felt more deprived, were more likely to justify the status quo and vice versa.Especially, level of education revealed a strong association with system-justifying attitudes.
In this subpopulation, the relative deprivation items did not correlate strongly with each other, and some even showed negative correlations.
In total, 7% of the full sample belonged to the education-based rejector subpopulation (French sample: 7%; Polish sample: 7%; Spanish sample: 8%; UK sample: 7%).We named this population Education-based rejectors because education had the strongest relation with system justification; however, note that for income and relative deprivation the results support SJT.
10990992, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2989 by EVIDENCE AID -BELGIUM, Wiley Online Library on [05/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License p ≤ .078),supporting system justification claims.System justification was not significantly related to relative deprivation (∼r = -0.07,p ≥ .121)(see Figure 1 for a network plot; see SOM Table S24 for correlation matrices).That is, people with higher income were more likely to justify the status quo, while no clear pattern for the other variables emerged.In total, 5% of the full sample belonged to the income-based rejector subpopulation (French sample: 5%; Polish sample: 3%; Spanish sample: 5%; UK sample: 7%).We named this population Income-based Justifiers because income had the strongest relation with system justification; however, note that for education and relative deprivation the results contradict SJT.
In sum, the CCA revealed that individuals differ in how social status is associated with system-justification.About half of our sample was characterized by a pattern which either strongly supported (i.e., Justifiers) or strongly rejected SJT assumptions (i.e., Rejectors).The Justifiers subpopulation supports SJT's claim that system-justifying attitudes are stronger/also prevalent in disadvantaged subpopulations, whereas the Rejectors subpopulation 2 contradicts SJT's claim and finds that higher status individuals were more likely to endorse system-justifying attitudes.Across all subpopulations, we find that 55% of people respond in ways that are (partly) consistent with SJT's claims (subpopulation 1, 4, 5 and 6), whereas the remainder of the sample is inconsistent with the theory (subpopulation 2 and subpopulation 3).
Claim 2. The association between social status and system justification is stronger when inequality is salient.
To test Claim 2, we conducted multinomial logistic regression analysis.We predicted subpopulation membership (a categorical variable) using subjective inequality in everyday life (SIEL) and perceived inequality in resources (PIR).A description of class composition is displayed in Table 2 and all results are displayed in Table 3.We chose the Justifiers subpopulation as the reference group because this tells us which characteristics are associated with a higher likelihood to deviate from the group that best matched SJT Claim 1.The results show that when SIEL and PIR are higher, individuals were more likely to belong to the Rejectors, Education-based Justifiers, Education-based Rejectors or Income-based Rejectors subpopulation compared to the Justifiers subpopulation.SIEL and PIR did not predict a difference between the Justifiers subpopulation and the Alternative subpopulation.The results contradict SJT's prediction.When inequality was salient, people were less likely to belong to the Justifiers subpopulation.In other words, the Justifiers perceived the least amount of inequality.

Who supports SJT?
We extended the multinomial regression used to test Claim 2 by entering sociodemographic characteristics (e.g., gender) as predictors of subpopulation membership (see Tables 2 and 3).Out of the sociodemographic characteristics, education turned out to be the most important predictor of subpopulation membership as higher-educated people were more likely to belong to any other subpopulation than the Jus-tifiers subpopulation.Additionally, people who had lower income and were older were more likely to belong to the Rejectors subpopulation compared to the Justifiers subpopulation.People with higher income were more likely to belong to the Alternatives compared to the Justifiers.People with lower income and less relative deprivation were more likely to belong to Income-based Rejectors compared to the Justifiers.
People with higher income, less deprivation, and who were male were more likely to belong to Education-based Justifiers compared to the Justifiers, and people who were male and less deprived were more likely to belong to Education-based Rejectors compared to the Justifiers.

Additional analyses: Exploring the association of identification and status anxiety for justifiers and rejectors
Two subpopulations-Justifiers and Rejectors-in combination comprised more than half the sample, emerged in each country, and while the associations between social status variables were similar, the association between social status and system justification was opposing.
Therefore, we conducted additional analyses to explore in more detail who is more likely to belong to the Justifiers or the Rejectors.We estimated logistic regressions predicting subpopulation membership by identification with status dimensions (education, income), status anxiety and their interactions with social status (see Table S41 and S42 and Figure S1-S4).The results indicate that for lower-educated people the probability of belonging to the Rejectors subpopulation decreased with higher identification with education, whereas for higher-educated people the probability of belonging to the Rejectors subpopulation increased with identification.For income the results were in the opposite direction.That is for people with lower income (−1 SD) the probability of belonging in the Rejectors subpopulation increased with higher identification with income, whereas for people with higher income (+1 SD) the probability of belonging to the Rejectors subpopulation decreased with identification.For status anxiety the result for both education and income were similar.That is for lower-status people the probability of belonging to the Rejectors increased with status anxiety, whereas for higher-status people the probability decreased with status anxiety.This suggests that psychological variables of status anxiety and group identification are differently associated with SJT support and may be important moderators of the extent to which SJT claims are upheld or not.We discuss these findings in the general discussion.

The role of what and how people think for legitimizing the system
To test the consequences of system justification, we ran multiple regression analyses predicting violence against the government, trust in institutions and support of democracy by subpopulation controlling for mean level of system justification, income, relative deprivation and education with fixed effects for country (results are displayed in Tables 4-6).The aim was to see if how people think (i.e., subpopulation Most notably, there was a consistent difference between Justifiers and the Rejectors.The rejectors had significantly lower trust and lower satisfaction with democracy but were also less supportive of violence against the government than the Justifiers.These findings are not fully as expected, and we will discuss them in the general discussion section.

DISCUSSION
We tested two claims which are key predictions of SJT (Jost & Banaji, 1994).The first claim we tested was if disadvantaged groups expressed more system-justifying attitudes than advantaged groups.
Using typical aggregate approaches, we find inconsistent results, a set of findings we share with past investigations of these questions.However, using a bottom-up approach, we find that how we answer the question depends on the subpopulation focussed on.In the Justifiers subpopulation, we identified a subpopulation in which disadvantaged people expressed more system-justifying attitudes than advantaged people, consistent with SJT.We also found a Rejectors subpopulation in which disadvantaged people expressed fewer system-justifying attitudes than advantaged people, contrary to SJT.Four subpopulations were identified that provided a mix of evidence supportive of or opposed to SJT predictions depending on the precise measure of social status.In total, these findings show that SJT does receive support, but that support is limited to a specific subpopulation.At the same time, these findings show that SJT does not receive support, but that rejection is limited to a specific subpopulation.These different subpopulations were not detected when considering the complete sample using regression analysis.
The second claim we tested was whether the negative association between social status and system-justifying attitudes was more likely in contexts with more perceived inequality.Although we found differences between the subpopulations in terms of perceived inequality, these differences were not as anticipated by the theory.Instead of finding that the Justifiers subpopulation perceived more inequality, we found that the Justifiers subpopulation perceived the least amount of inequality of all of the subpopulations we identified.That is, although the Justifiers subpopulation is consistent with SJT predictions, the perceived levels of inequality of this subpopulation are contrary to SJT predictions.This may suggest that other theoretical explanations are necessary to explain why the Justifiers have a negative association between social status and system-justifying attitudes.

Justifiers and rejectors
Our bottom-up approach revealed that in about one quarter of the sample the SJT prediction that social status is negatively associated with system-justifying attitudes held (i.e., Justifiers) and that for about one third of the sample the SJT prediction was rejected (i.e., the Rejectors).For the remainder of the sample, we found a mixed pattern.Given that the majority of people belonged to the Rejectors or one of alternative subpopulations, this can also explain why-when considering the whole sample-the effects of social status on system justification are small and/or non-significant.The presence of these different subpopulations may also explain why the results of prior investigations have been so inconsistent.
These two main subpopulations-Justifiers and Rejectors-were also found in each country.Justifiers consist of disadvantaged individuals who justify the status quo, and advantaged people who do not.
Rejectors consist of advantaged individuals who justify the status quo, and disadvantaged people who do not.They differ in how social status indicators are associated with system justification; however, the associations between the social status indicators were similar.That is, higher-educated individuals perceived that it is easier to live with their income and felt less deprived in both subpopulations, yet, only in the Rejectors subpopulation did the advantaged individuals defend the status quo.The Rejectors were slightly higher-educated on average and more likely to perceive inequality.Potentially, the positive association between social status and system-justifying attitude results from the need to justify the inequality and their own high position within the society.However, it remains unclear what exactly determines whether a lower-or higher-educated individual becomes a Justifier or Rejector.
Exploratory analyses showed that the association between people's status and probability of belonging to the Justifiers and Rejectors subpopulation depended on identification with status dimensions and status anxiety.Consistent with the assumptions of the social identity model of system attitudes (SIMSA, Owuamalam et al., 2018), we found that individuals who strongly identified with education were more likely to belong to a subpopulation in which their status group justified the system (i.e., higher-educated in the Rejectors subpopulation and lower educated in the Justifiers subpopulation).However, for income the pattern was reversed.For status anxiety, the results showed that both higher-and lower-status individuals were more likely to bolster the status quo when they felt secure in their status.However, when individuals worried about their status, they were more likely to reject the status quo (i.e., lower status individuals were more likely to be in the Rejectors subpopulation and higher-status individuals in the Justifiers subpopulation).This could indicate that contextual factors which

CONCLUSION
We suspect that some people will read about our findings and think that SJT is vindicated because we find support for the theory.We suspect that others will read about these same findings and think that SJT is rebuked because we find that most people do not support the theory.However, we want to push back against this dichotomous thinking.
Rather, our approach here suggests that SJT principles may be useful descriptions of people, but they are not descriptions of a majority of people.This is both a good sign for the theory (it works for some people), but also speaks against the theory as a broad description of how people reason about their political systems.In this way, it is limited and circumscribed.Considering a theory, such as SJT, in this way requires that scientists engage in dialectical thinking and that they consider the issue at hand from multiple, sometimes contradicting perspectives.More practically, it means that a key theoretical task for people interested in system justification processes is to a priori identify the subsamples in which different processes occur.Our conclusions here show the theoretical importance of expanding our methodological toolkits.By expanding our toolkits, we are able to better understand the conditional and contextual nature of system justification processes.
sample consists of 5157 respondents living in France (n = 1201), Poland (n = 1529), Spain (n = 1215) and the UK (n = 1212).The Polish sample was collected by the Pollster Research Institute and the France, Spain and UK sample was collected by IPSOS.Quotas were applied based on age, gender, education and region.Quotas for age, gender and education were crossed with each other.Region was not crossed with other quotas.The distribution of these quota variables is based on Eurostat 2020 figures.The sample excludes respondents who failed more than one of the three attention checks.Data collection was part of a larger project which collected data in nine countries.The combination of variables used in this research, however, were only available in four countries.Therefore, we only included these respective countries.
items adapted from European Social Survey Round 9. Participants indicated the extent to which they trust (1) [Country]'s parliament, (2) [Country]'s politicians and (3) the legal system on a scale from 0 (No trust at all) to 10 (Complete trust).Internal consistency was satisfactory, Cronbach's α = .86(France = .86;Poland = .86;Spain = .83;UK = .87).Satisfaction with democracy.Satisfaction with democracy was measured with one item adapted from European Social Survey Round 9. Participants indicated the extent to which they are satisfied with the way democracy is working in [Country] on a scale from 0 (Very dissatisfied) to 10 (Very satisfied).
Note.A significant b-weight indicates the semi-partial correlation is also significant.b represents unstandardized regression weights.sr 2 represents the semipartial correlation squared.LL and UL indicate the lower and upper limits of a confidence interval, respectively.**Indicates p < .01.
10990992, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2989 by EVIDENCE AID -BELGIUM, Wiley Online Library on [05/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License status indicators and system justification items.It is therefore a way of identifying the people in our sample whose pattern of correlations between survey responses is consistent with the first claim of SJT and those whose pattern is not.
That is, within a cluster, individuals are similar in terms of what they think (e.g., graphic characteristics.This approach allows us to test if interrelations supporting SJT Claim 1 are more likely to occur when inequality is salient.Finally, to explore whether how people think (i.e., CCA class membership) predicts political outcomes above and beyond what people think (i.e., their absolute scores on the measured attitudes), we predict violence against the government, trust in institutions and satisfaction with democracy by system-justification and subpopulation.
10990992, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2989 by EVIDENCE AID -BELGIUM, Wiley Online Library on [05/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Regression results using economic system justification as the criterion TA B L E 1 10990992, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejsp.2989 by EVIDENCE AID -BELGIUM, Wiley Online Library on [05/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License