Subjective status and perceived legitimacy across countries

Abstract The relationships between subjective status and perceived legitimacy are important for understanding the extent to which people with low status are complicit in their oppression. We use novel data from 66 samples and 30 countries (N = 12,788) and find that people with higher status see the social system as more legitimate than those with lower status, but there is variation across people and countries. The association between subjective status and perceived legitimacy was never negative at any levels of eight moderator variables, although the positive association was sometimes reduced. Although not always consistent with hypotheses, group identification, self‐esteem, and beliefs in social mobility were all associated with perceived legitimacy among people who have low subjective status. These findings enrich our understanding of the relationship between social status and legitimacy.

behaviors, beliefs, and perceptions in order to address this threat, and when the feelings of threat are assuaged they are less likely to alter their behaviors, beliefs, and perceptions (for reviews see Jonas et al., 2014;Proulx, Inzlicht, & Harmon-Jones, 2012). Factors that make social hierarchy appear less consequential in material or psychological terms (we discuss several factors below) should reduce the effects of low subjective status on the anxious arousal that results from status-based dissonance. Therefore, this reasoning suggests that when people's feelings of threat are addressed, the status-legitimacy link will be positive (i.e., low status people will be less inclined to legitimize the system as a way to assuage feelings of threat). Conversely, when people's anxious arousal is exacerbated, the status-legitimacy link will be negative (i.e., low status people will see the system as more legitimate).

| Moderators that reduce threat
There are potentially many factors that can mitigate anxious arousal. We focus on factors that have appeared in the system justification and social identity literatures because these literatures have been the focus of debates on this issue (for a recent example see Jost, 2019;Owuamalam, Rubin, & Spears, 2019). The most obvious factor is the identification with and salience of a valued group, which is one way social identity theorists suggest people address anxious arousal (Hogg, 2014) or cope with the threat of having a low status (McMahon & Watts, 2002;Tajfel & Turner, 1979). Consistent with this, the original paper proposing the status-legitimacy hypothesis argued that if group interests are salient and accessible (something that is likely to be correlated with identification), then support for the status-legitimacy hypothesis is unlikely to emerge (Jost et al., , 2004. Similarly, people with high self-esteem and positive self-views make plans, have high levels of personal agency, and high levels of self-certainty (e.g., Campbell, 1990;Harter, 1978), suggesting that they are less likely to be psychologically affected by uncontrollability and similar types of threats associated with low status (see Laurin, Kay, & Landau, 2018 or Schoel, Bluemke, Mueller, & Stahlberg, 2011 for a similar argument in different domains). This is consistent with the idea that self-esteem should push against system justification motivations for low status people (cf. Jost, Gaucher, & Stern, 2015, p. 330). 2

| Structural moderators that increase threat through dissonance
Other factors can exacerbate feelings of anxious arousal by increasing the amount of dissonance people experience. The most direct prediction comes from assessing the contradictory cognitions that could increase dissonance. The dissonance that people experience from being in a low status position might emerge "from the contradictory cognitions that (a) the system is putting me (and my group) at a disadvantage, and (b) through our acquiescence, my group and I are contributing to the stability of the system" (Jost et al., 2003, p. 16). Therefore, people who recognize that they are not doing enough to mitigate inequality should experience more dissonance. This recognition should increase perceived legitimacy for people who are low status and is part of the contradictory cognitions originally predicted to cause lower status people to see the system as more legitimate than higher status people do.
The amount of civil liberties, meritocratic culture, and inequality within a country could also affect feelings of dissonance. Countries and contexts with more civil liberties and more meritocratic cultures increase the amount of dissonance people low in status experience by implying they have choice and control over their outcomes (Jost et al., 2003, p. 17). Inequality may increase the conflict between self/group motivations and acceptance of the system for low status group members (e.g., Henry & Saul,p. 376). It may also create the impression that people have less control over their place within society. Both factors should increase the amount of anxious uncertainty and increase the necessity of rationalizing the system. No support was found for these predictions in some prior studies (Brandt, 2013;Caricati, 2017;Caricati & Lorenzi-Cioldi, 2012;Trump & White, 2018; for one exception on one measure see Vargas-Salfate et al., 2018).

| Structural moderators that affect threat
Structural factors, including levels of inequality or the stability of the hierarchy, influence what types of options people perceive that they have and their place within the society. These perceptions may affect people's perceptions of the whole system. For example, to the extent that inequality exacerbates feelings of hierarchy and the threatening feelings of low status, as well as making status differences more salient, this should further motivate challenges to the inequality by the low status group (Tajfel & Turner, 1979). This is the opposite of the inequality prediction above. When looking at the stability of the hierarchy, one of the traditional predictions of social identity theory is that low status group members are less likely to seek social change when the status hierarchy is perceived as stable, compared to when the hierarchy is perceived as unstable (Ellemers, van Knippenberg, & Wilke, 1990;Tajfel & Turner, 1979). That is, when the status hierarchy is stable, there is little scope and hope for social change (i.e., no cognitive alternative to the status quo), and people are more likely to accept the system as legitimate. This hypothesis is also consistent with work finding that system stability increases a system justification motivation (Laurin, Gaucher, & Kay, 2013).
2 Social identity theory and system justification theory have also both used self-esteem as an outcome variable. Although this is interesting, it is not the focus of our investigation.

| Summary
We have identified several potential moderators that might help us predict when we are more or less likely to find support for the status-legitimacy hypothesis. These moderators are expressed as individual hypotheses in Table 1. In the cross-national study that follows, we test these eight hypotheses.

| Approach 2: Finding the predictors of perceived legitimacy for people with lower levels of subjective status
In addition to searching for moderators, we also ask what predicts perceived legitimacy for people with low levels of social status. Put another way, assuming that people with low levels of

Variable
Moderator hypothesis

Group Identification
Identification-Moderation Hypothesis: The link between status and legitimacy will be negative for people low in group identification and positive for people high in group identification.

Self-Esteem
Self-Esteem-Moderation Hypothesis: The link between status and legitimacy will be negative for people low in self-esteem and positive for people high in self-esteem.

Increase threat through dissonance
Inequality contribution Inequality Contribution-Moderation Hypothesis: The link between status and legitimacy will be negative for people who see themselves as contributing to inequality and positive for people who do not see themselves as contributing to inequality.

Civil Liberties
Civil Liberties Hypothesis: The link between status and legitimacy will be negative in countries with high levels of civil liberties and positive for people in countries with low levels of civil liberties.

Meritocracy
Meritocracy Hypothesis: The link between status and legitimacy will be negative in countries with meritocratic cultures and positive for people in countries with less meritocratic cultures.

Inequality
System Justification Theory (SJT) Inequality Hypothesis: The link between status and legitimacy will be negative in countries with high levels of inequality and positive for people in countries with low levels of inequality.

Structural Factors That Affect Threat
Inequality Social Identity Theory (SIT) Inequality Hypothesis: The link between status and legitimacy will be negative in countries with low levels of inequality and positive for people in countries with high levels of inequality.

Stability
Stability-Moderator Hypothesis: The link between status and legitimacy will be negative when people see the status hierarchy as stable and positive when people see the status hierarchy as unstable.
TA B L E 1 Summary of the moderator hypotheses tested in this investigation (Approach 1)

Variable Predictor hypothesis
Social mobility Social Mobility-Legitimacy Hypothesis: People with low status who see the system as having high social mobility will be more likely to see the system as legitimate.

Stability
Stability-Legitimacy Hypothesis: People with low status who see the status hierarchy as stable will be more likely to see the system as legitimate.

Identification
Identification-Legitimacy Hypothesis: People with low status who have high group identification will be less likely to see the system as legitimate.
Self-esteem Self-Esteem Legitimacy Hypothesis: People with low status who have high self-esteem will be less likely to see the system as legitimate.

TA B L E 2
Summary of the predictor hypotheses tested in this investigation (Approach 2) subjective status vary in the extent to which they see the system  Table 2).
Some work suggests that people desire to live in social systems in which there is upward mobility and the system is relatively stable (Laurin et al., 2013;Martorana, Galinsky, & Rao, 2005). Consistent with this, people who perceived no possibilities to move up to a higher status group are less likely to justify the system and more likely to engage in collective action (Day & Fiske, 2017;Ellemers et al., 1990;Mandisodza, Jost, & Unzueta, 2006;Tajfel, 1981;Wright, Taylor, & Moghaddam, 1990). Similarly, systems with more stable social hierarchies are seen as more legitimate (Laurin et al., 2013) and are less likely to trigger efforts to change the system (Bettencourt, Charlton, Dorr, & Hume, 2001;Ellemers et al., 1990;Tajfel & Turner, 1979). This suggests that when people see the system as providing social mobility and stability, they are more likely to see the system as legitimate.
When people's group-and self-interests are prioritized, it may be less likely that people perceive the system as legitimate if they or their group are not benefitting from that system. That is, higher levels of group identification and self-esteem may both be associated with lower levels of perceived system legitimacy among people with low social status. This follows from the idea that group-and self-interest motivations are negatively related to system-related motivations among low status groups (Jost et al., 2004). It is also consistent with the idea that the effects of group-and self-interests-when sufficiently strong-may be more prominent than the effects of system-interests (Jost et al., , 2004(Jost et al., , 2011 and with the finding that people with low levels of group identification may be more likely to accept the current situation (Rubin & Hewstone, 2004;Spears, Jetten, & Doosje, 2001).
According to these ideas, people with lower subjective status who have high group identification or who have high self-esteem will be less likely to see the system as legitimate.
From our reading of the literature, the predictions in the prior paragraph seem most consistent with a straightforward extension of social identity theory's work on group identification to research on perceived legitimacy. However, it is important to note that recent work by some scholars (and co-authors of this article) has predicted the opposite, at least as a function of an additional qualifier. A series of papers by Owuamalam and colleagues have argued that high group identifiers and people under conditions of high group salience may be likely to see the system as legitimate to the extent that it is seen as one that affords the group collective social mobility in the long term and allows them to eventually achieve social change as a group (e.g., due to longer-term status instability; Owuamalam, Rubin, & Spears, 2016, 2019Owuamalam, Rubin, Spears, & Weerabangsa, 2017; for a countervailing view see Jost, 2019). In short, using the system rather than rejecting it, can be seen as a viable vehicle for group interests, especially for high identifiers. This idea can also be tested with our data and essentially predicts the opposite of the predictions in the prior paragraph.

| Summary
We identified several potential predictors of perceived legitimacy for people with lower social status that might help us understand the reasons some people with low social status perceive the social system as legitimate. Some predictors indicate that the system is fulfilling the person's goals and other predictors indicate that personal and group goals are prioritized. These hypotheses are specified individually in Table 2. In the cross-national study that follows, we test these four hypotheses.

| An international, multi-lab approach
We conducted a cross-country study on the association between subjective status and perceived legitimacy that allowed us to test for moderators of the status-legitimacy relationships, as well as to understand what predicts perceived legitimacy for people with lower levels of subjective status. 3 We consider three types of effects: To assess moderators (Approach 1): 1. We test the interactions between status and the proposed moderator variables. If the interaction is significant, we test whether the effect of subjective status is negative (consistent with the status-legitimacy hypothesis) at the predicted levels of the moderator variables (e.g., when identification is low).
The assessment of predictors of perceived legitimacy among people with low status will take two steps (Approach 2): 2a. We examine the main effects of the predictors on perceived legitimacy. Because the hypotheses about predictors of perceived legitimacy are not specific to people with lower levels of subjective status (e.g., social mobility could predict perceived legitimacy for people with both high and low status), this main effect analysis tells us if there is an average effect across the sample.
2b. When there is a significant interaction between the predictor and status, we examine whether the predictor still has the predicted significant effect (e.g., a positive effect of social mobility) for people with lower levels of status. This will tell us whether the effect of the predictor is specific to people with low subjective status.

| Participants and procedure
Sixty-six distinct samples were collected by researchers from 30 countries. We aimed for 150 participants per sample, so that we had at least 150 participants per country to give us approximately 80% power to detect a small to medium effect (i.e., r = .22) within each sample. Data collection was not continued after analysis. To ensure that respondents were part of the social and political system, we only included participants who indicated they were either born in the country or had lived in the country for six or more years and so not every sample resulted in 150 participants. 4 All exclusions are reported.
Countries, samples, type of sample, proportion of men, mean age, and sample size are presented in Table 3. Samples included a mix of student samples, community samples, and samples with both students and members of the community (the latter two were both considered non-student samples for the sake of analyses). Samples are primarily samples of university students, although some include community samples from Mechanical Turk (USA2), community email lists (e.g., USA6), or a representative sample (NLD3). Multiple samples per country helps guard against the possibility that results for any particular country are dependent on one particular sample. Research was conducted in accordance with APA and national guidelines. Data, code, and materials are available at the following link: https://osf.io/5uxc7 /.
Participating labs used materials designed by the first three authors. These labs translated the materials and adjusted them when necessary for their language and cultural context (e.g., replacing "United States" in the system justification measure). Participants

| Key predictor variable: subjective status
To measure subjective social status, we used the MacArthur Scale of Subjective Social Status (Adler, Epel, Castellazzo, & Ickovics, 2000), modified to capture people's sense of status within their country.
We chose this measure to allow easier comparisons across countries.
Participants were asked to rate themselves on a ladder that ranged from 1 to 10, where 10 was high status and 1 was low status. The

| Key outcome variable: perceived legitimacy
We included four measures of perceived legitimacy and system justification. Multiple research groups studying the status-legitimacy hypothesis have used all these measures (e.g., Brandt, 2013;Henry & Saul, 2006;Li et al., 2020). The 8-item general system justification scale (Kay & Jost, 2003) includes items like "In general, I find society to be fair" and "Society is set up so that people usually get what they deserve" (M α = .78, SD α = .06, α range [.66, .85]; Scale M = −0.64; Scale SD = 1.14; −3 = Disagree strongly to +3 = Agree strongly). 5 4 Six or more years was chosen to ensure that participants who were international or exchange students were not included in the sample.
5 For each scale, alpha or the correlation coefficient (for 2-item scales) was calculated in each country separately. Mean, standard deviation, and range of alphas and correlation coefficients across countries are reported in the text.

| Self-esteem
We measured self-esteem with the validated single-item measure (Robins, Hendin, & Trzesniewski, 2001). This measure reads "I have high self-esteem" with answers that ranged from 0 = Not at all to 6 = Very true of me (Item M = 3.50; Item SD = 1.57).

| Social mobility
We measured perceptions of social mobility with six items. They

| Contribute to inequality
We created five items to measure whether people felt that they contributed to inequality and the stability of the system. One item, "I could do more to change differences in power and status between groups in society", did not correlate as expected with the other items and substantially reduced the reliability of the scale (α = .58), so we omitted it. The remaining four items read "I contribute to keeping society the way it is", "I contribute to maintaining the current social hierarchy", "I don't do anything to change the current differences in power and status in society", and "I am not trying to change the current differences in power and status in society". All items were measured on a scale ranging from −3 = Disagree strongly to 3 = Agree strongly and the four remaining items created a reliable scale (M α = .71, SD α = .09, α range [.46, .84]; Scale M = −0.14; Scale SD = 1.12).

| Stability of the status hierarchy
We used a 2-item measure of the perceived stability of the status hierarchy. This was based on measures used in studies from a social identity perspective (Mummendey et al., 1999). The items were prefaced with "Differences in power and status between groups in [country] are..." and then participants rated the stems "… difficult to change" and "…will remain stable over time" on a scale ranging from −3 = Disagree strongly to 3 = Agree strongly (M r = .31, SD r = .14, r range [−.06, .76]; Scale M = 0.82; Scale SD = 1.14). 7

| Societal level measures of moderators
We included three measures to assess societal level conditions.
To capture the amount of civil liberties, we used the 2015 Civil Liberties subscale of the Democracy index created by the Economist Intelligence Unit (2016). This index uses a combination of survey data and expert ratings to estimate how democratically free individual countries are. It has been used in prior work testing similar questions (Brandt, 2013).
To assess the extent to which the culture holds meritocratic val-

| Covariates
We included three covariates for analyses that focused on the individual level: self-reported age, gender (0 = women, 1 = men), and type of sample (0 = student, 1 = non-student) to adjust for potential background influences. For the models that include predictors at the societal level of analysis, we included two countrylevel covariates. We included the GDP per capita for each of the countries to control for overall wealth (World Bank, 2016). To control for broader regional trends (Kuppens & Pollet, 2014), we also included contrast codes for each of the continents represented in our data. 7 The negative correlation (−.06) between the stability items is found in Malaysia. The next lowest correlation is .18 in Canada.

| A note on coding
We coded all of the variables to range from 0 to 1. Multilevel regression coefficients are then the proportion difference in the outcome variable as one goes from 0 (minimum) to 1 (maximum) on the predictor variable. For example, a coefficient of .05 is a 5% difference in the outcome between people scoring the lowest and the highest on the predictor variable.

| Preliminary analyses: the effects of subjective status on perceived legitimacy
We use multilevel models to account for participants' nesting within countries and within samples/labs. Models were estimated using lme4 (Bates, Mächler, Bolker, & Walker, 2015) and lmerTest (Kuznetsova, Brockhoff, & Christensen, 2016)  First, we tested the main effect of subjective status on perceived legitimacy. This test conceptually replicates many prior tests of these hypotheses (e.g., Brandt, 2013;Henry & Saul, 2006;. We regressed perceived legitimacy on country-mean-centered subjective status. We also included country-mean-centered sense of power in this analysis and all other analyses because it was part of our original analysis plan (see Appendix S1). People with higher subjective status are more likely to see the social system as legitimate compared to people with lower subjective status (b = .152, SE = 0.005, 95% CI [0.142, 0.162]). There is a 15% difference in perceived legitimacy between people with the highest and people with the lowest levels of subjective status. This is in contrast to the prediction of the status-legitimacy hypothesis.
To test whether these results are impacted by covariates, we included age (country-mean-centered), gender (country-mean-centered), type of sample (grand mean-centered), and the type of perceived legitimacy measure (contrast coded) as covariates, including the interactions between these contrast codes and status and sense of power.

| Approach 1: Moderators of subjective status on perceived legitimacy
We tested individual-level moderators and country-level moderators. Individual level moderators were country-mean-centered and included in separate models as fixed effects in the four-level multilevel model used above (see Figure 3 for correlations between these measures). 11 We built the country-level models using the same four-level multilevel models, with the addition of the country-level moderators and country-level covariates. 12 All countrylevel variables were grand-mean-centered, with the exception of the contrast codes for continents. 13 In the country-level models, 8 We also conducted the analyses with only the system justification scale because this scale is perhaps the most well-developed of the four scales and it was reliable in all of the countries. Conclusions in the main text are essentially unchanged. There are two exceptions. The negative interaction between stability and status was negative and non-significant and the positive interaction between civil liberties and status was positive and non-significant when only looking at the system justification scale.
Importantly, the sizes of these coefficients were very similar to the coefficients in the models including all legitimacy measures, suggesting that the reduction in error variance when using additional data is the reason for the different conclusions when using all legitimacy measures or only the system justification scale. 9 We used the simr package in R (Green & MacLeod, 2016) to see how much power our model and sample size had to detect the effects of subjective status when b = .01, .02, .05, and .10. These analyses showed that we had adequate power to detect effects of at least .02 (i.e. a 2% difference in the outcome variable between the minimum and maximum of the predictor variable; power ≈ 49%, 97%, 100%, 100% respectively).
Descriptively, there is more variation between countries than there is between labs. 11 We used the simr package in R (Green & MacLeod, 2016) to see how much power our model and sample size had to detect individual-level interaction effects when b = .02, .05, .10, and .20. We used the model with the identification moderator as our base model for these analyses. These analyses showed that we had adequate power to detect moderation effects of at least .10 (i.e., a .10 difference in the unstandardized slope of status/power at the minimum and maximum of the moderator variable; power ≈12%, 46%, 98%, 100% respectively). 12 We used the simr package in R (Green & MacLeod, 2016) to see how much power our model and sample size had to detect country-level interaction effects for both subjective status and sense of power when b = .01, .02, .05, .10, and .20 We used the model with the meritocracy moderator as our base model for these analyses. These analyses showed that we had adequate power to detect country-level moderation effects of at least .20 (power ≈6%, 10%, 45%, 87% respectively). 13 To check the robustness of our results to outliers at the country-level (cf. Ullrich, & Schlüter, 2012), we visually inspected histograms of the country-level predictors. There were clear outliers for both the Gini index and the measure of civil liberties. These models were run both with and without outliers. All models are in the figures summarizing results and the primary models discussed in the text include all data.
we included random slopes for subjective status and sense of power.
To interpret the interactions, we used marginal effects plots created with interplot (Solt & Hu, 2015

| Moderators that reduce threat
We tested the hypotheses that are based on the assumption that group identification and self-esteem reduce the experience of threat. If the hypotheses are supported, we should find positive interactions between the moderators and status, such that at high levels of the moderator variables the link between status and perceived legitimacy is positive and at low levels of the moderator variables the link is negative.
The results for group identification and self-esteem are in Figures 4 and 5, respectively. One significant positive interaction emerged between identification and subjective status (Figure 4a).
The positive marginal effect of subjective status is weaker at low levels of identification and non-significant at the lowest levels of identification ( Figure 4b). It is stronger and significant at high levels.
This interaction is in the direction predicted by the hypothesis; however, the effect of subjective status is never significantly negative.
Because this effect is not negative, it is only partially consistent with the full prediction that perceived legitimacy will be higher for low status than for high status groups among lower identifiers.
Self-esteem, however, did not significantly interact with subjective status ( Figure 5). In short, although group identification showed F I G U R E 2 Estimated slopes of subjective status for each country from the multilevel model. Dots are randomly jittered horizontally to help show the data.
partial support for the hypotheses in the status domain, there was no support for the hypothesis for self-esteem.

| Moderators that increase threat through dissonance
Inequality contribution, civil liberties, meritocracy, and inequality were all expected to increase the experience of threat through dissonance. If the hypotheses are supported, we should find negative interactions between the moderators and status, such that at high levels of the moderator variables the link between status and perceived legitimacy is negative and at low levels of the moderator variables the link is positive.
There were no significant negative interactions between any of the four moderator variables expected to increase the experience of threat (Figures 6-9). In the case of civil liberties, there was a positive interaction between civil liberties and subjective status. When probing the interaction, we found that the effect F I G U R E 3 Correlations (above the diagonal) and density plots (below the diagonal) between primary predictor variables and the individual level variables. All variables were country-mean centered before calculating the correlations. Density plots for each individual variable are on the diagonal. ID = Group identification; SelfEst = Self-esteem; SocMob = Social mobility; ConInqu = Contribution to inequality; Stable = Stability of the status hierarchy [Colour figure can be viewed at wileyonlinelibrary.com] of subjective status (Figure 7b) was positive and stronger at high levels of civil liberties. At lower levels of civil liberties, the effect of subjective status was still positive, but weaker and non-significant. This interaction is in the opposite direction predicted by the hypothesis. In short, in no cases did we find support for the hypotheses about moderators that could increase the experience of threat.

| Structural factors as moderators
The hypotheses suggested that inequality and perceived stability of the social system were structural factors that might moderate the association between status and perceived legitimacy. In contrast to the hypothesis tested in the prior paragraph, one hypothesis is that countries with high levels of inequality will have stronger, positive associations between status and perceived legitimacy because it makes status differences more salient. We did not find evidence for this ( Figure 9).
For stability, the hypothesis predicted that the link between status and perceived legitimacy would be negative when perceived stability was high compared to low. This hypothesis would be supported with a negative interaction effect. This effect did not emerge (Figure 10a). Instead, we found the opposite interaction effect (Figure 10b). Contrary to the hypothesis, the link between status and perceived legitimacy is strong and positive when stability is high and it is weaker or non-significant when stability is low.
In short, we do not find evidence that inequality may exacerbate status-based conflict. We also find that stability is a significant moderator, but in the opposite direction from that expected. It indicates that stability is interpreted differently in the context of real-life social inequality (see also Verkuyten & Reijerse, 2008), compared to stability that is manipulated in groups created in the lab (e.g., Ellemers et al., 1990). A possible interpretation of the pattern found in this study could be that the perception that social inequality will not change might be an additional reason for judging that inequality is unacceptable (assuming people perceive there to be inequality). At the same time, we note that the reliability of our stability measure could be improved.

| Approach 2: Predictors of perceived legitimacy for people with low levels of subjective status
We tested the hypotheses about the predictors of perceived legitimacy for people with low levels of subjective status using the same models described and presented above. To assess whether there was support for the hypotheses, we first examined the main effects of the predictors on perceived legitimacy. This main effect analysis tells us whether there is an average effect across the sample. When there is a significant interaction between the predictor and status, we examine whether the predictor still has the expected effect (e.g., a positive effect of social mobility) for people with lower levels of status (regardless of the pattern for those higher in status). This will tell us whether the effect of the predictor is specific to people with low social status or sense of power.
The hypotheses predicted that people would see the system as more legitimate when it fulfilled a social mobility or stability related goal. For social mobility, there was clear support for the hypothesis.
There was a large positive effect of social mobility, such that higher levels of perceived social mobility were associated with higher levels of perceived legitimacy (Figure 11a). This positive effect was moderated by subjective status (Figure 11a), suggesting that the positive effect is variable across levels of subjective status. At low levels of subjective status (Figure 11b), perceived social mobility remained a significant positive predictor of perceived legitimacy. This is consistent with the hypothesis.
For stability, results were not consistent with the hypothesis.
There was a negative effect of stability on perceived legitimacy ( Figure 10a). This negative effect was moderated by subjective status (Figure 10a); however, at low levels of subjective status ( Figure 10c) the effect of stability remained negative and significant.
In short, when people see social mobility as a possibility, people with low subjective status see the system as more legitimate; however, perceiving the system as more stable is associated with less legitimacy at lower levels of status, which is inconsistent with the hypothesis.
We also made the prediction that when personal or group interests are fulfilled, perceived legitimacy of the system is less necessary. The hypotheses predicted that identification and self-esteem
However, this is not our primary contribution. We sought to advance this debate beyond straightforward main effects by taking two approaches.

| Approach 1: Moderator of the association between subjective status and perceived legitimacy
Approach 1 examined how constructs that reduce threat, constructs that increase threat, and perceived structural factors may be moderators of the association between status and perceived legitimacy. We tested eight specific moderator hypotheses and found partial support for one: the identification-moderation hypothesis (see Table 4).
That is, the most common result was no clear support for the hypotheses. Sometimes the lack of support was due to non-significant interactions. The sensitivity analyses reported in footnotes 9, 11, and 12 indicate that we have substantial power to detect main effects and interactions at the individual level. For country-level interaction effects, we only had sufficient power to detect large effects. In the cases of null results, studies with even greater statistical power may find evidence in support of these hypotheses, in opposition to these hypotheses, or for a null effect (time will tell). In many other cases, the lack of support was due to a significant interaction that was the opposite of the prediction (e.g., stability-moderator hypothesis). In these cases (highlighted with ✗ op in

| Approach 2: Predictors of perceived legitimacy for people with lower levels of subjective status
Approach 2 examined the predictors of perceived legitimacy for people with low subjective status. We tested four specific hypotheses about these predictors and found support for one: the social mobility-legitimacy hypothesis (see Table 5). Most of the predictions were not confirmed and, in most cases, results were opposite to the predicted direction. As one example, the identification-legitimacy hypothesis predicted a negative association for members of low status/power groups. Yet, we found the opposite. Some members of our team have started to think through this type of effect (and have published some findings consistent with the data in this manuscript). As explained earlier, they have proposed that among people with low subjective status, identification will be positively associated with system legitimacy, especially when the system can be used as a vehicle for improvements to group status in the long term (Owuamalam, Rubin, & Spears, 2018). Our data lendsome credence to this hypothesis.
Although almost never in a direction supportive of the hypotheses (see Table 5), we find that group identification, self-esteem, and beliefs in social mobility are all associated with greater perceived legitimacy among people with low subjective status. In this way, we helped fulfill Jost's (2017) call to better understand what leads people who are oppressed to uphold the system.

| Strengths, limitations, and future directions
We used a design that leverages larger samples from multiple countries to understand how subjective status is associated with perceived legitimacy and related constructs. This helps us avoid low statistical power (Sedlmeier & Gigerenzer, 1989) and uses less-WEIRD sampling (Henrich, Heine, & Norenzayan, 2010). That said, hardly any of the samples were representative, and we were not able to recruit participants from all the regions of the globe. Most notably, we were unable to secure data from the continent of Africa, and Asia is underrepresented. However, given the consistency of our results with those of researchers who have used data from even more diverse and representative samples (Brandt, 2013;Caricati, 2017;Caricati & Lorenzi-Cioldi, 2012;Vargas-Salfate et al., 2018), we believe that our data provide a reasonable approximation at this time for the regions that we do cover. A related limitation is that although we report the range of Cronbach's α and r for all scales across countries, we have not formally tested measurement invariance (He & van de Vijver, 2012). It is encouraging that some of the scales we use have been shown to be invariant across at least some countries (e.g., Davidov & Coromina, 2013). Clearly, more work is needed on the social psychology of perceived legitimacy and its development in understudied regions; this future work will surely have both practical and theoretical benefits.
Many of our samples are student samples. This allowed us to use our resources efficiently, but necessarily implies that the samples are younger and better educated than the general population.
Although studies on similar topics (and social psychology more broadly) have relied on student and non-representative samples, it is always possible that samples with more objectively and subjec- Note: ✓ = indicates support and partial support for the hypothesis, ✗ = indicates no support for the hypothesis. ✗ op = indicates a significant interaction effect in the opposite direction.

TA B L E 5
Summary of predictor hypotheses (Approach 2) and whether they were supported

Predictor hypothesis
Social Mobility-Legitimacy Hypothesis ✓

Self-esteem legitimacy hypothesis ✗ op
Note: ✓ = indicates support and partial support for the hypothesis, ✗ = indicates no support for the hypothesis. ✗ op = indications a significant interaction effect in the opposite direction.
representative samples (e.g., Brandt, 2013), suggesting that our findings are consistent with data using other sampling techniques.

| Where to now?
Our project provides scholars with additional evidence that they can use to inform the direction of research on status and legitimacy. 14 This additional evidence comes in two forms. First, these hypotheses were inspired by both social identity and system justification theories and primarily made inaccurate predictions in our study.
This may be because the prediction was wrong, or because some auxiliary assumptions did not hold in our particular samples and measures. For example, nearly every moderator predicted to reduce feelings of threat or increase feelings of threat through dissonance were not supported. This may mean that these moderators are not associated with threat as expected (auxiliary assumption did not hold), or that threat is not a key mechanism linking status to legitimacy (theoretical prediction was wrong). Learning from theoretical failures and inaccurate predictions like these can often be more informative than learning from theoretical successes (Ferguson & Heene, 2012;Popper, 1959).
It is also plausible that some of the inaccurate predictions (and perhaps most of them) can be addressed by increasing the specificity of the theories and how theoretical concepts are translated into concrete operationalizations. For example, social identity theory has traditionally treated legitimacy as a moderator and used outcome measures such as group evaluations, allocation, and social change strategies. Here, we tried to extend the theory's predictions to legitimacy, but did not find consistent support. This may suggest that social identity theory's predictions about social change strategies do not extend to legitimacy in a straightforward way, which may also help to explain the limited support for predictions derived from social identity theory. Recent work has therefore investigated how a social identity perspective can be better applied to predict system legitimacy (Owuamalam et al., 2019).
Second, our data provide descriptive information about the precisely estimated associations between a large number of socially and politically relevant variables. That is, not only did our study fail to confirm a number of hypotheses about status and legitimacy, but it also provides the relevant associations and non-associations that relevant theories need to explain. For example, despite expectations of both perspectives, a negative association between status and legitimacy was not found at low levels of identification. Theories need to account for this pattern of results. Similarly, perceived stability was associated with lower levels of legitimacy, despite expectations. Updates will need to account for this pattern of results. Although post-hoc explanations are possible, we hope that theorists can build clearly specified models that can be used to explain the current data and to make new, falsifiable predictions (e.g., Muthukrishna & Henrich, 2019).

| Conclusion
We started this project because we noted that there is evidence both consistent and inconsistent with the status-legitimacy hypothesis. Our study found very little support for moderator hypotheses across 30 countries and several moderator variables. To the extent that the phenomenon predicted by the status-legitimacy hypothesis exists, this appears to be quite rare. That does not mean that people with low subjective status never see the system as relatively legitimate. Our data also shed light on when this is most likely to be the case. We find that group identification, self-esteem, and beliefs in social mobility are all associated with perceived legitimacy among people who are low status (as well as those with high status). We hope that our findings inspire scholars interested in this domain to pursue questions related to status and legitimacy to enrich our understanding of these constructs around the world.

ACK N OWLED G EM ENTS
All authors provided feedback on the drafts of the manuscript and

CO N FLI C T S O F I NTE R E S T
There are no conflicts of interest to report.