Discrimination and Political Engagement: A Cross-national Test

Abstract

What is the effect of personal discrimination on the political engagement of ethnic and racial minorities? Existing research theorizes increased engagement, but evidence is mixed. The discrimination and political engagement link is tested across six countries: Canada, Denmark, France, Germany, the United Kingdom, and the United States. Interest in politics and political actions (e.g. protest and donations) show constant relationships: people who have experienced discrimination have more interest in politics and take more political actions. There is no clear evidence of different effects of political vs social discrimination. However, the link between turnout and discrimination varies systematically across countries: a positive correlation in three separate American datasets, but mixed and null in other countries. This may be the result of the distinctive American conflict over voting rights for racial minorities. The conclusion discusses priorities for future research, including a focus on establishing causal relationships and testing mechanisms.

Discrimination against racial and ethnic minorities remains common in Western countries and has changed little in recent years (Lippens, Vermeiren, and Baert 2023; Quillian et al. 2017) despite some evidence that attitudes have become more positive (Besco 2021; Dennison and Geddes 2019; Hopkins and Washington 2020). Moreover, political and social conflict over immigration and racial issues seems to be growing, and norms excluding politicians who are explicitly racist and anti-immigrants are eroding (Koning 2023; Tesler 2020; Valentino, Neuner, and Vandenbroek 2018). While attitudes about race, immigration, and diversity are generally viewed as affecting party politics (Alonso and Fonseca 2012; Carmines and Stimson 1982; Golder 2016; Reny, Collingwood, and Valenzuela 2019), the reverse might also be true, as electoral politics such as the Brexit Referendum and election of Donald Trump (Forster 2016; Lichtblau 2016) led to increased attacks and discrimination. While much effort is put in to understand the behavior and motives of racist voters and politicians, the impact and response of minorities is also crucial.

What is the effect of experiencing discrimination on the political engagement of racial minorities? We might hope that the response would be contestation and mobilization, rather than alienation or withdrawal. This is not guaranteed, given that some research suggests discrimination results in a lower sense of efficacy (Broman, Mavaddat, and Hsu 2000) and disengagement from public visibility (Hobbs and Lajevardi 2019). Nonetheless, a substantial amount of research in political science, including theories of linked fate (Dawson 1994), group consciousness (Miller et al. 1981), and policy threat (Campbell 2003), suggest that the experience of discrimination will increase political engagement, mobilization, and participation.

Research examining the relationship between discrimination and political participation generally emphasizes that experience of discrimination makes people more likely to vote and be generally politically engaged. However, there have been many conflicting results, depending on the group (Lien 1994), source of discrimination (Bilodeau 2017; Oskooii 2016), or type of engagement (Lien 1994; Martin 2017; Mattila and Papageorgiou 2017; Schildkraut 2005).

This study aims to test the link between discrimination and political participation across multiple contexts and data sources. To do so, it uses survey data from six countries: Canada, Denmark, France, Germany, the United Kingdom, and the United States. The relationship between the experience of discrimination and three political engagement measures is examined: interest in politics, turnout, and political actions (protesting, donations, petitions, etc). The effects of political and social discrimination are also tested separately. While the available measures and cross-sectional data produce some limitations, the strength of the analysis is the breadth of the test across multiple countries.

The results show clear and straightforward results for political interest and political actions. Across all countries, racial minorities who experience discrimination have higher political interest. Moreover, they report taking more political actions, such as protesting, participating in boycotts, donating to parties and politicians, and signing petitions. Conversely, the results for voter turnout vary substantially across countries, with mostly null or even negative correlations. The exception is in the United States, where three separate datasets show positive and statistically significant correlations: respondents with higher reported discrimination also report higher turnout. These turnout results point to distinctive dynamics in American politics and may be related to longstanding political conflict and mobilization around voting rights for racial minorities. Finally, separating political and social discrimination is somewhat inconsistent, but generally shows similar (positive) rather than different relationships with political engagement. A range of alternative specifications and robustness checks produce similar results for all these findings. We conclude with a discussion about future directions for this research area, emphasizing the need for improvement in behavioral measure and causal designs, and testing of theoretical explanations by examining mediators such as perceived responsibility and efficacy.

Discrimination and Political Participation

Research on the effects of discrimination or attacks on minority groups often finds that experiencing discrimination increases political participation. Lien (1994) found that experiencing discrimination was positively correlated with both voting and non-voting political engagement for Mexican Americans, and Wrinkle et al. (1996) and Valdez (2011) show similar results for non-electoral political engagement among Hispanic Americans. Examining the effect of discrimination on voting and attitudinal engagement, Schildkraut (2005) shows that individual-level discrimination is related to higher levels of behavioral participation (voting and voter registration). Heath et al. (2013) report a similar finding in the U.K.: personally experiencing discrimination is correlated with political action (protest, signing petitions, joining boycotts), and Martin (2017) similarly finds that perceptions of Islamophobia are correlated with protest among Muslims in the UK. Bilodeau (2017) and Bilodeau et al. (2023) find that in Canada perceived discrimination is correlated with more protest and other forms of non-electoral participation. Perceived local discrimination increased political participation among Israeli-Arabs (Beeri and Saad 2014), and among immigrants to Germany, interest in politics is higher for those with perceived discrimination (Fischer-Neumann 2014).

Importantly, there is emerging research which demonstrates these effects are causal. White (2016) uses the staggered implementation of an immigration enforcement policy and a difference-in-difference method to show this increases registration among Hispanic Americans. Weiss, Siegel, and Romney (2023), also using difference-in-difference analysis, find that threat of exclusionary citizenship rules mobilized participation among Palestinians. Besco et al. (2022) use a survey experiment to show that racist campaign rhetoric increases turnout intentions among Hispanic Americans. Given the potential for endogeneity in cross-sectional measurements of self-reported discrimination and political engagement, the results from these causal designs provide some reassurance.

The theoretical basis for the effect of discrimination on political engagement supports both generalized and specific instances of discrimination, oppression, or threat. In generalized mechanisms, theories such as linked fate focus on how broadly defined oppression can politicize group identities and create a belief that the success or failure of minority citizens is connected to that of their fellow minority group members, which drives increased participation in politics (Chong and Rogers 2005; Dawson 1994; McClain et al. 2009; Miller et al. 1981). Importantly, Lu and Jones (2019) showed that the link between experience of discrimination and perceptions of linked fate holds for a range of different minority groups, but for White people these relationships might be quite different (Marsh and Ramirez 2019). Grewal and Hamid (2022) show the converse effect: that Muslims facing discrimination are less supportive of democracy and the “system” if they feel they face this discrimination alone—e.g. without group consciousness or group discrimination.

Specific situations of discrimination can also matter: the policy threat literature theorizes that when people are threatened by policy changes their political participation tends to increase (Barreto et al. 2008; Campbell 2003; Gutierrez et al. 2019). This includes new immigration enforcement programs mobilizing Latino Americans (White 2016) and the Patriot Act mobilizing (some) Arab Americans and (Cho, Gimpel and Wu 2006) and Hispanic Americans. Weiss, Siegel, and Romney (2023) and Besco et al. (2022) also test similar short-run effects of specific threats and find increases in political mobilization and engagement. Naturally, these general and specific effects can also intersect. For example, Pantoja, Ramirez, and Segura (2001) found that Latino immigrants socialized in a threatening anti-immigration context had high rates of voter registration and turnout, and Latinos with higher linked fate have stronger responses to threats like Donald Trump (Gutierrez et al. 2019).

Examining the circumstances by which discrimination might, or might not, affect political engagement, Page (2018) leverages the enlargement of the EU and argues that discrimination increases the participation of sexual minorities only if there is social and political support. Oskooii (2020, 2016) theorizes that the type and source of discrimination: political discrimination (e.g. laws or by government officials) tends to increase political participation. However, societal discrimination (e.g. by individuals on the street or private businesses) does not.

However, dispite the above findings, there are multiple studies find conflicting and mixed results. Lien (1994) found that experiencing discrimination was positively correlated with both voting and non-voting political engagement for Mexican-Americans, but not Asian-Americans. Martin (2017) found that perceived Islamophobia increased non-electoral participation, yet decreased rates of voting. Bilodeau (2017) also shows that discrimination is correlated with lower turnout among racial minorities in Canada, but increase in other forms of political activities. Besco et al. (2022) find effects on political interest and turnout intention, but not on other political actions. Other types of groups also produce similarly mixed results: Mattila and Papageorgiou (2017) show that perceived discrimination due to disability decreases voting but increases other political actions (protesting and contacting politicians). While Oskooii (2020, 2016) argues that political discrimination increased participation and social discrimination does not, Bilodeau et al. (2023) find political and social discrimination have quite similar relationship with mobilization. Given these mixed results, this study aims to test these relationships across a range of datasets and countries.

Analysis and Results

This analysis is a multi-country test of the relationship between discrimination against racial and ethnic minorities and the political engagement of members of these groups. The purpose is to see if the same relationship holds in multiple contexts, replicating the analysis in countries that are established democracies but which have substantially different political systems, ethnic groups, and economic contexts. The major constraints were data availability: sufficient sample size of racial minorities, and questions about discrimination. While many surveys ask about discriminatory attitudes, surprisingly few ask about the experience of discrimination. Nonetheless, suitable surveys were found in Canada, Denmark, France, Germany, the United Kingdom, and the United States. These surveys all include multiple different ethnic and racial groups. Table 1 provides an overview, with additional details in the appendix in the Supplementary material.¹

Table 1. Surveys

The population of interest is racial and ethnic minorities. This category is sometimes defined in different ways, but in this research the definition used is non-White, of non-European ancestry. This is because the effects of experiencing discrimination for White people might be quite different (Marsh and Ramirez 2019). It might be reasonable to include other (White) European national groups since they plausibly experience discrimination in a similar way to ethnic minorities, but this causes significant difficulties for comparisons to the United States and Canada, so these respondents are not included here. In the Canadian, Danish, American, and British surveys, there are questions specifically about the race or ethnicity of the respondents. In the French and German surveys, a variable is constructed using questions about the country of origin of the respondent and their parents.

The key independent variable is personal experience of discrimination. Most of the surveys ask a version of the question “have you personally experienced discrimination due to your race or culture” (see Appendix in the Supplementary material for question-wording details). Some of the surveys ask a series of question about discrimination in various places: on the street, at the office, visiting the doctor, visiting authorities, by the police, in applying for apartments, etc. These are combined into an additive scale for a general discrimination measure so that respondents reporting discrimination in multiple places score higher. These questions also enable the construction of separate political and social discrimination variables (see Appendix in the Supplementary material for coding details).

For the dependant variables, three measures of political engagement are used: interest in politics, turnout, and political actions (protests, donations to parties, etc.). The research discussed above suggests that all three measures will be impacted in a similar way by personal experience of discrimination. While interest in politics is not precisely a measure of political participation, it can be viewed as a measure of psychological engagement with politics, and therefore another useful measure of our key concept. It is also a dependant variable in many of the studies cited above. Notably, interest in politics is not subject to regulations which vary across countries in the way that voting or some other political actions are, and so in that way, it is especially useful measure for cross-national tests. More information on this measure is provided in the sections below, and details of the wording and scales are available in the Appendix, Supplementary material.

Not all surveys include all questions, and there is some variation in the measures available, such as questions about experiencing discrimination in general vs in specific locations, or in the exact political actions asked about. To address this there is a series of tests, including comparing analysis with the same and different questions. These produce very similar results, suggesting that differences in question-wording are not a significant issue, and they capture the same latent variables. This is discussed at greater length in the section on robustness checks below. The demographic correlates are mostly consistent across countries and measures, with education positively correlated and immigrant status and age (except in the USA) negatively correlated with the discrimination measures. Other demographics (gender and income) don’t have a consistent relationship with discrimination. Crucially, there are no systematic differences in correlates related to the type of question. In the Canadian and UK datasets, which have both the general and specific measures of discrimination, the correlates of the two measures are the same, again suggesting they are tapping the same factor. Nonetheless, the specific features of different types of variables are likely to be different, such as range and precision. For example, a scale of questions asking about discrimination is likely to be more precise, but possibly an underestimate relative to a general question, since it can’t ask about all situations. For that reason, the analysis here is on the direction and statistical significance of the estimates, rather than on comparing the precise size of the coefficients.

Control variables include age, education, income, gender, and immigrant status. Note that the main models do not use controls for variables such as partisanship or other political attitudes, for several reasons: they are post-treatment, plausible mediators, alternative measure of the DV (interest), and tend to vary in availability across countries and surveys. Nonetheless, alternative models with additional controls are estimated and discussed in the section on robustness checks below. To make interpretation easier, all variables are scaled to a 0-1 interval except age and political actions, the latter being a count of actions.

Interest in Politics

The first set of models examines the relationship between discrimination and political interest. The dependant variable is interest in politics and the independent variable is personal experience of discrimination. Models are estimated both with and without controls: age, income, gender, and immigrant status, which should contribute to avoiding results differing due to variation in composition of the across countries.

The results in Figure 1 show quite consistent results: in all countries, discrimination is positively correlated with interest in politics. That is, racial minority citizens who say they have personally experienced discrimination tend to be more interested in politics. The magnitude of the relationship for Canada, France, and the United States is very similar. The coefficients for Germany and Denmark are substantially larger. The reason for this isn’t clear—but notably these have the smallest sample size and the largest confidence intervals. The addition of controls makes little difference in the size or statistical significance of the effects.

Figure 1. Discrimination and interest in politics.

Note: Figure shows OLS coefficients from separate models.

Turnout

The second set of models examines the relationship between discrimination and voter turnout. The turnout variables are a scale composed of questions about voting at different elections as available: (national, local, first and second round presidential, etc.—see Appendix in the Supplementary material for specific questions). This provides a more precise measure than a single question, and also helps increase variation—an issue in that self-reported turnout is sometimes very high (see p.12 for analysis of separate items). Only citizens are included. Unfortunately, the German survey does not include a question about voting. Otherwise, the models are similar to the previous set.

In contrast to the political interest models, results for the turnout models in Figure 2 are quite inconsistent. While the discrimination variable is statistically significant in many of the models, in others it is not. In Denmark, discrimination is positively correlated with turnout, while in Canada, France, and the UK it is negatively correlated or non-significant, depending on the model. The clear exception is the United States, where three separate datasets, with and without controls, all produce positive and statistically significant relationship between discrimination and turnout.

Figure 2. Discrimination and turnout.

Note: Figure shows OLS coefficients from separate models.

Next, we turn to the relationship between discrimination and other political actions. The political action variables are scales composed of questions about substantive political actions, excluding voting. This includes participating in protests, petitions, boycotts, donating money to political causes, volunteering in politics or community affairs, and contacting officials, with some variations between surveys (see Appendix in the Supplementary material for details). Many of the surveys do not include detailed questions about political actions, and so models are only estimated for the Canadian, U.K., and U.S. surveys in 2016 and 2004. Since the political action variables are over-dispersed count variables (most people do not take any political actions, a few take a lot), the models use negative binomial regression, with predicted values of the number of actions generated for the figure.

As Figure 3 shows, the correlation of discrimination and political action is statistically significant and positive for all countries: Canada, the U.K., and the United States. The size of the effects ranges from an increase of roughly .5 actions in Canada and the U.K., and about .75 and 1 in the United States. ‘This means, for example, that in the 2016 U.S. survey, the respondents in the highest category of discrimination on average take about 1 more action than those in the lowest category. Given that very few citizens take any political actions at all (as measured by these questions), this is a very substantial difference with important political implications.

Figure 3. Discrimination and political actions.

Note: Figure shows predicted values generated from negative binomial models, showing the difference in number of actions for respondents at the top vs bottom of the discrimination scale.

Government and Social Discrimination

Finally, we examine the distinct role of social and political discrimination. Previous research suggested that discrimination by the government leads to increased political engagement, since this is a way to hold those responsible to account, and leads to change. Conversely, social discrimination by ordinary people does not lead to increased engagement because it is not clearly linked to politics. Political discrimination is usually defined broadly as state actions, including discrimination by state actors or commonly regulated areas, such as immigration officials, employment, or housing. Conversely, social discrimination includes situations such as discrimination by private businesses or on the street.

To test this, models with the three dependant variables are estimated: interest in politics, turnout, and political actions. Rather than a single general discrimination variable, separate measure of social and political discrimination are both included in each model. Following (Oskooii 2020), political discrimination is coded as reported discrimination by police, border guards, schools, work, housing, and social discrimination is by private businesses, in the street, or by friends and neighbors. See Appendix in the Supplementary material for question-wording and coding details. Some measure are not available in all datasets, so models are estimated where available. Otherwise, the models are identical to those above. Given the number of estimates, for clarity only models with controls are shown, but bivariate models are available in the Appendix, Supplementary material, and show similar overall results.

The results in Figure 4 show little evidence for effects from political but not social discrimination or for larger effects from political discrimination. The turnout results are mostly non-significant and of different signs, which is consistent with the mixed previous results in Figure 3. The interest results do show Canada and the UK with larger political discrimination than social discrimination for interest, but the other countries and measures do not show a clear pattern.

Figure 4. Political and social discrimination.

Note: Figure shows OLS and negative binomial (for political actions) coefficients from separate models with controls.

Robustness Checks and Additional Analysis

To test the robustness of these results, a number of alternative specifications are estimated. The turnout models were estimated with separate turnout variables from first-order elections and lower-order elections (national Parliament or presidential vs provincial or municipal, etc.) rather than a scale of all available turnout questions. The results are very similar, with no systematic differences between different types of elections, or relative to the scale used in the main analysis (Appendix Figure A1 in the Supplementary material). Turnout models with additional controls are also estimated where available, including interest, trust, efficacy, partisanship, and specific ethnic groups (Appendix Figure A2 in the Supplementary material). These also produce quite similar results as the models in the main text. More generally, the turnout variables in all countries are related as expected with other variables, such as positive correlations with age, education, and interest, suggesting there is nothing particularly unusual about the specific questions, coding, or data.

Separate models were estimated for racial and ethnic groups, where measures were available and sufficient sample sized permitted (Black-British, Asian-Americans, Arab-French, etc). For interest in politics and political actions, the results are positive and significant correlations across the board: essentially the same relationship for all groups, and similar to the main results (see Appendix Table A21, A22, and A25 in the Supplementary material). For turnout, other countries are mixed and null, as in the main results, including a number of the American groups (see Appendix Table A23 and A24 in the Supplementary material). Interestingly, the discrimination coefficients for Asian-Americans are the largest and most commonly statistically significant. Given modest sample sizes and the many models estimated it isn’t clear how robust this is, but may be worth future investigation.

Models using separate political actions (rather than the scale) were also estimated where data availability permits, including both those directed at conventional electoral-related actions (petitions, donations, contacting officials) and unconventional non-electoral actions (boycotts, protests). Nearly all are positively correlated with discrimination, statistically significant, and with coefficients of roughly comparable size. See Appendix Tables A12, A13, and A14 in the Supplementary material. Similarly, models with separate political actions and separate government and social discrimination measures show that both social and political discrimination variables are positive and statistically significant, with neither being systematically larger (Appendix Tables A15 and A16 in the Supplementary material). These are similar to Bilodeau et al. (2023)’s results for protest, but for a range of different political actions. Models are also estimated for social and political discrimination with additional controls (interest, trust, efficacy, partisanship, and specific ethnic groups (Appendix Figure A3 in the Supplementary material). These are also mixed: the political discrimination coefficients are larger for social discrimination coefficients for interest and turnout in the UK, but not in Canada and not for political actions in either country. The measures of social and political discrimination are correlated in general, but VIF scores are quite low (2 or lower), suggesting multicollinearity is not a serious issue. Finally, the political interest analysis above includes non-citizens (since unlike voting citizenship is not a requirement for interest in politics), but excluding non-citizens produces similar results (see Appendix Figure A4 in the Supplementary material).

Discussion and Conclusions

The ongoing discrimination experienced by racial and ethnic minorities across Western countries is both a product of politics and has important political effects. In light of conflicting and mixed results in previous research, this study examines the relationship between the experience of discrimination and political engagement across multiple countries.

The results are clear and consistent for interest in politics and non-voting political actions. Racial minorities who have personal experience of discrimination tend to have higher levels of political interest and take more political actions, and this holds across all countries examined. This evidence suggests that the relationship is broadly generalizable, rather than requiring explanations related to the particularities of immigration source countries, historical experiences, or political systems. This consistent relationship between discrimination and political participation holds despite some variation in measures and question-wording between surveys, which is point in favor of the robustness of the results.

However, the relationship between discrimination and voting (turnout) is quite different. In fact, the results are positive, negative, or null depending on the country and controls. While there is variation in questions across countries, even examining just the questions about voting in national elections does not seem to produce a clear cross-national pattern which suggests these questions are not the primary driver of inconsistent results. Some differences might be related to different modeling strategies, such as coding of scales, different controls, and so on. Since we do not know the true model or causal structure, one way to view these different results is as a form of robustness check, showing that some previous results may not be robust to alternative specifications. However, the clearly distinctive patterns in U.S. datasets suggest it isn’t simply a product of methodological choices.

Strikingly, all three of the American datasets show a positive and statistically significant relationship between turnout and discrimination—in contrast to most other countries. Moreover, the turnout results seem to be generally consistent with previous results, including negative effects in Canada and Europe (Bilodeau 2017; Martin 2017; Mattila and Papageorgiou 2017), and positive effects in the United States (Besco et al. 2022; Lien 1994; White 2016). Denmark doesn’t fit that pattern, but it was the only survey to ask about “any election” rather than the most recent election.

The restriction of voting rights in the United States, which generally falls most heavily on racial minorities (e.g. Kuk, Hajnal and Lajevardi 2022), has produced a robust civil society infrastructure dedicated to both contesting restrictions and mobilizing voters. Indeed, White (2016) credits the mobilization effects they find precisely to this kind of activism. Conversely, in Europe there is debate over citizenship regulations and non-citizen voting rights (e.g. Kayran and Erdilmen 2021), but there are rarely serious attempts to make it difficult for citizens to vote. Where voting rights are a major focus of demands for racial equality, a strong connection between voting and experience of discrimination is more likely to emerge. In places where discrimination is experienced in ways not directly related to voting, the relationship might be different. Other kinds of political engagement, such as protesting, signing petitions, or being interested in politics are not subject to significant government regulations, and therefore differences in these regulations between countries are not relevant. This might lead to the kind of between-country variation (and lack of variation) seen here.

Nonetheless, a puzzle remains. Even if the politics of voting rights in the United States produced a stronger relationship between discrimination and turnout than seen elsewhere, it seems intuitively strange that discrimination would make people interested in politics, more likely to donate or sign petitions, but not to vote. The analysis here seems to convincingly demonstrate a series of relationships, but there is clearly more to be done in explaining why these emerge.

The limitations of the methods used here point toward opportunities in future research—three are discussed here. This study, like many others in the area, uses self-reports of both discrimination and engagement. While these are useful, we need more work using validated vote measures or other administrative data. For costly behavior like protesting or donating the risk of self-reports being “cheap talk” are especially high. These issues make understanding which sub-groups are mobilizing particularly difficult, since for rare actions social desirability effects might be larger than real shifts. Although is already some work which uses administrative measures for political mobilization, such as White (2016) and Kuhn (2022), or uses threatening external events and media data as an IV rather than self-reports (Cho, Gimpel and Wu 2006; Weiss, Siegel and Romney 2023), more is needed. In particular, there is little such work outside the United States.

The use of cross-sectional observational data also raises issues of endogeneity. For example, perhaps people who are interested in politics perceive more discrimination—or perhaps it causes them to actually experience more discrimination as a backlash to their activism. Thankfully, emerging causal evidence suggests this is not entirely the case (Besco et al. 2022;Weiss, Siegel and Romney 2023; White 2016). However, it might be that certain types of engagement are affected more than others, or the size of effects is incorrectly estimated. Or, it could be that causal routes are different for different kinds of political engagement. Studying discrimination causally is difficult, given that random natural variation is uncommon, and experimentally inducing it requires careful ethical considerations. Nonetheless, causal evidence is crucial for understanding these important issues.

Another approach is to more carefully theorize and test the mechanisms and moderators. One useful set of concepts might be perceived responsibility and efficacy. The failure of the social vs political discrimination analysis to replicate is surprising, given evidence from two countries and groups, in addition to a highly plausible theoretical account (e.g. Oskooii 2020). One explanation might be that the different expected effects of these sources of discrimination rely on expectations about responsibility: that political participation is a good way to contest discrimination by government actors. However, it isn’t clear that people in all contexts will have similar perceptions of responsibility. Are governments responsible for discrimination in schools, workplaces, or housing? Perhaps, but these might also be different than more “typical” state actors like politicians, bureaucrats, or police officers. It isn’t obvious what people on average would think, or how much variation there would be across individuals, groups, or contexts. In addition, would people who experience discrimination view political mobilization as an effective way to contest discrimination? Even if the state isn’t the source of discrimination, it might well be the solution, and thus political engagement is an effective method. Measuring these factors might help explain why, and when, different forms of discrimination lead to political engagement.

Finally, the role of civil society and other types of organizations needs more attention. If mobilization by civic organizations plays a key role in creating the link between threat or discrimination and voting in the United States, as White (2016) and others suggest, does this not occur in other countries, and why not? There is research on immigrant organizations and political participation in Europe (e.g. Berger, Galonska and Koopmans 2004; Pilati and Morales 2016), but this is not well incorporated into the research on discrimination. In part, this is due to methodological difficulties, since work on individual-level behavior tends to focus on individual-level factors, and standard surveys can be difficult to use for studying organizations. Nonetheless, this may be an avenue of research for understanding these quite different results for turnout across countries.

In conclusion, the results here provide an important cross-national test of the relationship of personal discrimination on political participation. We might hope that experiencing discrimination causes people to take political action to resist, rather than becoming discouraged and alienated. This is partially borne out by the results presented here—racial minorities who experience discrimination are more interested in politics and take more political actions, but may not vote at higher rates—at least not everywhere. In a global political environment where discrimination is increasingly prevalent, this is at least partially encouraging news.