Turnout Turnaround: Ethnic Minority Victories Mobilize White Voters

n Western democracies, like the United States, the United Kingdom, and Germany, the number of ethnic minority representatives has been steadily increasing. How is this trend shaping electoral behavior? Past work has focused on the effects of minority representation on ethnic minorities ’ political engagement, with less attention to the electoral behavior of majority-group members. We argue that increased minorities ’ representation can be experienced as a threat to a historically white-dominant political context. This, in turn, politically activates white constituents. Using data from four U.K. general elections and a regression discontinuity design, we find that the next election ’ s turnout in constituencies narrowly won by an ethnic minority candidate is 4.3 percentage points larger than in constituencies narrowly won by a white candidate. Consistent with our argument, this turnout difference is driven by majority-white constituencies. Our findings have implications for intergroup relations and party politics and help explain recent political dynamics.


I
n Western democracies, the number of ethnic minority office holders is steadily increasing.In the United States, the United Kingdom, and Germany, for example, the share of ethnic and racial minorities in federal legislatures has increased two-to threefold in the past decade alone.1A large body of work has been devoted to understanding when such increased descriptive representation translates to greater political participation (Geese 2022;Henderson, Sekhon, and Titiunik 2016) and substantive representation (Griffin 2014) of members of ethnic minority communities.
However, less attention has been paid to the role that increased minority representation plays in shaping the political behavior of members of the majority dominant group.With few exceptions (e.g., Gay 2001;Keele and White 2019), past work has focused on the effects of minority representation on majority-group members' attitudes toward racial minorities (Jardina 2019) and immigrants (Grossman and Zonszein 2022), and on violent backlash (Jacobs and Wood 1999).In this letter, we assess instead the effects of minority representation on majority-group members' non-violent political response, in the form of turnout and vote choice.Studying the electoral responses of majority-group members to the increased representation of minorities is important not only because they have implications for intergroup relations, party politics, and electoral campaigns (Dancygier 2013), but also because it is a priori unclear whether these responses halt or accelerate the recent progress on minority representation gains.
Our theoretical expectations build on three literatures that have explored the role that contextual factors play in shaping political engagement.First, following Bobo and Gilliam (1990), one literature focuses on co-ethnicity between representatives and their constituents, exploring whether and how it empowers members of marginalized minority groups.We extend this work by focusing instead on its effects on dominant majority-group members.A second literature examines how certain contextual factors (e.g., an increase in the relative size of a minority group) trigger economic and political threat perceptions among the majority group (Blalock 1967). 2 Threat perceptions have been shown to increase white constituents' political engagement (Abrajano and Hajnal 2015).We broaden this literature by arguing that increased descriptive representation of ethnic minority voters can be experienced as a threat to a historically white-dominant political context, mobilizing, in turn, white voters aiming to revert their perceived disempowerment, at least when such reversal is plausible. 3Lastly, we build on a literature debating whether incorporation policies attenuate or exacerbate conflict between majority and ethnic minority groups (e.g., Gundelach and Manatschal 2017;Weldon 2006).We contribute to this work both substantively-with a focus on increased incorporation resulting from institutional agreements between political parties-and methodologically-by being more sensitive to causal identification.
Testing the causal effect of incumbents' characteristics is complicated because there are likely other correlated contextual factors that matter for political engagement.Indeed, as Fraga (2016, 100) forcefully argues, the majority of studies assessing the effects of candidates' and incumbents' ethnicity on political engagement-for example, those using redistricting for identification (e.g., Barreto, Segura, and Woods 2004;Whitby 2007)-have had a difficult time disentangling the effects of politicians' ethnicity from the effects of other contextual factors, like the ethnic composition of the constituencies electing minority representatives.To address these concerns, we use a regression discontinuity design (RDD) that allows us to identify the effects of ethnic representation by comparing constituencies that are otherwise identical (including their ethnic makeup), except for being represented by a minority Member of Parliament (MP).
Using returns from four U.K. general elections from 2010 to 2019, we first compare next election constituency-level turnout between constituencies where an ethnic minority candidate narrowly beat a majority-group candidate, and constituencies where an ethnic minority candidate narrowly lost to a majoritygroup candidate.We find that the ethnicity of the MP matters for electoral participation: in constituencies where the minority candidate narrowly won, turnout in the next election is 4.3 percentage points (pp) larger than in otherwise identical constituencies where the minority candidate narrowly lost.
Following our theoretical framework, we further distinguish between constituencies that are majority white and plurality minority.We find that the increase in turnout following a narrow minority victory is driven by majority-white constituencies, where turnout increases by 7.7 pp compared to otherwise similar constituencies where the minority candidate lost.In contrast, in plurality-minority constituencies, we find no discernible increase in turnout.These results are consistent with findings using individual-level postelection survey data. 4 While an increase in white turnout following a narrow victory of a minority candidate is consistent with a backlash effect, it can also be consistent with other explanations.We explore constituency-level vote choice to further assess potential white backlash.We find suggestive evidence, albeit limited in-sample size (and not significant), that at election t þ 1 voters in majority-white constituencies are more likely to support the party of the strongest white opponent to the winning minority candidate at election t.In contrast, voters in plurality-minority constituencies are somewhat more likely to support the party of the minority incumbent.
The increase in support for the party of the minority incumbent MP compensates for the increased mobilization and vote choice of white constituents, such that, overall, we find suggestive evidence of an increase in the vote share of the minority incumbent's party.Importantly, this modest incumbency advantage masks a polarization of the electorate with white voters increasing their support for the party of the strongest white candidate (arguably to restore lost power) and minority voters increasing their support for the party of the strongest minority candidate.This polarization is further supported by a negative (not significant) effect of minority victories on the effective number of parties in the next election, as both majority and minority voters seem to concentrate their votes on the most viable parties advancing candidacies of their co-ethnics.

Data
We use constituency-level results at the 2010-19 U.K. general elections from the Electoral Commission 5 to compute our main outcome and treatment variables.
For the treatment at time t, we use the 2010, 2015, and 2017 elections, and for the outcome at time t þ 1, we use the 2015, 2017, and 2019 elections. 6For supplementary outcomes, we use individual-level data from the 2015, 2017, and 2019 British Election Study, a survey fielded immediately after the election, and representative of U.K. eligible voters. 7

Outcomes
The main dependent variable is the official reported constituency turnout rate in U.K. general elections, defined as the share of registered voters that cast a valid vote.As mentioned, we supplement the official constituency (continuous) return outcome with a selfreported individual-level (binary) turnout measure from post-election surveys.To assess a possible participation (Barreto, Segura, and Woods 2004;Henderson, Sekhon, and Titiunik 2016). 4While survey data may suffer from self-reporting bias, they are nonetheless important for guarding against the possibility of ecological inference fallacy.
5 See https://www.electoralcommission.org.uk/who-we-are-and-whatwe-do/elections-and-referendums/past-elections-and-referendums/ukgeneral-elections. 6Our analysis starts in 2010 because it was not until this election that ethnic minority candidates started to participate and win elections at higher rates.This resulted from public commitments from the three biggest parties setting internal targets to increase minority representation (Sobolewska 2013).Appendix B.8 of the Supplementary Material shows that the results are robust to including the 2005 election. 7Data from voter files are not available in the United Kingdom.Electoral registers do not contain turnout information in electronic form.The polling day paper copies of the register are destroyed by regulation within 12 months of the election so that past information does not exist.backlash effect, we use as outcomes the incumbent's party vote share and reelection probability, and the vote share for the party of the strongest non-coethnic contestant at election t.To provide further evidence of voters' polarization along ethnic lines, we use the Laakso and Taagepera (1979) measure of effective number of parties.

Treatment
Our key independent variable indicates whether an ethnic minority candidate wins a seat in Parliament at election t.We rely on existing classifications of a parliamentary candidate's ethnicity as white or Black, Asian, and minority ethnic (BAME).We build on these classifications by identifying the ethnic origin of BAME candidates based on their self-identification, their country of origin, and the candidate's parents' and grandparents' countries of origin.Appendix A of the Supplementary Material provides details about this classification.

Covariates
We use data from the 2011 Census to compute constituency characteristics that may determine both an ethnic minority win and our outcomes: the population share that is ethnic minority, high socioeconomic status, and the population density. 8We also control for a candidate's incumbency and party affiliation.

Sample
Based on treatment at time t, the sample includes 465 constituency-election years in which an ethnic minority candidate competed against a white candidate in England, Scotland, and Wales in the 2010-17 elections.Table 1 presents summary statistics for our outcomes, treatment, and control variables.

Estimation Method
We use a sharp RDD that compares the turnout and vote choice at election t þ 1 between constituencies where minority candidates narrowly won and narrowly lost to a majority-group candidate in t.The focus on close elections is important because constituencies with and without minority representation differ from one another in many ways.In contrast, as we empirically demonstrate in Appendix B.1 of the Supplementary Material, constituencies where a minority candidate narrowly wins or loses to a majority-group candidate are, on average, otherwise identical.As such, this design allows to disentangle the effects of minority representation from other possible constituency effects.
We estimate the following linear equation: where Y itþ1 is an outcome in constituency i at election period t þ 1. VictoryMargin it is the running variable, which is defined as the difference between the vote share obtained by the strongest ethnic minority candidate and the vote share obtained by her or his strongest white opponent in constituency i and election t.EthnicMinorityVictory it is an indicator variable for whether the ethnic minority candidate wins a seat in Parliament at t.The quantity of interest is τ , which reflects the RD estimate of the effect of an ethnic minority victory on electoral behavior at the winning threshold.We estimate τ by local-linear regression, that is, we fit Equation 1 to a sample that includes only constituencyelection years with vote share winning margins within the symmetric mean-squared-error (MSE) optimal bandwidth around the winning threshold.We compute the MSE-optimal bandwidth using Calonico, Cattaneo, and Titiunik (2014)'s adaptive bandwidth selection algorithm.For efficiency, we control for predetermined characteristics of the candidates and constituencies (described above in the subsection "Covariates").We report conventional RD estimates and robust biascorrected inference.Because we are pooling observations across elections, τ does not necessarily recover the average of all the single-election minority victory effects, as the number of close elections changes over time.Instead, τ reflects the pooled effect (Sekhon and Titiunik 2012).

RESULTS
Table 2 presents key estimates of the effect of minority candidates' narrow victory in election t on the constituency-level turnout rate in election t þ 1. Representation by an ethnic minority MP increases turnout by 4.3 pp (column 2).This effect size is equivalent to an increase of 6.4% relative to the turnout rate in otherwise identical constituencies represented by a white MP.
A natural question stemming from our theoretical framework is whether minority representation has a heterogeneous effect on the turnout of majority-versus minority-group members.Here, we split the sample into majority-white constituencies (where white people make up more than 80% of the population)9 and plurality-minority constituencies, and estimate an RD effect for each group.Within each group, we compare constituencies narrowly represented by a minority MP to constituencies narrowly represented by a white MP.Therefore, within groups, constituencies around the winning threshold are otherwise identical except for their MPs' ethnic identity.Since the data we use to estimate the effects reported in Table 2 are at the constituency level and the sample size for each type of constituency is relatively small, we offer only suggestive evidence.
Focusing on column 4, we find that majority-white constituencies represented by an ethnic minority MP (who narrowly won the previous election) exhibit an increased turnout of 7.7 pp.This suggests that white voters residing in such constituencies are significantly more likely to vote than are white voters in otherwise similar constituencies represented by a white MP.In Appendix C of the Supplementary Material, we present individual-level survey evidence suggesting that this is indeed the case.
In contrast, representation by a minority MP does not seem to affect political engagement in plurality- Importantly, the estimated effect in majority-white constituencies (column 4) is statistically distinguishable from the effect in plurality-minority constituencies (column 6; the Z-statistic is 1.88 implying a two-sided p-value of 0.06).This result reinforces the possibility of differential effects of ethnic minority representation on the political engagement of majority versus minority group members, and suggests that minority victories mobilize white voters.

Validity Tests and Robustness Checks
The RD estimates would be invalid if candidates sort around the winning threshold, in which case observations on either side of the cutoff might not be comparable.Following common practice, we conduct density tests, which show that the approximate number of observations just above the cutoff is not significantly different from the number of observations just below it, and covariate tests that show null RD effects on relevant predetermined variables.To guard against other possible threats to the validity of our results, we conduct additional tests, which demonstrate that our results are not sensitive to the bandwidth choice or to the order of the fitted polynomials, and a series of robustness checks, including testing placebo cutoffs, assessing differential registration, robustness to alternative classifications of constituencies as majority white/plurality minority, and to including the 2005 election.Appendices B.1-B.8 of the Supplementary Material present these tests, which strongly support the validity of our results.

DISCUSSION
When an ethnic minority candidate beats an ethnic majority candidate, we argue, majority-group members feel threatened (Blalock 1967), and this threat can in turn increase their political engagement in subsequent elections to restore lost political power (Jardina 2019).Finding that voters in ethnic minority-representedmajority-white constituencies turnout to vote (in t þ 1Þ at higher rates for the party of the incumbent's strongest white opponent (in t), relative to voters in whiterepresented-majority-white constituencies would suggest efforts to restore a white-dominant equilibrium.
We report supporting evidence, albeit only suggestive, of such a dynamic in Appendix D of the Supplementary Material (See Supplementary Figure D.1 and Supplementary Table D.1.Supplementary Table D.2 presents consistent evidence from survey data).Specifically, in majority-white constituencies, the minority victory effect on the vote share for the party of the strongest white opponent is 10.6 pp (not significant; Supplementary Figure D.1a and row 8 of Supplementary Table D.1).In contrast, in pluralityminority constituencies (Supplementary Figure D.1b and row 9 of Supplementary Table D.1), the minority victory effect is negative (−7 pp, not significant).While these two estimated effects on the opponent's vote share are not significant, the difference between them is significant (the Z-statistic is 1.87, implying a two-sided p-value of 0.06), suggesting that white voters respond to minority representation by strengthening their support for parties advancing white candidacies, and that to a lesser extent, minority voters respond to white representation with increased support for parties advancing minority candidacies.This dynamic, in turn, suggests plausible polarization of the electorate along ethnic lines.
How does such a response from voters affect incumbency advantage?We find that increased turnout and plausible vote concentration on the parties of the minority incumbents' strongest white opponents are not sufficient to overcome the incumbency advantage of minority MPs parties.We present evidence of such incumbency advantage in Figure 1, which shows the incumbent's party vote share as a function of the margin between the ethnic minority candidate and her or his strongest white competitor.
To the right of the winning threshold (vertical line), an ethnic minority candidate narrowly won the seat in election t; to the left, a white candidate narrowly won the seat.The blue lines indicate the average vote share (in t þ 1) for the incumbent's party.The jump in the vote share at the threshold measures the ethnic minority victory effect on the MP party's incumbency advantage.The size of this effect is 9.4 pp (significant; row 4 of Supplementary Table D.1), and consistent with the effect on the winning probability (13 pp not significant; row 10 of Supplementary to 12 of Supplementary Table D.1).Supporting evidence from survey data, although not statistically significant, is consistent with this finding; a minority victory decreases white voters choice and increases BAME voters choice for the party of the minority incumbent (rows 4-6 of Supplementary Table D.2).
Finally, in constituencies represented by narrowly elected minority MPs, we observe a (not significant) reduction in the effective number of parties of 0.2 (equivalent to a decrease of 9% relative to the effective number of parties in majority-represented constituencies), reinforcing the possibility that minority victories contribute to polarize the electorate along ethnic lines; white and minority voters possibly concentrate their votes on parties advancing candidacies of their co-ethnics (rows 13-15 of Supplementary Table D.1).

CONCLUSION
Using official electoral returns from four U.K. general elections and an RDD, we show that when minority candidates narrowly win, members of the dominant ethnic group increase their participation in subsequent elections, arguably to restore symbolic and political power.This result is consistent with findings by Grossman and Zonszein (2022), who show that minority electoral victories can lead to increased hostility among white voters.Our results are also consistent with Jardina (2019, chap. 8), who finds that the positive association between white racial consciousness and political participation is in part a function of perceived group threat.
Our findings contribute to studies of political behavior that are rooted in the idea that contextual factorsfrom large-scale immigration and its media coverage (Abrajano and Hajnal 2015), to incorporation policies (Filindra and Manatschal 2020), to candidates' ethnic identity (Geese 2022)-matter for turnout beyond individual characteristics, like socioeconomic status.Substantively, unlike much of the existing literature, we focus on how minority office holders shape the political behavior of majority-group members.Our results suggest that findings from past studies documenting a reduction in white turnout in redistricted constituencies (e.g., Gay 2001) are likely a result of the ethnic makeup of majority-minority districts, and do not generalize to competitive constituencies in which white mobilization can plausibly restore a whitedominant political context.
We also make a methodological contribution to this literature.Our study is among the first to causally identify the effects of ethnic minority representation on dominant-group members' electoral participation, as the RDD separates constituency ethnic makeup from representatives' ethnicity effects.Notwithstanding, the RD effects are only representative of constituencies where ethnic minority candidates stand for Parliament, which differ from the average constituency (e.g., in their ethnic minority and immigrant population shares and population density; Supplementary Table E.1).However, the mobilization dynamics we document likely generalize to other multi-ethnic democracies where the majority ethnic group is also the dominant group.Future research in other countries and electoral contexts, with richer ethnic minority officeholding, and therefore larger sample sizes, like the U.S. case, can use the RDD and administrative electoral records to assess mobilization dynamics.The replication of this study would not only contribute to knowledge accumulation about dominant-group members' responses to minorities' political accession, but our research design can also be used to push the boundaries of power threat theory (Blalock 1967), by assessing responses of minoritygroup members to other growing minority groups' political accession.
We argue that increased political engagement of white voters is aimed at restoring symbolic and political power.Although these two forms of power are hard to separate, future work could engage in exploring these two mechanisms by investigating differences in the policymaking and constituency service of white and ethnic minority MPs.Lastly, while our findings point to attempts to reinstate a white-dominant political equilibrium, they also reveal ethnic minority incumbency effects that are larger than their white counterpart incumbency effects.Accordingly, future work could assess the role of minority victories on parties' candidate nomination strategies.

SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S000305542300103X.

TABLE 1 .
Summary Statistics Note: N is the number of constituency-election years.

TABLE 2 .
Ethnic Minority Representation Effects on TurnoutThe dependent variable is the turnout rate in general election t þ 1.Average treatment effect at cutoff estimated with local linear regression with triangular kernel and MSE-optimal bandwidth.In brackets robust bias-corrected 95% CI.Supplementary TableB.1 presents other relevant statistics, and Supplementary TableB.2 presents covariate coefficients.Table2is close to zero and not statistically significant.This result is consistent with findings from causally identified studies from the U.S. that do not find evidence that ethnoracial minority candidates and officeholders increase turnout among minority constituents (e.g., Fraga 2016; Keele and White 2019).

Table D
Lines show local linear regression without covariate adjustment, points averaged equally spaced mimicking-variance bins.Row 4 of Supplementary Table D.1 presents point and inference estimates. Note: