Se Habla Español: Spanish-Language Appeals and Candidate Evaluations in the United States

Political candidates use Spanish-language appeals in efforts to increase their support among Hispanic voters. We argue that candidates, Hispanic or not, can use Spanish to signal closeness to Hispanics and posit that the effectiveness of these appeals is conditional on proficiency. To test this, we run two experiments where participants listen to an audio clip of a hypothetical candidate’s stump speech. We vary the ethnicity of the candidate (Anglo or Hispanic) and the language of the speech (English, non-native Spanish, and native-like Spanish). We find that Hispanic support for the Anglo and Hispanic candidates is higher in the native-like Spanish condition compared with the English-only condition. Relative to the English condition, non-native Spanish does not increase support for the Anglo candidate, but it decreases support for the Hispanic candidate. We find mixed effects for Anglo participants. Our results suggest that candidates can effectively appeal to Hispanic voters using Spanish-language messages.


M
any political candidates and elected officials have used Spanish-language appeals to increase their support among Hispanic voters.They have done this by speaking to constituents in Spanish, creating campaign ads in Spanish, or, more recently, showcasing their Spanish-language skills on the debate stage.For example, in the first 2020 Democratic Party presidential debate, candidates Beto O'Rourke, Julián Castro, and Cory Booker each delivered part of their remarks in Spanish (Contreras and Anderson 2019).On the Republican side, Senator Marco Rubio and former Florida Governor Jeb Bush have interviewed on Spanish-language networks and did so while campaigning in the 2016 Republican Presidential primary (González and Nowicki 2015).Existing studies suggest that speaking Spanish may be a useful strategy for candidates hoping to appeal to a large group of Hispanic voters across national origins (Alamillo and Collingwood 2017;Flores and Coppock 2018;Lavariega Monforti, Michelson, and Franco 2013).However, as the 2020 Democratic presidential candidates made clear, politicians are not equal in their ability to speak the language.While Booker and Castro struggled with their Spanish, O'Rourke spoke the language with relative ease.The significant variation in Spanish-language ability with which candidates speak could change the effectiveness of these appeals.In this study, we ask whether Hispanic voters care if a political candidate speaks to them in Spanish and whether it matters how proficient the candidate sounds.
Spanish, the second most spoken language in U.S. homes (Gonzalez-Barrera and Lopez 2013), plays an important role in the lives of Hispanics.According to a report by the Pew Research Center, 73% of Hispanics in the United States spoke Spanish at home in 2015 (Krogstad and Lopez 2017).While scholars have long noted that Hispanic Americans are not a monolithic group, Spanish is spoken in nearly every Latin American country from which Hispanics in the United States may trace their ethnic heritage.1However, Hispanic Americans have historically faced discrimination, with accusations of refusal to assimilate into traditional American values (Alba and Nee 1997;de la Garza, Falcon, and Garcia 1996) and the questioning of one's immigration status and place in American society, in part due to the use of Spanish (e.g., Stack 2019).Experiences of language discrimination in the United States led many Hispanics to dissociate themselves from the group and the accompanying stereotypes of inferiority and foreignness by not speaking Spanish (García Bedolla 2005;Zou and Cheryan 2017).In this context, the use of Spanish on a public stage by political elites may help Hispanic voters feel more included in the political process.
While language discrimination toward Spanish and other minority languages still exists, political candidates are displaying an increasing willingness to make Spanish-language appeals.This is true for not only national debate stages, but also for campaign advertisements.According to work by Abrajano (2010), in the 2000 and 2004 House races, eight congressional districts in each race had at least one candidate using a Spanish-language ad.During these elections, a total of 22 candidates created a Spanish-language ad.To compare, we conducted a content analysis of all Spanish campaign advertisements in the House from 2010 to 2018. 2 In 2010, 13 House districts had at least one candidate who created a Spanish-language ad, and in the 2012 to 2018 elections, that number jumped to at least 18 districts with a high of 22 in 2018.In contrast to the 22 candidates in 2000 and 2004, the 2018 House election alone had nearly double the candidates, 40, who used a Spanish-language ad.Moreover, candidates using Spanish-language ads are not exclusively Hispanic nor do they speak Spanish with the same level of proficiency.During the 2010-2018 period, 62 of the 151 candidates who produced a Spanish-language ad were Anglo with most of the other candidates being Hispanic.While the expectation is that the Hispanic candidates will speak Spanish with native proficiency, 17% of the Hispanic candidates in our sample spoke the language with a non-native, American accent.On the other hand, Anglo candidates are not expected to be able to speak Spanish let alone with native proficiency, yet 79% spoke Spanish at one point in the ad with 3% of all Anglo candidates being able to do so with a native-sounding accent.Despite this diversity, particularly among Hispanic candidates, no study, to the best of our knowledge, has systematically studied how this variation in language proficiency shapes Hispanics' evaluations of the candidates who make this appeal.
We argue that a candidate's ability to speak Spanish acts as a community commitment signal (Collingwood 2020;Stout 2020;Wamble 2019), conveying cultural familiarity and time spent in a Spanish-speaking community.Language proficiency in Spanish indicates a significant investment of time and energy on the part of an Anglo candidate and highlights a Hispanic candidate's connection to their ethnic heritage.As a result, we expect Hispanic voters to have more positive evaluations of candidates who use Spanish-language appeals.We posit that the impact of these appeals, however, is conditional on the quality of the appeal as defined by the level of language proficiency.Whereas speaking with native-like proficiency acts as a strong signal, limited proficiency sends a weaker signal. 3For a Hispanic candidate, limited proficiency Spanish may indicate distance from their ethnic heritage, whereas in the case of an Anglo candidate, it may lead them to be perceived as disingenuous.
Existing studies do not address how a candidate's Spanish proficiency affects Hispanics' evaluations of that candidate nor do they fully isolate the effect of language from that of shared ethnicity.While some studies show that Anglo candidates can increase their support among Hispanic voters by emphasizing their ability to speak Spanish, respondents in these studies only read or hear about the candidate's language ability (Alamillo and Collingwood 2017;Collingwood 2020;Lavariega Monforti, Michelson, and Franco 2013).Research where respondents are exposed to several candidates speaking Spanish shows that not all candidates benefit from Spanish-language appeals (Flores and Coppock 2018).Furthermore, these studies have focused primarily on Anglo candidates, and in the cases when Hispanic candidates are included, the research design does not allow researchers to examine whether and how the impact of Spanish-language appeals varies by candidate ethnicity.
Across two studies using an experimental design with audio treatments, our paper examines the effect of Spanish-language appeals on Hispanic voters' evaluations of political candidates.Our first study is a 2-by-3 between-subject experiment where we present respondents with a hypothetical candidate running for state legislature.We vary the ethnicity of the candidate (Anglo or Hispanic) as well as the language used in part of the speech (all English, partially in non-nativeaccented Spanish, or partially in native-accented Spanish). 4 , 5This experimental design permits us to isolate the effect of language from the effect of shared ethnicity and test the role of language proficiency.We find that Hispanics prefer candidates who use Spanishlanguage appeals but only when the candidate speaks with native-like proficiency.This is true for both the Anglo and Hispanic candidates in our experiment.Since appeals toward racial and ethnic minorities may disrupt electoral coalitions that depend on moderate Anglo voters (see, e.g., Fraga and Leal 2004;Frymer 1999), we test the impact of Spanish-language appeals on Anglo respondents.Against our expectations, Spanish-language appeals do not have a consistent impact, positive or negative, on Anglos' candidate evaluations, although further exploratory analyses suggest that Democrat and liberal Anglos have better evaluations of Hispanic candidates overall and candidates who speak Spanish.In our second study, we successfully replicate these findings and find support for our proposed mechanisms.These studies contributes to the study of racial and ethnic politics by testing the role of language proficiency in Spanish-language appeals, demonstrating that a candidate's ability to speak Spanish, and speak it well, matters for Hispanics.More generally, these findings underscore the importance that language cues can have for candidates and their campaigns.Language proficiency, independent of the content of an appeal, has an impact of candidate evaluations and communicates relevant information to voters about a candidate's connection and commitment to their group.Because language serves a key function in the maintenance of group identity and as marker of group membership (García Bedolla 2003;Lippi-Green 2012;Milroy 1982), campaigns can look toward ways to use language-based appeals for different groups.

ETHNIC APPEALS AND THE HISPANIC VOTE
Scholars of Latino politics have examined how political parties and candidates seek to gain support among Latinos.One way this happens is by having Latino candidates on the ballot.Research shows that shared ethnicity matters: Latino voters are more likely to turnout to vote when there is a co-ethnic on the ballot and prefer co-ethnic candidates over non-co-ethnics (Barreto 2007;2010;Barreto, Villarreal, and Woods 2005;Manzano and Sanchez 2010;McConnaughy et al. 2010;Sanchez 2006b).Scholars posit that attachment to an ethnic identity, the shared experiences it communicates, and the belief that group members share common goals explains the significance of ethnic cues at the ballot box (Casellas, Gillion, and Wallace 2019;Casellas and Wallace 2015).This is supported by a number of studies showing that higher levels of Latino linked fate led to greater support for co-ethnic candidates (McConnaughy et al. 2010;Schildkraut 2013;Wallace 2014). 6Furthermore, as Manzano and Sanchez (2010) demonstrate, shared ethnicity can help Latino candidates overcome perceived deficits in qualifications: Latinos with higher levels of perceived discrimination, greater support for collective action, and whose primary language is Spanish are more likely to support a co-ethnic even when they are less qualified than their Anglo counterpart.
Despite its importance in shaping Hispanic political behavior, ethnicity is not the only cue Hispanic voters use when evaluating a candidate.Electoral campaigns often include other sources of information such as party labels, and research shows that party is typically a stronger predictor of vote choice than candidate ethnicity (Michelson 2005).In addition, shared ethnicity does not presume closeness to the group (Mansbridge 1999;McClain et al. 2009) nor does the lack of shared ethnicity preclude successful appeals to Hispanic voters.Alamillo and Collingwood (2017) argue that Anglo candidates can "tap into Latino identity" by showing their connection and respect to the group which they term "cross-racial mobilization" (534).They show that Anglo candidates can increase their support among Hispanics through this type of campaigning which involves, for example, emphasizing any Hispanic family the candidate may have, pointing toward Hispanics on their staff, and creating Spanishlanguage ads.In line with Alamillo and Collingwood (2017), we argue that a candidate's ability to speak Spanish serves as an important cue and seek to separate the impact of Spanish-language ability from that of other cues (e.g., ethnicity).While Hispanic Americans have different national origins, they share a common language.This makes Spanish a useful way to target Hispanics across national heritages. 7 Previous work has found that Spanish-language appeals are effective at increasing turnout and support for a candidate among Hispanics.Field experiments on get-out-the-vote efforts demonstrate that Spanishlanguage appeals can increase turnout, although they remain less effective than English ads (Abrajano and Panagopoulos 2011;Mann, Michelson, and Davis 2020;Panagopoulos and Green 2011).Spanish-language campaign ads also lead to higher turnout but only among individuals who are Spanish dominant (DeFrancesco Soto and Merolla 2006;Ramírez 2011).While scholars have found evidence suggesting that using Spanish-language appeals is an effective campaign strategy to increase Hispanic support for a candidate (e.g., Alamillo and Collingwood 2017;Collingwood 2020;Flores and Coppock 2018;Lavariega Monforti, Michelson, and Franco 2013), existing studies cannot separate the effect of language from that of shared ethnicity.In other words, we do not know if Hispanic and Anglo candidates obtain the same benefit when they speak Spanish.Furthermore, most studies have relied on telling respondents the candidate's Spanish-speaking ability rather than exposing them to these appeals (but see Flores and Coppock 2018).The distinction is important as voters may react differently to hearing the candidate speak Spanish, where they can evaluate the candidate's ability.While being proficient may send a signal that the candidate is close to the community, anything less than native-like Spanish may be considered a disingenuous attempt at attracting Hispanic voters and may result in less favorable evaluations of the candidate.We seek to address these concerns through an experimental design which considers how proficiency may condition the effectiveness of Spanish-language appeals and that allows us to 6 Linked fate is the degree to which an ethnic group member believes that their interests are dependent on the fate of the larger group.While this concept was first proposed to explain the political behavior of African Americans (Dawson 1994), scholars have examined whether other racial and ethnic minority groups exhibit similar levels of linked fate and group consciousness (McClain et al. 2009).Among Latinos, determinants of linked fate change with the political environment, but research finds that economic status, generational status, and perceptions of discrimination are the strongest predictors of Latino linked fate (Sanchez and Masuoka 2010;Sanchez, Masuoka, and Abrams 2019).In turn, these group orientations shape Latino political participation and policy attitudes (Sanchez 2006a). 7Because of the varieties of Spanish that are spoken throughout Latin America, it is possible that Hispanic Americans of a specific heritage only respond to candidates when they speak their own variant.
directly test how individuals react to the same Spanishlanguage appeal made by a Hispanic and an Anglo candidate.

LANGUAGE PROFICIENCY AS A SIGNAL OF COMMITMENT
We argue that candidates, Hispanic or otherwise, can establish rapport with Hispanic voters via the use of Spanish-language appeals.As Wamble (2019) explains, candidates may signal their commitment to minority communities by emphasizing any social connections they may have with the community or by pointing to actions of personal sacrifice they have taken to advance minority interests.Spanish-language proficiency may act as such a signal of social connection and commitment for Hispanic voters.Similar to how shared ethnicity fosters a connection between Hispanic candidates and voters (Barreto 2010;Casellas, Gillion, and Wallace 2019;Casellas and Wallace 2015), the ability to speak Spanish communicates the candidate's closeness to the group and understanding of what concerns them.However, whereas shared ethnicitybeing dependent on one's ascribed membership to an ethnic group (see McClain et al. 2009)-is an ethnic cue which only pertains to Hispanic candidates, language is a learned characteristic both co-ethnics and non-coethnics can use.
The use of Spanish by candidates during debates, campaign events, and political advertisements can serve to signal a candidate's ability and willingness to represent Hispanics and their interests.First, speaking Spanish may signal an ability to represent Hispanic interests because, whether learned as an adult or acquired as a child, speaking a group's language conveys a connection to that community.In the context of Spanish and the United States, being a native Spanish speaker implies growing up in a Hispanic household.For non-native speakers, learning Spanish signifies a substantial investment of time and energy and, in some cases, immersion in a Spanish-speaking community.Second, Spanish-language appeals may communicate to Hispanic voters a candidate is willing to represent them given the history of language discrimination in the United States.Because Spanish was seen as a hindrance to economic and social mobility, many secondand third-generation Hispanics were brought up with little to no Spanish (García Bedolla 2003;Lippi-Green 2012).Still today, one can find examples of elected officials and members of the public being criticized for speaking Spanish (e.g., Bever 2019;Stack 2019).The fact that some politicians are now doing so while campaigning, even at a nationally televised presidential debate, provides a stark contrast to long-standing practices of language exclusion in the United States.Furthermore, speaking Spanish in these contexts is a deliberate move away from the default of Englishlanguage campaigning and an unequivocal appeal to Hispanic voters.These explicit appeals may increase Hispanic support as they help inform the community's expectations for what the candidate will do in office (Stout 2020;Wamble 2019).
While both Hispanic and Anglo candidates may benefit from speaking Spanish, its use may convey different meanings to voters.For Hispanic candidates, the use of Spanish in public speech can help them reinforce their connection to Hispanic voters by emphasizing their shared ethnicity.Anglo candidates can use Spanish to signal commitment and bridge the social identity gap between them and Hispanic voters (see Alamillo and Collingwood 2017;Collingwood 2020).In other words, Anglo candidates could improve their image among Hispanic voters as the latter group may see these efforts as a sign of respect and "a desire to break down cultural barriers through reduction of linguistic dissimilarities" (Koslow, Shamdasani, and Touchstone 1994, 576).This leads to our first hypothesis.
Spanish-language appeals hypothesis (H1): Hispanic voters will give more positive evaluations to a candidate who speaks Spanish over one who does not.
Consistent with previous research on the importance of ethnicity as a cue, we hypothesize that, on average, Hispanics will prefer Hispanic candidates over Anglo candidates.Hispanic candidates should have closer ties to Hispanic voters and will have more cultural competency, making them more likely to have a better understanding of how to represent them (Barreto 2007;Wallace 2014).Furthermore, individuals typically value the in-group more so than the out-group as they seek value in belonging to something (Tajfel and Turner 1979).
Shared ethnicity hypothesis (H2): Hispanic voters will give higher evaluations to a Hispanic candidate than an Anglo candidate.
While previous research has shown that language can have a positive impact in how Hispanics view candidates, we argue that it is crucial to examine the candidate's Spanish-language proficiency.Individuals can show substantial variation in their command of the language and this may condition the effect of these appeals.On the one hand, native speakers, and individuals who achieve a similar level of proficiency, convey the strongest signal of commitment, ability, and willingness to Hispanic voters.On the other hand, whether because of poor syntax, pronunciation, vocabulary, or a combination of these, individuals with limited proficiency send a much weaker signal of commitment and their appeals may be seen as insincere.Thus, we expect the strength of Spanish as a signal of commitment to vary by speaker where limited proficiency is a weak signal and native-like proficiency is a strong signal.
It is important to consider the possibility that voters will evaluate candidates with lower Spanish proficiency differently based on whether the candidate is Hispanic or Anglo.If an Anglo candidate speaks non-native sounding Spanish, Hispanics may still reward them as there are no set expectations for them to be able communicate in Spanish, let alone with native-like proficiency.Since an appeal made in Spanish is clearly intended for Hispanics, even if the individual's language proficiency is poor and they are unable to hold a conversation, Hispanics may still reward the effort.By contrast, for a Hispanic candidate, lack of proficiency in Spanish may convey a level of removal from their ethnic heritage.Non-native sounding Spanish from a co-ethnic candidate can be interpreted as meaning that they do not have as many shared experiences and are therefore culturally different from a Spanish-speaking Hispanic.In short, these co-ethnic candidates may not be considered "real" Hispanics (see García Bedolla 2003), and Hispanic voters may feel less affinity toward them.While limited Spanish-language proficiency may not benefit an Anglo candidate, we hypothesize that it will actively hurt a Hispanic candidate.
Non-Native Spanish hypothesis (H3): Relative to the Anglo candidate speaking English, Hispanics will punish Hispanic candidates that speak non-native sounding Spanish.
As discussed above, however, not all Hispanics show a preference for co-ethnics.Those that highly identify as Latino look more favorably upon their in-group, making them even more likely to support a co-ethnic (McConnaughy et al. 2010;Schildkraut 2013;Wallace 2014).We hypothesize that the moderating role of ethnic group strength-that is, how central being part of the ethnic group is to one's identity-will extend to Hispanics' reaction to Spanish-language appeals.We expect that the positive impact of ethnicity and Spanish-language appeals, as well as the negative effect of non-native Spanish proficiency, will be larger among high identifiers compared with low identifiers.Having a candidate willing to represent Hispanic interests is a higher priority for high-identifying Latinos compared with low-identifying Latinos.Thus, Hispanics who see their ethnic identity as very important will prefer a candidate with native-like proficiency because it communicates an ability and willingness to represent them.
Group strength hypothesis (H4a): The effect of shared ethnicity will be stronger for Hispanics with higher levels of group strength.

Group strength hypothesis (H4b):
The effect of Spanishlanguage appeals will be stronger for Hispanics with higher levels of group strength.

AN EMPIRICAL TEST OF LANGUAGE-BASED APPEALS Experimental Design
To test our hypotheses, we conducted a pair of studies from two different samples where we varied the ethnicity and Spanish proficiency-via the accent-of a hypothetical candidate running for state legislature. 8 Participants listened to a brief audio clip of the candidate giving a part of their stump speech and responded to a battery of questions that measure evaluations of the candidate.For Study 1, we obtained a national sample of 503 Hispanic and 506 Anglo respondents through the survey firm Prolific. 9 , 10The survey was fielded from October 26-31, 2020, just before the 2020 November election.Overall, our Hispanic and Anglo samples were younger, had higher levels of education, were more liberal, and were more likely to identify as a Democrat compared with population estimates of their respective subgroups.Given the liberal skew of our Study 1 sample, from December 28, 2021 to January 11, 2022, we ran an additional study recruiting one thousand Hispanic and one thousand Anglo respondents through Lucid where the partisan composition of our Hispanic and Anglo samples corresponds to the two-party 2020 presidential vote share in their respective group.In addition to seeing if our results replicate, Study 2 allows us to tease out our proposed mechanisms and test the robustness of our findings to the inclusion of an in-party cue. 11 Prior to beginning both experiments, all participants were told that they were being asked to participate in a research study examining how people evaluate state legislature candidate campaign speeches.Participants were required to give consent to participate in the study before being allowed to continue.At the beginning of the survey, we asked several demographic questions including age, gender, ideology, and partisanship.In Study 1, participants were randomly assigned to one of six conditions where they were instructed to listen to a short audio clip from a campaign speech of a candidate running for a seat in a state legislature.We varied whether the candidate is Hispanic or Anglo, and 8 The hypotheses and the analysis plan were preregistered for both studies.The preregistration for Study 1 can be found at https:// osf.io/6ebav/ and at https://osf.io/g36q9for Study 2. In addition to the hypotheses included in the main text, we also preregistered hypotheses to test the role of ideology as a potential moderator among Hispanics and Anglos as well as how racial group strength moderates responses among Anglos.Due to space constraints, we include these analyses in Sections M and F of the Supplementary Material, respectively. 9Based on the results of a pilot study conducted in the 2020 spring and summer terms at the University of Texas El Paso, results from a power analysis suggested we collect about four hundred responses.We decided to collect five hundred responses for each group in order to account for missing data as well as boost our statistical power that would allow us to test for the moderating effects of group strength. 10Prolific is a crowdsourcing platform promising high-quality data for academic research.In a comparison with similar platforms, Peer et al. (2017) conclude that Prolific is a viable alternative to Amazon Mechanical Turk: Data quality-assessed through attention checks and ability to replicate findings of psychological research-is high and the subject pool is "more naive to common experimental research tasks" (161).whether part of the campaign speech is spoken in nonnative Spanish, native-like Spanish, or fully in English.To vary the ethnicity of the candidate, we gave a Hispanic-or Anglo-sounding name to the hypothetical candidate (Josué Martínez or Josh Martin) and explicitly stated the candidate's ethnicity or race.For the language treatment, respondents were assigned to one of three different audio recordings, all of which shared the same substantive content.The only differences between these audio files was in the language, English or Spanish, used in two sentences (shown in brackets below) and the candidate's accent when speaking Spanish.For Study 2, we also vary whether the candidate is running in a nonpartisan or partisan primary election for a total of 12 treatment conditions.Participants in Study 2 either did not know the candidate's partisanship or were told that the candidate was running in their self-identified political party primary (for more information on Study 2, see Section B of the Supplementary Material).The survey vignette for Study 1 (with the transcript of the audio clip) was as follows: In a recent campaign event, state representative [Josh Martin/Josué Martínez] talked about his plans for his district if he were to be reelected and why he believes he is the best choice to represent his constituents in the state legislature.Representative [Martin/Martínez], who is [White/Latino], currently represents a mostly rural district that is 70.8%Latino with a median income of $46,232.Below you will find an excerpt from [Josh Martin's/Josué Martínez's] remarks, where he makes his case to constituents.Please listen to the brief audio clip before continuing.
[AUDIO TRANSCRIPT] "Thank you all for being here.As your state representative I have worked tirelessly to represent your interests.[I am here to listen to your concerns and make sure the government addresses your needs.You can trust me for this job./ Estoy aquí para escuchar sus preocupaciones y asegurarme de que el gobierno atienda sus necesidades.Pueden confiar en mí para este trabajo.]Together, we can keep up the work of improving our schools, rebuilding our infrastructure, and bringing this great community closer together.If given the opportunity, I would be honored to continue representing you.I'm counting on each one of you so that we can forge ahead and make our vision for our district and our great state a reality." In the English conditions, the candidate spoke entirely in English.In both Spanish conditions, the candidate spoke mostly in English with two sentences in Spanish.In the non-native Spanish conditions, the candidate spoke Spanish with an American English accent.On the other hand, the candidate in the native-like Spanish conditions spoke with nativeaccented Spanish. 12It is important to note that both Spanish conditions use the same text, which is grammatically correct; the difference is in how it is spoken not what is spoken. 13To control for any effects of timbre or pitch (see Dietrich, Hayes, and O'Brien 2019), a single voice actor was used for all three recordings and the same recording was used regardless of candidate ethnicity. 14Furthermore, we decided against adding any substantive policy content or background information of the candidate (e.g., partisanship) other than ethnic identity to isolate the impact of language ability.
After receiving the treatment, we included four questions to measure our outcome of interest: evaluations of the candidate.We measure candidate evaluations by asking respondents how likely they are to vote for the candidate, whether they trust the candidate, how much they like or dislike the candidate, and if they would feel represented with the candidate in office.All items are measured on a 5-point scale. 15For our dependent variable, we construct an index of candidate evaluations by summing the scores of these four variables.Our index ranges from 0 to 1 where higher numbers indicate more favorable evaluations. 16 Following the candidate evaluation measures, we collected other relevant social and demographic information such as generational status and ability to write and speak in Spanish.In Study 1, we included questions about racial and ethnic identity, as well as importance of ethnic identification (our measure of group strength), after the treatment. 17We chose not to ask participants' questions regarding racial or ethnic identity prior to our experimental manipulation to avoid any potential for corrupting our treatments due to priming effects.We acknowledge that asking these questions after the treatment may introduce posttreatment bias in our estimates when we condition on these variables (Montgomery, Nyhan, and Torres 2018).However, as Klar, Leeper, and Robison (2020) discuss, when it comes to characteristics such as racial and ethnic identity, these questions have higher potential of inducing priming effects than to 12 Audio clips may be accessed at Zárate, Quezada-Llanes, and Armenta (2023).See Section A of the Supplementary Material for more detail. 13To examine whether our language treatments worked as expected, our manipulation checks asked respondents to indicate the languages they heard the candidate speak.In Study 1, 93% of respondents in a Spanish condition correctly identified that the candidate spoke Spanish.When asked to rate the quality on a scale from 0 to 10, those in the non-native condition gave an average rating of 4.55 for the Anglo candidate and 4.06 for the Hispanic candidate.Those in the nativelike Spanish conditions ranked the Anglo candidate's Spanish 8.09 and the Hispanic candidates' 8.55.See Section A.2 of the Supplementary Material for more detail on Study 1 or Section B.3 for this information on Study 2. The audio clips used in Studies 1 and 2 are the same. 14The voice actor featured is a native Spanish speaker and bilingual. 15See Section A of the Supplementary Material for question wordings. 16The Chronbach's alpha for these variables is high at 0.89 and 0.88 among our Hispanic samples from Studies 1 and 2, respectively.All of the items strongly load onto a single factor.For more information on our scale development, go to Section D of the Supplementary Material. 17We measure group strength by asking participants "How important is being (white/Latino) to your identity?"Participants are only asked about white identity if they marked being white and only asked about Latino identity if they marked being Latino.
show change as a result of an experimental manipulation (but see Egan 2020).While there is no one-size-fits-all solution to avoid both priming effects and posttreatment bias, we believe that in our particular case, the risk of corrupting our treatment was greater than that of conditioning on posttreatment variables.In Study 2, we ask for racial identification information at the beginning of the survey and randomly assign participants to receive the group strength questions either before or after the treatment.We find no differences in levels of ethnic group strength between those who answered these questions pre-and posttreatment. 18Finally, in both studies, all participants are debriefed at the end of the experiment.

Results
To examine whether Hispanics give higher evaluations to candidates who make Spanish-language appeals (H1), we use a linear regression with the candidate evaluation index as the dependent variable and our language conditions as a factor variable where English serves as the baseline category.Table 1 displays the results among our Hispanic samples for Studies 1 and 2. Models 1 and 3 of Table 1 show mixed support our Spanish-language appeals hypothesis (H1): Hispanics give higher evaluations to candidates when they speak Spanish, regardless of whether the candidate is Hispanic or Anglo.However, candidates only received better evaluations if they spoke Spanish with a native-like accent.Respondents' evaluation of the candidate in the non-native Spanish condition was statistically indistinguishable from their candidate evaluation in the English condition.On the other hand, native-like Spanish has a large, positive, and statistically significant impact (b ¼ 0:122 , p < 0:001 Model 1).The effect size is equivalent to 12% of the total range of the candidate evaluation index.We obtain a significant and positive, albeit smaller, result for nativelike Spanish in Study 2 (b ¼ 0:044, p < 0:001; Model 3).These results indicate that candidates, Hispanic or Anglo, can use Spanish-language appeals to foster more favorable evaluations among Hispanics but only if they have native-like proficiency. 19 We also find support for the shared ethnicity hypothesis (H2), successfully replicating a consistent finding in the Hispanic politics literature (e.g., Barreto 2010).Models 2 and 4 in Table 1 show the results of a linear model with candidate evaluation as the dependent variable and an interaction between the ethnicity treatment and the language treatment as our explanatory variables. 20The coefficient for Hispanic candidate, representing the difference in the candidate evaluation index between the Anglo and Hispanic candidates when both speak English, is positive and 18 In Appendix R of the Supplementary Material, we provide evidence that our treatments did not impact strength of ethnic identification in either sample. 19We ran a separate model where we regressed the candidate evaluation index on a binary indicator where 1 indicates being in either our non-native or native-like Spanish conditions.The coefficient for the grouped Spanish conditions was small but positive and statistically significant (b ¼ 0:056, p < 0:001).However, as we show, once we separate the Spanish conditions by accent, the results show that the positive impact is due entirely to the native-like Spanish treatment.We report these models for both Hispanic and Anglo participants in Section C of the Supplementary Material. 20We ran both models with and without controls for age, gender, education, ideology, partisanship, and income.Overall, the results in all models are virtually unchanged when we include these controls.See Table J.1 in Section J of the Supplementary Material for full results.
statistically significant (Study 1: b ¼ 0:118, p < 0:001; Study 2: b ¼ 0:047, p < 0:05).This effect can also be observed in the gap between the two leftmost estimates in each panel of Figure 1, where we plot the average candidate evaluation for each treatment condition for Studies 1 and 2. 21 Group means are shown with 84% confidence intervals around them to allow for visual inspection of statistically significant differences (Payton, Greenstone, and Schenker 2003;Schenker and Gentleman 2001). 22In Study 1, the average evaluation for the Anglo candidate in the English condition is 0.44 and goes up to 0.56 when the candidate is Hispanic.For Study 2, the average evaluation for the English-speaking Anglo candidate is 0.59 and the average evaluation for the Englishspeaking Hispanic candidate is 0.64.
Our experimental design allows us to account for how a candidate's Spanish-language accent and ethnicity affect their evaluation among Hispanic voters.We theorized that Spanish-language appeals may have different effects depending on whether they come from an Anglo or a Hispanic candidate.Overall, we find that Anglo candidates can benefit from Spanish-language appeals when speaking with native-like proficiency and Hispanic candidates with non-native Spanish are punished by their co-ethnics.For Study 1, our Hispanic sample evaluated the Anglo candidate speaking nativelike Spanish much more favorably compared with the Anglo candidate speaking exclusively in English.The average evaluation for the Anglo candidate increased from 0.44 in the English-only condition to 0.63 in the native-like Spanish condition (b ¼ 0:189).This effect size is larger than the effect size of shared ethnicity (b ¼ 0:118).Furthermore, this effect is replicated in our Study 2 where average evaluations of the Anglo candidate increased from 0.59 when only speaking English to 0.65 when they used a native-like Spanish appeal.Looking at the effect of non-native Spanish, the Anglo candidate does not receive better or worse evaluations when speaking non-native Spanish compared to the English condition.Although the impact is positive among the Study 1 sample, it is negative among respondents in Study 2, and neither has statistical significance.I.1 in Section I of the Supplementary Material.
21 Results for Study 2 are pooled by election type (nonpartisan and in-party primary).We report the results by election type in Section B.4 of the Supplementary Material.For results for the four individual variables, see Section I of the Supplementary Material. 22This is equivalent to using a t-test to compare two group means where a ¼ 0:05 (Payton, Greenstone, and Schenker 2003).The use of 95% confidence intervals around predicted values to assess whether the difference between two estimates is statistically significant represents a more conservative test than the a ¼ 0:05 standard (Schenker and Gentleman 2001).
When it comes to the evaluations of Josué Martínez, the Hispanic candidate, respondents give lower evaluations when the candidate speaks non-native Spanish than when the candidate speaks only in English (p ¼ 0:013 ).It is only when speaking Spanish with a native accent that the evaluation for Josué Martínez moves up from 0.56 to 0.61 (Study 1), but the difference is not statistically significant in either of our two Hispanic samples.In other words, Hispanic candidates are punished for speaking non-native sounding Spanish and may not necessarily be rewarded when making appeals with a native-like accent.That said, we do not find support for our Non-Native Spanish hypothesis (H3).Latinos do not evaluate the Hispanic candidate speaking non-native Spanish lower than the Englishspeaking Anglo candidate.
We can also analyze the data by looking within each language condition.The effect of shared ethnicity is only present in the English condition, where there is a gap between the candidates of about 12 and 6 percentage points in Studies 1 and 2, respectively.In the non-native Spanish and native-like Spanish conditions, Hispanic evaluations of the Anglo and Hispanic candidates are statistically indistinguishable.Furthermore, our results support the thesis that Anglo candidates can close the social identity gap between them and Hispanic voters by making Spanish-language appeals (Alamillo and Collingwood 2017).The Anglo candidate speaking native-like Spanish received similar or higher evaluations than the Hispanic candidate. 23Thus, it appears to be that Anglo candidates stand to benefit from Spanishlanguage appeals if they have native-level proficiency, whereas Hispanic candidates who speak Spanish with a non-native accent risk losing the advantage of shared ethnicity.
Previous work on ethnic appeals has shown that an appeal's impact is stronger among individuals with higher levels of ethnic group identification.To test this hypothesis (H4), we run a model where we interact our measure of ethnic group strength with the treatment conditions.Figure 2 shows predicted values of candidate evaluation for each treatment condition across levels of group strength in our Prolific sample. 24The left panel shows the results for the conditions with the Anglo candidate; the results for the Hispanic candidate  appear on the right.In general, as group strength increases, so too does the evaluation of the Hispanic candidate.Furthermore, moving from low to high on the ethnic group strength scale, candidate evaluations increase in all Spanish conditions.The only exception is when the Anglo candidate speaks in non-nativeaccented Spanish, but this difference does not reach statistical significance.Of note are the evaluations lowand high-identifying Hispanic respondents give to candidates based on Spanish-language proficiency.Lowidentifying Hispanics (i.e., group strength set at its minimum) do not give significantly different evaluations to candidates with non-native and native-like Spanish.By contrast, high-identifying Hispanics differentially evaluate candidates according to their accent, responding more favorably to Anglo and Hispanic candidates with native-like Spanish.
In addition to our stated hypotheses, we theorized that Spanish-language appeals increase candidate evaluations among Hispanics because these appeals signal the candidate's ability and willingness to represent them and their community's interests.We expect that this will be conditioned by the proficiency of the candidate.In Study 2, we included several items to tap into respondent's perceptions of candidate ability and willingness, and created an index to measure each of these variables that ranges from 0 to 1. 25 Figure 3 shows the mean ability score (left panel) and mean willingness score (right panel) for each treatment group.The candidate's ethnicity as well as their Spanish-language proficiency significantly impact perceived ability and willingness.Specifically, the Hispanic candidate is seen as better able and more willing to represent Hispanics compared with the Anglo candidate when both candidates speak only in English.The difference in perceived ability and willingness between these two conditions is of more than 0.2 points, which accounts for a fifth of the scale.When it comes to Spanish-language appeals, they appear to only be a credible signal when the candidate speaks native-like Spanish.Hispanic respondents perceive candidates with native-level Spanish as better able and more willing than candidates in the English-only conditions.Meanwhile, there is no difference in perceived ability or willingness between the Anglo candidate in the Englishonly condition, the Anglo candidate with non-native Spanish, and the Hispanic candidate with non-native Spanish.This evidence suggests that language, in addition to ethnicity, conveys information that shapes how voters perceive the candidate will act if elected.
In sum, the results from our experiment show that Hispanics have more positive evaluations of Hispanic and Anglo candidates who make appeals in Spanish.However, Hispanic voters react differently to these language appeals depending on the race and ethnicity of the candidate.While Anglo candidates benefit from speaking native-like Spanish, Hispanic candidates may be evaluated more poorly when speaking Spanish with a non-native accent.Lastly, our results confirm that ethnic group strength matters by showing that highidentifying Hispanics prefer Hispanic candidates over Anglo candidates in most cases. 26

HOW ANGLO VOTERS RESPOND
While Spanish-language appeals by candidates result in more positive evaluations among Hispanic voters, it is possible that they result in less positive evaluations among other groups (Hersh and Schaffner 2013).If this is true, then considerations about net negative electoral outcomes may lead candidates to refrain from making such targeted appeals.This may be especially true when the targeted group is in a state of electoral capture and politicians are not concerned with these voters jumping ship (Fraga and Leal 2004;Frymer 1999).Even though Hispanic voters may not be considered a captured group, over 60% of Hispanic registered voters identify as Democrats (Lopez et al. 2016).This reduces the incentives to make unequivocal ethnic appeals to Hispanics if there is an expectation that "appeals to the group will disrupt the party's electoral coalition" (Frymer 1999, 8, emphasis in original).While trying to appeal toward Hispanic Americans by speaking Spanish, politicians could be hurting their chances with Anglo voters.
We expect that Anglo voters will have more positive evaluations of an Anglo candidate compared with a Hispanic candidate for two reasons.First, similar to how Hispanics prefer members of their own group to represent them, Anglos for whom their racial identity is important and who perceive their group to face discrimination exhibit a preference for Anglo candidates (Schildkraut 2017). 27Second, Hispanic candidates may be perceived as less competent and more foreign than their Anglo counterparts (Zou and Cheryan 2017).Previous work has shown that anti-immigrant sentiment among Anglo voters is negatively associated with evaluations of competency and likelihood of vote for Hispanic candidates (McConnaughy et al. 2010).Without any information about the candidate's qualifications, Anglo respondents may believe that Hispanic candidates are less competent and unable to represent their interests as well as an Anglo candidate.
We also expect that Anglo voters will react negatively to candidates who use Spanish-language appeals.Accustomed to being the main target of campaign appeals, these individuals may feel as though their status is slipping.Their position in the racial hierarchy -and the resulting implicit attitudes regarding the status of their group (see social position theory; e.g., Bobo and Hutchings 1996;Masuoka and Junn 2013;Pérez 2016)-may impact their reaction to Spanishlanguage appeals in a political setting.That is, Anglo voters might be less likely to support a candidate who makes these ethnic appeals because they may be perceived as a cultural threat and as diminishing the existing racial hierarchy (Bellovary, Armenta, and Reyna 2020;de la Garza, Falcon, and Garcia 1996;Pérez 2016). 28In addition, research has also shown that linguistic ostracism, or the communication setting where an individual cannot understand the language being spoken, leads listeners to have negative evaluations of the speaker (Flores and Coppock 2018;Hitlan et al. 2006;2016).Assuming that the average Anglo individual is not fluent in Spanish, they may feel ostracized when politicians speak in Spanish because they cannot understand what is being said and feel left out of the conversation.
Table 2 shows the results of four models where we regress our candidate evaluation index on the language treatments among Anglo respondents in Studies 1 and 2. Parallel to our analysis of the Hispanic sample, Models 1 and 3 pool the Hispanic and Anglo candidate treatment conditions, whereas Models 2 and 4 include an interaction between the language and ethnicity conditions.We hypothesized that Spanish-language appeals would result in backlash from Anglo voters.Our results, however, do not support this.In Study 1, Anglo respondents in the native-like Spanish condition, regardless of candidate ethnicity, give more positive evaluations than respondents in the English condition (b ¼ 0:066, p < 0:01; Model 1).Against our expectations, Anglo respondents also give more positive evaluations to the Hispanic candidate when both candidates speak English (b ¼ 0:093, p < 0:01; Model 2).That said, respondents in this sample were more liberal and more likely to identify as a Democrat compared with Anglos in the general population. 29In Study 2, where our sample matches the population on these characteristics, we find no significant effects for ethnicity or language in either direction (see Models 3 and 4).
For easier interpretation of the effect of language and candidate ethnicity on candidate evaluations, Figure 4 shows the mean candidate evaluation score in each condition from our two samples of Anglo

DISCUSSION
Our results suggest that if Hispanic or Anglo candidates want to court Hispanic voters, putting to use their Spanish-language skills while campaigning can be an effective strategy.The ability to speak native-like Spanish acts as a signal to the Hispanic community that the candidate has the ability and willingness to represent Hispanic interests.However, only native-level proficiency appears to act as a credible signal of this understanding and commitment.When the candidate moves from speaking only in English to speaking Spanish with a non-native accent, Hispanics' evaluations of the Anglo candidate remain the same but are significantly lower for the Hispanic candidate.In other words, speaking non-native Spanish as a Hispanic candidate seems to remove the connection that Hispanic voters feel based on shared ethnicity.This has important implications for how candidates choose to reach out to Hispanic candidates.Hispanic candidates without a native-like accent should avoid memorized and practiced lines in Spanish.Anglo candidates, not carrying the burden of being expected to speak Spanish, stand to gain from language-based appeals.At the highest level of proficiency, Anglo candidates can close the gap among Hispanics in overall evaluation between them and Hispanic candidates.
The results from our experiment may provide context for our content analysis of congressional Spanish ads.Although our content analysis suggests that it is rare for Hispanic candidates to speak Spanish with a non-native accent, if Hispanic candidates know they will be punished for poor Spanish-language skills, they may be avoiding it altogether.While it is even rarer for Anglo candidates to have a native-like accent when speaking Spanish, the experimental results suggest that honing their language skills is a fruitful path for appealing to Hispanics. 31ur work does not mean to imply that Hispanic voting behavior can be explained solely by ethnic appeals, whether through shared ethnicity or language.While Hispanic voters use ethnicity as a cue (Barreto 2010), it does not supersede other potential sources of information such as party label (Casellas, Gillion, and Wallace 2019;Michelson 2005).32Furthermore, Hispanic voters rate policy issues as more important than language ability.In a 2019 poll of eligible Latino voters by the advocacy group UnidosUS, 61% of respondents said that they thought it was "very important" that a presidential candidate had plans for issues of interest to the Latino community, whereas only 33% said the same about a candidate's ability to speak Spanish (UnidosUS and Decisions 2019).Our results show, however, that while Hispanics may not explicitly care about a candidate's Spanish-language ability, and often rank it low on their list of important qualities in a candidate, Hispanics' evaluations of candidates are shaped by such efforts nonetheless.
In addition to examining how Hispanic Americans react to Spanish-language appeals, we also looked at the reactions of Anglo voters.Contrary to our expectations that Anglo voters would evaluate candidates making these appeals less favorably, Anglo respondents in Study 1 gave more positive evaluations to the Hispanic candidate and to candidates that spoke native-like Spanish.Results from Study 2 also go against our expectations, as we find no differences in candidate evaluations among Anglo respondents based on candidate ethnicity or Spanish-language appeals.Based on exploratory analyses of the moderating role of ideology and partisanship, we are inclined to attribute our original results to our Anglo sample being, on average, more liberal and more Democratic compared to Anglos in the general population (see Appendix N of the Supplementary Material).Previous work shows that self-identified liberals tend to score higher on the openness to experience personality trait, which could result in those individuals being more likely to want diverse representation.Individuals high in openness to experience as well as other psychological propellers of change (i.e., individual difference variables that facilitate one's craving for change) may seek further change in their environment being accustomed to demographic change (see, e.g., cultural inertia theory; Bellovary, Armenta, and Reyna 2020;Zárate, Reyna, and Alvarez 2019).In sum, in contrast with earlier studies that found negative effects for Spanish-language appeals among Anglos (Flores and Coppock 2018;Lavariega Monforti, Michelson, and Franco 2013), our study provides evidence that candidates can appeal to Hispanic voters by speaking Spanish without fear of alienating Anglo voters from their electoral coalition.
Judging by the increasing proportion of minority elected officials in the last two decades, the U.S. electorate appears to be more accepting of minority candidates than before.In Congress, for instance, while only 11% of individuals in the House and Senate identified as a racial or ethnic minority in 2001, this number had increased to 23% in 2021 (Schaeffer 2021).In addition, recent work shows that when minority candidates run, they are just as likely to win compared to white candidates (Juenke and Shah 2016).In the context of these developments and the census projections that the U.S. population will be majority-minority by 2045 (Frey 2018), politicians may be willing to try and appeal to these groups by speaking their ethnic language more often in the coming years.Our findings on the importance of language proficiency point to some crucial aspects of campaigning that warrant further attention from scholars and practitioners.First, the quality of a campaign appeal impacts their effectiveness.Politicians may attempt symbolic appeals that do not have the intended effect.For example, Ben Carson running a rap campaign advertisement or Gerald Ford eating a tamal with the corn husk on did not connect with their target audience.Not all appeals are executed equally and scholars should pay attention to how these appeals, such as speaking Spanish in a non-native accent, affect voter evaluations.Second, the language a candidate speaks, and the accent with which they speak it, conveys information to the voter.Individuals "make assumptions about the speaker on the basis of language markers that signal alliance to certain social groups, primarily those having to do with race, ethnicity and economic factors" (Lippi-Green 2012, 15).Considering the myriad of accents within the United States, scholars should look at how a candidate's accent affects their evaluation among different groups of voters.

CONCLUSION
The proportion of Hispanic voters is growing in the United States making their vote increasingly more sought-after.One method that politicians have used to try to sway Hispanic voters is speaking Spanish at public events.Previous work suggests that Hispanic support for Anglo and Hispanic candidates may be higher when candidates speak Spanish, but it relies on telling respondents about the candidate's language ability (Alamillo and Collingwood 2017;Lavariega Monforti, Michelson, and Franco 2013).Other work exposes respondents to real-life campaign ads but finds an effect for only one of two Anglo Republican candidates (Flores and Coppock 2018).We argue that variation in language proficiency affects how effective language-based appeals are as a signal of ability and willingness to represent Hispanic interests.Our experimental design allows us to vary the ethnicity of the candidate and their language proficiency while holding other characteristics such as prosody (i.e., the rhythm and intonation of speech) and substantive content constant.Results show that on average, Hispanics give more favorable evaluations for co-ethnic candidates and candidates who speak Spanish at a high level of proficiency regardless of ethnicity.However, when the Hispanic candidate has non-native-accented Spanish, their evaluation is similar to that of an Anglo candidate speaking English.This study suggests that Hispanic candidates who have limited Spanish proficiency may lose the positive evaluation that shared ethnicity would otherwise give them.
While we are confident in our ability to isolate the effect of Spanish-language appeals and ethnicity, we chose to focus on how Spanish words are pronounced as the source of variation in language proficiency rather than on what is said.Our hypothetical candidate pronounced the same grammatically correct sentences but with a different accent.In reality, individuals also show varying levels of proficiency with a language's grammar.Future work should explore how poor grammar impacts the effectiveness of language-based appeals whether they are delivered on the debate stage or through advertisements, social media messages, or campaign websites (Rodriguez 2019).In addition, Hispanic Americans have different national origins and corresponding Spanish-language variants.Scholars could study whether differences in word choice and pronunciation, or signaling a specific national identity, matter when making campaign appeals to the Hispanic community.Furthermore, it is worth noting that our findings are obtained in the context of a lowinformation environment.The effect of these Spanish-language appeals may be conditioned by the candidate's stances on policy issues of interest to the Hispanic community.Candidates who speak Spanish with a strong non-native accent but who emphasize issues that Hispanics care about may see an overall positive impact, whereas a candidate with native-like speech but who advocates for harmful policies might not see any benefit.
Considering that the United States is a multiracial society and in a couple of decades there will be no racial or ethnic majority, scholars should continue to explore how politicians campaign in a diverse society (Collingwood 2020).In addition to studying other languages and accents, future research should address how targeted appeals toward Hispanics affect the candidate evaluations among other racial minorities such as African Americans and Asian Americans, as well as how the effectiveness of targeted appeals toward different groups is conditioned by the quality of the appeal.Through this, we can develop a more complete understanding of how citizens form their impressions of candidates.

FIGURE 1 .
FIGURE 1. Mean Candidate Evaluation among Hispanics by Treatment Group

FIGURE
FIGURE 2. Predicted Candidate Evaluation among Hispanics by Treatment Group across Levels of Ethnic Identity (Study 1)

FIGURE 3 .
FIGURE 3. Mean Perceived Candidate Ability and Willingness among Hispanics by Treatment Group

FIGURE
FIGURE 4. Mean Candidate Evaluation among Anglos by Treatment Group 11 Table A.4 in Section A.2 of the Supplementary Material contains the demographic breakdown of our Prolific sample.See Table B.5 in Section B.3 of the Supplementary Material for full demographic characteristics and comparisons with population estimates among our Lucid sample.For information on the compensation of our sample, go to Section H of the Supplementary Material.
2. Predicted Candidate Evaluation among Hispanics by Treatment Group across Levels of Ethnic Identity (Study 1) Note: Larger values on the x-axis indicate greater importance of ethnicity to an individual's identity.Predicted values are plotted with 84% confidence intervals.Dependent variable ranges from 0 to 1. Full model results are shown in Model 1 of TableK.2 in Section K of the Supplementary Material.
4. Mean Candidate Evaluation among Anglos by Treatment Group Results from the Prolific sample are shown on the left; Lucid sample is shown on the right.Group means are plotted with 84% confidence intervals.Dependent variable ranges from 0 to 1. Results in tabular form are shown in Table I.3 in Appendix I of the Supplementary Material. Note:

TABLE 2 .
Effects of Language and Ethnicity on Candidate Evaluations among Anglos