Latinos in the United States and in Spain: the impact of ethnic group stereotypes on labour market outcomes

ABSTRACT This is the first harmonised correspondence study on the recruitment behaviour of employers in the US and in Spain. By comparing the call-back rates of Latino minority and majority group applicants, we measure the labour market discrimination that Latinos experience in these two national contexts. Due to their proximity in terms of culture and language, Latinos are expected to experience less discrimination in Spain than in the US. This is supported by our findings, as the level of discrimination against Latinos in the US is high and statistically significant, while we find no evidence of statistically significant discrimination against Latinos in Spain. In line with research on the intersection between ethnicity and gender in stereotyping, we find gender differences regarding discrimination in both countries, though in opposite directions. While Latino males are more discriminated than Latino females in the US, Latino females experience more discrimination than their male counterparts in Spain, who are not treated differently from Spanish native men. Our results indicate that ethnic group stereotypes are country-specific and different for males and females of the same ethnicity. Moreover, we find partial evidence that ethnic group stereotypes can be counteracted when favourable information on warmth and competence is provided.

market discrimination, as well as on the stereotypes that natives hold towards the main immigrant minorities, and on whether those differ by gender. While there are studies analysing Latinos' employment outcomes in Spain (e.g. Aysa-Lastra and Cachón 2012; Rodríguez-Planas and Nollenberger 2016; Stanek and Veira 2012), there is yet no experimental evidence on whether Latino men and women are discriminated against by employers.
In this paper we use data from a harmonised comparative correspondence study to compare the successes and failures of Latinos' job applications relative to those sent by majority group applicants. To do so, we send out comparable applications of fictitious applicants to job adverts, only varying the applicants' names and information about their country of origin in the cover letter (see Lancee 2021 in this issue). In addition, we identify the drivers of employers' discriminatory behaviours by varying the amount of information included in the application regarding two key dimensions of social judgement: warmth and competence (Fiske et al. 2002). Differences in call-back rates across groups reveal unequal treatment and serve as evidence of discrimination The contribution of our paper is threefold: first, we estimate the extent to which discrimination shapes the labour market positions of both Latino men and women in the US and Spain. Latinos are one of the main ethnic minorities in both countries, though there are differences in how they are perceived in terms of cultural foreignness, which is likely to produce different levels of discrimination. Second, we contribute to the scholarly debate on the intersection between ethnicity and gender, as we systematically compare the discrimination experienced by Latino men and women in a range of occupations that vary in skill level. Finally, we investigate the extent to which the prevalent stereotypes associated to Latino men and women can be counteracted by including positive information in the application about the warm personality and high competence of candidates.

The immigrant origin of Latinos in the US and Spain
Latin American migration flows to the US and Spain Despite their different migration histories, the US and Spain have become the most preferred destinations for Latino migrants (Connor and Massey 2010). While the US has long been an immigration country, Spain was primarily an emigration country for much of the twentieth century. Latin American migration flows to the US, particularly from neighbouring Mexico, increased exponentially after the abolition of national and ethnic quotas in 1965. Currently, 51.5 per cent of the foreign-born population in the US was born in a Latin American country, and 17.8 per cent of the American population self-identifies as Latino or Hispanic (ACS 2016). The increase in undocumented immigration since the 1960s has also been driven by Latin Americans, mostly Mexicans, who today represent over three quarters of undocumented residents in the US (Tienda and Sanchez 2013). The discourse on irregular migration has become more aggressive in recent years, which has manifested itself in a growing number of anti-immigration policy shifts, and the increased arrests and removals of unauthorised immigrants (Migration Policy Institute 2017). Recent investigations have shown that these legal changes, and the prominence of the irregular migration debate in the American media have increased anti-immigrant sentiments in general and negative attitudes towards Latinos in particular (Chávez 2013;Flores 2017), who have become an increasingly criminalised minority (Menjívar and Abrego 2012).
In contrast to the US, Spain first became an attractive destination for international migrants at the end of the 1990s (Cebolla and Gónzalez-Ferrer 2008). The growing demand for workers in the construction, care and cleaning sectors led to labour shortages that acted as a pull factor for international migrants during the first half of the 2000s (Reher and Requena 2009;Stanek and Veira 2012). The foreign-born population rapidly increased from 2.9 per cent in 1998-13.1 per cent in 2007 (Spanish Statistical Office 2011). Latin American inflows, particularly from Ecuador, were responsible for the extraordinary migration boom that Spain experienced until 2007 (Bertoli and Fernández-Huertas 2013;Cebolla-Boado and González-Ferrer 2013). Despite their lack of social networks, the costs of migrating to Spain were lower than to the US for many Latino migrants, not only due to the cultural/linguistic proximity, but also due to the visa waiver that applied to many South American countries (Bertoli 2010;Cebolla-Boado and González-Ferrer 2013).

Geographic and gender composition of Latinos in the US and Spain
Latin American migration flows to the US and Spain differ greatly in terms of their geographic and gender composition. Mexicans have always dominated the inflows to the US, though the share of Central Americans has increased during the last decades (Tienda and Sanchez 2013). The Mexican origin population currently comprises 63.2 per cent of Americans that self-identify as Latino or Hispanic (ACS 2016). In contrast, Mexican and Central American migrants have been a negligible minority among the Latin American inflows to Spain. In 2016, Latinos represented 30.7 per cent of the foreign-born population (Spanish Statistical Office 2016a), the main origin countries being Ecuador, Colombia, Argentina, Peru and Bolivia.
Another element that clearly differentiates Latin American inflows to Spain from those to the US is the phenomenon of female pioneer migration. While many Latino women came to Spain attracted by the high demand for workers in the care and cleaning sectors (Sánchez-Domínguez and Fahlén 2017), Mexican migration to the US has historically been male-dominated, with women more frequently arriving as reunified spouses (Donato, Wagner, and Patterson 2008).

Labour market participation and occupational segregation of Latinos in the US and Spain
The labour market position of the Latino population in both countries is not entirely comparable due to the different shares represented by the second and third generations and the length of stay of the first generation. Nonetheless, their labour market outcomes are likely to inform the stereotypes that employers may have about Latino men and women in the two countries.

Labour market participation
In the US, the labour force participation rate of Latinos in 2016 was 65.8 per cent and was thus slightly higher than that of whites, who had a participation rate of 62.9 per cent (BLS 2017). This percentage does not, however, reflect the existing gap in terms of job participation between Latino men and women; while 76 per cent of Latino men were employed or seeking employment in 2016, it was only 55.8 per cent for Latinas. This gender gap of 20 percentage points is, in fact, the highest among all racial groups in the US (BLS 2017). Moreover, Latinos in the US have the lowest educational attainment across all racial groups both in terms of secondary school diplomas and tertiary education (ACS 2016). This partially accounts for their over-representation in unskilled jobs and their higher unemployment rates.
The employment rate of Latino men in Spain was only slightly below that of Spanishborn men in 2008, before the start of the economic recession. This gap notably widened in the following years due to the high concentration of Latino men in the construction sector, which collapsed during the economic crisis (2008)(2009)(2010)(2011)(2012)(2013). Surprisingly, Latino women in Spain have always had a high employment participation rate, even higher than that of native women, and their unemployment did not rise as much as that of Latino men during the crisis (Fernández-Reino, Radl, and Ramos 2018). Latinos in Spain are also less likely to hold tertiary qualifications compared to natives, 25 per cent versus 36 per cent in 2014 (Spanish Statistical Office 2016a), but they tend to be overqualified for their jobs, especially the first generation (Fernández and Ortega 2008).

Occupational distribution
Latinos are occupationally segregated both in Spain and in the US, as they are concentrated in the blue collar sectors of the economy and they are more likely to work in unskilled jobs with non-standard working arrangements (Alonso-Villar, Gradín, and Del Río 2013;BLS 2017;Stanek and Veira 2012). In the US, Latino men predominantly work in the construction and maintenance sector (agricultural workers and construction labourers), while Latino women predominantly work in the service sector (maids, food preparation, and cleaning) (BLS 2017).
Latino women are more segregated than Latino men in the Spanish labour market, in part due to the clear gender divide across sectors of activity (Spanish Statistical Office 2016b). In Spain, Latinas have satisfied a growing demand for domestic workers (Da Roit, González Ferrer, and Moreno-Fuentes 2013) and they are highly concentrated in jobs related to care and cleaning, most of them in the shadow economy (Sánchez-Domínguez and Fahlén 2017). Latino males, in contrast, have found employment in the construction sector, which was one of the motors of the Spanish economy until 2008 (Stanek and Veira 2012).

Theoretical background and hypotheses
Stereotyping and the position of ethnic minorities Stereotypes are mainstream beliefs that are widely recognised by the members of a society, including by minority groups themselves. The fact that members of a society are aware of these stereotypes has implications for their attitudes and behaviours, regardless of whether they support them or not (Bargh, Chen, and Burrows 1996;Fiske 1998). Stereotypes about group-specific characteristics are used to justify emotions (i.e. prejudice) towards members of certain groups, and they are also used to legitimise hierarchical intergroup relations and discriminatory behaviours (Crandall, Eshleman, and O'Brien 2002;Kay and Jost 2003;Sidanius and Pratto 1999). Given the size and the visibility of the Latino community in both Spain and the US, employers in both countries are likely to hold distinct beliefs about the general characteristics of this group, as well as about their performance in the labour market.
Extensive research in the field of social cognition demonstrates that people categorise each other along the core dimensions of warmth and competence (Fiske et al. 2002). Warmth reflects the extent to which we like and trust a person or a group. A person or a group that is perceived as low in the warmth dimension is seen as threatening, distant and unreliable. Competence indicates the agency and capability attributed to a person or a group (Fiske 1998). A group that is stereotyped as incompetent is considered to be unskilled and of low ability and, thus, inferior to the majority group (Ridgeway 2014). In the US, Latinos tend to score low on both warmth and competence, receiving similar ratings to other low status groups such as poor people or undocumented migrants (Lee and Fiske 2006). In addition to warmth and competence, a recent study has added a separate dimension that is also relevant from a theoretical and an empirical perspective to understand how ethnic minorities are stereotyped: the degree of cultural foreignness, which indicates the extent to which a minority is seen as alien and culturally distant from the majority group (Zou and Cheryan 2017). The cultural foreignness dimension is particularly relevant when it comes to the relative position of ethnic minorities with an immigrant background. Existing research provides compelling evidence of the importance of this dimension, showing that the prejudice towards groups perceived as less culturally-distant is small or non-existent (Turner et al. 1987;Vanbeselaere 1996).

Stereotypes associated with Latinos in the US
In the US, the most common stereotypes associated with Latinos describe them as poor, uneducated, lazy, gangsters and of illegal status (Ghavami and Peplau 2013), all of which reflect the perception that they are less capable (competence dimension), more threatening (warmth dimension) and less American (cultural foreignness dimension) in comparison to whites. Prior studies have also shown that individuals with accents are perceived as less competent (Ruscher 2001), which is relevant considering the common stereotype that Latinos have poor English language skills (Dovidio et al. 2010;Huynh, Devos, and Smalarz 2011). Some of these negative associations also apply to other minorities such as African Americans, who are also heavily stereotyped, in ways similar to Latinos with regard to work ethic, intelligence and perseverance (Park et al. 2015) as well as criminal nature (Eberhardt et al. 2004). However, Latinos experience an additional penalty by also being stereotyped as aliens and illegal immigrants (Zou and Cheryan 2017). The passing of new laws targeting undocumented migrants at the state and federal level as well as the media's attention to this issue further reinforced the image of Latinos as the perpetual foreigners (Chávez 2013;Jiménez 2008).

Stereotypes associated with Latinos in Spain
Compared to the US, we know less about the stereotyping and prejudice towards ethnic minorities in Spain, both in terms of theory and empirics. To our knowledge, the only exception is a survey from 2007 (CIS 2007), which specifically asks respondents to mention the three ethnic minorities who have the hardest time integrating into Spanish society, as well as the groups that will more successfully integrate into society. Respondents reported more trust in Ecuadorians and/or South Americans than in Romanians, Moroccans or Chinese, and perceive Latinos to be easier to integrate than the other three minorities. In addition, there is some empirical evidence from non-representative samples showing that Latinos are perceived as more similar to natives and more likeable than immigrants from other countries (Rodríguez-Pérez et al. 2011). López-Rodríguez et al. (2014) also show that Ecuadorians, who are the major national group among Latinos in Spain, are perceived as more capable (competence dimension), more alike (cultural foreignness dimension), and less threatening (warmth dimension) than other immigrant minorities such as Moroccans or Romanians.
Despite the limited amount of data and the few studies on stereotyping, the evidence suggests that Latinos experience lower levels of discrimination than other stereotyped immigrant groups because of their cultural closeness in terms of shared language and historical ties with the Spanish population.
These considerations lead us to our first hypothesis: H1: Due to the perceived lower cultural foreignness of Latinos in Spain than in the US, Latinos in the US experience higher levels of employment discrimination than Latinos in Spain.

Stereotyping and the intersection between ethnicity and gender
The literature on the intersectionality of different axes of inequality stresses the fact that individual identities are constituted of multiple dimensions, such as gender, class and ethnicity, all of which need to be considered when theorising about social inequalities (Hancock 2007;Hurtado and Sinha 2008;Salgado de Snyder, Cervantes, and Padilla 1990). For example, there is a large body of literature arguing that minority women hold a subordinate position in relation to majority group women and minority men (Almquist 1975;Landrine et al. 1995;Sidanius and Pratto 1999). The existence of a gender divide in minority group stereotyping should also be reflected in employers' attitudes towards males and females of the same ethnic minority. These attitudes are also likely to be occupation-specific, depending on how the stereotypes associated with each gender and ethnicity match the qualities that employers' value in each occupation.

The stereotype of Latino men as threatening and aggressive
The stereotypes attributed to Latinos in the US have been negative for both genders, though those associated with Latino men are generally more negative. Ghavami and Peplau (2013) find that Latino men are perceived as aggressive and threatening as well as sexist and 'macho'. This is in line with previous research showing that Latino boys are often stereotyped by popular media as gang members, dangerous and delinquent (López 2003). Furthermore, there seems to be a relationship between traditional masculine gender roles and ethnic belonging among Latino men, who tend to use their masculinity to show their status and instil respect through a tough image (Abreu et al. 2000). Latino men are also more frequently identified with illegal immigration than Latino women, largely because Mexican migration flows to the US have historically been male dominated (Donato, Wagner, and Patterson 2008). These stereotypes about Latino men can arguably be considered more negative at the workplace than those associated to Latinas, which are, overall, less problematic in terms of their employability, e.g. docile, family-oriented, and unskilled. In consequence, we expect Latino men to be equally penalised across both skilled and unskilled occupations.
In the Spanish case, the scarcity of data on ethnic minority stereotyping prevents us from having clear expectations about the gender-specific stereotypes attributed to Latino men and women. In consequence, we examine whether employers' behaviour towards Latino men in Spain is driven by the same stereotypes as in the US.
To test whether there is a Latino male penalty driven by employers' perception of them as more threatening and aggressive than white men, half of the job applications included positive information about the candidates' warm personality in the CV and cover letter. These details are expected to soften the influence of the stereotype and thereby, diminish the gap in positive call-backs with respect to the majority group more strongly for Latino men than for Latino women.
From the previous arguments, we derive the following hypotheses: H2: Latino men experience more discrimination than Latino women because of their stereotyping as threatening and aggressive, which is more negative for Latino men's employability than the stereotyping of Latino women is for theirs.
H3a: If employers' discriminatory behaviours against Latino men are driven by their stereotyping as threatening and aggressive, there should be a decrease in the level of discrimination when information on warmth is included in the application. The effect of including this information should be smaller or non-existent for Latino women.
H3b: The level of discrimination against Latino men does not vary across occupations, as being aggressive and threatening is negative in low, medium, and high-skilled occupations alike.
The stereotype of Latinas as docile, incompetent and not career-oriented Latinas in the US have the lowest female labour force participation rate of all racial groups. In contrast to white women, they are also highly concentrated in 'typically feminine' lowskilled and low-status jobs. This is reflected in their weekly median earnings, which are significantly lower than those for both white women and Latino males with the same level of education (BLS 2017). Though the common stereotypes associated with Latino women in the US make them appear as less aggressive or threatening than Latino men, they are in turn perceived as unskilled, traditional and less career-oriented than white women (Williams, Alvarez, and Andrade Hauck 2002). Despite being frequently hypersexualised in the media, they are also portrayed in traditional female occupations, such as maids or cooks (Ghavami and Peplau 2013), in which they are in fact overrepresented. A study by Williams, Alvarez, and Andrade Hauck (2002) shows how Latino women must fight the prevalent stereotype of the 'domesticated Latina', i.e. family-oriented and with low educational and occupational expectations. These stereotypes are also confirmed by López and Chesney-Lind (2014), who find that school counsellors working with young Latinas described them as 'passive' 'docile' and 'subservient', all of which are in agreement with a traditional gender ideology. Latino women in Spain have a high employment participation rate, though they also stand out for being overrepresented as domestic workers in the cleaning and care sectors relative to Spanish native women. Moreover, Latino women face difficulties finding employment outside of this sector and tend to get stuck in low-skilled jobs for which they are overqualified (Fernández-Reino, Radl, and Ramos 2018). In 2011, 34 per cent of Ecuadorian women (the largest nationality among Latinas) worked as cleaners and helpers compared to only 9 per cent of Spanish women (Spanish Statistical Office 2011). In contrast, Ecuadorian men were not as overrepresented as Ecuadorian women in low skilled occupations in the construction sector, where many of them were employed (22 per cent of Ecuadorian men versus 10 per cent of Spanish men). The high presence of Latinas in such typically female and elementary occupations may have generated negative stereotypes about their abilities and competences for performing other types of jobs.
In contrast to Latino male applicants, we expect to find variation in the levels of discrimination experienced by Latinas across occupations. Because of their stereotyping as domestic workers, Latinas are likely to face more discrimination when applying to medium and high-skilled jobs outside of typically feminine occupations in the care and cleaning sectors. The inclusion of a competence indicator in the application may therefore diminish their discrimination by counteracting their stereotyping as unskilled traditional women who lack career ambitions. Although Latino men in the US are also stereotyped as incompetent, the negative attributes that are most frequently mentioned are their aggressive and threatening behaviours (Ghavami and Peplau 2013). From that it follows that the inclusion of a competence indicator should have a smaller effect for Latino men than for Latino women.
H4a: Latino women experience more discrimination when applying for jobs in medium-and high-skilled occupations than in low-skilled occupations due to their stereotyping as incompetent and not career-oriented.
H4b: If employers' discrimination against Latino women is driven by their stereotype as incompetent and not career-oriented, this should decrease when information on competence is included in the application. For Latino men, the effect of including this information should be, in contrast, smaller.

Data
We used data from a harmonised correspondence study with an unpaired design (one application per vacancy and employer) that was conducted between June 2016 and June 2017 in the US, and between November 2016 and May 2018 in Spain. 1 All vacancies were collected via common job searching websites in Spain and in the US. In total we send out 1,547 applications in Spain and 804 applications in the US.
The educational qualifications and labour market experience of candidates, both obtained in the destination country, were kept constant for each occupational profile to ensure that all applicants were equally qualified for the advertised position. In addition, all candidates specified that they hold American or Spanish citizenship. This is particularly relevant for the American case, where suspicions about Latinos' legal status are far more common than in Spain (Mukherjee, Molina, and Adams 2012). We therefore ensured that there were no doubts about applicants' citizenship, which could have led to employers' immediate rejection of an application. 2 The Spanish field experiment also included a photo in the CV, which is a common practice in the Spanish labour market. 3

Ethnicity and names
In the Spanish study, most Latino applicants had Ecuadorian origins, as they represent the largest minority of Latin American descent in Spain (Spanish Statistical Office 2016a), though Mexican 4 and Dominican applicants were also included. In the US study, applicants were of Mexican, Salvadorian, Cuban or Puerto Rican descent, since they are the largest groups among Latinos (Migration Policy Institute 2016). We chose country distinctive names to signal applicants' ethnic origin (see Appendix Table 1), making sure that they were all neutral in terms of social class and religion. 5

Treatments: warmth and competence
In addition to ethnicity and gender, there were other characteristics that varied randomly across applications and changed the amount of information given to the employer: first, applicants' past job performance and academic achievement 6 (competence); and second, applicants' friendly and cooperative personality (warmth). The paragraph signalling applicants' warm personality appeared in the cover letter and reads as follows: My colleagues and friends think I am a pleasant, trustworthy and warm person who gets along with people from all walks of life. I am a team player who values a positive work environment and that is why I am always friendly and attentive to other people's needs.
The paragraph that signalled applicants' high productivity appeared in the cover letter and reads as follows: I am a fast learner who can take on a wide variety of tasks and rise to new challenges. My current employer has been extremely satisfied with my work and has given me more responsibility since I started. Since month/year, I have served as the orientation leader for all new staff members.
Both the competence and the academic achievement indicators were randomly and independently assigned to half of the applications. However, we only show three possible combinations in our results, namely: low competence (no indicators of academic achievement or high work performance included); medium competence (either indicator of academic achievements or of high work performance included); and high competence (both indicators of academic achievements and of high work performance included).

Occupations
We sent applications to six different occupations, two of which had lower entrance and skill requirements (cook, store assistant), two had medium entrance and skill requirements (payroll clerk/secretary, receptionist), and two had higher entrance and skill requirements (sales representative and IT developer). Since we used a harmonised design, we specifically chose occupations that were comparable across countries and had enough variation both in academic and skills requirements.

Outcome variable and estimation strategy
Employers' call-backs are the outcome variable. We coded an invitation to a job interview or other signs of interest in the applicant with the value of 1, and no reply or a rejection of the application with the value of 0. Since all applicants were equally qualified for the job offer, a significant (p < 0.05) differential in employers' positive call-back to Latinos in comparison to the majority group is interpreted as evidence of discrimination. The dependent variable is dichotomous, and we used a linear probability model, which allowed us to estimate the model with an OLS regression. 7

Results
The sample consists of 2,351 applications: 1,547 in Spain and 804 in the US. The average positive call-back rate is similar in both countries: 20 per cent in Spain and 22 per cent in the US. Appendix Table 2 gives a detailed overview of the sample and treatment distribution in Spain and in the US.

Discrimination against Latinos in the US and in Spain
We find evidence in favour of our first hypothesis (H1). Figure 1 shows that the call-back rate for Latinos in the US is ten percentage points lower than that of white applicants and statistically significant. In contrast, we find no evidence of statistically significant discrimination against Latinos in Spain. Their positive call-back rate is four percentage points lower than that of Spanish natives, though not statistically significant different from zero ( Figure 1).

Do Latino men experience more discrimination than Latino women?
We expected that the stereotyping of Latino males as threatening and aggressive would have a more negative impact on their employability than the stereotypes associated to Latinas. Figure 1 shows that this is true for the US, where compared to white men, Latino men are significantly discriminated against. Their call-back rate is 13 percentage points below that of white men, and statistically significant, while the call-back rate for Latino women in the US is only seven percentage points lower than that of white women and not statistically significantly different from zero. Thus, H2 is confirmed for the US case.
In contrast, Latino women in Spain experience a higher penalty than Latino men when we compare them to majority group candidates of the same gender (the negative interaction between Latino and female is statistically significant). The call-back rate for Latino women in Spain is twelve percentage points below that of Spanish native women and statistically significant. Counter to our expectations, Latino men are not treated differently from majority group men in the Spanish labour market. In fact, their call-back rate is three percentage points above that of Spanish native men, though not statistically significant. Thus, H2 is not confirmed for the Spanish case.

Does the inclusion of a warmth indicator diminish the discrimination against Latino men?
In a second step, we examine whether the discrimination against Latino men is driven by the prevalent stereotypes as more threatening and aggressive than majority group men. Figure 2 shows that the penalty for Latino men in the US decreases from −0.24 to −0.01 and becomes statistically insignificant when we include positive information about the applicants' warm personality in the job application. The positive interaction between Latino male and the warmth treatment is statistically significant. Though including the warmth treatment for Latino women in the US reduces the negative coefficient from −0.10 to −0.02, it needs to be noted that the initial difference in call-back rates is not statistically significant ( Figure 2). Nevertheless, the positive impact of the warmth indicator in the application documents for Latinos in the US leads to lower gaps in call-back rates for the overall sample, and as a result Latinos in the US do not experience higher levels of employment discrimination than Latinos in Spain anymore.
In Spain, adding a warmth indicator has a negative effect for both Latino men and women compared to Spanish native men and women. In fact, the overall call-back rate for Latinos decreases from −0.02 to −0.07 and becomes statistically significant when the warmth indicator is included (Figure 2).
Our results only confirm H3a for the US and are in accordance with prior research showing that members of the majority group see Latino men as more threatening and less warm than white men. These negative perceptions are not attributed to Latino women in the US to the same degree and, thus, the penalty experienced by Latino men in the US is higher. However, the results for the Spanish sample not only show that Latino women experience more discrimination than Latino men, but also that indicating warmth in the application is penalised when the applicants are Latinos (and ignored when the applicants are Spanish natives). In this regard, we cannot confirm that Latino men are discriminated against in Spain or that they are stereotyped as more threatening than native Spanish men.

Does the level of discrimination vary across occupations for Latino men?
In Spain, Latino men have higher call-back rates than Spanish native men when they apply for payroll clerks/secretaries (+0.14) and IT developers (+0.17), though due to the low number of observations, none of the coefficients are statistically significant. Across all other occupations, the differences are similar and comparably low, varying from −0.01 for cooks to −0.03 for receptionists and store assistants. In sum, we find no statistically significant variations in call-back rates across occupations for Latino men, thus H3b is not confirmed in the Spanish case ( Figure 3).
In the US, there is more variation in the level of discrimination across occupations, as Latino men experience a significantly higher level of discrimination when they apply for vacancies as cooks (−0.32) and IT developers (−0.21), followed by store assistants, sales representatives and payroll clerks/secretaries. We find the lowest level of discrimination against Latino men when they apply as receptionists (−0.02) (Figure 4). Regarding H3b, we therefore cannot confirm that the level of discrimination against Latino men is similar across occupations. Moreover, the mixed results across countries and occupations indicate that there is no clear pattern regarding the skill level of the occupations.

Does the level of discrimination vary across occupations for Latino women?
We hypothesised that Latino women would be more discriminated against than Latino men in medium-and high-skilled occupations due to their high level of occupational segregation in female dominated occupations and their stereotyping as domestic workers (H4a). Regarding call-back rates by gender and occupation in Spain (Figure 3), we find that the penalty for Latino women is significantly higher when they apply as IT developers (−0.25), sales representatives (−0.34), payroll clerks/secretaries (−0.23), and store assistants (−0.17). In contrast, Latino women experience no statistically significant discrimination when applying for jobs as cooks or receptionists.
The results shown in Figure 3 confirm H4a for the Spanish case, as Latino female applicants mainly experience discrimination when applying to non-typically female dominated and skilled occupations (e.g. IT developers, sales representatives, payroll clerks/secretaries). Figure 4 shows that the results for the US are not as clear cut. Here we find that Latino women do not have significantly lower call-back rates across occupations, though the coefficients show that Latino women have similar call-back rates to white women when they apply for female dominated vacancies as cooks or payroll clerks/secretaries, and much lower call-back rates when they apply as IT developers (−0.17), which is a male dominated occupation. Therefore, we cannot confirm H4a in the US.
Does the inclusion of a competence indicator diminish the discrimination against Latino women?
In Spain, adding information on job productivity and good academic performance (high competence) slightly improves the call-back rates of Latinas relative to Spanish native women, though their penalty remains statistically significant. The positive effect of indicating high competence for Latino men in Spain is also small and not statistically significant ( Figure 5). In the US, adding information on high competence compared to no information about competence leads to lower call-back rates for Latino women. What is even more surprising is that the call-back rate for Latino women in the US is lowest and statistically significant when they only indicate either good grades or high work performance (medium competence) ( Figure 5). This stands in contrast to the results for Latino men in the US, as for this group the negative coefficient of −0.17 (when no competence indicator is included) drops to −0.15 when they indicate a medium level of competence and further drops to −0.03 when they indicate a high level of competence ( Figure 5). Therefore, we find mixed evidence for H4b, because the inclusion of a high competence indicator only seems to benefit Latino women in Spain, albeit the effect is very small, and their penalty remains statistically significant. In contrast, including information on competence and academic achievements only helps Latino men in the US, and has, if at all, a negative effect for Latino women.
Does the inclusion of a both a competence and a warmth indicator diminish discrimination against Latinos?
The combination of both treatments ( Figure 6) confirms the findings described earlier, when the treatments were analysed separately in Figures 2 and 5. In the absence of information on competence, indicating warmth has a statistically significant negative effect in Spain, particularly for Latino females. In contrast, including information about warmth has a strong and positive effect for both Latino males and females in the US, especially when information about competence is absent from the resume. Although the competence treatment alone has a positive effect on the overall call-back rates in both countries, the discrimination against Latino women in Spain and Latino men in the US remains statistically significant. Overall, the combination of both treatments has virtually no effect in Spain, while in the US the discrimination rate decreases and becomes statistically insignificant when both treatments are included. However, in the US this positive effect is mainly driven by the warmth indicator.

Robustness checks
In our experimental design we have considered only one majority group per country, which was signalled by the applicants' names. However, given the current strength of the secessionist movement in Catalonia, we were concerned that the native reference group in this region could have become the population with Catalan ethnicity (i.e. with Catalan-sounding names and surnames). The data analysis for the complete Spanish sample indicates that using Spanish/Castilian names could potentially bias the discrimination estimates downwards, as applicants with Spanish/Castilian names receive significantly lower call-back rates in Catalonia than in the other regions of Spain. Since the Catalan vacancies account for roughly 30 percent of the Spanish sample, this could pose a serious bias in our data. We therefore estimated all models excluding the Catalonian subsample and find no substantial changes in our estimates (Appendix Figures 1-5).

Conclusion
This is the first harmonised field experiment comparing the levels of employment discrimination against Latino men and women in two heterogeneous national settings, and the first field experiment on ethnic employment discrimination against an immigrant minority in Spain. Thus, our study builds on and extends previous research on ethnic discrimination in several important ways: First, we estimated the extent to which employers discriminate against Latinos in Spain and the US, two countries where Latinos are one of the largest ethnic minorities. In line with previous studies, we confirmed that Latinos are highly discriminated against in the US labour market. In Spain, however, the overall level of employment discrimination against Latinos appears to be low. These contrasting findings are likely to be related to two phenomena: first, Latinos are perceived as more culturally distant in the US than in Spain. And second, Latinos are often perceived as undocumented migrants in the US, while this is not the case in Spain. In the US, the narrative that has emerged amidst ongoing debates over Mexican immigration has frequently portrayed Mexican immigrants (and Latinos in general) as a threat to American society due to their perceived unwillingness or lack of capacity to integrate (Chávez 2013). Even though the racial divide has historically shaped majority-minority group relations in the US, the issue of undocumented migration, which particularly affects Latinos, has increasingly structured the political debate in the past years. This differs from the situation in Spain, where the tension between central and peripheral nationalism has dominated the politics over identity issues and continued to do so even during the last economic recession (Alonso and Rovira Kaltwasser 2015). Consequently, the native-immigrant divide has been less salient in the Spanish media and political debate, which has mostly focused on the traditional centre-periphery cleavage (Pardos-Prado 2012).
Second, our study provides compelling evidence of an intersection between ethnicity and gender regarding stereotyping in both countries. In the US, discrimination against Latinos is, above all, a male phenomenon, driven by the portrayal of Latino men as 'threatening foreigners'. The Trump era may have reinforced these negative stereotypes of Latinos, particularly those attributed to Latino men, e.g. during Trump's inauguration speech in 2015 Mexican immigrant males were depicted as drug dealers, criminals and rapists. This stands in stark contrast to the Spanish case, where only Latino women face discrimination, mainly due to employers' preference for Spanish native women in medium-and high-skilled occupations.
Third, the discrimination against Latino men in the US disappears when they include information about their warm and friendly personality in the resume, which underscores that the discrimination they experience is mainly driven by employers' stereotyping of Latino men as threatening and aggressive. Surprisingly, and in contrast to the US, including information about the applicants' warm and friendly personality leads to lower callback rates for both Latino men and women in Spain.
Fourth, although Latino women in Spain have a high employment participation rate, they are also highly overrepresented as domestic workers in the cleaning and care sectors compared to native women. We believe that the discrimination against Latino women in Spain is shaped to a large extent by their structural position in the labour market. Employers may stereotype Latinas as domestic workers, which has a negative impact on the labour market prospects of high skilled Latino women. We have examined whether the penalty against Latinas could be, in fact, attributed to their stereotyping as unskilled workers who are mainly suited for female dominated elementary occupations. We do not find strong evidence showing that this is the case, as the positive effect of indicating high levels of competence in their application documents is relatively small. This effect does not, however, occur in the US. There, Latino women do not benefit at all from indicating competence despite being also overrepresented in low-skilled and female dominated occupations. In the US context, it seems that positive information on both warmth and competence benefits Latino men the most, while Latino women have no additional bonus when they present themselves in ways that are inconsistent with their stereotyping as incompetent and not career-oriented. However, it is important to note that the overall discrimination rate against Latino women in the US was not statistically significant in the first place.
Fifth, our results indicate that stereotypes are country and gender specific, which supports prior research on the intersectionality of ethnic stereotyping. Moreover, we show that the prevailing gender specific stereotypes against Latino men and women in each national context can be partially counteracted when favourable information about the applicant's warm personality and/or competence is stressed in the application.
One shortcoming of our study is that we cannot explain why employers reward stereotype inconsistent information for Latino men in the US and (to a smaller extent) also for Latino women in Spain, while they appear to ignore stereotype inconsistent information for Latino women in the US. There is evidence from the US showing that women who demonstrate themselves as successful and competent suffer from bias in evaluative judgement, as this is inconsistent with their stereotype as warm and communal (Heilman et al. 2004). Future research is needed to see if this finding also holds true for ethnic minority women, and whether there are differences between male and female dominated occupations.