Previous Research and Hypotheses

One of the critical issues regarding immigrant integration is differential treatment based on race. In this regard, some ethnic penalty studies employed second-generation non-white immigrants and compared their integration outcomes with respect to natives or Westerners in order to control for the legitimate penalty or standardization process of migration. This is because, unlike with first or 1.5 generations (children of the first generation who are not born in the destination countries), the second generation would have similar levels of human capital such as language and education in the destination countries compared with native peers. They also do not have negative experiences that could stem from the migration process, particularly found in the case of refugees (Bakker et al. Reference Bakker, Dagevos and Engbersen2017; Crul and Vermeulen Reference Crul and Vermeulen2003; Gracia et al. Reference Gracia, Vázquez-Quesada and Van de Werfhorst2016; Heath and Cheung Reference Heath and Cheung2007). In a related vein, EU/Western immigrants showed higher performance regarding job quality as they were employed more in R&D, IT, real estate, and similar areas, while non-white ethnic groups were investigated to experience a lower level of employment rate and unskilled service or manufacturing positions (Khattab and Johnston Reference Khattab and Johnston2013, Reference Khattab and Johnston2015; Kogan Reference Kogan2007).

In terms of low job quality outcomes, ethnic penalties could become more consolidated over time. From the perspective of career trajectories, if employees enter an unskilled position at an earlier career stage, future employers may receive a negative impression regarding the productivity or returns of employees (Lawrence Reference Lawrence1984; Verbruggen et al. Reference Verbruggen2015). Therefore, the result of ethnic penalty regarding prevalent low job quality when controlling education levels could lead to a chronic low job status, so it needs to be re-highlighted as the systematic penalty “trap” for non-white migrants. In this regard, a study in the Netherlands revealed similar results in which first-generation migrants mostly experienced a skill shortage, while negative skill mismatch was prevalently found for the second generation of non-white migrants (Belfi et al. Reference Belfi2021).

Nevertheless, in terms of the migrant penalty research which combines ethnicities as “non-white” immigrants, the ethnic difference in relation to racial or ethnic discrimination could often be ignored under the binary categorization of natives and non-white or third-country migrants (Ballarino and Panichella Reference Ballarino and Panichella2013, Reference Ballarino and Panichella2017; Belfi et al. Reference Belfi2021; Kogan Reference Kogan2007; Lee Reference Lee2022). Therefore, beyond otherness as non-white, the condition of racial or ethnic difference is required within migrant integration analyses to be able to discern differentiated treatment depending on race (Blank et al. Reference Blank, Dabady and Citro2004; Siebers and Dennissen Reference Siebers and Dennissen2015, Zschirnt and Ruedin Reference Zschirnt and Ruedin2016). In line with this, there are several studies that analyzed ethnicity differences in labor market outcomes.

After controlling human capital, immigrants from Europe, North America, India, Pakistan, and Bangladesh in the UK were compared in terms of employment rates. As expected, white immigrants from Western countries were less penalized, indicating 10% higher employment rates with unemployment rates similar to those of white British. However, among non-white ethnicities, Indians were revealed to be a lesser-penalized ethnicity compared with the other ethnic groups (Modood and Khattab Reference Modood and Khattab2016). Similarly, Heath and Cheung (Reference Heath and Cheung2006) found an ethnic penalty with respect to natives and white migrants, particularly for African, Caribbean, Pakistani, and Bangladeshi migrants in terms of unemployment, income, and occupations. Furthermore, they importantly uncovered how ethnic penalty continued for second generations, which turned out to be at a similar level compared with the first generation in the UK. Gracia et al. (Reference Gracia, Vázquez-Quesada and Van de Werfhorst2016) found Moroccan and Turkish’ immigrants were significantly penalized with respect to native Dutch in employment and job quality, despite equivalent education and skill possession. Furthermore, cultural influences on labor market integration were uncovered through the low performance of Turkish females when compared with Moroccan females due to the stricter conservative manner of Turkish culture concerning women.

Beyond single-country cases, 11 Western European countries’ ethnic wage penalty was investigated using four different ethnicities. Regardless of gender, a large wage penalty was found in Southern Europe since migrants’ positive selection occurs more in the UK, Ireland, and Continental European countries due to higher income thresholds to obtain a visa, unlike Southern Europe in which migrants usually entered without jobs. Besides, no greater advantage for Eastern Europeans compared with African and Middle Easterners, Asians, and South Americans was uncovered (Cantalini et al. Reference Cantalini, Guetto and Panichella2023). Although meaningful results were found in the study regarding ethnic wage penalty under the cross-country perspective, the important distinction between MENA and sub-Saharan Africans was not achieved by combining both ethnicities. When discussing racialization in the labor market, the distinction between MENA and SubAf would be imperative since they not only have different racial markers but also predominantly embed different religious cultures largely reflecting Islam and Christianity, respectively. As these racial characteristics including color, religion, and economic concerns are bound to be the fundamental factors for ethnic penalties (Goldberg Reference Goldberg2009; Heath and Cheung Reference Heath and Cheung2006), the difference between the two ethnic groups should be recognized in large-scale comparative studies.

Meanwhile, correspondence experiment and meta-analysis studies uncovered employers’ hiring discrimination with regard to ethnic migrants’ entrance or access to labor markets by measuring callback rates with detailed ethnicity distinctions. In terms of the cross-country perspective, the GEMM (Growth, Equal Opportunities, Migration and Markets) project was conducted in five European countries (the UK, Spain, Germany, Norway, and the Netherlands) by accounting for socioeconomic backgrounds, including religion, through treatment with eight different phenotypes (see Lancee et al. Reference Lancee2019). According to the findings, Black and Middle Eastern men were penalized more in receiving callbacks in male-typed jobs, which is consistent with previous single-county case research (Arai et al. Reference Arai, Bursell and Nekby2016; Bursell Reference Bursell2014; Dahl and Krog Reference Dahl and Krog2018; Derous et al. Reference Derous, Ryan and Serlie2015). Asian men were less discriminated against in female-typed jobs due to their perceived nonthreatening and feminine character, whereas Black women regarded as dominant were also more penalized compared with White and Asian women who are perceived as more competent (Di Stasio and Larsen Reference Di Stasio and Larsen2020).

On the other hand, a meta-analysis that synthesizes the correspondence experimental studies over time and investigates the consistency of results between countries uncovered the tendencies of ethnicity-related hiring discrimination. The discrimination levels hardly changed over time across Europe and North America, while the penalty toward MENA increased throughout two decades (Quillian and Lee Reference Quillian and Lee2023). Although some European countries’ discrimination decreased compared with the United States, such as France, due to notably higher initial discrimination, age discrimination intersecting with ethnicity is consistently high in European countries (Lippens et al. Reference Lippens, Vermeiren and Baert2023; Quillian et al. Reference Quillian2017). Despite the significant findings from correspondence studies securing causality and meta-analysis tracing ethnic penalty trends, they have limitations such as the omission of post-callback situations and publication bias. This is because callbacks from the initial stage of the employment process do not necessarily mean the successful employment of ethnicities, so while it is useful to discern racialization in recruiting processes, it could not provide the final reality of ethnic penalty in the labor market. Moreover, publication bias, indicating a higher probability of not publishing non-statistically significant results in experiments, can be expected despite the careful approach of meta-analyses (Quillian and Midtbøen Reference Quillian and Midtbøen2021).

Therefore, the current research undertakes a comparative study regarding five ethnicities’ labor market outcomes with respect to natives across 16 Western European countries. This overcomes the confinement of previous research and uncovers ethnic penalty in employment and job quality alongside the distinction between MENA and sub-Saharan Africans. Accordingly, this study developed a framework that helps analyze how ethnic penalty in the labor market can be orientated based on labor demand (native) and supply (ethnicity) sides, respectively (see Figure 1).

Figure 1.

A framework regarding ethnic penalty processes in the labor market

Source: Author’s elaboration.

Theories regarding ethnic penalty processes are presented in relation to the hypotheses and organized in two ways (Figure 1). First, how culture-based racial or ethnic discrimination could happen in Western European countries is explored based on the concept of homogeneity and secularism, which stems from natives’ particular perspective on migrants in relation to the perceived illegitimacy of destination countries’ sociocultural backgrounds. These theories are expected to be particularly associated with employment levels among the five ethnicities, which can be compared with the findings on hiring discrimination. Second, the socioeconomic backgrounds of the ethnicities are suggested as another significant factor affecting different labor market outcomes. This is explained by mixed embeddedness (Kloosterman et al. Reference Kloosterman, Van der Leun and Rath1999) and locational inequality (Milanovic Reference Milanovic2016) in relation to employment and job quality, respectively.

Methodology

The 16 Western European case countries included Austria, Belgium, Switzerland, Germany, Denmark, Spain, Finland, France, Greece, Ireland, Italy, Portugal, the Netherlands, Norway, Sweden, and the United Kingdom. In order to test the four hypotheses, the analysis proceeded in two ways. First, ethnic penalty regarding employment and job quality at the European level was analyzed alongside the total ethnicity figures of the 16 Western European countries according to gender and labor market outcomes in employment and job quality. Second, ethnic penalty at the country level proceeded according to each country, gender, and labor market outcomes.

A linear probability model (LPM) was employed to compare the labor market outcomes between natives and the five ethnicities. Dichotomous dependent variables need to be used in LPM so that employment and job quality are distinguished through binary conditions such as employed or unemployed and skilled or unskilled, respectively. Probit or logit models are also binary variable methods by presenting odd ratios rather than probabilities like LPM, which is much more straightforward to interpret the results and regarded to be realistic (Greene Reference Greene2011; Holm et al. Reference Holm, Ejrnæs and Karlson2015; Long Reference Long2009). This investigation drew upon realistic measurements, such as the EU Labor Force Survey (LFS) (registered number: RPP 373/2020-LFS), which is survey data reflecting the current work status rather than projections. Hence, LPM is employed to convey ethnic penalty with respect to natives.

The binary data was organized into two dependent variables (employment and job quality) by setting “1” as employed/skilled and “0” as unemployed (including inactivity)/unskilled, respectively. In terms of job quality, the International Standard Classification of Occupations (ISCO) from the International Labour Organization (ILO), which distinguishes four occupational levels based on nine occupational codes, was used to distinguish skilled and unskilled jobs in a binary variable. Specifically, unskilled positions consist of the bottom two levels, the first (elementary occupations (9)) and second (plant, craft, skilled agriculture, service workers (5–8)), while skilled jobs are comprised of the two higher levels including technicians (3) for the third level and professionals (1 and 2) for the fourth level alongside the top-skilled jobs such as doctors and managers, excluding the military (0). To note, clerical support staff (4) belongs to the second lowest level alongside codes 5–8. However, it pertains to public administration and bankers, which cannot be regarded as unskilled jobs, so it was classified as among skilled jobs.

With regard to the independent variable, migrants were divided into five ethnicities coded 1–5 according to country of birth based on regions such as Eastern Europe (EE), the Middle East and North Africa (MENA), sub-Saharan Africa (SubAf), Asia, and South America (SA).Footnote ¹ As ethnic penalty is measured with respect to natives, natives are coded as 0 for the reference category. Immigrants from the EU-15 and European Free Trade Association (EFTA), as well as North America including Oceania, were treated as a residual category that has a contribution but without the results being shown in the statistical analysis since they have the same status as natives within the European labor market (see Ballarino and Panichella Reference Ballarino and Panichella2013, Reference Ballarino and Panichella2017; Khattab and Johnston Reference Khattab and Johnston2013, Reference Khattab and Johnston2015; Kogan Reference Kogan2007; Modood and Khattab Reference Modood and Khattab2016; Zwysen et al. Reference Zwysen, Di Stasio and Heath2021).

There were four control variables in this study, including education, age, marriage status, and year terms. Education was divided into three levels based on ISCED-97: 1 (up to lower secondary), 2 (upper secondary), and 3 (tertiary). Three age blocks were presented from 25–34 to 35–44 and 45–66. Marriage status was categorized as divorced/widowed, single, and married. The year term was employed to control for specific time effects because the employed LFS data comprised an 11-year timespan from 2005 to 2015 in 16 countries. Accordingly, Models 1 and 2, which show ethnic penalty at the European level, included the $COUNTRY\# YR$ term to control for the 16 countries’ different contingencies interlinked with years. Meanwhile, Models 3 and 4, which identify ethnic penalty at the country level, included only the $YR$ term as it was separately analyzed according to each country case (full regression tables can be found in appendix Tables 10–13 according to gender and labor market outcomes).

Model Specification

Model 1. Male ethnic penalty (Europe average)

$$\small {E(X|Y) = \alpha + {\beta _1}METH + {\beta _2}EDU{\rm{\;}} + {\beta _3}MRRG{\rm{\;}} + {\beta _4}AGE{\rm{\;}} + {\beta _5}COUNTRY\# YR + {\rm{\;}}\varepsilon} $$

Model 2. Female ethnic penalty (Europe average)

$$\small{E (X|Y) = \alpha + {\beta _1}FETH + {\beta _2}EDU{\rm{\;}} + {\beta _3}MRRG{\rm{\;}} + {\beta _4}AGE{\rm{\;}} + {\beta _5}COUNTRY\# YR + {\rm{\;}}\varepsilon} $$

Model 3. Male ethnic penalty by country

$$E(X|Y) = \alpha + {\beta _1}METH + {\beta _2}EDU{\rm{\;}} + {\beta _3}MRRG{\rm{\;}} + {\beta _4}AGE{\rm{\;}} + {\beta _5}YR + {\rm{\;}}\varepsilon $$

Model 4. Female ethnic penalty by country

$$E(X|Y) = \alpha + {\beta _1}FETH + {\beta _2}EDU{\rm{\;}} + {\beta _3}MRRG{\rm{\;}} + {\beta _4}AGE{\rm{\;}} + {\beta _5}YR + {\rm{\;}}\varepsilon $$

Note: METH (male ethnicity), FETH (female ethnicity), EDU (education), MRRG (marriage), AGE (age), COUNTRY#YR (interaction between country and year), YR (year).

The total sample of the 16 European countries from the LFS was 16,413,798. In terms of ethnicity, among the five ethnicities migrants from EE is the highest population, standing at 628,797 subjects, and the second largest population turned out to be MENA (290,987). They were followed by Asians (180,444), SubAf (162,949), and SA (147,280), respectively. This demographic profile shows how geographical adjacency concentrates particular ethnicities in receiving countries, and this would also indicate the negative selection of migrants in the main analysis. On the other hand, the residual category including EU-15 and EFTA (Norway, Switzerland, Iceland, and Lichtenstein) alongside North America and Oceania accounted for 3.13% (see Table 1 and a detailed descriptive analysis regarding variables by country found in appendix Tables 1–8).

Table 1.

Demographic breakdown according to ethnicity

Discussion and Conclusion

The four hypotheses were examined through the level and pattern of ethnic penalty in the labor market with the five ethnicities across 16 Western European countries. The trade-off pattern regarding high employment and low job quality was occupied most prominently by EE for both genders. On the one hand, MENA was found as the most penalized ethnicity in both patterns regarding double penalty and trade-off in low employment and high job quality. Here, female MENA migrants showed a greater dominance than male counterparts by occupying half of the plots allocated in these patterns, in contrast to the male cases of about 30% across 16 countries. Therefore, the first hypothesis which expected an EE preference in employment and the highest discrimination for MENA based on homogeneity and secularism was supported. Furthermore, the second hypothesis regarding gender difference based on the mixed embeddedness assumption was also confirmed since the level of employment penalty regarding MENA was much more prominent among female cases.

Meanwhile, the locational inequality hypothesis based on the origin countries’ socioeconomic backgrounds was diverged into two hypotheses. First, the locational inequality assumption regarding the lowest penalty for Asians thanks to the economic development of the origin countries (hypothesis 3) is half supported through the male case. This is because, in the female case, the less penalty pattern was comprised mostly of SA and the second-best ethnicity in this regard was revealed to be not only Asians but SubAf as well. Hypothesis 4 regarding how the nearest sending countries’ larger immigrant population might lead to a significant penalty in job quality was supported. As can be seen in the descriptive analysis regarding this point, EE and MENA were the first and second largest ethnicities in the 16 subject countries, so EE revealed the most penalty in terms of job quality, followed by MENA (except MENA female cases).

Furthermore, the two hypotheses (1 and 4) are complementarily linked since these two ethnicities dominantly occupied the ethnic demographic composition in Europe, so the trade-off pattern regarding low job quality and high employment could have happened for both ethnicities. However, only EE turned out to be the prominent ethnicity in this pattern, which means that employers’ preferences were clearly placed on EE rather than MENA. This is because the unskilled sectors do not require higher formal qualifications, so employers’ impressions of ethnicity are more important than those of skilled job positions. Therefore, in the end, racialization regarding the double penalty indicates not only how low job quality but also how low employment is strongly occupied by MENA, representing a majority Muslim demographic with strong religious norms maintained from their origin countries in contrast to host society practices, alongside host’s Islamophobia.

Overall, three implications emerge from this study. First, the comprehensive statistics of Western European countries’ ethnic penalties were provided according to gender so that the different strategies not only for ethnicities but also for gender can be applied to policymakers to ameliorate ethnic penalties at country and European levels. Second, the consistency between employers’ hiring discrimination and the labor market outcomes was discovered as well. Specifically, MENA females were less discriminated against compared with males in the correspondence experiments (Arai et al. Reference Arai, Bursell and Nekby2016; Bursell Reference Bursell2014; Derous et al. Reference Derous, Ryan and Serlie2015; Di Stasio and Larsen Reference Di Stasio and Larsen2020), while the employment penalty was much higher for females than that for males. The discrepancy can be explained through hypothesis 2 regarding the cultural effect of the labor supply, rather than demand side when considering findings between previous research and this study.

Moreover, Asian women in the experiments showed callback rates similar to white minorities (Di Stasio and Larsen Reference Di Stasio and Larsen2020), whereas this study revealed their second highest penalty in employment at the European average level after MENA females. This can arguably be like the MENA case associated with mixed embeddedness since Asians may not be discriminated against by employers as severely as women of other ethnicities; however, conservative male breadwinners or Confucian (in the case of East Asian) cultures could influence on the low performance in employment.

In line with this, SubAf revealed notably interesting results regardless of gender by holding an in-between position regarding ethnic penalty patterns throughout the 16 case countries: a double penalty and trade-off regarding low employment and high job quality (MENA), the trade-off penalty regarding low job quality and high employment (EE), and less penalty (Asian or SA). Besides this, the least penalty in job quality was found for both genders of SubAf at the European average level. Hence, this result clearly supports once more how the use of a categorization distinguishing between North and sub-Saharan Africa is required as the penalty patterns are clearly different between them. Moreover, this fact also can be linked to the result of hiring discrimination, which showed less callback rates for Black women than white minorities and Asians (Di Stasio and Larsen Reference Di Stasio and Larsen2020), but in the labor market outcome, they are less penalized compared to Asian women in employment and Asian and EE women in job quality.

Lastly, the ethnic penalty pattern showed another interesting tendency that classifies subject countries according to similar regional or institutional blocks. Apart from the less penalty pattern, most of the other three penalty patterns’ cases are predominantly clustered in relation to Mediterranean countries (trade-off penalty 2 with EE), Continental European countries (trade-off penalty 1 with MENA and SubAf), and Northern Europe (double penalty with MENA). This result well-matches previous analysis’ findings which revealed how labor market characteristics and relevant institutions’ differences affected migrant penalty patterns across Western European countries, although specific ethnic distinctions were not treated within these studies (see Ballarino and Panichella Reference Ballarino and Panichella2013, Reference Ballarino and Panichella2017; Kogan Reference Kogan2007; Lee Reference Lee2022).

Hence, why the ethnic penalty phenomenon represents a great complexity regarding migrant labor market outcomes, which are compatible with not only racialization but also prospects for comparative institutional connectivity in the European context. This should be revealed further in detail alongside time effects including cohort differences because past penalties connected with institutions and racialization can affect contemporary ethnic penalties (Lang and Spitzer Reference Lang and Spitzer2020; Lippens et al. Reference Lippens, Vermeiren and Baert2023; Small and Pager Reference Small and Pager2020). Therefore, these interlinked factors should be complementarily investigated further to provide better solutions for migrant integration into European society.

Although this study revealed the association between racialization processes and ethnic penalty in the 16 European labor markets based on the sociocultural and economic backgrounds of both native and ethnicity sides, there remain areas that can be considered for future research. First, more detailed differences within the ethnicities could be investigated where there is available data. For example, although it is expected that Asian migrants can be differently clustered according to East Asian and South (East) Asian based on culture and economic development, these specific categorizations for migrant status have not been available in EU-LFS data.

To be specific, there are categories in EU-LFS covering East Asian and South Asian or South American and Caribbean countries as origin countries, but the categories of the source questionnaire are necessarily merged as Asia and South America in the end. This is because there are multiple options that migrants could choose for their country of origin so that South Asian or East Asian could answer to Asia rather than choosing a specific regional category. Therefore, in order to secure statistical power and prevent substantial missing values, the higher regional-level category was considered to be the better option by merging the subcategories. Accordingly, this study suggests that survey questionnaires would be better should they distinguish more clearly the country of origin, benefiting future migrant research in measuring ethnicity penalties in greater detail.

Second, individual characteristics such as education, age, and marriage status were controlled in this study in order to highlight ethnic differences associated with labor market penalty with respect to natives. Hence, in future studies, the detailed interaction between ethnicities and individual socioeconomic conditions should be analyzed according to gender. As can be seen in the appendix, Figures 1–12 visualized how the employment and job quality of each ethnicity can be differentiated according to the condition of education, age, and marriage status at the European average level. Interestingly, although education is similarly effective for both genders by revealing that the higher the education, the better employment and job quality outcomes, in terms of age and marriage conditions, differences between genders were able to be found.

Specifically, the lowest employment was uncovered in older (45–65) and younger (25–34) ages for male and female migrants, respectively, while married migrant men showed the highest performances in both measures, but married migrant women revealed the lowest employment rates. Moreover, the probability of being in the highest employment and lowest job quality categories turned out to be divorced/widowed female migrants. This critical information can arguably be discussed regarding how ethnic women could experience more conservative cultures that hinder their employment, particularly when married so that, in case of divorced or widowed situations, they could experience substantial livelihood issues as the results clearly stated (see appendix Figures 6 and 12).

Last but not least, an unobservable area that may also highlight ethnic penalty could be considered. When it comes to ethnic penalty, the umbrella concept is referred to as there are many aspects that could be difficult to disentangle (Panichella et al. Reference Panichella, Avola and Piccitto2021). In this sense, this study also considered multiple factors at the macro-level, including labor supply and demand sides alongside socioeconomic and cultural backgrounds, while different levels of analyses alongside other factors could be developed as well. In line with this, micro- and meso-levels could be assessed which are relevant to individual’s motivations or experiences regarding migration processes and community-level ethnic networks or similar. Furthermore, how colonial histories and the relevant linguistic proximity may influence labor market ethnic penalty could be meaningfully analyzed through country-specific research.