Which Degree for Which Occupation? Vertical and Horizontal Mismatch Among Immigrants, Their Children, and Grandchildren in France

Abstract

Prior research has found that immigrants are often overeducated: their educational attainment is higher than required or commonly observed in their occupation. Yet, less is known about the education–occupation mismatch among immigrants’ children and grandchildren (the second and third generations). Using the French Trajectories and Origins 2 (TeO2, 2019–2020) survey, we test theoretically grounded hypotheses on the level of vertical (educational attainment) and horizontal (field of study) mismatch in the first, second, and third generations as well as on the mechanisms underlying the mismatch. Results indicate that vertical mismatch is substantially lower in the second and third generations than in the first, supporting the hypothesis that vertical mismatch is largely the result of imperfect international transferability of credentials. By contrast, higher levels of horizontal mismatch persist in the second and third generations among men of non-European descent. Differences in horizontal mismatch between immigrants’ and natives’ descendants are largely accounted for by initial sorting into fields of study.

Introduction

A rich body of research has studied the education and labor market outcomes of immigrants and their descendants (Drouhot and Nee 2019; Heath et al. 2008). In many destination countries, children of immigrants—the so-called second generation—tend to have similar levels of educational attainment as natives once differences in parental background are considered (Algan et al. 2010; Dustmann et al. 2012; Hermansen 2016).¹ By contrast, differences in employment and earnings persist for some origin groups in certain destination countries (Algan et al. 2010; Platt and Nandi 2020), such as among the descendants of immigrants from North Africa in France (Meurs 2018; Meurs and Pailhé 2010; Primon et al. 2018). However, fewer studies have addressed the extent to which the educational qualifications attained by immigrants’ descendants match their occupational positions in the labor market.

Overeducation is common and consequential in Europe (Quintini 2011), particularly among immigrants who are more often overqualified for their occupations relative to natives (Andersson Joona et al. 2014; Visintin et al. 2015). Overeducation has been connected to lower wages, accounting for a large share of the immigrant–native wage gap and explaining some of the wage disadvantages ethnoracial minorities experience (Andersson Joona et al. 2014; Lu and Li 2021). Wage penalties are especially high for immigrants who are mismatched in terms of both educational attainment (vertical mismatch) and field of study (horizontal mismatch) (Banerjee et al. 2019). When an individual's education level or field of study differs from the norm in the occupation, this mismatch can also incite feelings of frustration or perceptions of injustice (Clark and D'Ambrosio 2015; Godechot and Senik 2015). Educational mismatch is related to poorer self-rated health and lower psychosocial well-being, especially among immigrants (Clark and Senik 2014; Dunlavy et al. 2016; Espinoza-Castro et al. 2019). Therefore, assessing the extent to which higher levels of educational mismatch persist among immigrants’ children and grandchildren, as we do in this article, illuminates an important mechanism that can drive socioeconomic assimilation and ethnoracial inequalities.

Among immigrants, the commonly observed higher levels of educational mismatch have generally been interpreted as the result of imperfect transferability of qualifications between the origin and destination countries (Aleksynska and Tritah 2013). Given that the descendants of immigrants are educated in the destination society, speak its language, and know its institutions, we would expect a decrease in educational mismatch across generations. Indeed, there are no a priori reasons why their educational degrees should be valued differently in the labor market than natives’ degrees. However, racial discrimination, residential segregation, and unfavorable tracking decisions despite high educational aspirations (Safi 2017; Shahrokni 2021) could lead to persistently high levels of mismatch across generations, especially among members of racialized minorities. To date, the small but growing literature on the education–occupation mismatch among the children of immigrants remains inconclusive. Some studies have found limited evidence of mismatch in the second generation (Khoudja 2018; Pecoraro 2011), whereas others have concluded that mismatch is actually strong, especially among the descendants of non-Western immigrants (Belfi et al. 2021; Dahlstedt 2015; Falcke et al. 2020).

Using the French Trajectories and Origins 2 (TeO2, 2019–2020) survey, this study assesses the prevalence of educational mismatch across generations, regions of origin, and gender in France. We contribute to the literature in three ways. First, our study is the first to assess educational mismatch among the third generation, or immigrants’ grandchildren. Prior studies have focused on first-generation immigrants (Andersson Joona et al. 2014; Visintin et al. 2015) and have increasingly studied the second generation (Belfi et al. 2021; Falcke et al. 2020; Khoudja 2018). However, we are not aware of any studies on the third generation, even though the demographic and social significance of this group is increasing in established countries of immigration, such as the United States (Tran 2018) and France (Lê et al. 2022). Second, we assess vertical and horizontal mismatch. A considerable literature has examined vertical mismatch—the level of divergence between an individual's educational attainment and occupation. By contrast, less attention has been paid to horizontal mismatch—differences between an individual's field of study and their occupation. Third, we go beyond describing the extent of educational mismatch and test theoretically grounded hypotheses on the mechanisms underlying observed mismatches. We distinguish between structural and individual mechanisms and assess whether they explain the observed differences between groups.

Theoretical Background and Previous Literature

Vertical and Horizontal Mismatch

Educational degrees, indicating the level of attainment and field of study, are used as credentials to enter the labor market and as signals for employers. Although specific credentials are strictly required for some closed occupations (Weeden 2002), such as medical doctors and lawyers, most occupations are more open, resulting in within-occupation heterogeneity in workers’ educational degrees.

Vertical mismatch is generally categorized as undereducation (lower educational attainment than is required or observed in a given occupation), overeducation (higher educational attainment than is required or observed in a given occupation), and no mismatch (similar educational attainment to what is required or observed) (Quintini 2011). Undereducation tends to increase with labor experience because individuals can show their worth on the labor market and progress within their company (Groot and Maasen van den Brink 2000; INSEE 2021). By contrast, overeducation indicates lower returns to individuals’ human capital investments and is generally linked to lower wages (Andersson Joona et al. 2014; Li and Lu 2023; Lu and Li 2021). Overeducation can occur when too many applicants have the same education level. In these situations, securing a job might entail having higher educational attainment than required. Indeed, from this perspective, overeducation is understood as an imbalance between labor supply and demand that results from a surplus of graduates relative to job openings.

Some recent work notwithstanding (Di Stasio 2017; Falcke et al. 2020; Li and Lu 2023), prior studies have paid less attention to horizontal mismatch. In certain situations, individuals can be expected to experience both vertical and horizontal mismatch simultaneously. Indeed, individuals who lack employment opportunities and whose career choices are constrained are more likely to accept work that requires lower qualifications (vertical mismatch) and is in a different field (horizontal mismatch) (Falcke et al. 2020). Similarly, overeducation might compensate for a lack of occupation-specific training (Di Stasio 2017), which would result in both vertical and horizontal mismatch for individuals with unspecialized academic degrees.

Still, vertical and horizontal mismatch do not necessarily coincide because they have distinct determinants. Higher levels of horizontal mismatch could indicate difficulties in finding a job matching one's field of study. In this case, horizontal mismatch is said to be demand-related (Somers et al. 2019) because the job corresponding to the field of study is unavailable or difficult to secure (Betts 1996). This type of horizontal mismatch is perceived as undesirable (Kucel and Vilalta-Bufí 2013), leading to lower salaries for such mismatched individuals because employers factor in a “matching cost” (Bruyère and Lemistre 2005). However, in some instances, horizontal mismatch might indicate achievement. For instance, successful individuals might be recruited or promoted to occupational positions outside their field of study, especially as they gain experience in the labor market (Somers et al. 2019). Li and Lu (2023) distinguished between horizontal undermatch—in which individuals are employed in out-of-field occupations that pay less than matched occupations—and horizontal overmatch—in which individuals are employed in out-of-field occupations that pay more than matched occupations. They found that immigrants more often experience horizontal undermatch and less often horizontal overmatch than natives in the United States.

Educational Mismatch Among Immigrants and Their Descendants

Among first-generation immigrants, educational mismatch is a well-established empirical finding in Europe (Aleksynska and Tritah 2013) and the United States (Lu and Li 2021). Following human capital theory, imperfect international transferability of degrees and skills might account for overeducation (Chiswick and Miller 2010) and horizontal mismatch among immigrants (Li and Lu 2023). Overeducation has been deemed to be a form of apparent mismatch, given that skills could be poorly correlated with qualifications among newly arrived immigrants (Flisi et al. 2017; Prokic-Breuer and McManus 2016). For instance, language difficulties and limited institutional knowledge might increase the educational mismatch experienced among immigrants during their first years in the destination country. Consistent with this reasoning, educational mismatch decreases as immigrants settle in the country (Banerjee et al. 2019; Larsen et al. 2018; Nielsen 2011), leading us to expect educational mismatch to disappear by the second generation.

Prior evidence on the educational mismatch of the descendants of immigrants is more limited. Studies suggest that the descendants of immigrants experience lower mismatch than immigrants and sometimes even fully converge toward natives (Khoudja 2018; Pecoraro 2011), despite considerable differences across regions of origin (Belfi et al. 2021; Dahlstedt 2015; Falcke et al. 2020). For instance, children of non-Western immigrants are more likely to experience vertical mismatch than those from Western countries (Belfi et al. 2021; Dahlstedt 2015). Falcke et al. (2020) found that Western and non-Western second-generation graduates are more likely to experience horizontal mismatch than natives, but only non-Western groups are also more likely to experience vertical mismatch.

For immigrants and their descendants, educational mismatch as it relates to their educational and labor market incorporation can be viewed through the lens of assimilation theories (Khoudja 2018), which point to competing expectations regarding educational mismatch among the descendants of immigrants. According to the neoclassical assimilation theory (Alba and Nee 2003), which reframes the classical assimilation theory (Park and Burgess 1921; Warner and Srole 1945), natives and immigrants’ descendants gradually converge with time and across generations. Despite differences between the United States and Western Europe and notable variations between groups, studies point to an overall pattern of intergenerational assimilation in terms of educational attainment and labor market outcomes (Drouhot and Nee 2019; Heath et al. 2008). As a result, the link between the level and field of the degree obtained and the occupation held should become increasingly similar to that of natives over generations. Considering education–occupation mismatch as an additional indicator of socioeconomic incorporation, the neoclassical assimilation framework and research findings lead us to expect a gradual decline in both vertical and horizontal mismatch over generations among immigrants’ children and grandchildren.

However, integration trajectories might also differ across ethnic groups. The segmented assimilation theory (Portes and Zhou 1993; Zhou and Gonzales 2019) and research on racialization (Telles and Ortiz 2008) assume that different ethnic groups follow distinct pathways and that immigrants’ descendants might converge with natives in some areas but not others (White and Glick 2009). The theories posit that the segmented character of assimilation results from an interaction between individuals’ human capital, parental socioeconomic status, and the characteristics of the coethnic community on the one hand and the policies, values, and prejudice in the receiving society on the other (Zhou and Gonzales 2019). For instance, negative beliefs against certain ethnic minorities can lead to a discounting of their skills (Esses et al. 2006). Although the segmented assimilation theory originates from studies of the United States, research suggests that it may also apply to the French setting (Safi 2008). In France, studies underscore the impact of its colonial history and ethnic discrimination as sources of segregating dynamics against certain immigrant-origin groups (Drouhot and Nee 2019; Silberman et al. 2007; Silberman and Fournier 2006). Overall, the descendants of non-European immigrants have lower socioeconomic outcomes and are more at risk of being outside the labor market (Meurs 2018; Meurs and Pailhé 2010; Primon et al. 2018). Therefore, we expect that the descendants of non-European immigrants are likely to experience more durable education–occupation mismatch than the descendants of European immigrants.

Education–Occupation Mismatch as a Gendered Process

Men and women might experience different levels of educational mismatch, given that gender differences in fields of study and occupations persist (Dahl et al. 2023; Humlum et al. 2019). Women continue to be overrepresented in the arts and humanities and in health, whereas men are overrepresented in STEM fields and occupations (Hermansen and Penner 2022). Recent work has indicated heterogeneity in the gendered patterns of educational mismatch across educational groups, with college-educated women more likely to experience educational mismatch than college-educated men (Addison et al. 2020).

The few prior studies that have assessed gender differences in the educational mismatch of immigrants provided heterogeneous findings (Birgier and Bar-Haim 2023; Pecoraro 2011). However, in general, ethnic minority men are often stereotyped as threatening and therefore are more likely to experience discrimination than minority women (Sidanius and Pratto 2001). This pattern has been verified in hiring discrimination in Denmark (Dahl and Krog 2018) and Sweden (Arai et al. 2016). Wage disadvantages also appear greater for minority men in the United States (Mandel and Semyonov 2016) and in France (Gueye and Ceci-Renaud 2022). Therefore, because they are more likely to experience discrimination in the labor market, male descendants of ethnic minority immigrants are expected to undergo higher levels of education–occupation mismatch than women from the same ethnic group.

Mechanisms Underlying Differences in Educational Mismatch

To gain a better understanding of what explains gaps in vertical and horizontal mismatch between natives, immigrants, and their descendants, we identify three structural and three individual mechanisms. Structural mechanisms—occupational closure, vocational training, and fields of study—refer to the structure of the educational system and labor market and, specifically, the allocation of individuals into distinct educational tracks and occupations. First, accessing closed occupations requires holding a specific degree type (Weeden 2002). Therefore, occupational closure almost guarantees the absence of any educational mismatch. These degrees often imply considerable human capital investment. For instance, obtaining a medical degree in France takes 9–12 years after high school. An underrepresentation of immigrants and their descendants in these low-mismatch occupations (Lancee and Bol 2017) could help explain the higher mismatch among these groups.

The second mechanism concerns vocational training. Because vocational training provides individuals with occupation-specific skills, individuals with vocational degrees tend to experience lower educational mismatch than individuals with academic degrees (Levels et al. 2014; Wolbers 2003). In France, vocational degrees are often acquired through an apprenticeship, increasing the likelihood of a good match between one's qualifications and job (Rose 2005). Although the descendants of immigrants often attend vocational tracks, they tend to be underrepresented in apprenticeships (Palheta 2012: chap. 6). In general, a population group's proportion of vocational graduates likely helps explain its observed educational mismatch level.

The third mechanism pertains to differential sorting into fields of study. In particular, graduates from broad fields of study tend to experience greater mismatch. For example, arts and humanities graduates (Robst 2007; Wolbers 2003) and graduates of other programs with a general orientation (Verhaest et al. 2017) are at greater risk of educational mismatch than health graduates (Somers et al. 2019). Evidence on the fields of study of the first and second generations in France remains scant. However, using Norwegian data, Borgen and Hermansen (2023) showed that children of immigrants are overrepresented in some fields, such as business administration and law, science and engineering, and health. A greater likelihood of mismatch among these fields of study could help explain greater mismatch in these groups.

Individual-level mechanisms, all relating to individuals’ strategies and biographies, might also account for observed differences in educational mismatch levels between groups. We propose three mechanisms, with the first two concerning the length of the job search: recent unemployment spells and the school-to-work transition. We expect that individuals who experience long periods of unemployment, either recent or directly after graduating, widen their job search and hence experience greater educational mismatch when they find a job (Ordine and Rose 2015; Sam 2018), a pattern that is particularly present for early-career individuals in France (Nauze-Fichet and Tomasini 2002; Rose 2005).

Our third proposed individual-level mechanism relates to experiences of discrimination. We expect that past experiences of discrimination at work will lead affected individuals to change their employment sector, increasing educational mismatch. Ethnic, religious, and racial discrimination in hiring and in the labor market are persistent from the first to the second generation among certain groups (Pager et al. 2009) and are particularly widespread in France, according to a meta-analysis of field experiments in Europe and North America (Quillian et al. 2019). Immigrants and their descendants who are perceived as Muslim are commonly targeted in France (Adida et al. 2010, 2016). This mechanism could impact employment searches among non-European immigrants and their descendants.

Hypotheses

The theoretical and empirical literature leads us to hypothesize that vertical and horizontal mismatch levels differ across generations and origin groups. Following human capital theory, we expect immigrants to be overeducated as a result of the imperfect international transferability of degrees and skills, but we expect the educational mismatch levels of the second and third generations to be similar to those of natives.

Alternatively, we might expect a more gradual process. The neoclassical assimilation framework suggests that immigrants’ integration is a slow intergenerational process of convergence with the native population. Higher educational mismatch might not fully disappear by the second or even the third generation.

Segmented assimilation and related theories lead us to expect that convergence patterns differ markedly across ethnic groups and that mismatch is more persistent among the descendants of racialized non-European immigrants.

Gendered discrimination in the labor market leads us to expect differences in men's and women's mismatch levels. Male descendants of ethnic minority immigrants are expected to have higher education–occupation mismatch than women from the same ethnic group because racialized men are more likely to experience discrimination in the labor market.

Regarding the mechanisms, holding a closed occupation, having vocational training, and graduating from a general or specific field of study lead to different levels of educational mismatch and can explain observed gaps in educational mismatch levels with natives.

At the individual level, experiences of unemployment and discrimination will likely affect individuals’ job search strategies and lead them to accept jobs below their qualifications or outside their field.

Data and Methods

TeO2 Survey and Sample Restrictions

We use data from the TeO2 survey (Beauchemin et al. 2023). This large-scale (N = 27,500; more precisely, the data files provided to us in September 2023 included 27,422 individuals) nationally representative, face-to-face survey oversamples immigrants and their descendants. In France, an immigrant is an individual who was born abroad without French nationality at birth. We distinguish between individuals who are immigrants (G1), those who were born in France to at least one immigrant parent (G2), and those who were born in France to parents born in France (or arrived prior to age 3) and have at least one immigrant grandparent (G3). We compare these generational status groups with natives (G4+) with no immigrant ancestry (up to the grandparents’ generation).

Individuals with an immigrant background come from a diverse set of geographic origins. The most common origins in the survey are North Africa and Southern Europe (especially Portugal and Italy). Other common regions are sub-Saharan Africa and the Middle East. As a result of the history of immigration to France, the three generations also differ in their national composition (see Tables A1 and A2; all tables and figures designated with an “A” appear in the online appendix). Given that the distinction between European and non-European origins is salient in French society (Beauchemin et al. 2018), as well as to garner enough statistical power, we distinguish between European and non-European origins in our analyses of whether patterns of assimilation are segmented. However, we acknowledge the internal heterogeneity within these two broad categories. We therefore also present results with a more detailed categorization of regional origins (see Figures A1 and A2). The findings are consistent with the results presented in the main analysis and justify the use of the European versus non-European distinction as a relevant dividing line.

We make several successive sample restrictions. We exclude respondents born in France's overseas territories and their descendants (1,538 observations), as well as French foreign-born individuals and their descendants (3,255 individuals). Being neither immigrants nor natives, the excluded groups are quite specific to the French context and fall beyond the scope of this study. We also exclude 20 individuals who seem to be falsely identified as the third generation.

Then, because education–occupation mismatch can be measured only among individuals who have an occupation and educational qualifications, we restrict our analytic sample to individuals who have an occupation at the time of the survey (excluding 1,837 individuals who are studying and 5,181 who are not employed) and individuals who have educational qualifications (excluding 2,462 individuals who attended school but never obtained a degree). This selection could be viewed as a source of bias in estimating the association between immigrant origin and mismatch in the general population. However, if we consider that our analysis of education–occupation mismatch is only valid for and applies to the population in which mismatch can—theoretically and practically—be measured (those who have educational qualifications and an occupation), then these sample restrictions should not be a cause for concern.²

We further exclude individuals for whom we have no information on educational attainment (excluding 2 observations), field of study (excluding 81 observations), school-to-work transition (excluding 23 observations), or occupation (excluding 69 observations) (see Table A3).³ Our final analytic sample comprises 12,954 individuals: 6,554 men and 6,400 women.

Measures of Educational Mismatch

The literature commonly uses three types of measures to capture mismatch: expert assessments of job requirements, subjective measures based on workers’ perceptions of mismatch, and empirical measures based on the observed distribution of educational degrees in each occupation. Even though these indicators operationalize the same concept (Chiswick and Miller 2010), they can provide different results (Flisi et al. 2017; Groot and Maasen van den Brink 2000). Expert assessments of the degree required for an occupation are based on detailed job analysis. This measurement strategy implies exhaustive lists of standardized occupations, such as O*Net codes in the United States. Because of the costliness of the exercise, these measures are infrequently updated and often refer to broad occupational categories (Addison et al. 2020; Banerjee et al. 2019). This approach also adopts an employer's perspective and is objective only to the extent that it is not assessed by employees themselves. Subjective measures of employees’ self-assessed mismatch provide more up-to-date information and may be more accurate in referring to the actual work individuals do. However, workers tend to overstate their job requirements and their level of overeducation (Hartog 2000).

The present study uses empirical measures and follows an inductive approach of realized matches (Clogg and Shockey 1984), which is widely used (Li and Lu 2023). This measure is based on the empirical distribution of educational attainment and fields of study within each occupation. The empirical approach constructs mismatch measures relative to the general population and has several strengths: it relies on up-to-date information, uses detailed occupation categories, and neutralizes any bias related to individual interpretations of the educational requirements for different occupations. However, we recognize that it also has some disadvantages compared with other measurement strategies. In particular, as an inductive approach relying on observed distributions, the empirical measure identifies deviations from the general population. If overeducation—as measured by expert analysis or workers’ subjective perceptions—is common in an occupation, then it becomes the statistical standard and will not be considered a mismatch by the empirical measurement approach. Our measure can therefore be considered downwardly biased and provides conservative estimates of the prevalence of mismatch (Groot and Maasen van den Brink 2000; Johansson and Katz 2006).

To measure vertical mismatch, we categorize educational attainment into 11 levels: primary school; middle school; short vocational; vocational baccalauréat; technological and academic baccalauréat; Bac +2, referring to shorter studies, having passed the baccalauréat and completed two years of university studies; Bac +3, equivalent to having completed a bachelor's degree; Bac +4, a one-year master's program; Bac +5, a two-year master's program; elite schools (grandes écoles), which are highly competitive and comparable to U.S. Ivy League institutions; and doctorate (PhD). Using the 2013–2020 pooled French Labor Force surveys, we compute the distributions of educational attainment in each occupation at the most disaggregated occupational level (four-digit level of the French PCS nomenclature amounting to almost 500 occupations). We then locate the educational attainment of each TeO2 respondent within the educational distribution of individuals in the same detailed occupation. We calculate the “ridit” of this relative position (Bross 1958; Ichou 2014). The ridit is a scaled percentile that indicates the proportion of the population within the same occupation holding a degree below that of the respondent plus half the proportion holding the same degree as the respondent. We z-standardize this variable so that a value above 0 indicates overeducation and a value below 0 indicates undereducation.

We measure horizontal mismatch using the distribution of fields of study in the International Standard Classification of Education, consistent with prior research (DiPrete et al. 2017). This procedure provides us with 11 fields of study: generic programs and qualifications; education; arts and humanities; social sciences, journalism, and information; business, administration, and law; natural sciences, mathematics, and statistics; information and communication technologies; engineering, manufacturing, and construction; agriculture, fishery, forestry, and veterinary; health and welfare; and services.

Similar to vertical mismatch, our measure of horizontal mismatch relies on reference distributions from the 2013–2020 French Labor Force surveys. For each of the 500 detailed occupations, we compute the distribution of fields of study. For each TeO2 respondent, we assess the proportion of individuals in the same occupation who completed a different field of study. The variable is z-standardized. A lower value indicates that the individual's field of study is common among people with the same occupation (low horizontal mismatch), whereas a higher value indicates that their field of study is rare (high horizontal mismatch). We complement this measure of horizontal mismatch with an indicator of the frequency of the modal field of study in the occupation to capture how closely the occupation is linked to a specific field of study. We include this additional variable in all analyses on horizontal mismatch.

Multivariate Analyses

We estimate linear regressions to assess differences in the vertical and horizontal mismatch levels experienced by immigrants (G1), children of immigrants (G2), grandchildren of immigrants (G3), and natives (G4+). Our two dependent variables of educational mismatch are continuous, enabling us to assess undereducation and overeducation (i.e., vertical mismatch) and how common or uncommon the field of study is in the occupation (i.e., horizontal mismatch). We run all analyses separately for men and women because we expect educational mismatch patterns to be gendered.

We include several control variables to account for sociodemographic differences between the groups that likely contribute to differences in educational mismatch. We control for age and age squared because the age distribution of the groups differs, and age (or birth cohort) can impact mismatch levels (Addison et al. 2020). Given that parts of the TeO2 survey were conducted during the COVID-19 pandemic, we also include a dummy variable indicating whether the interview was conducted before or after March 2020 (during the first lockdown in France). The models further control for income tercile to compare individuals who hold broadly similar positions in the labor market. We control for income rather than occupation or educational attainment because we use the latter two to construct the dependent variables.⁴

We provide the coefficients and standard errors of the control variables in Tables A4–A7. These tables also include a second model that controls for years of education to compare individuals with similar amounts of education. Controlling for this variable does not change the general patterns observed.

Main Findings

Descriptive Results

Table 1 presents descriptive results across generations and origin groups. The most common education levels are short vocational training (26%) and Bac +2 (18%). Non-European immigrants and their descendants are somewhat less likely to have short vocational training than the other groups (18%), whereas non-European and European G1 are less likely to have a Bac +2 (13% and 12%, respectively). The most common fields of study, completed by 24% and 27% of respondents, are (1) business, administration, and law, which are particularly common among non-European and European G2; and (2) engineering, manufacturing, and construction, which are more prevalent among natives and European G1, G2, and G3.

Regarding occupations, non-European and European G1 are more often manual workers than the other groups (23% and 20% vs. the average of 18%). However, these two groups are also somewhat more likely to be executives and higher professionals (23% vs. the average of 20%). In comparison, among non-Europeans and Europeans, G2 and G3 are more often intermediate professionals and nonmanual workers than G1.

Vertical mismatch is highest for G1, especially among non-Europeans (0.30). Mean values of vertical mismatch are also high among European G1 (0.20) relative to natives (−0.04). Vertical mismatch decreases between G1 and G2 for both non-Europeans and Europeans. However, it increases again somewhat for G3, although this difference is significant only for Europeans. This pattern of mismatch decline between G1 and G2 and the subsequent increase for G3 is also observed for horizontal mismatch. The G3 increase is significant among both non-European and European descendants.

Vertical Mismatch

We begin by assessing whether immigrants’ vertical mismatch relative to natives diminishes across generations. Figure 1 shows results from linear regressions on vertical mismatch, with negative values indicating undereducation and positive values overeducation. G1 tends to be overeducated relative to natives, but this difference disappears by G2: we observe no evidence of overeducation among G2 and G3 relative to natives. The patterns are similar for men and women, although vertical mismatch is slightly higher among G1 women than among G1 men. In short, G1 is overeducated relative to natives, although we find no statistically significant evidence of persistent differences among G2 and G3. However, we note a borderline significant difference in overeducation among G3 women relative to natives, which could signify a subsequent increase in overeducation for G3.

Given that average results by generation might conceal substantial heterogeneity between origin groups, Figure 2 presents results from similar models on vertical mismatch disaggregating generations between non-European and European origins. Both non-European and European G1 experience greater vertical mismatch than natives. Among non-European G1, we observe similar relative levels of overeducation among men and women. By contrast, European G1 women have somewhat higher relative levels of overeducation than European G1 men. Among G2 and G3, we observe no statistically significant evidence of overeducation in either origin group relative to natives. Consistent with H1, these findings indicate that immigrants are overeducated relative to natives but that this difference is not perceptible by G2, regardless of geographic origin.

To investigate the connection between higher vertical mismatch among G1 and the limited international transferability of foreign degrees, we focus on G1 and differentiate between individuals who obtained their highest degree in France versus abroad (see Figure A9). These analyses reveal that gaps with natives are smaller among G1 who obtained their degree in France than among those with a foreign degree. Thus, higher relative levels of vertical mismatch among G1 are mainly the result of overeducation among those with a foreign degree. This evidence corroborates the idea that overeducation results from limited international transferability of foreign degrees among G1 and therefore disappears by G2, as H1 suggests. We observe this pattern among European immigrants of both sexes and among non-European immigrant women. However, non-European immigrant men who obtained their degree in France experience only slightly lower vertical mismatch than those with a foreign degree, suggesting that they face barriers beyond having a foreign degree recognized.⁵

Horizontal Mismatch

Figure 3 shows greater horizontal mismatch among G1 than natives, but by G2 and G3, the horizontal mismatch is similar to that of natives. Patterns are similar for men and women, although G1 men experience somewhat higher relative levels of horizontal mismatch than G1 women. As in Figure 1 for vertical mismatch, horizontal mismatch shows signs of increasing again for G3 relative to natives, but these differences are not statistically significant.

In Figure 4, we distinguish between non-European and European immigrants and their descendants, revealing striking differences between origin groups. Among men, non-Europeans experience greater horizontal mismatch relative to natives across the three generations analyzed. In this group, horizontal mismatch decreases somewhat between G1 and G2 but remains significantly higher among G2 relative to natives. Among G3, horizontal mismatch increases again. Still, large confidence intervals indicate that the estimate for non-European G3 is relatively imprecise. Among European men, we observe higher relative levels of horizontal mismatch among G1 but no evidence of higher horizontal mismatch in subsequent generations. Among women, we similarly observe higher levels of horizontal mismatch among non-European and European G1 relative to natives but no significant differences among G2 or G3, even though the gap between G3 and natives appears larger than that between G2 and natives. In sum, our findings highlight an exception to the overall pattern of intergenerational convergence: sons and especially grandsons of non-European immigrants experience significantly greater horizontal mismatch than natives.

These findings are consistent with H3, which posits the existence of enduring differences among the descendants of non-European immigrants, rather than with H2, which posits convergence with natives. They are also consistent with H4, which predicts greater inequality among men. Further, occupations with the highest average horizontal mismatch require a lower skill level and pay less than occupations with the lowest horizontal mismatch (see Table A9), indicating that greater horizontal mismatch is generally linked to a penalty in France.

Mechanisms

To understand the potential processes underlying persistently greater horizontal mismatch among non-European men, we incorporate two sets of theoretically motivated mechanisms discussed earlier: (1) structural mechanisms refer to occupational closure, vocational training, and fields of study; and (2) individual-level mechanisms regard recent unemployment spells, difficulties experienced in the school-to-work transition, and belonging to a group that experiences discrimination. In this analysis, we focus on horizontal mismatch among men, the only case in which we observed persistent differences across generations. For completeness, we provide estimates for women in Figure A10.

Figure 5 provides results from linear regressions on horizontal mismatch, adjusting for each mechanism one by one. The full set of coefficients and standard errors indicate that the coefficients on most mechanisms are sizable, significant, and in the expected directions (see Table A10).⁶

Differences do not substantially decline when we control for occupational closure in panel a of Figure 5. For all groups, the coefficients remain close to those provided in the baseline model (indicated by the hollow black circles). By contrast, differences diminish for non-European and European G1 when we control for having completed vocational training. Still, the differences between non-European and European G1 and natives remain significant. For G2 and G3, controlling for vocational training does not lead to commensurate declines in horizontal mismatch. Importantly, when we control for sorting into fields of study, differences decline for all groups and are no longer statistically different from 0. Thus, differences in fields of study between groups are a major mechanism for the observed gaps in horizontal mismatch for men of non-European origin. These findings corroborate H5, which posits that group-specific sorting (here, in the educational system) helps explain differences in educational mismatch. An assessment of the distribution of the origin groups across fields of study reveals that non-European men are particularly overrepresented in generic programs, which, along with the natural sciences, have the highest average levels of horizontal mismatch (see Table A12). Non-European men are also more frequently in the social sciences and arts and humanities, which are fields with high horizontal mismatch. Conversely, these men are underrepresented in engineering, which is tied with health and welfare for having the lowest mismatch.

The models presented in panel b of Figure 5 control for individual-level mechanisms. Contrary to H6, we find that differences do not considerably decline when we incorporate recent unemployment spells, difficulties experienced in the school-to-work transition, or belonging to a group that experiences discrimination. Still, belonging to a group experiencing discrimination explains a nontrivial part of the differences observed among non-European G1, G2, and G3. The difference between non-European G2 and natives observed in the baseline model becomes nonsignificant, whereas the difference between non-European G1 and G3 relative to natives remains significant.

Sensitivity Analysis

We ran sensitivity analyses excluding generic programs and qualifications, which are concentrated at the lowest education level and do not provide a specialization. For vertical mismatch, including individuals with generic programs and qualifications might yield lower estimates because this group cannot experience overeducation. By contrast, for horizontal mismatch, including such individuals might yield higher estimates, because these individuals did not specialize in a particular field and might search for a job in different fields of study. In the main analysis, we include generic programs to study the full working population with a degree. Table A13 indicates similar patterns to those presented in the main analysis, although differences in horizontal mismatch are no longer significant for G2 or G3.

We also ran the main analyses excluding small occupations—that is, occupations with fewer than 50 individuals in the French Labor Force survey—to assess whether our results are driven by outliers. The results reveal that the patterns for vertical and horizontal mismatch are similar to those in the main analysis (see Table A14). Again, point estimates remain similar but are no longer significant for G2.

Discussion and Conclusion

In this study, we analyze how closely individuals’ educational attainment and field of study match those of others in the same occupation. We assess whether the levels of vertical and horizontal mismatch experienced by immigrants, children of immigrants, and grandchildren of immigrants differ from those of natives in France. Linear regression models estimated on data from the TeO2 survey reveal that overeducation is primarily experienced by first-generation immigrants, whereas the descendants of immigrants have similar levels of vertical mismatch as natives. However, men of non-European descent experience persistently greater horizontal mismatch, even in the third generation.

These findings indicate distinct patterns for vertical and horizontal mismatch. Our results on vertical mismatch are consistent with H1, which posits that educational mismatch results from imperfect international transferability of degrees and skills. The human capital theory highlights obstacles in obtaining training accredited in the destination country, language difficulties, and incomplete institutional knowledge as the main factors underlying immigrants’ greater educational mismatch (Aleksynska and Tritah 2013; Chiswick and Miller 2010; Li and Lu 2023). Because these mechanisms are specific to the first generation, mismatching is expected to disappear for the second and third generations, who were born and educated in France. In line with past research (Khoudja 2018; Li and Lu 2023; Lu and Li 2021), our findings corroborate this expectation: higher relative levels of vertical mismatch in the first generation mainly result from overeducation among immigrants with a foreign degree. However, the descendants of immigrants have jobs that are consistent with their educational attainment—at least to the same extent as natives. Still, employed descendants constitute a selected group, given their specific obstacles to entering paid employment (Meurs 2018), including racial discrimination (Quillian et al. 2019) and gender discrimination (with women more likely to be out of the labor market in France) (Meurs and Pailhé 2010). Analyzing the education–occupation mismatch among individuals with a degree and an occupation cannot capture the full degree of educational and labor market inequality disadvantaged groups experience.

Our finding of convergence in vertical mismatch by the second generation corroborates prior evidence from the Dutch and Swiss settings based on similar empirical measures of mismatch (Khoudja 2018; Pecoraro 2011). In contrast, prior work based on subjective measures paints a more pessimistic picture, noting persistently greater vertical mismatch among non-European G2 in Sweden (Dahlstedt 2015) and the Netherlands (Belfi et al. 2021). These differences might reflect that self-reports often overestimate the level of mismatch experienced, whereas realized matches provide more conservative estimates (Groot and Maasen van den Brink 2000).

For horizontal mismatch, we observe persistent differences among non-European men, supporting H3, which posits the existence of enduring disadvantages among the descendants of non-European immigrants. This result qualifies the gradual convergence scenario (H2) following the neoclassical assimilation framework (Alba and Nee 2003) and suggests lasting penalties for racialized men. We also find evidence supporting H4, which posits that men of immigrant descent will experience a more persistent educational mismatch than women because ethnic minority men are more often stereotyped as threatening than ethnic minority women (Sidanius and Pratto 2001).

By contrast, for vertical mismatch, we do not observe substantial gender differences. Hence, although both male and female immigrants experience imperfect international transferability of degrees, having a job outside of one's field of expertise is more common among men than among women.

How do we interpret the finding of a persistent disadvantage in horizontal but not vertical mismatch among men of non-European descent? Racial discrimination, the main mechanism underlying H3 and H4, might contribute to this pattern. Indeed, this durable disadvantage might, to some extent, be the result of discrimination experienced in the educational system and the labor market, as segmented assimilation theory (Portes and Zhou 1993; Zhou and Gonzales 2019) and theories of racialization (Telles and Ortiz 2008) suggest. Occupations with the highest average horizontal mismatch have lower skill levels and pay less than occupations with the lowest horizontal mismatch (Somers et al. 2019:567). However, racial discrimination is unlikely to be the only mechanism, or even the main one, at play. Indeed, if racial discrimination were the exclusive mechanism, we would also expect it to manifest in greater vertical mismatch among the descendants of non-European immigrants, which we do not observe. Therefore, specific mechanisms related to the choice of study fields likely operate as well.

Our analysis of the potential mechanisms underlying different levels of horizontal mismatch across generations and origin groups indicates the centrality of structural sorting in the educational system (H5). Sorting into fields of study explains the persistently greater horizontal mismatch among non-European men, but having vocational training and belonging to a discriminated group also play a role. Further, in line with past research (Rose 2005; Somers et al. 2019:583), we find that the descendants of non-European immigrants are overrepresented in fields of study that are not directly linked to occupations, which can explain persistently higher levels of horizontal mismatch.

The specialized literature on the educational trajectories of descendants of immigrants in France provides evidence of differentiated educational aspirations across origins and generations. Descendants of European immigrants (mostly from Southern Europe, especially Portugal) tend to be successfully pushed by their parents to acquire shorter specialized vocational qualifications and quickly enter the labor market. By contrast, descendants of non-European immigrants (especially from North Africa) often internalize their parents’ aspirations for more extensive studies in (undifferentiated) academic tracks (Brinbaum and Kieffer 2005; Palheta 2012). However, their lower socioeconomic background and poorer school performance, especially among boys, frequently prevent them from realizing these ambitions. They are therefore overrepresented among high school and university dropouts and in having fields of study that are less selective and more broad, such as the social sciences and humanities (Beaud 2013; Brinbaum and Cebolla-Boado 2007; Ichou 2018).

Discrimination and choices regarding fields of study are not necessarily mutually exclusive mechanisms. A recent correspondence test suggests that students with North African–sounding names are more likely penalized in their master's degree admissions (Chareyron et al. 2023) and that the level of discrimination increases with the prestige of the program. Thus, non-European men are potentially relegated to economically unviable fields of study, ultimately increasing the likelihood of experiencing horizontal mismatch. In other words, we suggest that both internalized academic aspirations and experiences of discrimination play a role in accounting for the durable disadvantage in horizontal mismatch experienced by men of non-European origin.

Finally, we observe tentative signs of a resurgence in vertical and horizontal mismatch levels in the third generation relative to natives. Given that these mismatch gaps among G3 never reach conventional levels of statistical significance (except among non-European men), we refrain from interpreting this pattern or speculating about its causes. Nevertheless, if future research confirms this trend, it would further challenge the optimistic conclusions suggested by assimilation theories.

Our study has some limitations that future research could help address. First, we compare generations whose parents (for G2) and grandparents (for G3) migrated to France at different periods, which might affect mismatch patterns across these groups. To address this concern, we control for observable differences in sociodemographic characteristics across groups. However, comparing educational mismatching across actual family generations would also be worthwhile. Second, we provide suggestive evidence that greater horizontal mismatch is negatively related to achievement in the labor market, likely leading to disadvantages among non-European men. However, further analysis is needed to address the potential negative labor market consequences of horizontal mismatch in the second and third generations. Finally, we are aware of the internal heterogeneity within the broad categories of European versus non-European origins. Larger samples will be needed to analyze this heterogeneity.

In conclusion, we make three key contributions. First, we extend educational mismatch research to the third generation. Second, we analyze vertical and horizontal mismatch, revealing a specific persistence in horizontal mismatch among grandsons of non-European immigrants. Finally, in addition to uncovering descriptive patterns of educational mismatch, we show that a major mechanism explaining persistently high horizontal mismatch among immigrants’ descendants is their overrepresentation in specific fields of study in the French educational system.

Acknowledgments

We are grateful to Xavier St-Denis, Lucas Drouhot, and Ugo Palheta for their comments on earlier versions of this article. We also thank three anonymous reviewers, participants at the 2023 annual meeting of the Population Association of America in New Orleans, the 2022 RC28 Spring meeting at the London School of Economics, and 3GEN-project team members for their helpful input. This work was supported by the Agence Nationale de la Recherche (grant ANR-20-CE41-0001) and Swedish Research Council for Health, Working Life and Welfare (Förskiningsrådet för Hälsa, Arbetsliv och Välfärd, grants 2021-00026 and 2023-00609).

Notes

References