Abstract
How does education–occupation mismatch shape racial/ethnic labor market inequality among highly educated workers? Bridging the literatures on racial/ethnic discrimination and labor market signaling, we propose a new concept, “racialized signaling,” to explain inequality in the college-to-work transition, operationalized through education–occupation mismatch. We then use longitudinal data to examine the labor market consequences of racialized signaling, analyzing vertical and horizontal dimensions of mismatch. We find that Black and Hispanic graduates experience the negative consequences of mismatch most strongly at the point of occupational allocation relative to their White peers, whereas Asian graduates experience the greatest negative consequences of mismatch regarding wage penalties. Advanced degrees, STEM degrees, and degrees from more selective institutions have some moderating effects, but they do not fully level the playing field for minority graduates. Overall, our findings suggest education–occupation mismatch is a powerful, although heterogeneous, mechanism reproducing racial/ethnic inequality among the most educated segment of the U.S. population.
Keywords
Over the past half-century, the rate of bachelor’s degree completion among Black and Hispanic individuals has grown substantially, doubling for each group and outpacing the rate of growth for White individuals (Snyder, de Brey, and Dillow 2018). However, substantial wage inequality by race and ethnicity remains (Cheng et al. 2019; Daly, Hobijn, and Pedtke 2017). In fact, the wage gap is largest and has expanded most among highly educated workers (Autor, Katz, and Kearney 2006). In 2015, the wage disparity between White and Black college graduates was 18 percent, compared to 8.8 percent in 1980 (Wilson and Rodgers 2016). During the same period, the wage gap between Hispanic and White college graduates rose from 24.5 percent to 34.2 percent (Mora and Dávila 2018).
These paradoxical trends raise questions about why the expanded rate of bachelor’s degree completion among racial minorities has not translated into greater equalization in the labor market. Scholars have typically studied racial/ethnic wage inequality by focusing separately on education (Huffman and Cohen 2004; Tomaskovic-Devey, Thomas, and Johnson 2005) or occupations (Kaufman 2002). However, research focusing on education or occupations alone has not provided an adequate explanation for increased racial/ethnic wage inequality among highly educated workers (Silva et al. 2021). Evaluating the (mis)match between education and occupation in the college-to-work transition rather than distinct educational qualifications or occupations is an important next step for research.
In the present study, we pursue this approach by examining how differences in the translation of educational credentials into labor market positions shape racial/ethnic labor market inequality among highly educated workers. Specifically, we analyze mismatch between educational qualifications and occupational positions. We advance the growing literature on education–occupation mismatch (Attewell and Witteveen 2023; Bol et al. 2019; Lu and Li 2021) by providing a comprehensive conceptual framework for understanding racial/ethnic differences in mismatch. This framework builds on insights from the literatures on racial/ethnic discrimination and labor market signaling, yielding a new concept, “racialized signaling,” to explain racial/ethnic differences in the experience and consequences of mismatch among highly educated workers.
Racialized signaling describes how race/ethnicity pervasively shapes the signaling value of college credentials and the stigma attached to negative employment experiences (i.e., mismatched employment). It may manifest functionally in concrete racial/ethnic differences in education–occupation mismatch across three employment stages: (1) initial occupational allocation (incidence of mismatch), (2) subsequent occupational trajectory (persistence of mismatch), and (3) economic returns (wage penalties of mismatch). We explore the extent to which several educational factors, including terminal bachelor’s degree versus advanced degrees, STEM versus non-STEM fields, and selective versus less selective institutions, moderate the racial/ethnic differences in education–occupation mismatch.
We pool two decades of longitudinal data from the Survey of Income and Program Participation (SIPP; panels from 1996, 2001, 2004, and 2008) to examine the extent to which minority college graduates are disproportionately channeled into mismatched occupations, confined in such positions, or penalized when mismatched relative to their White counterparts. Here, we move beyond a two-race dichotomy to include four main ethnoracial groups (non-Hispanic White, non-Hispanic Black, Hispanic, and Asian), allowing us to uncover how different racial/ethnic groups experience mismatch distinctively. In addition, our analytic strategy addresses two significant measurement and data challenges facing existing research: the imposition of arbitrary cutoff points to quantify mismatch and reliance on supply-side data. In all, our approach provides a comprehensive understanding of how race/ethnicity shapes labor market inequality among highly educated workers using the transition from college to work as our primary explanatory site.
The Context and Dimensions of Mismatch among College Graduates
Conventional wisdom assumes that highly educated workers match easily with commensurate labor market positions. This view, however, has come under scrutiny due in large part to the slowdown in high-skill employment opportunities since 2000 (Beaudry, Green, and Sand 2016). This slowdown has created an imbalance between the supply and demand for highly educated workers. In such a context, education–occupation mismatch among college graduates, both vertically and horizontally, has become more pervasive (Bol et al. 2019; Li and Lu 2023).
“Vertical mismatch” refers to mismatch between workers’ educational credentials and the level of education typically required for their occupation. In the context of this study, we are most concerned with vertical “undermatch,” or the underutilization of general human capital given educational level (Verhaest, Sellami, and van der Velden 2017). This circumstance often leads college graduates to take jobs that typically do not require a college degree (e.g., college graduates working as retail sales associates). Previous research across data sets and time periods suggests that such vertical undermatch—which we refer to as “vertical mismatch” throughout the article 1 —affects 20 percent to 35 percent of college graduates in the United States (Leuven and Oosterbeek 2011; McGuinness and Redmond 2018). Vertical mismatch limits a worker’s ability to convert human capital into productivity and commensurate economic rewards, resulting in wage penalties (Lu and Li 2021; McGuinness 2006).
“Horizontal mismatch” refers to mismatch between workers’ fields of study and the type of education required for their occupation. Horizontal mismatch captures the extent to which a worker’s knowledge and skills are congruent with the demands of their occupation. It occurs when college graduates work in occupations that are not closely related to their field of study (e.g., engineering majors working as accountants). Horizontal mismatch may arise from macro-level imbalances between supply and demand of specific types of skills (Verhaest et al. 2017) or the occupational specificity of a field. Horizontal mismatch may seem more relevant for workers in vocationally oriented fields (e.g., computer science), which have more clearly defined occupational pathways (Bol et al. 2019), than those in generally oriented areas of studies (e.g., humanities), which tend to be connected to a wider spectrum of professions. However, we argue that the concept of horizontal mismatch is meaningful when comparing college graduates within the same field of study, even in general fields—that is, whether individuals follow the common occupational pathways for their field of study more than others with degrees in the same field. Our analysis demonstrates that horizontal mismatch is prevalent and has notable wage implications for graduates across vocational and general fields of study. Hence, there is substantive value to identifying (mis)match even for more general fields of study.
Unlike vertical mismatch, horizontal mismatch is nominal. Its wage consequences depend on its specific character. “Horizontal undermatch” can result in wage penalties (e.g., engineering majors working as accountants). Undermatch is largely driven by involuntary factors, occurring when horizontally matched jobs are unavailable or when workers prioritize other job attributes, such as job prosociality (Kim 2024; Wilmers and Zhang 2022), over skill match. “Horizontal overmatch,” in contrast, can indicate that workers are employable outside a field and thus could pursue more remunerative career paths (e.g., social science majors working as corporate managers). Overmatch is typically voluntary, often initiated by workers to achieve career advancement (Bender and Heywood 2011; Robst 2007), especially through entry into managerial positions (Ishida, Spilerman, and Su 1997). Table 1 presents common examples of mismatch among college graduates in the United States.
Common Examples of Education–Occupation Mismatch, Survey of Income and Program Participation, 1996 to 2011.
Note: The term before the arrow (→) indicates field of study (major); the term following the arrow denotes occupation.
Existing research typically focuses on a single dimension of education–occupation mismatch (i.e., vertical or horizontal). We examine vertical and horizontal mismatch together, which allows us to identify the respective role of each dimension for racial/ethnic labor market inequality. Given our research interests, we do so by comparing experiences of mismatch among college graduates with degrees in similar fields of study. 2
Why Signaling of College Degree and Mismatch is Racialized in the Labor Market
To provide a framework for understanding how education–occupation mismatch shapes racial/ethnic inequality, we integrate two broad literatures, the first on racial/ethnic discrimination—particularly through the mechanism of social closure—and the second on signaling. This integration offers a rich new perspective on the reasons for racial/ethnic inequality in mismatch. We refer to this perspective as “racialized signaling.”
Extensive research on labor market discrimination by race/ethnicity highlights explicit, taste-based or implicit, stereotype-driven biases that disadvantage minority workers in the labor market (Pager, Bonikowski, and Western 2009; Quillian et al. 2017). Such discrimination extends to highly educated workers (Attewell and Witteveen 2023; Gaddis 2015). This literature draws on the concept of “social closure,” which posits that individuals tend to extend opportunities to those who share their own demographic characteristics (Ray 2019). A central mechanism in this literature is employers’ prioritization of “cultural fit” with the dominant (White) cultural systems prevailing in elite professions (Kanter 1977; Rivera 2012). Accordingly, White employers, consciously or unconsciously, may discriminate against non-White job candidates because of a perceived inability to “work well” with them. Such a perception allows White employers to rationalize their default to hiring same-race candidates.
Parallel research on the relationship between structural oversupply of education in the labor market and the signaling power of credentials suggests that oversupply and signaling interact to amplify racial discrimination among highly educated workers. Scholars have argued that workers can powerfully signal their value in competitive job markets through educational credentials (Bills 2003; Spence 1973). The pursuit of higher education is, in part, motivated by this logic—individuals seek college degrees to increase their competitiveness in the labor market. Yet this condition may create structural oversupply, meaning there are more highly educated workers than the labor market demands. In a saturated labor market, credentials may become less powerful signals of workers’ value (Horowitz 2018) because employers perceive a declining correlation between college education and productivity. This circumstance may lead employers to seek alternative mechanisms to make hiring decisions, such as defaulting to informal selection criteria (e.g., cultural matching). Insofar as White employers view race/ethnicity as a marker of fit and by extension, productivity (Moss and Tilly 2001; Ridgeway 2011), they may be less inclined to hire minority workers—especially when more workers hold higher education credentials than structurally supported by the labor market.
Combining these insights, we theorize that the value of labor market signals via credentials is racialized in today’s high-skill labor market—a process we term “racialized signaling.” Racialized signaling encapsulates two mechanisms that contribute to a racialized opportunity structure among highly educated workers. First, the positive signaling power of higher-education credentials may diminish disproportionately among minority workers due to structural oversupply and White-dominated hiring networks in a context of social closure. This process may result in less favorable treatment of minority graduates in the labor market than that experienced by White graduates. Functionally, this first mechanism likely manifests in heightened incidence of education–occupation mismatch because employers funnel minority graduates into lower level or less relevant labor market positions than White graduates.
Second, employers may interpret mismatched employment experiences as a labor market signal of low productivity. Previous research demonstrates that workers’ employment history shapes employers’ perceptions of their quality (Connelly et al. 2011; Weisshaar 2018). Because structural oversupply of college graduates reduces the positive signaling value of a bachelor’s degree, education–occupation mismatch may serve as a negative signal for college graduates (Pedulla 2016). Critically, employers may interpret this negative signal of mismatched employment differently by race/ethnicity, producing distinct labor market effects (e.g., employers may perceive mismatch as more damaging for minority workers than for White workers). The functional manifestation of this second process of racialized signaling can be differential persistence or wage penalties of mismatch, with minority workers likely experiencing more prolonged persistence or greater wage penalties than similar White workers.
In the present study, we do not directly test for racialized signaling because such a test is not feasible given our data. Instead, we follow previous research on signaling, which relies on analyzing its functional manifestation in labor market outcomes—in our case, how the incidence, persistence, and economic returns of mismatch vary by race/ethnicity.
Three Processes by Which Education–Occupation Mismatch Shapes Racial/Ethnic Inequality
Figure 1 summarizes our conceptual framework, which focuses on occupational allocation (incidence of mismatch), subsequent occupational trajectory (persistence of mismatch), and economic returns (wage penalties of mismatch).
Conceptual framework.
Incidence
Education–occupation mismatch naturally emerges with the structural oversupply of college graduates; a key question, then, is whether graduates are differentially allocated to mismatched positions along ethnoracial lines (Figure 1, Stage I). We hypothesize that racialized signaling of college credentials operates through employers’ differential evaluation of graduates based on race/ethnicity, especially in highly educated hiring contexts where most hiring managers are White. This process disadvantages minority applicants while favoring White applicants, manifesting in differential incidence of mismatch. 3
Hypothesis 1a: Highly educated minority workers are more likely than similarly educated White workers to experience initial mismatch in the labor market, entering positions below their educational level (vertical mismatch) or outside their field of study (horizontal undermatch).
Hypothesis 1b: Minority workers are less likely than White candidates to experience positive mismatch (i.e., horizontal overmatch) because they may experience greater scrutiny and are perceived as less culturally “fit.”
We additionally investigate factors that may alleviate racial/ethnic inequality in the incidence of mismatch, focusing on three theoretically important variables: degree level, field of study (i.e., STEM vs. non-STEM), and college selectivity (see Part A of the online supplement for a comprehensive discussion of the related theoretical and empirical literatures). We hypothesize that these three variables may moderate the greater incidence of vertical and horizontal mismatch among minority graduates.
Hypothesis 1c: Minority workers who have (1) achieved postbaccalaureate education (master’s degree or higher), (2) majored in a STEM field, or (3) attended a selective college are less likely than other minority workers to experience vertical or horizontal mismatch.
Persistence
Accumulating evidence suggests that mismatch tends to persist beyond its initial incidence (Dolton and Vignoles 2000; Lu and Li 2021; Pedulla 2016). This pattern emerges because employers perceive mismatched employment as a negative signal about competence and skill deterioration (Cockx and Picchio 2013). These perceptions make mismatched workers less desirable than fully matched workers in future employment opportunities, consigning them to such positions for an extended period of time.
It remains unclear whether there are racial/ethnic differences in the persistence of mismatch (Figure 1, Stage II) above and beyond the cumulative effects potentially resulting from initial racial/ethnic differences in incidence. Minority graduates, who are more vulnerable to employers’ less favorable evaluations due to discrimination via social closure and negative competence signals, may face greater challenges in securing matched jobs throughout their careers. One result would be persistent mismatch. White graduates may face fewer obstacles recovering from mismatch, but research shows a strong and consistent negative signaling effect of mismatch overall (Pedulla 2016). This research suggests mismatch is an enduring challenge for all college graduates: Once they experience it, they may struggle to realign regardless of their characteristics. We therefore propose two contrasting hypotheses:
Hypothesis 2a: Education–occupation mismatch is more likely to persist among minority graduates.
Hypothesis 2b: Education–occupation mismatch demonstrates high persistence across all ethnoracial groups.
Wage Penalty
The third process through which education–occupation mismatch may affect ethnoracial wage inequality is differential wage penalties associated with mismatch by race/ethnicity (Figure 1, Stage III). The wage penalty measures forgone wages associated with mismatched employment, that is, the wage differences between mismatched and matched workers who hold similar credentials. Prior work documents a general wage penalty of mismatch (Leuven and Oosterbeek 2011; McGuinness 2006), but little is known about whether the wage penalty varies by race/ethnicity.
Existing research on negatively stereotyped groups provides some insight into how ethnoracial differences may affect wages. For members of negatively stereotyped minority groups (i.e., Black and Hispanic workers; Moss and Tilly 2001; Pager et al. 2009), multiple negative status categories may reinforce one another and thus result in multiplicative, negative effects on wages. These effects may occur because mismatch reaffirms employers’ deeply held negative stereotypes about the competence and commitment of minority workers (Karren and Sherman 2012). Such a process would amplify the negative signal sent by mismatched employment, leading minority graduates to receive steeper wage penalties than similarly mismatched White graduates.
In contrast, other research suggests that stereotypes associated with multiple negative status categories often overlap; when this occurs, it can diminish the amplification of effects (Correll and Ridgeway 2003). We consider this scenario, often referred to as “muted congruence” (Pedulla 2018), to be more likely among highly educated workers because the negative stereotypes associated with mismatch tend to be highly congruent with those associated with Black and Hispanic workers. Also, because mismatch is more prevalent and normative among minority workers, the overlap between these statuses becomes even more likely. In this scenario, mismatch may bear limited additional penalties for Black and Hispanic workers, limiting the wage penalties for these groups.
The wage consequences of mismatch may unfold differently for positively stereotyped minority groups (i.e., Asian workers). The positive stereotypes associated with being Asian may conflict with the negative stereotypes associated with mismatch, signaling their deviation from the “model minority” norm (Lee and Zhou 2015). Mismatched employment thus may send a stronger negative signal about Asian graduates’ qualities and commitment, leading to higher wage penalties than their White counterparts experience.
Hypothesis 3: Education–occupation mismatch should result in similar or weaker wage penalties for negatively stereotyped ethnoracial groups (i.e., Black and Hispanic graduates, overlapping statuses) but larger wage penalties for positively stereotyped minority groups (i.e., Asian graduates, conflicting statuses) compared to White graduates.
We attend to potential heterogeneity in the incidence and consequences of mismatch by workers’ specific minority identification (i.e., Black, Hispanic, or Asian) in Part B of the online supplement.
Data and Methods
Data
For the main analyses, we used longitudinal data representative of the U.S. population from the 1996, 2001, 2004, and 2008 panels of the SIPP. Each SIPP panel contains three to five years of observations. Households were interviewed every four months (a wave), resulting in a total of 9 to 15 waves. 4
SIPP has several merits for our research purpose. First, it provides a large sample and detailed employment information for an ethnically diverse sample. Second, SIPP panels cover multiple birth cohorts rather than a single birth cohort. The drawback, however, is that SIPP panels cover short- to medium-term outcomes, meaning we cannot directly examine longer-term mismatch patterns.
The analytic sample includes individuals with at least a bachelor’s degree who were ages 23 to 55 during the observation window. 5 We apply the age restriction to retain sufficient sample size and to capture the experience of workers across a wide age spectrum. Following previous research, we include individuals under age 55 to focus on people who are strongly attached to the labor market (Di Stasio, Bol, and Van de Werfhorst 2016). Although we control for age and year in all analyses, we do not stratify by age or survey year—both because we wish to retain a sufficiently large sample size and because additional analyses reveal relatively stable age and period trends.
We use several other restrictions, retaining only workers who are continuously employed and who have complete information on occupation throughout the first eight waves of each panel. We eliminate workers with unemployment spells because this phenomenon is relatively uncommon among educated workers and sensitivity analyses suggest no clear relationship between unemployment and mismatch. We retain cases with complete information through eight waves because a large proportion of the sample diminishes beyond that point. 6 We exclude individuals who are in school, disabled, or in the military at any time during the panel because they are weakly attached to the labor market. We also exclude self-employed people to focus on wage workers.
We combine men and women in the main analyses to increase the sample size, but we control for sex in all analyses (additional analyses show similar ethnoracial differences in mismatch across men and women). We pool all four panels of SIPP, including the 2008 panel that overlapped with the Great Recession. We conducted a sensitivity analysis dropping the 2008 panel, which yielded consistent results.
The final analytic sample contains 91,544 observations from 11,443 individuals across four SIPP panels. About 84.7 percent of the sample (college graduates) identified as White, followed by 6.2 percent Asian, 5.6 percent Black, and 3.5 percent Hispanic. This distribution aligns with the distribution of workers with bachelor’s degrees reported in the American Community Survey.
Measuring Education–Occupation Mismatch
We construct categorical and continuous mismatch measures at the individual level using matching standards derived from the American Community Survey (ACS; supply side) and job-posting data from Burning Glass Technology (demand side). We distinguish 465 occupations based on the 2000 Census Occupational Classification System, seven educational levels, and 22 ISCED fields of study.
We begin with categorical measures, which are commonly used in the literature because of the simplicity in conceptualizing and interpretating them. The categorical measures require first determining the typical or “matched” educational requirements (level or field) of an occupation and then comparing each respondent’s education with the matched education. We use supply- and demand-side measures to provide a more comprehensive picture.
For the main measure, we use the “realized match” approach to first determine the matched educational qualifications (level or field) for an occupation (we provide our rationales for choosing this approach in Part C of the online supplement). The realized match approach entails observing the educational distribution of workers in occupations in the ACS data (see Part D of the online supplement) and then comparing respondents’ (in SIPP) actual education with the matched education for their occupation. The matched educational qualification for an occupation is derived from what workers in an occupation have typically attained, namely, the mode or mean of that distribution. Mode is preferred over mean because it is less sensitive to outliers and technological change (Kiker, Santos, and de Oliveira 1997).
Using the matched education criteria derived from the pooled ACS data set, we then construct individual-level mismatch measures by merging the matched education qualification to individual-level data in SIPP based on an individual’s actual education (level or field) and occupation in each wave. For vertical mismatch, we use the mode educational level as the matched level for each of the 465 occupations. College graduates who reported working in occupations that typically require subbaccalaureate education are classified as vertically mismatched. We do not further distinguish occupations requiring different baccalaureate and postbaccalaureate degree levels (i.e., bachelor’s, master’s, and PhD or professional degrees) because few occupations (5.3 percent) require an advanced degree.
For horizontal mismatch, we use the two most common fields of study as the matched fields for each occupation, following previous research (Bol et al. 2019). We classify individuals whose field is outside the top two matched fields as horizontally mismatched. We conducted sensitivity analyses defining match with varying numbers of matched fields of study for an occupation and obtained similar results.
We further distinguish horizontal undermatch (negative) and overmatch (positive) using median occupational income derived from the ACS, comparing the median income of the respondent’s actual occupation with that of matched occupations for their field of study in the ACS. In horizontal undermatch, workers hold out-of-field occupations that typically have lower wages than matched occupations. In horizontal overmatch, workers hold out-of-field occupations that typically have higher wages than matched occupations.
The main measure of mismatch is derived from supply-side data (using the realized match approach) and is categorical in nature. We supplement this conceptually simpler categorical mismatch measure with two additional measures—a categorical measure based on data from the demand side and a continuous measure based on a new approach to studying school-to-work linkage (DiPrete et al. 2017). The demand-side measure relies on educational requirements in online job postings. The continuous measure avoids arbitrary cutoff points and allows us to study racial/ethnic differences in the degree of mismatch along a continuum. More details about the demand-side and continuous measures are in Parts D and E, respectively, of the online supplement.
Other Variables
In the incidence analysis, the dependent variables are the categorical or continuous measures of mismatch. The key independent variable is race/ethnicity. We distinguish four ethnoracial categories: non-Hispanic White, non-Hispanic Black, Hispanic, and Asian. We drop a small percentage of respondents (1.5 percent) in the “other” category. Other demographic control variables include age (and age squared), gender, education (bachelor’s degree, master’s degree, and doctoral degree), 7 nativity (native-born, immigrants with U.S. degree, and immigrants with foreign degree), marital status, whether an individual has children, survey year, and geographic variables (metropolitan area and region). We also include a number of independent variables that enable us to make granular comparisons across individuals with specific educational and labor market experiences (e.g., field of study and work experience). More details are provided in Part F of the online supplement; the descriptive statistics are presented in Table 2.
Descriptive Statistics of Variables Used in the Analysis, Survey of Income and Program Participation, 1996 to 2011.
Note: The analysis is based on individuals with at least a bachelor’s degree. Percentages are shown for categorical variables. Means and standard deviations (in parentheses) are shown for continuous variables.
Methods
Incidence of mismatch
We estimate longitudinal random-effects models, both linear and logistic, to examine the incidence of mismatch while taking into account correlation of within-person observations. We study the different dimensions of mismatch by estimating separate models for vertical and horizontal mismatch. We control for horizontal mismatch in the model predicting vertical mismatch, and vice versa, to account for the possibility that both types of mismatch may co-occur. For moderation analyses, we include interaction terms between race/ethnicity and each potential moderator (advanced degree, STEM field, and selective institution). We present average marginal effects (AMEs) in all models to enhance clarity of interpretation (Mize 2019). To account for the possibility that racial/ethnic differences in the incidence of mismatch are partially attributable to unobserved heterogeneity, we conduct sensitivity analyses using longitudinal system GMM (generalized method of moments) models. The results are largely similar, increasing our confidence in the main results. Further details of the GMM models are provided in Part G of the online supplement.
Persistence of mismatch
To examine persistence of mismatch over time, we use sequence analysis (Aisenbrey and Fasang 2010) for categorical mismatch measures and group-based trajectory analysis (Nagin et al. 2018) for continuous mismatch measures. We first construct short- to medium-term mismatch trajectories during the observation window and then classify these trajectories into common patterns. Details are in Part H of the online supplement.
To investigate racial differences in the persistence of mismatch, we use the mismatch trajectories identified in sequence analysis or group-based trajectory analysis as the outcome variable. We estimate multinomial logit models predicting mismatch trajectories based on race/ethnicity and the covariates noted above.
Wage consequences of mismatch
We estimate longitudinal random-effects models to examine racial differences in the wage consequences of education–occupation mismatch. The models predict log hourly wage based on mismatch, race/ethnicity, and other covariates. We also include an interaction term between mismatch status and race/ethnicity to capture racial differences in the wage penalties of mismatch. This formulation allows us to examine whether the forgone wages associated with mismatch vary by race/ethnicity. As robustness checks, we also estimate GMM models that adjust for potential endogeneity bias due to unobserved productivity-related characteristics.
Results
Notable Racial/Ethnic Differences in the Incidence of Mismatch
Table 3 presents descriptive statistics on mismatch status by race/ethnicity across all waves of SIPP. Both vertical and horizontal mismatch are common among highly educated workers, but Black and Hispanic college graduates experience a greater degree of mismatch than do White and Asian graduates. Interestingly, there are marked differences by horizontal overmatch (positive) and undermatch (negative). White graduates are most likely to achieve horizontal overmatch, whereas Black and Hispanic college graduates and to a lesser degree, Asian graduates are more likely to experience horizontal undermatch.
Percentage of Mismatch among College Graduates by Race/Ethnicity, Survey of Income and Program Participation, 1996 to 2011.
Note: The sample is individuals with at least a bachelor’s degree. Mismatch is measured using the supply-side categorical approach with the realized match method.
We examine the racial/ethnic differences in the incidence of mismatch more systematically in a regression framework. The first column in Figure 2 presents results regarding vertical mismatch among highly educated workers with the same degree level and field of study. The results point to noteworthy differences by race/ethnicity. Black and Hispanic graduates experience a higher risk of vertical mismatch than do similarly educated White college graduates. The Asian-White difference is not statistically significant. Across all models, the difference between Black and Hispanic graduates is not statistically significant. We also conducted an analysis examining the implications of structural oversupply of college graduates for vertical mismatch, finding, in line with our concept of racialized signaling, that vertical mismatch is larger for the three minority groups of graduates when structural supply increases (see Part I of the online supplement).
Incidence of mismatch by race/ethnicity, Survey of Income and Program Participation, 1996 to 2011.
Turning to horizontal mismatch (second column, Figure 2), Black college graduates appear most vulnerable to overall horizontal mismatch: The AME is statistically significant for Black workers but not for other racial groups. Because these results do not separate horizontal overmatch from undermatch, it is impossible to tell if the greater incidence of horizontal mismatch among Black graduates derives from overmatch, undermatch, or both.
We gain greater clarity on this front by examining the results for horizontal overmatch and undermatch (third and fourth columns, Figure 2). Among individuals with similar educational credentials, Black and Hispanic graduates are more likely to experience horizontal undermatch than their White peers, meaning they are disproportionately mismatched to out-of-field occupations with lower pay than matched occupations. Hispanic graduates are also less likely to experience horizontal overmatch than their White peers. The Asian-White difference is not statistically significant. Because racial differences in horizontal mismatch depend on the type of mismatch, we focus on the results that separate horizontal undermatch from overmatch in the following.
The coefficients for the main models are presented in Part J of the online supplement. The main findings are consistent across both demand-side and continuous measures of mismatch (see Figures K1 and K2 in the online supplement). We also carried out an analysis restricting the sample to terminal bachelor’s degree holders only (see Figure K3). The results are largely similar, especially for Black and Hispanic graduates, who continue to experience higher rates of vertical mismatch and horizontal undermatch.
In an additional analysis (see Figure K4 in the online supplement), we developed a combined mismatch variable that incorporated both vertical and horizontal mismatch statuses; this analysis yielded similar results. Specifically, minority graduates, especially Black and Hispanic graduates, are more likely to experience vertical mismatch alongside horizontal undermatch (the most negative combination) and less likely to experience vertical match alongside horizontal overmatch (the most positive combination). In addition, Black workers are more likely to experience vertical match alongside horizontal undermatch. Based on the similarity of these results, we continue to use two separate variables for vertical and horizontal mismatch in the main analyses because of the greater interpretability of the results.
Overall, the findings demonstrate that Black and Hispanic college graduates, compared to White graduates, face greater difficulties in translating bachelor’s degrees and particular fields of study into commensurate occupational positions (consistent with Hypotheses 1a and 1b). These findings are robust to all the different measures of mismatch we incorporate. Robustness checks using the GMM models (see Figure L1 in the online supplement) show largely consistent results with the corresponding random-effect models (Figure 2), suggesting the findings are robust to endogeneity bias.
Limited Moderating Effects
Figure 3 presents the moderating effect of several educational factors. We find that degree level plays a limited moderating role in education–occupation mismatch. The Black-White disparity in vertical mismatch and horizontal undermatch is notably less pronounced among advanced degree holders (nonsignificant) compared to bachelor’s degree holders (positive and significant). This finding suggests advanced degrees provide Black graduates with some protection against vertical mismatch and horizontal undermatch. The Hispanic-White gap in horizontal undermatch is significant for bachelor’s degree holders and nonsignificant for advanced degree holders, but the difference between the two groups is not significant. In other words, the racial gap in different types of mismatch is similar among individuals with bachelor’s degrees and those with advanced degrees. Among Hispanic graduates, the difference in vertical mismatch between individuals with bachelor’s degrees and those with advanced degrees is also not significant.
Moderation role of advanced degree, STEM degree, and college selectivity on the incidence of mismatch by race/ethnicity, Survey of Income and Program Participation, 1996 to 2011.
With respect to fields of study, the racial gap in vertical mismatch for Black and Hispanic graduates is larger among STEM majors than among non-STEM majors. Despite a growing number of Black and Hispanic STEM graduates, such qualifications do not necessarily offer them the same leverage as their White peers in obtaining college-level positions. This finding may result from workplace discrimination, lack of networking, and limited diversity initiatives in STEM fields. Yet the potential equalizing effect of STEM fields is more evident with respect to horizontal undermatch: Black and Hispanic STEM graduates fare similarly to their White peers regarding matched employment, whereas their counterparts in non-STEM fields face significantly higher risk of horizontal undermatch. Taken together, these results mean Black and Hispanic STEM graduates are just as likely as their White counterparts to secure jobs related to their majors, although some of these positions may not require a college degree (e.g., graduates with computer science and engineering degrees working as mechanics).
Finally, we find only a limited moderating effect of college selectivity on occupational mismatch. Among graduates from nonselective and selective colleges, the Black-White gap in vertical mismatch is statistically similar. For horizontal undermatch, the racial gap is significant for graduates from nonselective colleges and nonsignificant for those from selective colleges, but the difference between the nonselective and selective categories is not statistically significant. 8
Overall, these results suggest that advanced degrees, STEM degrees, and degrees from selective colleges do not fully level the playing field for Black and Hispanic college graduates regarding education–occupation mismatch, providing limited support for Hypothesis 1c. We found no statistically significant differences between Asian and White graduates across various dimensions of mismatch regardless of degree type, field of study, or institutional selectivity.
Persistence of Mismatch
General persistence of mismatch
To assess the general persistence of mismatch, we use sequence analysis for the categorical mismatch variables. Panel A in Figure 4 displays the sequence index plot for the four general patterns of vertical mismatch. Individuals are arrayed along the y-axis, with each horizontal line representing one individual sequence. The x-axis is time (first to the eighth wave). The figure shows a large share of highly educated workers continuously holds positions commensurate with their education level (Cluster 1; 70.5 percent), but a notable proportion of workers consistently experience vertical mismatch during much of the observation window (Cluster 2; 24.3 percent). Only a small fraction of mismatched workers (2.6 percent, Cluster 3) are able to transition out of a mismatched position during the three years under study. Another 2.7 percent of workers (Cluster 4) start out matched but later fall into vertical mismatch. 9
Sequence index plots of (mis)match trajectories, Survey of Income and Program Participation, 1996 to 2011.
Panel B in Figure 4 displays the sequence index plot for horizontal mismatch. Horizontal over- and undermatch both appear to be highly stable, representing about 27.1 percent (Cluster 2) and 27.4 percent (Cluster 3) of the college-graduate sample, respectively. In contrast, transitions from overmatch to undermatch or vice versa occur infrequently, representing only 1.6 percent (over- to undermatch; Cluster 4) and 1.8 percent (match or undermatch to overmatch; Cluster 5) of college graduates. Another 1.8 percent of graduates transition from match to undermatch (Cluster 6). 10 Overall, the results underscore a high level of persistent mismatch during the short- to medium-term under study. Results for continuous mismatch measures using group-based trajectory analysis are consistent (see Part M of the online supplement).
To shed light on the longer-term persistence of mismatch, we construct a synthetic cohort of educated workers using the 1996 and 2008 panels. We compare the mismatch patterns of workers 23 to 35 years old in the 1996 panel with those ages 35 to 47 in the 2008 panel. Over the 12-year period, the percentage of workers in the persistent vertical mismatch cluster decreases from 26.3 percent to 24.8 percent, a small reduction of 1.5 percent. The percentage of workers in the persistent horizontal mismatch cluster in fact increases from 53.6 percent to 57.9 percent; such an increase occurs for horizontal overmatch and undermatch. These findings provide suggestive evidence that mismatch tends to persist into the longer term.
Limited racial/ethnic differences in persistence
To answer the question of whether there are racial/ethnic differences in the persistence of mismatch, we use multinomial regressions to estimate mismatch cluster membership based on race/ethnicity and other covariates (persistent match, either vertical or horizontal, is the base category). Results for the categorical mismatch measures based on sequence analysis clusters are presented in Model 1 (vertical mismatch) and Model 2 (horizontal undermatch and overmatch) in Part N of the online supplement.
The results suggest there is no racial/ethnic difference in the persistence of mismatch, which supports Hypothesis 2b and not Hypothesis 2a. Although we see racial differences in the clusters representing persistent vertical mismatch and horizontal undermatch in the top panel of Figure N in the online supplement, these differences disappear after controlling for first-wave mismatch status (see the bottom panel of Figure N). This pattern holds across all mismatch measures. We also find a high degree of state dependence of mismatch, as shown in the positive and highly significant coefficient of graduates’ first-wave mismatch status in all result panels. The GMM models in Figure L2 in the online supplement offer additional evidence in support of this conclusion.
In summary, the results demonstrate the substantial difficulties of securing a matched occupation once graduates have entered a mismatched position, regardless of race/ethnicity. The disadvantages facing highly educated minority college graduates (especially Black and Hispanic graduates) mainly stem from the fact that they face a higher risk of being channeled into mismatched positions in the first place (Stage I).
Racial Differences in the Wage Penalties of Mismatch
Figure 5 shows results regarding ethnoracial differences in the wage penalties of mismatch. The main effects of mismatch in these models confirm the findings in previous research documenting notable wage penalties attached to vertical and horizontal mismatch. The wage penalties can be interpreted as forgone wages for mismatched college graduates compared to what they would have earned in a matched position. Specifically, the wage penalty for vertical mismatch among White college graduates is large, almost 12 percent. For horizontal mismatch, undermatch carries a penalty of 10.1 percent for White graduates, and overmatch carries a wage premium of 2.7 percent. 11
Wage penalties of mismatch by race/ethnicity, Survey of Income and Program Participation, 1996 to 2011.
Turning to interactions between race/ethnicity and mismatch, which are our focus, the results point to some racial differences in the wage penalties of mismatch. The interaction terms between race and different types of mismatch are generally nonsignificant, except for Asian (vertical and horizontal mismatch) and Hispanic (horizontal mismatch) graduates. We interpret these results to mean that highly educated Asian workers receive more severe wage penalties from vertical mismatch than do similarly situated White workers and that Asian and Hispanic graduates are more adversely penalized by horizontal undermatch. Statistically, there appears to be no additional penalty for mismatched Black graduates relative to mismatched White graduates and no racial difference with respect to the wage premiums associated with horizontal overmatch (except for Hispanic graduates, who receive a lower wage premium than do White graduates; but the difference becomes nonsignificant in the GMM models).
The corresponding GMM results are largely similar, especially with respect to the notably greater wage penalties for Asian college graduates (see Figure L3 in the online supplement). That said, the disadvantage experienced by Hispanic graduates seems to disappear after accounting for unobserved heterogeneity in GMM models. These results indicate that unobserved heterogeneity is unlikely to fully explain the observed differences in wage penalties of mismatch with the exception of horizontal mismatch for Hispanic workers. For Hispanic graduates, a negative selection into horizontal mismatch may explain the previous result.
Overall, the results consistently show that mismatched Asian college graduates are more severely penalized than their White peers with similar educational credentials and in similar mismatched occupations (consistent with Hypothesis 3). This pattern does not hold for mismatched Black and Hispanic graduates, potentially because mismatch offers a weaker negative signal about Black and Hispanic workers’ productivity—consistent with literature suggesting overlapping effects of negative status categories. 12 Conversely, Asian graduates, who do not face heightened risks of mismatch (during Stage I), are penalized particularly severely when they fall into mismatched positions. This result perhaps occurs because mismatched occupational status challenges the “model minority” norm and thus amplifies the negative signaling effect of mismatch.
Discussion and Conclusions
This study investigates education–occupation mismatch as a potential source of racial/ethnic inequality among highly educated U.S. workers. We identify two dimensions of mismatch, vertical and horizontal, as important but previously underexplored drivers of racial/ethnic inequality among college graduates. Vertical and horizontal mismatch are prevalent features of the U.S. labor market, affecting, respectively, one-quarter and more than half of college graduates. More importantly and alarmingly, the experience and effects of mismatch are not uniform: Mismatch disproportionately disadvantages all main minority groups, either through a higher incidence of initial mismatch or through greater wage penalties associated with mismatch.
By examining the functional manifestation of racialized signaling processes in the incidence (allocation), persistence (mobility), and wage penalties of mismatch (economic return), we find heterogeneous inequality-generating processes in the high-skill labor market. In the initial allocation stage, Black and Hispanic graduates are less likely to translate their educational gains into commensurate occupational positions: These graduates are disproportionately channeled into occupations that are either below their educational level or are not closely related to their field of study (and less lucrative); they are also less likely to attain out-of-field but more lucrative positions. Several educational factors, such as earning advanced degrees or STEM degrees or graduating from selective institutions, play a limited moderating role but are unable to close racial/ethnic inequalities in the college-to-work transition.
Concerning graduates’ subsequent labor market trajectories, our study demonstrates that although different types of education–occupation mismatch are generally persistent, the level of persistence does not differ systematically by race/ethnicity. Even so, because Black and Hispanic graduates face greater risks of mismatch in the initial allocation, a disproportionate number of them are caught in mismatch in the longer term and adversely affected. In the economic returns process, minority college graduates are similarly, if not more penalized by mismatch than their White peers, but there is notable heterogeneity depending on how multiple social status categories (i.e., race/ethnicity and mismatch) combine to shape the wage returns attached to mismatch. Here, Asian workers are most penalized not only compared to White graduates but also relative to their Black and Hispanic peers. Taken together, racial/ethnic inequality among highly educated workers is largely a result of differential placement into mismatch early in one’s career compounded by the general persistence of mismatch and differential wage penalties driven by racialized signaling processes. The fact that differential placement early in a career is the primary driver of racial/ethnic inequality among college graduates underscores the crucial role of structural oversupply at labor market entry in shaping career trajectories and outcomes. Without structural oversupply, the patterns of initial mismatch we observe may not prove as prevalent.
This study makes several contributions to the literature. First, we demonstrate that the translation of educational credentials into labor market positions can serve as a central source of labor market inequality among the growing and diversifying college-educated workforce. For low-skill workers, racial inequality often stems from unemployment or nonstandard employment (Kalleberg 2009). Among college graduates, in contrast, inequality centers on education–occupation mismatch and associated economic costs. In this way, education–occupation mismatch represents a crucial mechanism that effectively reproduces racial/ethnic inequality in the high-skill labor market.
Second, we integrate theories of racial/ethnic discrimination, social closure, and labor market signaling to introduce a new concept, racialized signaling. We argue that the signaling strength of educational credentials and the stigma attached to mismatched employment are shaped by widely held perceptions of specific racial/ethnic groups—especially in labor market contexts characterized by structural oversupply. With more credentialed workers available than structurally needed, employers are likely to default to other hiring strategies than the evaluation of credentials alone. Existing literature on hiring inequality indicates that taste-based hiring, an important form of social closure, is one such strategy frequently used by (White) employers, resulting in advantages for White job applicants. This process, although often implicit, amplifies discrimination against minority workers and generates path dependencies that reinforce existing patterns of racial/ethnic inequality. These patterns persist due to the stigma attached to initial occupational mismatch, disadvantaging minority workers in the long run.
Third, we study all main ethnoracial minority groups in the United States. This approach allows us to show that although all three groups of minority college graduates are disadvantaged relative to their White peers, the inequality-generating process is not uniform. Instead, each minority group is disadvantaged to different degrees and through different mechanisms. One of the most consistent findings is that Black and Hispanic college graduates are more likely to experience vertical mismatch and horizontal undermatch and are less likely than their White peers to experience horizontal overmatch. Yet these graduates are not more adversely penalized economically by mismatched employment than are White graduates. Hence, the disadvantage facing highly educated Black and Hispanic workers largely stems from their higher risk of initial placement into mismatched positions, which tends to produce persistent mismatch and ongoing wage penalties. In contrast, the Asian-White disparity largely occurs due to the steeper wage penalties experienced by mismatched Asian workers: Asian graduates, who do not encounter heightened incidence of mismatch, face disproportionate forgone wages when they become mismatched. Hence, although Asian workers are often portrayed as excelling in the labor market, our results make clear that they still do not achieve outcomes on par with their White counterparts, especially when they deviate from positive stereotypes and normative expectations.
At the methodological level, we make systematic distinctions across different dimensions of education–occupation mismatch (vertical vs. horizontal and undermatch vs. overmatch within horizontal mismatch). Such distinctions are important because different dimensions of mismatch produce different labor market consequences. Horizontal mismatch is more common but tends to be less (economically) costly than vertical mismatch. Also, although we see little racial/ethnic difference in overall horizontal mismatch, we find notable group differences when separating different qualities of horizontal mismatch (i.e., undermatch vs. overmatch). For Black and Hispanic graduates, horizontal mismatch most often manifests in undermatch (downgrade). For White graduates, horizontal mismatch often indicates an overmatch (upgrade).
Another methodological contribution is our use of dichotomous and continuous and supply- and demand-side measures of education–occupation mismatch. These measures have unique strengths and together provide complementary information. Our main findings are quite consistent across measures, suggesting the patterns and consequences of mismatch arise with respect to both the typical occupational pathways and the degree of mismatch. This consistency strengthens our confidence in our substantive conclusions.
We note several limitations of this study. First, we conceptualize mismatch as a mechanism through which racialized signaling disadvantages highly educated minority workers. The data do not offer direct evidence on this process. But our results, particularly regarding patterns of mismatch in conditions of structural oversupply, align with the theoretical predictions of the racialized signaling process. Future research should investigate racialized signaling directly, especially through studies that focus on employer decision-making in regard to intersections between race/ethnicity, educational credentials, and employment history in the context of structural oversupply.
Also, we focus on matching between education and occupations. Whereas occupations have been a main locus of inequality research, even the most detailed occupational categories used in survey data tend to conceal heterogeneity in jobs within occupations. Ideally, we would like to study mismatch at the more refined job level. Yet this approach is not feasible because large-scale national longitudinal data do not provide information on jobs. Despite this limitation, the racial/ethnic differences should hold if conducting analyses at the job level. In fact, disparities between White and minority graduates may become even more pronounced insofar as the same racialized signaling process disproportionately channels minority college graduates into lower-paying firms or lower-level jobs within an occupation (Grodsky and Pager 2001; Storer, Schneider, and Harknett 2020).
Despite these limitations, this study represents a step toward understanding labor market inequality within the increasingly diverse educated workforce of the United States. The finding that highly educated minority graduates face greater difficulties than their White peers in translating educational credentials into commensurate occupational positions and rewards suggests that obtaining a college degree is insufficient to fully level the playing field. Our results also demonstrate high persistence of education–occupation mismatch, raising concerns about the common wisdom that selecting into mismatched positions to accumulate training and work experience may enable future upward mobility. This strategy may prove ineffective because the cycle of mismatched employment can be difficult to break.
As a final, cautionary note, it is important to emphasize that the problem of mismatch should not lead us to dismiss the value of a college education. We echo extensive research that documents the pecuniary and nonpecuniary value of higher education (Hout 2012). In the future, it will be even more difficult to secure a good job without a college degree. Even when minority college graduates fail to find a matched job, they still retain a competitive edge in the labor market over their less educated co-ethnic peers. In this respect, racial/ethnicity inequality in education–occupation mismatch may aggravate overall racial inequality: As highly educated minorities move down the occupational ladder, they squeeze out their less educated co-ethnic peers who have traditionally occupied the lower-level jobs, pushing these workers even further down the ladder or out of the labor force all together. Our findings suggest that closing racial/ethnic economic gaps should not end at advocating for diversity in higher education: Closing racial/ethnic economic gaps will require institutional supports to smooth transitions from college to work and addressing inequality in college-to-work pathways more broadly.
Supplemental Material
sj-docx-1-soe-10.1177_00380407251356275 – Supplemental material for Underemployed and Penalized: Education–Occupation Mismatch and Racial/Ethnic Inequality among Highly Educated Workers
Supplemental material, sj-docx-1-soe-10.1177_00380407251356275 for Underemployed and Penalized: Education–Occupation Mismatch and Racial/Ethnic Inequality among Highly Educated Workers by Yao Lu, Xiaoguang Li and Christina Ciocca Eller in Sociology of Education
Footnotes
Acknowledgements
The authors appreciate the valuable feedback from seminar participants at New York University and the University of Haifa and from the editor and anonymous reviewers.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors gratefully acknowledge funding support from the Washington Center for Equitable Growth, the Columbia Population Research Center, and the Institute for Empirical Social Science Research (IESSR) at Xi’an Jiaotong University.
Research Ethics
This study utilizes secondary data sets and does not involve human subjects. It was exempt from Institutional Review Board review.
Supplemental Material
Supplemental material for this article is available online.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
