Understanding Inequity in Graduation Rates at Hispanic-Serving Institutions (HSIs): An Intersectional Analysis by Race, Gender, and First-Generation College Status

Abstract

We examine complex inequities that emerge when race, gender, and first-generation college status are treated as interdependent, rather than independent statuses, for assessing student outcomes at Hispanic-Serving Institutions (HSIs). Drawing on student-level administrative data from two public HSIs in the U.S. Southwest, we analyze four-year graduation and placement in developmental English and mathematics. Using continuing-generation college white women as the reference group, we estimate marginal effects and then construct linear combinations for twenty intersectional social locations defined by race, gender, and first-generation college status. Our findings show that first-generation American Indian men, first-generation college Black men, and first-generation college Hispanic men experience some of the largest achievement gaps in both graduation and developmental placement, gaps that would remain obscured in conventional reporting by race, gender, or class alone. We argue that quantitative intersectionality, grounded in critical race and intersectionality scholarship, offers a value-added approach to state-based institutional analytics that can inform equity metrics, accountability systems, and resource allocation at HSIs and beyond. We conclude with recommendations for redesigning data infrastructures, reporting practices, and equity initiatives to better align HSI servingness with the lived realities of structurally marginalized students.

1. Introduction

As Collins and Bilge explain, “intersectionality is a way of understanding and analyzing the complexity of the world, of people, and of human experience” (Collins and Bilge 2020). Rather than treating race, gender, or class as separate axes of inequality, intersectionality directs attention to how these dimensions work together in specific contexts to shape power relations, opportunities, and outcomes. Intersectionality is critical inquiry and praxis (action and reflection) that provides a metaphor, heuristics and paradigmatic assumptions that serve as analytic tools that communities use to diagnose and respond to the inequities they face (Collins 2009).

Hispanic-Serving Institutions (HSIs) are federally recognized two- and four-year nonprofit institutions where at least 25% of full-time equivalent undergraduates identify as Hispanic and at least half of enrolled Hispanic students are low-income (Garcia 2023; Martinez and Garcia 2020; Laden 2004). HSIs are central to the educational trajectories of Latinx students in the United States, yet long-standing concerns remain about whether they truly serve students equitably in ways that transform structures rather than simply enrolling numerically sufficient numbers of Hispanic students (Garcia 2013, 2016, 2019; Garcia et al. 2019).

Most institutional reports and national statistics present student outcomes by race alone, gender alone, or class alone. Although such disaggregation marks an important step beyond aggregate reporting, it does not considerthe simultaneity of race, gender, and class origin for assessing educational outcomes. Only a small number of higher education studies have applied inferential quantitative methods to explicitly examine intersectional inequities in admissions, retention, developmental course placement, or graduation (e.g., López et al. 2018a; Van Dusen and Nissen 2020). In their systematic review of intersectionality in higher education research, Nichols and Stahl (2019) conclude that there is considerable work to be done to address intersecting and co-constructed systems of inequality that shape participation and outcomes.

This article responds to this call by examining how quantitative intersectionality can be used to reveal inequities that would otherwise remain invisible in conventional analyses at HSIs. We focus on two public institutions in a Southwestern state in the United States (U.S.) where roughly half of residents identify as Hispanic/Latinx. These universities, which we refer to as Southwest University (SU) and Borderlands University (BU), provide a compelling context in which to examine intersectionality as a potential new standard for accountability and equity metrics.

The central aim of this paper is to demonstrate how place-based quantitative intersectional analysis of student outcomes can inform equity-focused policy and practice. We treat race, gender, and first-generation college status as interdependent social locations rather than independent variables to be controlled. We then examine how intersections of these statuses are related to four-year college graduation and developmental English and mathematics placement.

We address three questions. The first asks who is being served and underserved within HSIs once race, gender, and first-generation college status are considered simultaneously. The second asks how four-year graduation and developmental placement vary across twenty intersectional social locations. The third asks how integrating quantitative intersectionality as critical inquiry and praxis into conventional institutional analytics might reshape administrative data for equity metrics, accountability practices, and resource allocation at HSIs.

The article proceeds in five parts. The next section reviews quantitative intersectional studies in education and situates our work in relation to QuantCrit and HSI scholarship. This is followed by a materials and methods section that describes institutional context, author positionality, data and measures, and quantitative methods. We then present results for four-year graduation and developmental course placement, followed by a discussion of what quantitative intersectionality reveals about complex inequities at HSIs and a set of policy implications. The article concludes with reflections on limitations and directions for future research.

2. Materials and Methods

2.1. Quantitative Intersectional Studies in Education

Although conceptual and qualitative work on intersectionality in education is extensive (Castro and Cortez 2017; Covarrubias and Vélez 2013; Fernández 2002; López 2003; Lynn and Dixson 2013; Pérez Huber 2010; Solórzano 1998), quantitative applications remain comparatively scarce. A growing body of research has begun to address this gap using descriptive statistics, multivariate models, or innovative measures of race and class (Garcia et al. 2018; Gillborn et al. 2018; Irizarry 2015; Jang 2018; Latino et al. 2020; López et al. 2018b; Zerai and Banks 2002).

Covarrubias (2011) uses the Current Population Survey (CPS) produced by the Census Bureau and the Bureau of Labor Statistics to document how educational attainment varies within Latino groups by gender and nativity, demonstrating that certain combinations of race, ethnicity, and nativity are systematically devalued. Sáenz et al. (2024) similarly employ American Community Survey (ACS) produced by the Census Bureau to reveal heterogeneity in educational outcomes by ethnicity, nativity, gender, and race among Latina/os in the United States (Sáenz et al. 2024). These studies highlight the importance of disaggregating diverse Latinx experiences through intersectionality rather than treating all Latinos as a homogeneous category.

López and colleagues apply logistic regression to examine six-year graduation rates using an intersectional framework (López et al. 2018a). Contrary to narratives that treat income as a simple proxy for race or gender, they find that low-income white and Asian women are only modestly less likely to graduate than high-income white women, whereas low-income Hispanic and Black men and women experience more substantial gaps. Ferrare (2016) shows similar complexity in intergenerational educational mobility by race and gender), and Lord and colleagues document patterned inequities in the trajectories of engineering students by race and gender (Simmons and Lord 2019). Together, these studies suggest that analyzing inequalities by race, gender, or class alone yields only a partial picture of educational stratification.

A key challenge concerns how to operationalize class origin or relational advantage and disadvantage in intersectional analysis. One promising approach focuses on building data infrastructures that routinely capture detailed parental educational attainment as an indicator of class origin, permitting institutions to identify first-generation college and continuing-generation college students as distinct intersectional social locations and categories of experience (Beattie 2018; Ives and Castillo-Montoya 2020; Nguyen and Nguyen 2018; Morgan et al. 2022). Our study extends this emerging body of work by concentrating on two HSIs in a state with high percentages of Hispanic resident, treating race, gender, and first-generation college status as a joint category of experience, and explicitly linking quantitative findings to institutional accountability and equity metrics.

2.2. Institutional and Historical Context

SU and BU are public universities located in the same Southwestern state, both designated as HSIs and enrolling high proportions of Latinx students. SU is a research-intensive institution, and BU was a land-grant university classified as a Research 2 institution at the time of publication. Both have long histories that are entwined with settler colonialism, racialized labor markets, and educational inequities affecting Indigenous, Black, and Latinx communities (Dancy et al. 2018; Glenn 2015; Gonzales 2001). The state consistently ranks near the bottom on indicators such as adult literacy and K–12 educational performance, with stark gaps in degree attainment between white and Hispanic adults.

These structural conditions shape the student populations and missions of SU and BU. Prior research on HSIs suggests that designation alone does not guarantee institutional practices that prioritize Latinx or other marginalized students in their policies, resource distribution, or symbolic practices (Castillo et al. 2025; Contreras et al. 2008; Garcia 2013, 2016, 2018, 2019; Greene et al. 2012; Vargas and Villa-Palomino 2019). Our analysis situates inequities in graduation and developmental placement within what Collins (2009) calls the matrix of domination, which consists of a particular arrangement of systems inequality as well as a particular arrangement of power at the structural/institutional levels, disciplinary/surveillance-level, interpersonal/individual level and cultural/hegemonic level (Collins and Bilge 2020).

2.3. Author Positionality and Intersectional Praxis

Following intersectionality’s emphasis on reflexive knowledge production (Combahee River Collective 1983; Collins 2007; Yuval-Davis 2011), we briefly note our own social locations and interpretative and ethical commitments. The co-authors occupy different positions in grids of power, including a continuing-generation college white man, a first-generation elementary school Black Latinx Dominican woman, a first-generation Chicana with mixed race, ethnicity and country of origin and a first-generation college white woman of mixed ethnicity. Although our lived experiences differ, we share ethical and political commitments to interrogating our ontologies, epistemologies, axiologies and specifically how race, gender, class, and first-generation college status shape educational opportunities and to advancing equity lifts for those who have historically been marginalized (Zambrana 2018). Ongoing self-implicating critical reflexive praxis about positionality informs our interpretation of the results and our policy recommendations (Boveda and Annamma 2023).

2.4. Data and Sample

We analyze institutional administrative data on first-time, full-time first-year college students entering SU and BU between academic years 2014 and 2019. This focus reflects both analytic and conceptual considerations. Four-year graduation is most clearly interpretable for first-time, full-time cohorts. Transfer and part-time students often follow different trajectories, such as arriving with accumulated credits or enrolling less than full-time, which complicates comparisons. Developmental courses are typically taken in the first year of enrollment, so developmental placement is well defined for the cohorts we study.

Data are drawn from university offices of institutional analytics and include demographic information (race, ethnicity, gender, and parental educational attainment), high school codes, cohort year, developmental coursework in English and mathematics, and degree completion date. Because we observe cohorts only through 2019, we focus on four-year completion rates. Six-year completion rates are fully observable only for the earliest cohort, which would limit statistical power and comparability if used as the primary outcome. This limitation is partially offset by including developmental course placements, which occur within the first year.

Table 1. Descriptive statistics for incoming first-time, full-time freshmen resident students, Southwest University (SU) and Borderland University (BU), 2014 to 2020 cohorts.

	(1)	(2)
Variable	Southwestern Public University	Borderlands University
First-Generation College Student	0.275	0.376	***
College Graduation:
Within 4 Years	0.082	0.041	***
Within 5 Years	0.093	0.070	***
Within 6 Years	0.094	0.078	***
Developmental Course:
Mathematics Required	0.238	0.105	***
English Required	0.056	0.131	***
Female	0.575	0.562	*
Ethnicity:
Hispanic	0.591	0.635	***
Race:
White	0.310	0.317
American Indian	0.036	0.027	***
Asian	0.046	0.011	***
Black	0.018	0.010	***
Observations	12,269	6354

Source: Institutional Analytics Division at Southwest University (SU) and Borderlands University (BU). SU and BU are R1 and R2 public universities in the U.S. Southwest. Graduation rates are smaller than one would normally find because the sample is of incoming cohorts of first-time, full-time freshmen from academic years 2014–2020, so many students had not been enrolled long enough to achieve graduation within six years, for example. First-generation college students are those where neither one of their parents attended college. *, and *** denote statistical differences at the ten and one percent levels, respectively, and are reported for two-sample t-tests assuming data are unpaired with unequal variances.

presents descriptive statistics for first-time, full-time resident freshmen at SU and BU from 2014 to 2020. SU students are more likely to graduate within four, five, and six years than BU students. SU students are also more likely to be placed in developmental mathematics but less likely to be placed in developmental English than BU students. BU students are more likely to be first-generation college and Hispanic, whereas SU enrolls relatively higher proportions of American Indian, Asian, and Black students. These patterns caution against treating HSIs as interchangeable.

2.5. First-Generation College Status

First-generation college students are defined as those who reported that no parent or guardian attended college. This choice reflects the need to harmonize data across SU and BU. At BU, parental education is drawn from FAFSA items that ask, for each parent, about the highest level of schooling completed, with response options ranging from middle school to college or beyond. At SU, data come from both FAFSA and undergraduate applications, which contain more detailed categories such as no high school, some high school, high school diploma or GED, some college, associate degree, bachelor’s degree, and graduate or professional degree. The FASFA form asks about parent Educational Status with the following question: Did either of the student’s parents attend college or complete college? Neither parent attended college; one or both parents attended college but neither parent completed college; one or both parents completed college; don’t know.

We code first-generation college status as one if no parent or guardian attended college and zero otherwise. This measure is relatively blunt and obscures important differences between students whose parents completed some college with those students whose parents eared college degrees schooling, high school diplomas, or graduate degrees. It nonetheless provides a feasible common definition of first-generation college students as those whose parents never attended college across institutions and allows us to identify students who are likely to be navigating college as the first in their family.

A substantial share of students are missing information on parental educational attainment. Approximately 8% of students at SU and 32% of students at BU lack parental education data. Because FAFSA completion is voluntary and may be patterned by socioeconomic status, citizenship, and other factors, this missingness raises concerns about sample selection bias. We therefore compare descriptive statistics for students with and without observed first-generation college status and report these comparisons in Section 3.4. The results suggest that students missing first-generation college status information often differ systematically from those with observed data, which implies that our estimates likely understate the magnitude of intersectional inequities in developmental course placement and graduation.

2.6. Race, Ethnicity, and Sample Restrictions

To examine race and ethnicity intersectionally while also protecting student confidentiality, we restrict the analytic sample to students who identify as Hispanic (of any race), white, American Indian, Asian, or Black. Students categorized as Native Hawaiian or Pacific Islander, two or more races, non-resident alien, or unknown are excluded because small cell sizes for these groups could risk re-identification. This reduces the full sample by 4.9% at SU and 3.7% at BU.

At BU, detailed Hispanic origin data are not consistently available, which limits our ability to disaggregate Latinx students by ethnic identity, such as Mexican, Puerto Rican, Cuban, etc. We are also unable to disaggregate Hispanic by race, precluding our ability to examine outcomes for Hispanics that identify as racially White, Native American, Black, etc. in a manner that would be comparable across the two institutions. This represents an important limitation in light of research findings that illustrate that street race (e.g., how one believes they are preceived racially based on a conglomeration of skin color, facial features and hair texture) is associated with experiences of discrimination and opportunity within Latinx communities (Allen et al. 2000; López et al. 2018b; Telles 2014). We return to this issue in the discussion and in our policy recommendations.

With two generation college status categories (first-generation college status and continuing-generation college status), two genders, and five racial and ethnic groups, we conceptualize twenty intersectional social locations that reflect distinct configurations of structural racism, classed opportunity, and gendered expectations.

2.7. Quantitative Methods: A Quantitative Intersectionality Approach

Intersectionality challenges what has been termed business-as-usual quantitative approaches that treat race, class, and gender as separate, additive variables to be controlled (Collins 2007; Landry 2006). For our purposes, quantitative intersectionality has two components. Conceptually, it requires treating race, gender, and first-generation college status as a joint category of experience rather than as independent predictors. Statistically, it requires modeling both main effects and interactions in a way that allows us to recover estimates for each of the twenty social locations.

This study extends earlier work on quantitative intersectionality by applying this framework to institutional administrative data from two Hispanic-Serving Institutions and linking results directly to servingness and accountability metrics (López et al. 2018a). For each outcome—four-year graduation, developmental mathematics placement, and developmental English placement—we estimate mixed-effects logistic regression models. These models include fixed effects for race and ethnicity, gender, and first-generation college status, as well as all possible interactions among these variables. Cohort fixed effects capture time-varying characteristics of incoming classes. We also include random intercepts for high schools to reflect the clustering of students within feeder schools.

Using the latent-response formulation, the model can be written as1

$$y_{ij}^{*}={\alpha }_0+\mathit{X}\mathit{\beta }+\mathit{Z}\mathit{\gamma }+{\delta }_C+{\xi }_j+{\epsilon }_{ij}$$

(1)

$${\xi }_j~N(0,\psi )$$

(2)

where i indexes students, j indexes in-state high schools, and y is one of the three binary outcomes. The vectors X and Z contain main and interaction effects, ${\delta }_C$ are cohort fixed effects, ${\xi }_j$ is a high-school-level random intercept, and ${\epsilon }_{ij}$ follows a standard logistic distribution. The observed outcome is defined as

$$y_{ij}=1if{y}_{ij}^{*}>0,0otherwise$$

(3)

We use likelihood ratio tests to assess whether the mixed-effects models provide a statistically significant improvement over standard logistic models that ignore high school clustering. Within-school clustering is summarized using the intraclass correlation coefficient,

$$\rho ={\displaystyle \frac{\psi }{\psi +\phi }}$$

(4)

which can be interpreted as the proportion of variance attributable to differences between high schools.

We pool observations from SU and BU. This increases cell sizes at the social-location level and improves statistical power, while also enhancing external validity within the state by identifying patterns that persist across two institutions with similar demographic profiles, funding formulas, and regional contexts.

Our analytic strategy proceeds in two steps. In the first step, we compute marginal effects of each main variable and interaction term, which answer questions such as how being Hispanic, relative to non-Hispanic white, changes the probability of graduating within four years while holding other factors constant, or how being first-generation college versus continuing-generation college matters net of race and gender. These estimates are not causal. Instead, they are a form of social accounting that summarizes patterns likely shaped by historical and contemporary exposure to structural, interpersonal, and institutional racism and classed processes rather than innate group differences.

In the second step, we construct linear combinations of marginal effects to estimate logistic probabilities for each of the twenty intersectional social locations, such as first-generation college Black men or continuing-generation college American Indian women. These estimates report standard errors calculated using the delta method. This procedure is numerically equivalent to fitting models with twenty dummy variables for social locations but has the advantage of preserving interpretability of main effects and allowing us to see whether inequities are primarily driven by single statuses or by their intersections.

3. Results

3.1. Four-Year Graduation Rates

Table 2. Mixed effects logistic models of 4-year completion rates, marginal effects.

Variable	Coefficient (Standard Error)
Male	−0.025 *** (0.007)
First-generation college	−0.017 ** (0.007)
Hispanic	−0.012 *** (0.005)
American Indian	−0.039 *** (0.014)
Black	−0.033 * (0.020)
Asian	−0.013 (0.010)
Male × Hispanic	0.002 (0.007)
Male × Black	0.010 (0.030)
Male × Asian	0.020 (0.014)
Male × American Indian	0.035 * (0.020)
Male × First-generation college	0.011 (0.011)
First-generation college × Hispanic	0.010 (0.009)
First-generation college × Black	0.031 (0.036)
First-generation college × Asian	0.020 (0.020)
First-generation college × American Indian	0.019 (0.020)
Male × First-generation college × Hispanic	−0.011 (0.014)
Male × First-generation college × Black	-
Male × First-generation college × Asian	−0.058 (0.042)
Male × First-generation college × American Indian	−0.067 (0.051)
Cohort Fixed Effects	YES
$\rho$	0.038
LR ${\chi }^2$(1)	33.64 ***
Observations	13,949

Source: Offices of Institutional Analytics at Southwest University and Borderlands University. First generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). *, **, and *** de-note statistical differences at the ten five, and one percent levels, respectively. Missing coefficients indicate that this group perfectly identified a particular outcome regarding 4-year graduation rates.

reports marginal effects from the mixed-effects logistic model of four-year graduation. Accounting for clustering by feeder high school modestly improves model fit, although the intraclass correlation coefficient is relatively small. American Indian students are about 3.9 percentage points less likely, and Black students approximately 3.3 percentage points less likely, to graduate within four years compared to white students, all else equal. Hispanic students are about 1.2 percentage points less likely to graduate in four years than white students. Men are about 2.5 percentage points less likely than women to graduate in four years, and first-generation college students are about 1.7 percentage points less likely to graduate than continuing-generation college students. These differences appear modest in absolute terms but are large relative to the overall four-year graduation rate of roughly seven percent in the sample.

Table 3. Logistic probabilities of graduation within four years, by social location.

Social Location	Coefficient (Standard Error)	Cell Size
Continuing-generation college White Women	(Reference)	2044
First-generation college White Women	−0.017 ** (0.007)	431
Continuing-generation college White Men	−0.025 *** (0.007)	1740
First-generation college White Men	−0.031 *** (0.011)	276
Continuing-generation college Hispanic Women	−0.012 *** (0.005)	2914
First-generation college Hispanic Women	−0.020 *** (0.006)	1884
Continuing-generation college Hispanic Men	−0.036 *** (0.007)	2214
First-generation college Hispanic Men	−0.042 *** (0.009)	1326
Continuing-generation college American Indian Women	−0.039 *** (0.014)	178
First-generation college American Indian Women	−0.037 * (0.019)	94
Continuing-generation college American Indian Men	−0.029 * (0.015)	134
First-generation college American Indian Men	−0.083 * (0.044)	45
Continuing-generation college Asian Women	−0.010 (0.010)	188
First-generation college Asian Women	−0.008 (0.018)	76
Continuing-generation college Asian Men	−0.015 (0.015)	140
First-generation college Asian Men	−0.060 ** (0.028)	62
Continuing-generation college Black Women	−0.033 * (0.020)	99
First-generation college Black Women	−0.019 (0.033)	23
Continuing-generation college Black Men	−0.049 ** (0.022)	81
First-generation college Black Men	-	0
Observations	13,949

Source: Offices of Institutional Analytics at Southwest University (SPU) and Borderlands University (BU). First generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). *, **, and *** denote statistical differences at the ten five, and one percent levels, respectively.

translates these marginal effects into social-location estimates using continuing-generation college white women as the reference group. We selected this reference group because most other social locations have significantly lower probabilities of four-year graduation than this group. The largest gaps are observed for first-generation college American Indian men, who are about 8.3 percentage points less likely to graduate within four years than continuing-generation college white women, as well as for first-generation college Asian men, continuing-generation college Black men, continuing-generation American Indian college women, and first-generation college Hispanic men. These patterns are largely driven by main effects of race, gender, and first-generation college status, but the intersectional estimates highlight how these relational disadvantages differ for particular groups.

3.2. Developmental Mathematics Placement

Table 4. Nonlinear models of developmental course mathematics placement, marginal effects.

Variable	Coefficient (Standard Error)
Male	−0.053 *** (0.014)
First-generation college	0.005 (0.018)
Hispanic	0.042 *** (0.011)
American Indian	0.066 ** (0.026)
Black	0.106 *** (0.039)
Asian	0.018 (0.025)
Male × Hispanic	0.006 (0.016)
Male × Black	0.060 (0.062)
Male × Asian	−0.002 (0.036)
Male × American Indian	0.002 (0.037)
Male × First-generation college	−0.006 (0.033)
First-generation college × Hispanic	0.026 (0.020)
First-generation college × Black	−0.065 (0.089)
First-generation college × Asian	−0.032 (0.064)
First-generation college × American Indian	0.029 (0.039)
Male × First-generation college × Hispanic	−0.025 (0.037)
Male × First-generation college × Black	0.142 (0.127)
Male × First-generation college × Asian	−0.083 (0.085)
Male × First-generation college × American Indian	−0.035 (0.067)
Cohort Fixed Effects	YES
$\rho$	0.094
LR ${\chi }^2$(1)	375.84 ***
Observations	18,623

Source: Offices of Institutional Analytics at Southwest University and Borderlands University. First generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). ** and *** denote statistical differences at the five and one percent levels, respectively.

presents marginal effects from the developmental mathematics model. The intraclass correlation coefficient is larger than in the graduation model, at approximately 0.094, which indicates that high schools substantially shape college-level mathematics readiness. Men are about 5.3 percentage points less likely to be placed in developmental mathematics than women. Black students are roughly 10.6 percentage points more likely, American Indian students 6.6 percentage points more likely, and Hispanic students 4.2 percentage points more likely to be placed in developmental mathematics than white students. First-generation college status is not a strong predictor of developmental mathematics placement net of other covariates.

Table 5. Logistic probabilities of developmental course mathematics placement, by social location.

Social Location	Coefficient (Standard Error)	Cell Size
Continuing-generation college White Women	(Reference)	2608
First-generation college White Women	0.005 (0.018)	580
Continuing-generation college White Men	−0.053 *** (0.014)	2233
First-generation college White Men	−0.054 ** (0.024)	395
Continuing-generation college Hispanic Women	0.042 *** (0.011)	3965
First-generation college Hispanic Women	0.072 *** (0.012)	2567
Continuing-generation college Hispanic Men	−0.006 (0.013)	2966
First-generation college Hispanic Men	−0.006 (0.012)	1783
Continuing-generation college American Indian Women	0.066 ** (0.0266)	244
First-generation college American Indian Women	0.100 *** (0.029)	136
Continuing-generation college American Indian Men	0.015 (0.036)	167
First-generation college American Indian Men	0.008 (0.054)	69
Continuing-generation college Asian Women	−0.0004 (0.024)	260
First-generation college Asian Women	−0.009 (0.050)	104
Continuing-generation college Asian Men	−0.038 (0.031)	191
First-generation college Asian Men	−0.153 * (0.084)	77
Continuing-generation college Black Women	0.106 *** (0.039)	125
First-generation college Black Women	0.047 (0.070)	32
Continuing-generation college Black Men	0.113 *** (0.040)	105
First-generation college Black Men	0.189 ** (0.088)	16
Observations	18,623

Source: Offices of Institutional Analytics at Southwest University (SPU) and Borderlands University (BU). First generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). *, **, and *** denote statistical differences at the ten, five, and one percent levels, respectively.

reports developmental mathematics placement probabilities by social location. Approximately 19% of the sample is placed in developmental mathematics. Continuing-generation college white men are about 5 percentage points less likely than continuing-generation college white women to be placed in developmental mathematics. First-generation college Hispanic women are about 7.2 percentage points more likely to be placed in developmental mathematics. The largest disparity is for first-generation college Black men, who are about 18.9 percentage points more likely to be placed in developmental mathematics than continuing-generation college white women, despite a small cell size. American Indian women, particularly first-generation college American Indian women, also face elevated rates of developmental mathematics placement. Asian men are less likely than the reference group to be placed in developmental mathematics.

3.3. Developmental English Placement

Table 6. Nonlinear models of developmental course English placement, marginal effects.

Variable	Coefficient (Standard Error)
Male	−0.004 (0.010)
First-generation college	0.046 *** (0.013)
Hispanic	0.048 *** (0.009)
American Indian	0.107 *** (0.013)
Black	0.072 *** (0.020)
Asian	0.053 ** (0.023)
Male × Hispanic	0.008 (0.011)
Male × Black	−0.003 (0.032)
Male × Asian	−0.026 (0.029)
Male × American Indian	−0.006 (0.020)
Male × First-generation college	−0.033 (0.020)
First-generation college × Hispanic	−0.007 (0.014)
First-generation college × Black	0.029 (0.033)
First-generation college × Asian	−0.009 (0.033)
First-generation college × American Indian	−0.027 (0.019)
Male × First-generation college × Hispanic	0.018 (0.022)
Male × First-generation college × Black	0.051 (0.043)
Male × First-generation college × Asian	0.052 (0.049)
Male × First-generation college × American Indian	0.054 * (0.031)
Cohort Fixed Effects	YES
$\rho$	0.197
LR ${\chi }^2$(1)	461.40 ***
Observations	18,623

Source: Offices of Institutional Analytics at Southwest University and Borderlands University. First generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). *, **, and *** denote statistical differences at the ten, five, and one percent levels, respectively.

displays marginal effects for developmental English placement. The intraclass correlation is notably higher than in the graduation or developmental mathematics models, at approximately 0.197, which indicates strong clustering by high school. American Indian students are about 10.7 percentage points more likely, Black students 7.2 percentage points more likely, Hispanic students 4.8 percentage points more likely, and Asian students 5.3 percentage points more likely to be placed in developmental English than white students. First-generation college students are about 4.6 percentage points more likely than continuing-generation college students to be placed in developmental English. The marginal effect of being male is small and not statistically significant overall.

Table 7. Logistic probabilities of developmental course English placement, by social location.

Social Location	Coefficient (Standard Error)	Cell Size
Continuing-generation college White Women	(Reference)	2608
First−generation college White Women	0.046 *** (0.013)	580
Continuing−generation college White Men	−0.004 (0.010)	2233
First−generation college White Men	0.010 (0.017)	395
Continuing−generation college Hispanic Women	0.048 *** (0.009)	3965
First−generation college Hispanic Women	0.088 *** (0.009)	2567
Continuing−generation college Hispanic Men	0.053 *** (0.008)	2966
First−generation college Hispanic Men	0.078 *** (0.009)	1783
Continuing−generation college American Indian Women	0.107 *** (0.013)	244
First−generation college American Indian Women	0.127 *** (0.015)	136
Continuing−generation college American Indian Men	0.098 *** (0.017)	167
First−generation college American Indian Men	0.139 *** (0.020)	69
Continuing−generation college Asian Women	0.053 ** (0.023)	260
First−generation college Asian Women	0.091 *** (0.024)	104
Continuing−generation college Asian Men	0.024 (0.026)	191
First−generation college Asian Men	0.080 *** (0.017)	77
Continuing−generation college Black Women	0.072 *** (0.020)	125
First−generation college Black Women	0.148 *** (0.027)	32
Continuing−generation college Black Men	0.066 ** (0.022)	105
First−generation college Black Men	0.160 *** (0.029)	16
Observations	18,623

Source: Offices of Institutional Analytics at Southwest University (SPU) and Borderlands University (BU). First-generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). ** and *** denote statistical differences at the five and one percent levels, respectively.

provides social-location probabilities for developmental English placement. Approximately 8.2% of the sample is placed in developmental English, yet almost all social locations, except continuing-generation college white men and continuing-generation college Asian men, have higher probabilities of developmental English placement than continuing-generation college white women. The largest disparities are observed for first-generation college Black men, first-generation college Black women, and first-generation college American Indian men, who are estimated to be 16.0, 14.8, and 13.9 percentage points more likely, respectively, to be placed in developmental English.

3.4. Missing First-Generation College Status and Sample Selection

Table 8. Comparison of descriptive statistics by whether students are missing first-generation college status.

Variable	First-Generation College Present	First-Generation College Missing	Difference
Graduation within 4 years	0.068	0.043	0.025 ***
Developmental English	0.082	0.143	−0.061 ***
Developmental Mathematics	0.192	0.176	0.016 **
Female	0.570	0.575	−0.004
White	0.312	0.253	0.060 ***
Hispanic	0.606	0.674	−0.068 ***
Black	0.015	0.014	0.001
Asian	0.034	0.027	0.007 **
Observations	18,623	4136

Source: Offices of Institutional Analytics at Southwest University (SPU) and Borderlands University (BU). First-generation college student status was based on voluntary information on the Free Application for Federal Student Aid (FASFA). ** and *** denote statistically significant differences in means at the five and one percent levels, respectively.

compares descriptive statistics for students with and without observed first-generation college status. Students missing this information are less likely to graduate within four years, more likely to be placed in developmental English, and slightly less likely to be placed in developmental mathematics. They are less likely to be white and more likely to be Hispanic. Because first-generation college status is coded as missing when both parents’ education levels are unknown or when one parent is known not to have attended college and the other is missing, many single-parent households and lower-income students may be disproportionately excluded from the main analyses. This pattern suggests that our estimates, while conservative, likely understate the true magnitude of intersectional inequities in graduation and developmental placement.

4. Discussion

The results indicate that inequities in higher education at HSIs cannot be fully understood when race, gender, or first-generation college status are considered in isolation. Instead, intersectional social locations such as first-generation American Indian men, first-generation college Black men, and first-generation college Hispanic men experience patterned disadvantages that are obscured in conventional reporting. Several themes emerge from the analyses.

One theme concerns the concentration of disadvantage at specific social locations. Although white and continuing-generation college students generally have higher four-year graduation rates and lower rates of developmental placement, the largest gaps are concentrated among first-generation college American Indian, Black, and Hispanic men, as well as among some women in these groups. This pattern aligns with intersectional paradigms that emphasize that racism, sexism, and classed inequalities do not operate independently but are simultaneous and co-constructed social statuses that are associated with distinct outcomes and categories of experiences (Crenshaw 1991; Collins 2009; McCall 2005).

A second theme concerns the role of developmental coursework as an equity bottleneck. Developmental mathematics and English placement disproportionately affect students from historically marginalized racialized groups and first-generation college students, particularly Black and American Indian students. Because these courses add time and cost without always leading to successful progression, they can function as structural bottlenecks that reproduce inequality (Pell Institute for the Study of Opportunity in Higher Education 2011; Hurtado et al. 2015; Hernández et al. 2024). The patterns observed in developmental placement likely contribute to the graduation gaps documented in this article.

A third theme concerns the limits of single-axis equity metrics. Standard institutional dashboards that report graduation by race alone, gender alone, or first-generation college status alone may suggest that HSIs are making progress in closing gaps, while masking substantial inequities for particular intersectional groups. The findings provide empirical support for scholarship, arguing that equity metrics must move beyond single-axis categories if they are to guide meaningful change (Garcia et al. 2018, 2019; Gillborn et al. 2018).

Finally, the results have implications for how HSIs understand and enact servingness. An institution can meet the numerical thresholds for HSI designation while still producing or tolerating large intersectional inequities in graduation and developmental placement. Quantitative intersectionality offers a way to align claims of servingness with the lived realities of students who are most affected by structural racism and classed inequalities.

4.1. Policy Implications for Hispanic-Serving Institutions

The analyses point toward several policy directions for HSIs and other institutions committed to equity. These implications flow directly from the empirical patterns and from the intersectional framework.

One implication is the need to build data infrastructures that support intersectional equity analysis. HSIs can take concrete steps by collecting parental educational attainment for all applicants, undergraduate and graduate, as a required field. The question should clearly specify that it refers to schooling of parents or guardians when the student was approximately age sixteen, and response options should capture the full range of schooling from no formal education to graduate or professional degrees. It is also important to distinguish U.S.-based from foreign degrees without conflating parental education with English language proficiency. Harmonizing data collection across institutional forms such as applications, FAFSA interfaces, and internal surveys would reduce missingness and improve comparability over time. These steps would allow institutions to reliably identify first-generation college students and to distinguish between different degrees of educational advantage among continuing-generation college students.

A second implication is that equity dashboards and accountability reports, whether internal, state-level, or federal, should routinely report applications, enrollment, retention, developmental placement, and graduation by intersectional social location rather than by race or gender alone. Reports should provide disaggregated statistics at least for the types of groups considered in this article, including combinations such as first-generation college Black men and continuing-generation college American Indian women. Reliance on a single composite measure such as all Hispanic students may hide important heterogeneity in outcomes. States and accreditors could support this work by updating reporting requirements and providing technical assistance to institutional research offices.

A third implication is the need to use intersectional evidence to guide the allocation of resources and the design of programs. HSIs can conduct equity audits that overlay intersectional outcome data with information about the reach of existing initiatives such as pipeline programs, bridge programs, learning communities, advising, and tutoring. This exercise can help identify social locations that are persistently underserved or excluded from equity initiatives and can inform decisions about which groups should be prioritized for targeted interventions. Redesigning developmental education models, for example, through co-requisite approaches or accelerated learning programs, can reduce delays and improve progression for groups that are disproportionately placed in remediation (Latino et al. 2020). Intersectional metrics help institutions move beyond broad categories such as all first-generation college students or all Latinx students and instead ask which specific groups are not being served and why.

A fourth implication is the importance of investing in more nuanced measures of race and racialization (Omi and Winant 2015). In addition to standard self-identified race and ethnicity questions, institutions could incorporate a street race item that asks students how they believe others typically perceive their race in public (López et al. 2018b; Vargas et al. 2021). Collecting more detailed Hispanic origin data and, where feasible, distinguishing between Latinx students racialized as white, Black, Brown, or Indigenous would support a more accurate understanding of how racialized experiences, not just formal categories, shape educational trajectories within HSIs. Such practices would align more closely with intersectional theoretical commitments to examining how power operates through racialization rather than solely through static race labels.

4.2. Limitations and Future Directions

The study has several limitations that also point to directions for future research. One limitation is the relatively short time horizon of four-year graduation, which is constrained by the availability of data on recent cohorts. Future research should examine six-year and longer-term outcomes as more cohorts mature. Doing so would provide a fuller picture of degree attainment and might reveal additional intersectional patterns.

A second limitation is that the measure of first-generation college status is coarse and does not capture meaningful variation within both first-generation college and continuing-generation college categories. Linking institutional data to richer survey information or systematically expanding parental education questions would permit more refined measures of educational advantage and disadvantage.

A third limitation is that the race and ethnicity measure conflates self-identified race and ethnicity and does not capture street race or finer-grained racialization within Latinx communities. Given the literature documenting how skin color and perceived race shape discrimination and opportunity (Allen et al. 2000; Hogan 2017; Telles 2014), improving race measures is an important area for methodological and institutional development.

A fourth limitation is that missing data on parental education and the resulting exclusion of many students likely biases the estimates. Descriptive comparisons suggest that missingness is not random; students missing first-generation college status are less likely to graduate and more likely to be placed in developmental English. The estimates presented here therefore likely understate the true magnitude of intersectional inequities.

Finally, the analysis is purely quantitative and does not incorporate qualitative data on student experiences or institutional practices. Mixed-methods case studies could deepen understanding of how specific institutional decisions, practices, and narratives contribute to the patterns observed. Future work could combine intersectional quantitative analysis with interviews or focus groups that explore how students in particular social locations navigate HSIs, and how institutional actors interpret and respond to equity metrics.

5. Conclusions

Landry (2006) argues that the old approach of studying race or gender or class individually should be replaced with attention to all three in a given project (Landry 2006). This study operationalizes that insight in the context of HSIs by treating race, gender, and first-generation college status as an intersectional category of experience and by examining how this joint status shapes graduation and developmental course outcomes. The results indicate that conventional single-axis metrics can obscure substantial inequities experienced by specific intersectional social locations, especially first-generation college American Indian, Black, and Hispanic men and women.

Quantitative intersectionality does not, by itself, eliminate inequality. It does, however, provide a more accurate and ethically grounded lens for understanding how systems of power are reproduced in higher education. For HSIs, the implications are clear. To live up to their promise of truly serving Latinx and other marginalized students, institutions must build data systems that support intersectional analysis, report and act on intersectional equity metrics, target resources to the social locations most affected by structural inequities, and expand data collection practices to capture racialization and street race.

These findings raise a broader question about the purpose and beneficiaries of institutional data practices. If reporting systems continue to rely on aggregate or single-axis categories, they risk obscuring the very inequities that HSIs aim to address. Intersectional inquiry—both quantitative and qualitative—creates opportunities to render such inequities visible and guide institutional action.