Abstract
Reported hate crimes in the United States have increased rapidly in recent years, alongside an increase in general racial animus. Scholars have shown that the larger sociopolitical environment can directly impact the campus climate and experiences of all students, particularly students of color. However, little is known about how reports of hate crime incidents relate to college enrollment levels of students of color. This lack of evidence has especially troubling implications for Black people, the most frequent targets of reported hate crimes. This paper helps to fill in that gap by exploring the association between the number of reports of hate crimes within states and Black students’ college enrollment. We examine a comprehensive dataset of institutional enrollment and characteristics, reported hate crimes, and census data on state racial demographics from 2000 to 2017 using several techniques, including institution fixed effects. We find that a 1 standard deviation increase in reports of state-level hate crimes predicts a 17 to 22 percent increase in Black first-time student enrollment at historically Black colleges and universities. As the number of reported hate crimes is almost assuredly an undercount of the actual number of incidents, we explore the implications of what these results mean.
1. Introduction
Since 2010, the number of reported hate crimes across the United States has increased, with the highest recorded number of these crimes occurring in 2021 (Smith 2021).1 This could be because people have become more aware of and more likely to report hate crimes. Evidence suggests, however, that there has been a genuine increase in the number of hate crimes, not simply an increased awareness and reporting of these types of crimes, at a time when rates of other types of crime were falling (Edwards and Rushin 2018).2 Concurrent with this escalation in reports of hate crimes, racial animus has increased in the United States (Pew Research Center 2019).
Scholars have shown that the larger racial and sociopolitical environment has an influence on college campuses and that campus climate is associated with postsecondary outcomes, particularly for students of color (Hurtado et al. 1998; Wessler and Moss 2001; Museus et al. 2008; Van Dyke and Tester 2014; Stokes 2020). Even with a robust body of literature available on national or institutional climates of racial animus and college climate, little is known about how reports of incidents, such as hate crimes, relate to the college enrollment decisions of students, particularly the enrollment of students of color. These types of relationships are critical to understand, as they provide evidence of intersections between larger sociopolitical structures that dictate hate crime reporting, racism, and college enrollment.
Part of the reason that there is less scholarship on this relationship is due to the difficulty of measuring the number of hate crimes or similar incidents. Nevertheless, the history and number of hate crimes in a geographic area influence the living conditions and perceptions of those conditions for citizens (Williams et al. 2021). For example, King et al. (2009) found that counties with more lynchings prior to the 1930s were less likely to have police forces that complied with federal hate crimes laws and had fewer police reports of hate crimes seventy years later. Hate crimes, particularly in places with histories of racial violence, remain underreported (Pezzella et al. 2019). One reason this might be the case is the populations targeted. Levin and Reitzel (2018) found that the majority of racial hate crimes in the United States target Black people.
Hate crimes can impact living conditions of the targeted population in varied ways, from the physical to the psychological. For example, anti-racial and anti–sexual orientation crimes were more likely to be directed against people compared with anti-religious hate crimes, which were usually directed against property (Cheng et al. 2013). Nevertheless, all hate crimes impact an individual's sense of welcome and safety in a community. People feel the influence of hate crimes on living conditions in both social and educational settings.
Since 2015, college campuses saw a 25 percent increase in the number of reported hate crimes (Bhattacharya 2018). This paper explores whether increases in reports of hate crimes in a state are associated with Black students’ college enrollment. We focus on Black students because (a) Black individuals are the most frequent target of reported racial hate crimes and (b) there is a sector of higher education explicitly focused on helping these students succeed, historically Black colleges and universities (HBCUs). We do this using the lens of racial homophily, the seeking out of individuals who have the same race or spaces that are welcoming to one's race. To this end, we explore whether Black students are more likely to enroll in HBCUs in states where there are more reports of hate crimes when compared with college enrollment in states in which there are fewer reported hate crimes. This focuses on the ways that environmental racial animus may play a role in college enrollment decisions.
The college selection process is complex. Student enrollment decisions are based on numerous familial, individual, institutional, and structural factors. Some familial and individual factors are race and ethnicity, gender, prior academic performance, knowledge of the college process, and family income (Kim and Núñez 2013). Institutional aspects encompass the cost of attendance at a particular institution, the availability of financial aid, and distance of the institution from the student's home (Perna 2000; Scott-Clayton 2015; Dache-Gerbino 2018; Skinner 2019). Some of the structural factors include the neighborhood poverty levels of prospective students; whether a secondary school is in an urban, suburban, or rural locale; the resources available and quality of the secondary education and college counseling received; and state financial and equity support for higher education (Perna 2006; Ananat et al. 2017; Baker 2019).
Perceptions of campus climates also play a role in the college enrollment decision. Rankin and Reason (2005) found that while students of all races and ethnicities report a similar level of knowledge of racial harassment incidents on campus, demographic differences exist with respect to the beliefs about whether a campus is racist or welcoming. Further, Black students are more likely to report racial–ethnic conflict on campus when compared with White students. There is, however, limited evidence as to whether either perceptions of campus climate or an increase in the number and visibility of hate crimes on campuses relate to changes in decisions about where to enroll for Black students (Ancis et al. 2000).
There are a number of reasons that an increase in reported hate crimes in a community might increase the likelihood of enrollment in HBCUs for Black students. One reason that students of color attend HBCUs is to learn in racially supportive environments (Laird et al. 2007; Palmer et al. 2010). When Black students feel that there is an increased likelihood of racial animus–related incidents or a less welcoming climate at non-HBCU institutions, they are more likely to apply to and enroll in an HBCU (Johnson 2019). In these cases, the increase in the number of racist incidents in campus communities could have increased the salience of racism and its occurrence on college campuses. This effect, in turn, presumably encouraged a search for a more racially welcoming and homogenous climate.
This study measures whether an association exists between the changes in the number of reported hate crimes in a community and the college enrollment decisions of Black students. It builds upon qualitative work that finds that a link exists between the “Missouri effect”—an increase in racial harassment of Black students attending historically White institutions (HWIs)3—and student decisions to enroll in HBCUs (Williams and Palmer 2019). To our knowledge, however, ours is the first study to measure the quantitative association between racial animus within a state or institution and college enrollment decisions. We find that a 1 standard deviation increase in reports of state-level hate crimes predicts a 17 to 22 percent increase in Black first-time student enrollment at HBCUs. In the pages that follow, we provide an overview of the literature around hate crimes and racial incidents on college campuses; summarize our theoretical framework of racial homophily; present data, empirical models, and results; and conclude by discussing our findings and future directions for research.
2. Literature Review
Hate Crimes
The United States Department of Justice (DOJ) website (2020) defines a hate crime as a crime committed because of “race, color, religion, national origin, sexual orientation, gender, gender identity, or disability.” Hate crimes receive a special designation under federal legal statutes because of the disproportionate harm caused to both individuals and communities by these acts. According to estimates from the DOJ, there were over a quarter of a million hate crimes committed between 2004 and 2015; over half of these crimes were never reported to law enforcement. Much of the current data around hate crimes is a result of the passage of the Hate Crime Statistics Act in 1990 (P.L. 101-275), which mandated that the U.S. Attorney General collect data on hate crimes. There are two primary sources of data on hate crimes in the United States, the Bureau of Justice Statistics National Crime Victimization Survey and the Federal Bureau of Investigations (FBI) Uniform Crime Reporting (UCR) on Hate Statistics Data. These data sources show that certain types of hate crimes are more prevalent than others. For example, approximately half of all hate crimes were related to racial bias between 2011 and 2015 (Masucci and Langton 2017). Black people are historically the most targeted racial or ethnic group (Levin and Reitzel 2018).
In part due to the voluntary nature of agency reporting, hate crimes are underreported. Hate crimes remain underreported for a number of other reasons. Pezzella et al. (2019) found that victims of bias crimes are less likely to report them due to the communities’ lack of confidence in the police. These findings underscore the reality, delineated by King et al. (2009), that people were less likely to report hate crimes in places that had more lynchings in the 1920s. Thus, a strong link exists between current perceptions of police protection and historical relationships in which police chose not to protect the rights of Black citizens. Further, Black, Latinx, and White victims are less likely to report any crime perpetrated by White people as compared with those committed by people of other racial and ethnic groups (Powers et al. 2020).
Hate crime reporting also varies by state and jurisdiction. One barrier to accurate reports is that officers must recognize and report a hate crime as such, but many have relatively limited training in recognizing this unique type of crime (Davis and O'Neill 2016).4 Even if an individual describes or explicitly states that a crime is a hate crime, the officer ultimately decides whether the crime will be classified as one. Because of the subjective nature of the reporting of hate crimes, the DOJ (2020) website recommends that targets of hate crimes both report the crime to local police and call the FBI to report. Another barrier to accurate reporting is that violent hate crimes have been historically less likely to end in arrest when compared with violent crimes that were not deemed hate crimes. Masucci and Langton (2017) report that 28 percent of violent crimes not related to hate led to arrests from 2011 to 2015, as compared with 10 percent of violent crimes with hate motivations. Victims could also be leery of having a hate crime designated as such by the police in light of the lower arrest rate for hate crimes. This could also be why half of all reported hate crimes were not reported to the police during the same time period (Masucci and Langton 2017). Thus, any research focused on reported hate crimes must accept the limitation that these crimes are underreported and that the counts must be viewed as the bare minimum number of incidents.
Racist Incidents and College Campuses
Racism on college campuses has a long history in the United States. Examples begin before the founding of institutions focused on the education of Black students under the second Morrill Act of 1890 (P.L. 51-841) and include efforts to avoid the integration of state land-grant institutions, anti-Jewish quotas in the 1920s, and incidents of hate because of ethnicity, race, and religion in the present era (Reder 2000; Albritton 2012). At various time periods, the number and type of racist incidents on college campuses have increased and decreased, though official reporting on race-based hate crimes only began in 2008. In each year after 2008, the majority of crimes on campus were related to race and ethnicity as compared to religion, sexual orientation, gender, or disability status.
Racist incidents are more likely to occur in certain institution types. Prior research has generally found that a larger share of students of color on a campus predicts fewer reports of race-based hate crimes. Stotzer and Hossellman (2012) examined the ten-year average of race-based hate crimes on college campuses from 1998 to 2008 using the FBI's UCR data. They found that institutions with a larger share of Black and Latinx students were associated with fewer reported race-based hate crimes when compared with the number of reports at institutions with smaller fractions of Black and Latinx students. Analyzing the same data from 2002, Van Dyke and Tester (2014) found that racist incidents on college campuses are reported more frequently on campuses with more fraternities. The authors also found some evidence of a relationship between the share of students of color and prevalence of reports of race-based hate crimes, though it was nonlinear, with institutions with the smallest (less than 10 percent) and largest (greater than 17 percent) share of students of color associated with fewer reports.
The most recent research on racial incidents and campus climate comes from Stokes (2020), who qualitatively analyzed a subset of news articles focused on racist campus incidents from 2013 through 2018. In this dissertation, Stokes found that an institution's state, urbanicity, level, and size were related to the likelihood of having an article published about a campus racial incident. Similar to prior research on national hate crime reports, Stokes found that the majority of incidents on campuses included some anti-Black bias. We wish to be clear that these findings, and our own, can only examine the likelihood of incidents being reported; it is likely that additional incidents have gone unreported.
Racist incidents can impact the educational experiences of all students on campuses, but have particularly negative effects on the experiences of minoritized students (Johnson-Ahorlu 2012). Moore and Bell (2017) described how racist incidents that are defended under the banner of free speech pair with neoliberal educational policies—in particular the belief that education is race-neutral—to define college campuses as institutional spaces that are designed for and safe for White students. Defining these spaces as White serves as a form of both exclusion and othering for students of color.
Paradoxically, Reynolds et al. (2010) found that Black and Latinx students depend less on extrinsic and more on intrinsic motivation and family and community support in order to excel academically in the face of racism on college campuses. The implication of this finding is that for some students of color, negative campus events could be less influential on academic achievement in college than other factors. However, it is important to recognize the heterogeneity in the relationship between racism on campuses and student academic engagement and performance. While students in one study depended on intrinsic motivation in order to persevere academically, numerous research studies also demonstrate how racism is deleterious to academic achievement, mental health, and social outcomes for students (Keels et al. 2017). When students of color encountered more racial microaggressions, they were less likely to believe that they would successfully complete college (Durkee et al. 2021). Although research exists on the negative relationship between general racial animus and racist incidents and student learning and environment, there is limited evidence on how reported incidents, especially those where individuals were targeted due to their race, are related to students’ college enrollment decisions. This study seeks to begin filling this empirical gap.
Predictors of and Trends in HBCU Enrollment
Students enroll in HBCUs for numerous reasons: the supportive academic and campus environments for students of color, for example, or the relatively lower costs of some HBCUs when compared with other institution types (e.g., Laird et al. 2007; Wilson 2007; Palmer et al. 2010). At HBCUs, students learn in more racially homogenous and welcoming environments. HBCUs also serve a disproportionate share of students from low-income backgrounds and first-generation students. The student population at HBCUs has a larger fraction of Pell Grant–eligible students (72 percent) when compared to other institution types (38 percent) (Johnson et al. 2019). Thus, HBCUs serve as a point of access to higher education for many students with less familial knowledge of the college enrollment process and fewer monetary resources.
Larger societal policies and politics play a significant role in HBCU enrollment levels. HBCUs saw declines in the rate of growth of Black students enrolling after the passage of federal desegregation laws, particularly for institutions of higher education in the southeastern United States where the largest share of the Black populace resides (Allen et al. 2007). One study found that the number of Black students enrolling in HBCUs has historically increased under Republican presidents (Sissoko and Shiau 2005). The authors argue this was likely a result of less rigorous enforcement of desegregation laws. Several studies have also shown that larger state-level policy decisions, such as bans on affirmative action, are associated with Black students’ enrollment (e.g., Backes 2012; Fryar and Hawes 2012; Hinrichs 2012). Over the past thirty years, in line with national trends, the number of students enrolled at HBCUs has generally increased. In 1990, there were 257,152 persons enrolled in HBCUs (NCES 2019). By fall of 2018, enrollment was 291,767, a decline from a total of 326,614 in the fall of 2010. While enrollment trends at HBCUs have generally aligned with non-HBCUs, enrollment rates have not increased at HBCUs at the same rate as in other sectors of higher education. Between 1976 and 2010, enrollment increased by 47 percent at HBCUs and 91 percent in non-HBCU degree-granting institutions.
There could be a relationship between racial animus and Black students’ enrollment decisions. For example, while HBCUs have seen an overall decline in enrollment from 2010 to 2018, these institutions also saw an increase in enrollment during the campaign and election of Republican president Donald Trump. In the fall of 2017, HBCU enrollment was 298,138 students, a 2 percent increase from the prior year, at a time when enrollment was decreasing across institutions of higher education (National Student Clearinghouse Research Center 2017; NCES 2019). This increase represented a marked change in enrollments for HBCUs compared with prior years. As explained earlier, there is some qualitative evidence that Black students decide to enroll in HBCUs during times when racism and racist incidents are on the rise or are at least made more visible by increases in reported hate crimes (e.g., Williams and Palmer 2019). However, there is less research that quantitatively investigates this relationship, especially across the entire country. In the current study, we contribute to the broader body of knowledge on larger sociopolitical trends and their relationships with college enrollment by exploring whether an increase in racial animus was associated with changes in institutional enrollment.
Theoretical Framework
Guided by prior research on hate crimes, racist incidents, and HBCU college choice, we use the theoretical framework of racial homophily to guide the current study. Sociologists Lazarsfeld and Merton (1954) formalized the conception of homophily, seeking persons with similar characteristics. An extensive body of literature subsequently used this concept to explore how individuals seek out others similar to themselves to create social, romantic, and work networks in both schools and communities (e.g., Kandel 1978; McPherson et al. 2001; Kossinets and Watts 2009). Much of the work on racial homophily studies how this phenomenon relates to the social ties and work networks formed as well as the relationship between population demographics and homophily-seeking behaviors (e.g., Joyner and Kao 2000; Park and Bowman 2015).
Numerous studies explore the relationship between homophily and how individuals choose their social groups. Wimmer and Lewis (2010) found that racial homophily is one of the most important factors for tie formation for Black students in an elite HWI when compared with other racial and ethnic student groups. Stearns et al. (2009) found that interracial friendships tend to decrease for Black students when they move from secondary to collegiate settings. These findings suggest that Black students seek friendship with other Black students, particularly in settings where they constitute a relatively small fraction of the population or feel threatened by racism.
Thus, individuals who have witnessed an increase in racist rhetoric and reported hate crimes might be more likely to seek spaces that are more racially homophilous. Black students in their junior and senior year of high school are more likely to stay in their secondary schools, even if there is an increase in racist incidents, given the difficulty in transferring schools within or between districts. However, applying to and enrolling in college presents an opportunity to deliberately choose a place to learn. A component of the decision, particularly in a climate of heightened racial tension, is the perception of how racially welcoming and safe a campus is. Black students could intentionally seek out campuses with a clear, welcoming, and positive racial climate for students like themselves in order to avoid the psychological stressors related to racial animus, as well as physical instances of hate and aggression. It stands to reason, then, that Black students would seek racial homophily through the choice of a more racially homogenous college campus—an HBCU—in response to an increase in racial tensions in schools and/or broader communities. Although we analyze institutional data (as opposed to student-level data) in the current study, this aggregate data reflects a set of students’ individual choices. If we see increases in HBCU enrollment when there are more instances of racial animus, this might suggest that a correlation exists between expressions of racial animus and a collective set of students’ choices for a more homophilous campus environment.
Shifts in racial animus can affect college enrollment in several different ways. We have primarily discussed how students who planned to enroll in higher education shift their enrollment to an HBCU instead of a non-HBCU. However, it is important to point out that increased racial animus could motivate students not to pursue higher education altogether. For example, Black individuals could decide to forgo higher education and enter full-time collective action and organizing work in response to increased reports of hate crimes, or choose to enter the workforce directly, perhaps having calculated that the benefits of higher education are not worth the potential costs. All of that said, in the current study we are primarily interested in exploring how institutional enrollments shift between HBCUs and non-HBCUs. This is one of the driving reasons we chose to analyze institutional data.
Prior work finds that the campus racial climate is associated with student learning and well-being (e.g., Hurtado et al. 1998; Johnson-Ahorlu 2012; Eschmann 2020). Mwangi et al. (2018) expound on how Black students experience college campus climates as reflections of the national racial climate. In light of an increase in racist rhetoric and racist actions across the United States over the past several years, as measured by reported hate crimes, HBCUs may experience increased numbers of Black students enrolling as students seek racial homophily and a more racially welcoming climate that affords them some protection from racist incidents.
3. Research Method
Data
We compile data from the Department of Education's Integrated Postsecondary Education Data System (IPEDS), FBI hate crime data, population estimates from the U.S. Census Bureau, and employment data from the Bureau of Labor Statistics to create the analytical data for the current study (see table A1, available in a separate online appendix that can be accessed on Education Finance and Policy’s website at https://doi.org/10.1162/edfp_a_00400, for an overview of data sources and how they align). We use IPEDS as our primary data source since it contributes a census of all institutions receiving Title IV funds in the United States. We include all U.S. not-for-profit four-year institutions from 2000 to 2017 that serve first-time full-time students, but exclude for-profit institutions because of the high share of online learners at these institutions.5 We also exclude two-year institutions because the majority of HBCUs are four-year institutions (and the decision-making process for deciding whether to enroll in a two-year versus four-year institution differs from the decision process of choosing between four-year institutions).6 Similarly, we remove the U.S. service schools because of their special admissions requirements. Finally, because all analyses rely on the institution having enrollment data, we remove the limited number of institutions missing total first-time undergraduate enrollment.7 These restrictions result in an analytical sample of approximately 35,000 institution-year observations for the entire panel and between 1,860 and 2,050 institutions per year.
We supplement the IPEDS data with FBI data on reported hate crimes. We use Kaplan's (2019) concatenated UCR program data for 1999 to 2017, which includes reported hate crimes at the agency level. The UCR data provide figures on crimes voluntarily reported by police agencies each year. Importantly, not all states have UCR programs. For example, Indiana and Mississippi do not have these programs (Davis and O'Neill 2016). Still, agencies within the states can choose to report these crimes, and the vast majority of research focused on colleges and hate crime reporting relies on UCR data (e.g., Janosik 2004). We aggregate these data up to the state level to create state annual counts of reported hate crimes and merge it with IPEDS enrollment information so that hate crimes occurring during the prior academic year (August through July) are matched with fall enrollment for the following year.8 For example, fall enrollment in 2015 is linked to reports of hate crimes that occurred from August 2014 to July 2015. This creates close to two decades’ worth of annual state-level reports of hate crimes.9
We also include state population intercensal estimates from the U.S. Census Bureau.10 Due to the focus of the current study, we collect the number of college-eligible residents (age fifteen and above) and Black resident share (percentage; regardless of ethnicity or age) at the state level from 1999 to 2016.11 Intercensal estimates are generally collected in July. Therefore, we link these population estimates to the following year's enrollment since the states’ share of residents who are Black likely influences the number of reported hate crimes. Continuing with the 2015 example, fall enrollment in 2015 is merged with the July 2014 intercensal estimates. Finally, we include state annual local area unemployment rates from the Bureau of Labor Statistics; in our example, this means merging 2015 fall enrollment data with the 2015 unemployment rate.
Measures
Our primary measures of interest are enrollment, reported hate crimes, and institutional characteristics. From IPEDS, we collect information on first-time fall enrollment by race (all students, Black students).12 Though it is true that racial homophily could predict shifts in first-time enrollment, it could also create shifts in whether students persist or choose to transfer to a different institution. In order to evaluate this possibility, we analyze the relationship between reported hate crimes and total overall enrollment. Due to space constraints, we focus on the first-time enrollment estimates as the main results and discuss the total undergraduate enrollment results as sensitivity analyses.
We create state-level annual counts of the number of hate crimes from 1999 to 2016, which matches to 2000 to 2017 fall enrollment, respectively. The hate crimes are categorized as (1) any bias, (2) race-based, and (3) anti-Black.13 If a state is missing reported hate crime data, we record this as a value of 0 for each institution in that year since we are interested in studying how reported hate crimes relate to enrollment; “missing” means that the state or institution did not have reported hate crimes. We recognize that federal hate crime reports are one of many potential measures for racial animus that are likely to undercount actual incidents. We reflect on the strengths and weaknesses of selecting this measure in the Limitations and Discussion sections further below.
Guided by prior research (e.g., Stokes 2020), we include the following institutional characteristics: HBCU status, sector, region, urbanicity, in-state cost of attendance (COA), percentage of undergraduates receiving federal grants, state political ideology, affirmative action ban, total state college eligible population, Black share of state residents, and state unemployment rate. HBCU status, sector, region, urbanicity, in-state tuition, and the percentage of undergraduates receiving federal grants are linked to the likelihood of racist incidents on college campuses (e.g., Stokes 2020). These measures all come from IPEDS.14 We supplement this institutional data with several state measures. We use the Nominate measure of state government ideology (Berry et al., 2010). We also include an indicator for whether a state has a ban on affirmative action, given the relationship between statewide bans on affirmative action and college enrollment, and we link this measure to the first fall affected by an affirmative ban based on the years outlined in Baker (2019). If a state removes its ban, like Texas did after the 2003 Supreme Court cases, we allow the binary variable to return to 0. Total college eligible population (residents aged fifteen years and older) is included to allow us to control for the population of residents who are eligible to attend college, which is related to demand for college enrollment. We include the percentage of residents, regardless of age, who are Black since that share should be correlated with the number of hates crimes that are reported. This relates to the total number of reported hate crimes regardless of the bias motivation, since Black individuals are the most frequent targets of hate crimes. We also include the state unemployment rate due to its relationship with college enrollment: during economic downturns, students are more likely to enroll in college (e.g., Long 2014).
Empirical Strategy
We then investigate how this relationship shifts by including an indicator of HBCU status (=1 if institution is an HBCU) and an interaction between the annual number of hate crimes and HBCU status (model 2 above).
Model 3 adds a vector of covariates, , which includes sector (=1 if four-year private), region (New England, Mid East, Great Lakes, Plains, Southwest, Rocky Mountains, and Far West, with Southeast as the reference group), urbanicity (suburb, town, and rural, with city as the reference group), in-state COA, federal grant percentage, state government ideology, statewide affirmative action ban (=1 if state has an active ban), and continuous annual measures of the number of college-eligible persons by state, Black resident share (regardless of age), and state unemployment rate. For any model that does not have the total reported hate crimes as the measure of interest (i.e., race-based or anti-Black), we include a continuous measure of total hate crimes. As an example, when the measure of reported hate crimes is race-based, we include a control for the count of total hate crimes. is a year fixed effect that has a range of 2001 to 2017 (with 2000 as the reference year).
Thus, model 4 adds , an institution fixed effect, to model 3. Because we are exploring within-institution variance, time invariant institutional characteristics (including the indicator of whether an institution is an HBCU) cannot be included.15 Therefore, the vector of covariates for this model, , includes the in-state COA, federal grant percentage, state government ideology, statewide affirmative action ban indicator, state college eligible population, state share of Black residents, and state unemployment rate in all models, adding the appropriate total hate crimes reported measure when the hate crimes of interest for the model are race-based or anti-Black.
In all models, we compare to in order to analyze the difference in the relationship between reported hate crimes and enrollment at non-HBCUs and HBCUs. It could be that, in any given year, there are increases in the number of reported incidents, non-HBCU enrollments, and HBCU enrollments, which could be driven by the increase of the college-eligible or Black resident population. This is one reason that our interpretation of the models focuses on comparing the two estimates to each other and why we include a measure of the total college eligible population and Black resident share in the state. Standard errors are clustered at the state level for all models.
Limitations
First, extensive evidence suggests that hate crimes are underreported at the campus, state, and national levels (Pezzella et al. 2019). Two theories for the underreporting are that victims are not as likely to report hate crimes as other crimes and agencies are more likely to misclassify this type of crime, particularly in places with histories of racial animus. Given the underreporting, our estimates likely underestimate the relationship between hate crimes and the college enrollment decisions of Black students. Nonetheless, policy makers at the federal, state, and institutional levels only have reported hate crimes available as a measure for crafting policies to support students and create equitable campus climates. Therefore, while we acknowledge that reported hate crimes are an undercount, they are also the most policy-relevant measure that can be used by stakeholders to improve conditions for students.
Second, the current research focuses on HBCUs and considers non-HBCUs with a significant share of the undergraduate enrollment who identify as Black as similar to HWIs. We chose to focus on HBCU enrollment due to the significant body of research that shows that simply enrolling a large share of students of color does not automatically mean that an institution is focused on serving those students (e.g., Garcia et al. 2019). Therefore, we sought to explicitly examine how reported hate crimes are associated with enrollment at institutions whose missions directly call for the uplifting of the Black community.
Third, we acknowledge that higher education enrollment consists of choices made both by students and by higher education institutions. Our theoretical framework is primarily focused on understanding students’ responses to racial animus, which can create shifts in enrollment within higher education at the institution level. However, institutions play a clear role in determining who enrolls in their institution, even if that is by choosing to be an open-access institution.
Finally, we are unable to explore whether changes in enrollment at HBCUs involve in-state and/or out-of-state students. IPEDS does not consistently collect the number or percentage of students enrolled from other states. Also, when this information is reported, there is no way to link changes in enrollment to the students’ state of residence, nor are there racial demographic breakdowns of the residency data. If the growth is from students in other states, then it is possible that it could be driven either by rates of hate crimes in their home states or other factors. If increases in enrollment are driven by in-state students, then this might serve as evidence of the relationship between the number of hate crime reports and HBCU enrollment. Future research using a single state's administrative data might be able to explore this question further.
4. Results
Descriptive Statistics
We provide an overview of the descriptive statistics on the analytical sample, with attention to patterns in the reporting of hate crimes, to provide context for the main results. Table 1 includes summary statistics of institutional characteristics and the measures of interest. The All States columns (1–3) show variable means from 2000, 2008, and 2017. Panel A provides summary statistics of institutional characteristics. Four-year private institutions are the dominant sector in the sample. Regardless of the time period, the majority of institutions are located in cities, in states without an affirmative action ban, and in the southeastern region of the United States. Over our analytical panel, in-state COA rose from approximately $16,000 to $33,000, a 106 percent increase. The percentage of students receiving federal grants held fairly steady but has risen in recent years (from 35 percent in 2000 to 42 percent in 2017). State governments became more ideologically liberal toward the middle of our panel, then rebounded to become even more conservative than at the start. The college-eligible population increases by approximately 22 percent from 2000 to 2017. During the same time period, the average share of state residents who identify as Black in states where institutions are located remains fairly static at nearly 13 percent. Panel B shows the summary statistics for the varying definitions of reports of hate crimes and first-time enrollment. Over the analytical panel, all statewide reports of hate crimes have decreased. At the same time, there has been a general increase in average first-time enrollment for all students, and Black students in particular, though there is a dip in average enrollment around the Great Recession.
Summary Statistics of Institutional Characteristics and Key Variables
. | All States . | HBCU States . | ||||
---|---|---|---|---|---|---|
. | 2000 . | 2008 . | 2017 . | 2000 . | 2008 . | 2017 . |
Panel A: Institutional Characteristics | ||||||
HBCU | 0.05 | 0.04 | 0.04 | 0.10 | 0.09 | 0.09 |
4-year private | 0.69 | 0.68 | 0.66 | 0.66 | 0.65 | 0.63 |
New England | 0.09 | 0.09 | 0.08 | 0.00 | 0.00 | 0.00 |
Mid East | 0.21 | 0.20 | 0.19 | 0.21 | 0.20 | 0.18 |
Great Lakes | 0.17 | 0.16 | 0.15 | 0.10 | 0.10 | 0.10 |
Plains | 0.11 | 0.11 | 0.10 | 0.06 | 0.06 | 0.06 |
Southwest | 0.07 | 0.07 | 0.07 | 0.12 | 0.13 | 0.13 |
Rocky Mountains | 0.03 | 0.03 | 0.03 | 0.00 | 0.00 | 0.00 |
Far west | 0.10 | 0.11 | 0.12 | 0.00 | 0.00 | 0.00 |
Suburb | 0.25 | 0.21 | 0.25 | 0.25 | 0.20 | 0.23 |
Town | 0.19 | 0.20 | 0.19 | 0.19 | 0.21 | 0.22 |
Rural | 0.06 | 0.11 | 0.07 | 0.07 | 0.12 | 0.07 |
In-state COA | 15,977.54 | 24,243.93 | 32,718.31 | 14,863.45 | 23,048.76 | 30,768.16 |
(7,605.78) | (11,613.92) | (16,976.85) | (7,121.22) | (10,720.10) | (15,796.46) | |
Federal grant recipients (%) | 34.57 | 34.13 | 42.34 | 35.48 | 36.92 | 45.01 |
(20.79) | (20.28) | (21.07) | (21.40) | (21.36) | (21.31) | |
State government ideology | 47.42 | 51.12 | 42.28 | 45.90 | 44.94 | 31.84 |
(13.34) | (14.77) | (18.09) | (13.06) | (14.82) | (11.95) | |
State affirmative action ban | 0.15 | 0.15 | 0.21 | 0.16 | 0.08 | 0.14 |
State college eligible (thousands) | 7,934.50 | 8,657.24 | 9,696.13 | 7,118.89 | 8,016.67 | 9,245.66 |
(6,419.81) | (7,148.47) | (8,116.66) | (4,146.15) | (4,877.04) | (5,892.31) | |
State Black resident share (%) | 12.88 | 12.81 | 13.09 | 17.50 | 17.53 | 17.86 |
State unemployment rate (%) | 4.13 | 4.62 | 4.80 | 4.16 | 4.59 | 5.00 |
Panel B: Hate Crime Reports and Enrollment | ||||||
Hate crime reports | ||||||
Total | 311.14 | 282.92 | 261.71 | 138.64 | 135.83 | 134.76 |
(448.90) | (344.53) | (289.55) | (98.85) | (95.90) | (114.48) | |
Race-based | 191.11 | 164.61 | 140.72 | 101.36 | 92.83 | 89.99 |
(275.27) | (200.39) | (149.39) | (69.42) | (64.87) | (83.40) | |
Anti-Black | 105.93 | 93.53 | 69.88 | 61.81 | 55.49 | 43.66 |
(139.29) | (107.74) | (70.90) | (42.93) | (34.96) | (32.69) | |
First-time enrollment | ||||||
Total | 715.41 | 838.56 | 921.05 | 723.23 | 847.68 | 951.97 |
(980.30) | (1,173.73) | (1,362.76) | (960.55) | (1,165.41) | (1,412.49) | |
Black student | 78.07 | 97.47 | 105.23 | 117.62 | 143.53 | 156.74 |
(168.50) | (191.93) | (206.87) | (221.32) | (246.30) | (269.62) |
. | All States . | HBCU States . | ||||
---|---|---|---|---|---|---|
. | 2000 . | 2008 . | 2017 . | 2000 . | 2008 . | 2017 . |
Panel A: Institutional Characteristics | ||||||
HBCU | 0.05 | 0.04 | 0.04 | 0.10 | 0.09 | 0.09 |
4-year private | 0.69 | 0.68 | 0.66 | 0.66 | 0.65 | 0.63 |
New England | 0.09 | 0.09 | 0.08 | 0.00 | 0.00 | 0.00 |
Mid East | 0.21 | 0.20 | 0.19 | 0.21 | 0.20 | 0.18 |
Great Lakes | 0.17 | 0.16 | 0.15 | 0.10 | 0.10 | 0.10 |
Plains | 0.11 | 0.11 | 0.10 | 0.06 | 0.06 | 0.06 |
Southwest | 0.07 | 0.07 | 0.07 | 0.12 | 0.13 | 0.13 |
Rocky Mountains | 0.03 | 0.03 | 0.03 | 0.00 | 0.00 | 0.00 |
Far west | 0.10 | 0.11 | 0.12 | 0.00 | 0.00 | 0.00 |
Suburb | 0.25 | 0.21 | 0.25 | 0.25 | 0.20 | 0.23 |
Town | 0.19 | 0.20 | 0.19 | 0.19 | 0.21 | 0.22 |
Rural | 0.06 | 0.11 | 0.07 | 0.07 | 0.12 | 0.07 |
In-state COA | 15,977.54 | 24,243.93 | 32,718.31 | 14,863.45 | 23,048.76 | 30,768.16 |
(7,605.78) | (11,613.92) | (16,976.85) | (7,121.22) | (10,720.10) | (15,796.46) | |
Federal grant recipients (%) | 34.57 | 34.13 | 42.34 | 35.48 | 36.92 | 45.01 |
(20.79) | (20.28) | (21.07) | (21.40) | (21.36) | (21.31) | |
State government ideology | 47.42 | 51.12 | 42.28 | 45.90 | 44.94 | 31.84 |
(13.34) | (14.77) | (18.09) | (13.06) | (14.82) | (11.95) | |
State affirmative action ban | 0.15 | 0.15 | 0.21 | 0.16 | 0.08 | 0.14 |
State college eligible (thousands) | 7,934.50 | 8,657.24 | 9,696.13 | 7,118.89 | 8,016.67 | 9,245.66 |
(6,419.81) | (7,148.47) | (8,116.66) | (4,146.15) | (4,877.04) | (5,892.31) | |
State Black resident share (%) | 12.88 | 12.81 | 13.09 | 17.50 | 17.53 | 17.86 |
State unemployment rate (%) | 4.13 | 4.62 | 4.80 | 4.16 | 4.59 | 5.00 |
Panel B: Hate Crime Reports and Enrollment | ||||||
Hate crime reports | ||||||
Total | 311.14 | 282.92 | 261.71 | 138.64 | 135.83 | 134.76 |
(448.90) | (344.53) | (289.55) | (98.85) | (95.90) | (114.48) | |
Race-based | 191.11 | 164.61 | 140.72 | 101.36 | 92.83 | 89.99 |
(275.27) | (200.39) | (149.39) | (69.42) | (64.87) | (83.40) | |
Anti-Black | 105.93 | 93.53 | 69.88 | 61.81 | 55.49 | 43.66 |
(139.29) | (107.74) | (70.90) | (42.93) | (34.96) | (32.69) | |
First-time enrollment | ||||||
Total | 715.41 | 838.56 | 921.05 | 723.23 | 847.68 | 951.97 |
(980.30) | (1,173.73) | (1,362.76) | (960.55) | (1,165.41) | (1,412.49) | |
Black student | 78.07 | 97.47 | 105.23 | 117.62 | 143.53 | 156.74 |
(168.50) | (191.93) | (206.87) | (221.32) | (246.30) | (269.62) |
Notes: Means are reported for all variables. Standard deviations are included in the parentheses for the continuous variables. All States columns include all four-year not-for-profit institutions (excluding U.S. service schools). HBCU (historically Black college and university) States columns includes the same sample of institutions but trims out any institutions physically located in a state without at least one HBCU. COA = cost of attendance.
Total Reports of Hate Crimes by State (for states with at least one HBCU [historically Black college and university])
Total Reports of Hate Crimes by State (for states with at least one HBCU [historically Black college and university])
Columns 4, 5, and 6 of table 1 show the descriptive statistics for institutions in states with at least one HBCU. The sectors and urbanicity of institutions have qualitatively similar trends. However, this sample includes no institutions in the New England, Rocky Mountains, or Far West regions of the United States. While the general trend is similar, fewer institutions in states with at least one HBCU are subject to an affirmative action ban (14 percent in 2017 compared with 21 percent for all institutions). States with at least one HBCU have a fairly steady trend in state government ideology becoming more conservative. These states also have a higher share of residents who identify as Black, approximately 18 percent on average, and a less precipitous decline in reporting of hate crimes over time, which may be due, in part, to the fact that these states had fewer reported hate crimes at the beginning of the analytical time period. In contrast to the All States sample, the enrollment trends in HBCU states are steadily increasing over time with no dip around the Great Recession. The descriptive differences between the All States and HBCU States samples provide additional evidence that it would be worthwhile to trim the analytical sample and focus primarily on states with at least one HBCU, which we do with the main estimates.
Main Estimates
Table 2 provides estimates from the preferred models, including total hate crimes (all models included in online appendix table A2). Panel A contains models with the outcome of total first-time enrollment. Panel B includes models with the outcome Black first-time enrollment. Column 1 includes model 3, including all covariates, year fixed effects, and prior year enrollment, while column 2 includes model 4, which adds an institution fixed effect to model 3 and removes all time-invariant institutional characteristics. We primarily focus on the estimates for the reported hate crimes measure and that variable's interaction with HBCU status, since the coefficients report the correlational relationship between reported hate crimes and enrollment for non-HBCUs and HBCUs, respectively. We include a test of joint significance in order to explore whether the interactions, in addition to the indicator of HBCU status, are jointly significant even when individual estimates are not. When applicable, we use the standard deviation of reported hate crimes and average number of first-time students at non-HBCUs (since these institutions are the reference group) to translate the estimates into more practically relevant numbers.17
Total Reports of Hate Crimes Predicting First-Time Enrollment
. | (1) . | (2) . |
---|---|---|
Panel A: Total First-Time Enrollment | ||
Total reported hate crimes | 0.016 | −0.037 |
(0.017) | (0.025) | |
HBCU | −4.560 | |
(4.118) | ||
Total reported hate crimes * HBCU | −0.008 | 0.204** |
(0.028) | (0.065) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.108 | 0.006 |
Panel B: Black First-Time Enrollment | ||
Total reported hate crimes | −0.004 | −0.011 |
(0.003) | (0.009) | |
HBCU | 10.918** | |
(3.481) | ||
Total reported hate crimes * HBCU | 0.036 | 0.214** |
(0.018) | (0.074) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.001 | 0.010 |
. | (1) . | (2) . |
---|---|---|
Panel A: Total First-Time Enrollment | ||
Total reported hate crimes | 0.016 | −0.037 |
(0.017) | (0.025) | |
HBCU | −4.560 | |
(4.118) | ||
Total reported hate crimes * HBCU | −0.008 | 0.204** |
(0.028) | (0.065) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.108 | 0.006 |
Panel B: Black First-Time Enrollment | ||
Total reported hate crimes | −0.004 | −0.011 |
(0.003) | (0.009) | |
HBCU | 10.918** | |
(3.481) | ||
Total reported hate crimes * HBCU | 0.036 | 0.214** |
(0.018) | (0.074) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.001 | 0.010 |
Notes: Panel A include estimates from models with total first-time enrollment as the outcome. Panel B include estimates from models with Black student first-time enrollment as the outcome. Covariates include prior year's enrollment, sector, region, urbanicity, in-state COA (cost of attendance), federal grant recipient percentage, state government ideology, statewide affirmative action ban, state college-eligible population, state Black resident share, and state unemployment rate. Indicator for HBCU (historically Black college and university) status is not included in column 2 since it includes institution fixed effects (FE). Joint significance test includes the p-values from joint significance test of total reported hate crimes, HBCU, and their interaction in model 1 and total reported hate crimes and its interaction with HBCU in model 2. Sample only includes institutions in states with at least one HBCU. Standard errors clustered at the state level. *p < 0.05, **p < 0.01.
In table 2 panel A, the estimates show that there is little evidence of a consistent relationship between the total reported hate crimes within a state and the first-time enrollment for both HBCUs and non-HBCUs. We find statistically insignificant estimates of the relationship between total reported hate crimes and first-time enrollment at non-HBCUs (first row). This means that we do not have any clear evidence that the estimate is statistically different from 0. We find evidence of a positive relationship between total reported hate crimes and first-time enrollment at HBCUs only when we include institution fixed effects. Said another way, when we remove institution-specific, time-invariant trends, we find that each additional reported hate crime at the state level is associated with a 0.2-student increase in first-time enrollment at HBCUs.
Turning to panel B, the estimates for non-HBCUs (first row) provide little evidence of a relationship between reports of hate crimes and Black student enrollment—0.04 to 0.01 predicted decrease in the number of students, with neither estimate statistically significant. The estimates for HBCU enrollment, similar to panel A, are statistically insignificant at conventional levels until the final column (true across all models as shown in online appendix table A2). Column 2, our preferred specification, shows that, when controlling for time-invariant characteristics of institutions and time-varying state characteristics, an increase in states’ total reported hate crimes predicts a 0.21 increase in Black student first-time enrollment at HBCUs. This means that a 1 standard deviation increase in total state-level hate crimes in 2017 (114.48) is associated with a twenty-four student increase in Black student enrollment at HBCUs (multiplying the standard deviation by the point estimate). Using the average number of Black first-time students attending non-HBCUs (120.87) in 2017 as a baseline since they are the reference group, we predict a nearly 20 percent increase in students (twenty-four additional students from a baseline of 120.87) enrolling in an HBCU in a given year—a modest, yet practically significant, increase.
Table 3 presents the race-based hate crimes estimates. When focusing on total first-time enrollment and the institution fixed effects models (panel A), an additional race-based hate crime in the state predicts a decrease in overall enrollment at non-HBCUs—0.25 to 0.39, which is approximately 3 percent fewer total first-time students for a one standard deviation increase. One additional report of a race-based hate crime also predicts an 0.33 increase in overall enrollment at HBCUs, approximately 23 percent more first-time students per standard deviation increase in reported race-based hate crimes. Panel B shows a more consistent positive relationship between an increase in reported race-based hate crimes and Black students’ enrollment at HBCUs—22 percent or twenty-seven more Black students per standard deviation increase in the institution fixed effects model.
Race-Based Reports of Hate Crimes Predicting First-Time Enrollment
. | (1) . | (2) . |
---|---|---|
Panel A: Total First-Time Enrollment | ||
Total reported hate crimes | −0.253** | −0.382** |
(0.080) | (0.119) | |
HBCU | −5.249 | |
(4.133) | ||
Total reported hate crimes * HBCU | −0.002 | 0.328** |
(0.042) | (0.085) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.005 | 0.000 |
Panel B: Black First-Time Enrollment | ||
Total reported hate crimes | −0.041 | −0.105 |
(0.049) | (0.054) | |
HBCU | 10.423** | |
(3.496) | ||
Total reported hate crimes * HBCU | 0.060* | 0.318** |
(0.027) | (0.098) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.001 | 0.001 |
. | (1) . | (2) . |
---|---|---|
Panel A: Total First-Time Enrollment | ||
Total reported hate crimes | −0.253** | −0.382** |
(0.080) | (0.119) | |
HBCU | −5.249 | |
(4.133) | ||
Total reported hate crimes * HBCU | −0.002 | 0.328** |
(0.042) | (0.085) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.005 | 0.000 |
Panel B: Black First-Time Enrollment | ||
Total reported hate crimes | −0.041 | −0.105 |
(0.049) | (0.054) | |
HBCU | 10.423** | |
(3.496) | ||
Total reported hate crimes * HBCU | 0.060* | 0.318** |
(0.027) | (0.098) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.001 | 0.001 |
Notes: Panel A include estimates from models with total first-time enrollment as the outcome. Panel B include estimates from models with Black student first-time enrollment as the outcome. Covariates include total reports of hate crimes, prior year's enrollment, sector, region, urbanicity, in-state COA (cost of attendance), federal grant recipient percentage, state government ideology, statewide affirmative action ban, state college-eligible population, state Black resident share, and state unemployment rate. Indicator for HBCU (historically Black college and university) status is not included in column 2 since it includes institution fixed effects (FE). Joint significance test includes the p-values from joint significance test of total reported hate crimes, HBCU, and their interaction in model 1 and total reported hate crimes and its interaction with HBCU in model 2. Sample only includes institutions in states with at least one HBCU. Standard errors clustered at the state level. *p < 0.05, **p < 0.01.
Table 4 focuses on hate crimes with an anti-Black bias motivation. The results show a positive correlation between reports of anti-Black hate crimes and both overall and Black first-time enrollment at HBCUs in the institution fixed effect models. We find that an additional report of an anti-Black hate crime predicts an increase in Black students’ first-time enrollment at HBCUs of 17 percent, or twenty more per standard deviation increase. Interestingly, we also find a positive relationship between total reported anti-Black hate crimes and enrollment at non-HBCUs in the full covariate model that does not appear in the institution fixed effect model. We explore this shift by reestimating table 4, column 1 panel A via removing institutions with a single year of data (since these observations are dropped when including institution fixed effects). We obtain qualitatively similar estimates to the coefficients from models including those institutions. When we explore other elements of the models, we find that the time-variant characteristics of federal grant recipient percentage, in-state COA, and whether a state has an affirmative action ban lose their statistically significant relationship with first-time enrollment when we include institution fixed effects. Therefore, it may be that something unique about institutions that relates to their affordability partially explains the shift in estimates from column 1 to column 2 in table 4.
Anti-Black Reports of Hate Crimes Predicting First-Time Enrollment
. | (1) . | (2) . |
---|---|---|
Panel A: Total First-Time Enrollment | ||
Total reported hate crimes | 0.212* | 0.112 |
(0.099) | (0.144) | |
HBCU | −3.017 | |
(3.895) | ||
Total reported hate crimes * HBCU | −0.061 | 0.511** |
(0.076) | (0.177) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.016 | 0.003 |
Panel B: Black First-Time Enrollment | ||
Total reported hate crimes | 0.002 | −0.059 |
(0.041) | (0.054) | |
HBCU | 10.566** | |
(3.548) | ||
Total reported hate crimes * HBCU | 0.105 | 0.618** |
(0.057) | (0.194) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.001 | 0.004 |
. | (1) . | (2) . |
---|---|---|
Panel A: Total First-Time Enrollment | ||
Total reported hate crimes | 0.212* | 0.112 |
(0.099) | (0.144) | |
HBCU | −3.017 | |
(3.895) | ||
Total reported hate crimes * HBCU | −0.061 | 0.511** |
(0.076) | (0.177) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.016 | 0.003 |
Panel B: Black First-Time Enrollment | ||
Total reported hate crimes | 0.002 | −0.059 |
(0.041) | (0.054) | |
HBCU | 10.566** | |
(3.548) | ||
Total reported hate crimes * HBCU | 0.105 | 0.618** |
(0.057) | (0.194) | |
Covariates & Year FE | X | X |
Institution FE | X | |
Joint significance test | 0.001 | 0.004 |
Notes: Panel A include estimates from models with total first-time enrollment as the outcome. Panel B include estimates from models with Black student first-time enrollment as the outcome. Covariates include total reports of hate crimes, prior year's enrollment, sector, region, urbanicity, in-state COA (cost of attendance), federal grant recipient percentage, state government ideology, statewide affirmative action ban, state college-eligible population, state Black resident share, and state unemployment rate. Indicator for HBCU (historically Black college and university) status is not included in column 2 since it includes institution fixed effects (FE). Joint significance test includes the p-values from joint significance test of total reported hate crimes, HBCU, and their interaction in model 1 and total reported hate crimes and its interaction with HBCU in model 2. Sample only includes institutions in states with at least 1 HBCU. Standard errors clustered at the state level. *p < 0.05, **p < 0.01.
Overall, an increase in reports of hate crimes predicts increases in Black students’ first-time enrollment at HBCUs once we control for time-invariant characteristics of institutions (1 standard deviation increase in reports of hate crimes predicts between a 17 to 22 percent increase in Black students) in this study. While the strongest evidence comes from models where we control for time-invariant characteristics of institutions, there is suggestive evidence that this relationship exists even when we do not include institutional fixed effects. In column 1 panel B of tables 2–4, while we do not consistently see statistically significant estimates at conventional levels, they are marginally significant (p < 0.10). Therefore, we have suggestive evidence in all of the main models that an increase in reports of hate crimes is associated with an increase in Black student first-time enrollment. When we only focus on total enrollment with the same controls, we find a consistent predicted increase in first-time students at HBCUs when reports of hate crimes increase, regardless of the type of hate crime.
Sensitivity Analyses
Note: HBCU = historically Black college and university.
Note: HBCU = historically Black college and university.
Note: HBCU = historically Black college and university.
Note: HBCU = historically Black college and university.
Note: HBCU = historically Black college and university.
Note: HBCU = historically Black college and university.
We prefer the models using first-time enrollment as an outcome, as this allows us to focus on students who are choosing to enroll in higher education. However, our theoretical framework of homophily could influence students’ decisions on persisting at HBCUs at a different rate than at non-HBCUs. In order to investigate this, we estimate the main models using the outcome of total overall undergraduate enrollment and total Black undergraduate enrollment (online appendix tables A5–A7). We consistently find a positive relationship between total reported hate crimes and Black students’ overall enrollment at HBCUs regardless of the type of hate crime. These findings provide evidence that the relationship between reports of hate crimes and Black student enrollment endure whether we focus on first-time or overall undergraduate enrollment (though the magnitude of the estimate for overall enrollment is larger).
We explore the sensitivity of our findings to expanding the analytical sample to include institutions in all states (online appendix tables A8–A10). All estimates for total reported hate crimes’ relationship with Black student enrollment at HBCUs are qualitatively similar to the main estimates, with a few exceptions. Those exceptions relate to the estimates for enrollment at non-HBCUs, which align with our concerns that the comparison group of colleges that students are using may be fundamentally different for those living in states with an HBCU compared to those living in states without an HBCU.
Lastly, we explore whether our estimates are being driven by a select set of HBCUs. We reestimate the main models excluding the five “most highly ranked” HBCUs.18 We find qualitatively similar estimates in magnitude, direction, and statistical significance.
5. Discussion
In this descriptive study, we explore the relationship between an increase in racial animus, as measured by reported hate crimes within a state, and changes to the enrollment patterns of Black students. We primarily explore this relationship by focusing on HBCUs, as considerable prior research demonstrates how HBCUs create welcoming environments for Black students. We find consistent, though modest, relationships between institutions’ Black student enrollment and reported hate crimes. An increase in the number of reported hate crimes predicts an increased number of Black students enrolled at HBCUs. We also find evidence that the overall enrollment at HBCUs increases contemporaneously as total reports of hate crimes increase. There are several reasons one might expect to see a relationship between reports of hate crimes and total enrollment at HBCUs. For example, Maramba et al. (2015) investigated why Asian and Latinx students choose to enroll at HBCUs, highlighting the supportive learning environment.
We do not consistently see declines in enrollment at non-HBCUs when there are increases in the number of hate crimes. It is not clear what options these students had in their choice sets when determining to enroll at an HBCU. One explanation for these findings is the complexity of the college choice process and the high price of college. It could be the case that students who have additional financial resources are more likely to enroll in HBCUs when the number of hate crimes increase. However, many students might need to take the best financial package available and thus, even though students would prefer to enroll at an HBCU, they must attend a non-HBCU. To deepen scholars’ understanding of this relationship it could potentially be useful to explore more granular student-level data, such as state administrative data, to better explore what these students might have been doing had they not chosen to enroll at an HBCU. Ultimately, this study provides new evidence that a relationship exists between the prevalence of reports of hate crimes within a state and college enrollment decisions for Black students at HBCUs.
Prior research suggests that campus racial climates reflect those at the state and national levels and that Black students make decisions about their social ties on campus based, in part, on the seeking of racial homophily (Stearns et al. 2009; Wimmer and Lewis 2010; Stokes 2020). Students’ ties to racially homophilous groups on campus are stronger than their ties to the campus in general (Gilkes Borr 2019). The findings from this study suggest that Black students are not only making decisions about their social ties on campuses based on perceptions of racially safe spaces, but also that these students and their families might be making choices about where to enroll based on these factors. It is not clear what mechanisms may be creating this relationship. Future research should explore the role that organizational structures within institutions play in mediating the relationship between reports of hate crimes and Black student enrollment.19
We focus our analyses on institutions located in states with at least one HBCU. While the data in this study do not allow for an exploration of the demographic make-up of students’ home communities, for historical reasons, most HBCUs are located in states with relatively larger Black populations when compared with states that do not host these institutions (see table 1). It could be that students coming from more racially homophilous home communities seek to replicate these spaces, and the feeling of increased safety they provide, in their new college community.
As the number of hate crimes reported to the police in the state increases, especially those that are racially motivated, it seems reasonable that students’ awareness of sociopolitical tensions around race and racism could increase as well. As a result, potential college students and their parents who hear about reports of hate crimes could choose to enroll in college partially in response to a desire for spaces with fewer racial incidents. Reports of race-based hate crimes that occur both off- and on-campus are associated with greater stress and less optimal academic outcomes for young persons (Keels et al. 2017). In contrast, spaces perceived to be more racially welcoming could have a positive relationship to academic success, with a higher likelihood of college persistence and graduation. Thus, the choice of a more racially homophilous space for Black students at an HBCU serves as a strategy to increase college success (Laird et al. 2007; Palmer et al. 2010).
Policy Implications
There are policy implications for both states and campuses in response to the increase in racial animus. One of the limitations of this study is the underreporting of hate crimes by individuals as well as misclassification of hate crimes by police (Masucci and Langton 2017). For states and campuses, there is a need for additional training for potential reporters with respect to hate crimes. One possibility is to provide more information to the public about how hate crimes are defined and require police reports taken on- and off-campus to consider whether the offense could constitute a hate crime, namely, providing authorities with a checklist of the possible ways in which the crime could be classified. More accurate data are needed in order to understand the full extent of the number of hate crimes committed so that this information can be used to measure the relationship between the number of hate crimes and health and social outcomes.
When it comes to college campuses, we find that the number of hate crimes reported in a state are associated with overall enrollment and Black student enrollment rate in particular. To help mitigate the potential negative effects perceptions of hate crime can have on students’ mental health and academic performance, all college campuses should provide additional support to students, especially Black students. Further, it is incumbent on non-HBCU campuses to clearly communicate how they are providing support to prospective Black students and their families. Prior work demonstrates that one of the challenges to fostering a supportive environment in higher education is that free speech laws can be (and have been) used to defend racist hate speech (Moore and Bell 2017). Thus, a seeming lack of repercussions for hate speech, or even defenses for it, could potentially render campuses more open to hate crimes. Our research makes clear that if the goal of institutions is to increase racial equity in their enrollment, they need to focus on creating truly welcoming climates that demonstrate a clear commitment to the success of Black individuals. HBCUs are at the forefront of constructing these environments.
6. Conclusion
We find that a 1 standard deviation increase in the number of reported hate crimes, regardless of the bias motivation, is associated with an approximately 20 percent increase in first-time Black students enrolled at HBCUs. This study focused on the changes in enrollment at HBCUs for Black students following an increase in reports of hate crimes. However, larger sociopolitical animus has been on the rise for many individuals—animus based on gender and religion, for example.
There is currently a dearth of work on the relationship between an individual's life choices and increases in animus—whether the targets are selected on the basis of their race, gender, religion, or some other characteristic—in the United States. Though we explore college enrollment, it is likely that the increase in these crimes might not only change where people go to college but also where people ultimately choose to live and the professions that they enter. It is important to explore the correlation between incidences of hate speech and hate crimes both on campuses and in society. In the world of higher education, it is important to understand how an increase in macro-level intolerance and oppression is related to institutional enrollment decisions. Accordingly, higher education institutions must grapple with these realities if they truly want to create welcoming campus environments where all students can thrive while feeling safe.
Acknowledgments
The authors would like to thank Benjamin Skinner, Denisa Gándara, Sondra Barringer, and Christopher Bennett for their helpful comments. The authors would also like to thank the seminar participants at New York University and the University of Pennsylvania for their constructive feedback and suggestions. The authors bear sole responsibility for the content of this article.
Notes
We use the federal government's definition of hate crime: crimes committed based on a victim's race, national origin, ethnicity, religion, sexual orientation, gender, or disability.
While the number of reported hate crimes in the United States increased, particularly in large cities, where the number of hate crimes reported in the ten largest cities rose by 12.5 percent in 2017 (Levin and Reitzel 2018), rates of other types of crime were falling. Gramlich (2020) notes that the violent crime rate fell 49 percent between 1993 and 2019 and the property crime rate fell 55 percent during the same time period.
Some education policy and higher education research uses the term HWI rather than predominantly White institution (PWI) because the changing demographic profile of the college-going population means that a number of institutions that had a majority of White students at inception are no longer PWIs.
Recognizing includes both police officers understanding and acknowledging that a hate crime has occurred. These are two different processes that we wish to explicitly name since one assumes ignorance and the other intent.
We use ft_ftug from IPEDS to restrict the sample of institutions to those that have ever served first-time, full-time undergraduate students in at least one year of the panel.
We do also estimate the models for all two- and four-year institutions and find qualitatively similar results, although, since the theory is weak for keeping two-year institutions in the sample, we do not rely on those models for the current study.
We find that approximately sixty institutions never report first-time enrollment (out of approximately 2,300 institutions). None of these institutions are HBCUs and, given that there are no practically significant differences between these institutions and the rest of the sample, we remove them.
We do not include annual counts of hate crime reports below the state level (i.e., county, city) due to the significant deterioration in quality in the data for smaller geographic units over the years in this study.
We include reports from the summer in light of research affirming that the college decision process often continues into the summer before college for students of color and the prevalence of college melt (e.g., Rall 2016).
Census data from 2010 to 2017 can be accessed at https://www.census.gov/data/tables/time-series/demo/popest/2010s-state-detail.htm. We use sc-est-2019-alldata6, which includes age, sex, race, and Hispanic origin. Census data from 2000 to 2009 can be accessed through the datasets archive https://www2.census.gov/programs-surveys/popest/datasets/. We use st-est00-alldata. We collected census data for 1999 from http://www.census.gov/popest/data/intercensal/.
In older years of intercensal data, the age groups do not allow us to include only residents who are eighteen and older (since the categories are 15–19 and 20–24 years). Therefore, we chose to operationalize the college-eligible population as any resident who is at least fifteen years old.
Over time, IPEDS has several different measures for Black student enrollment. We use efrace04 and efrace03 combined from 1999 to 2001 (these are gender disaggregated enrollments that we combine to create a total measure), efrace18 from 2002 to 2007, dvefbk from 2008 to 2009, and efbkaa from 2010 to 2017. We start with 1999 to create a lag measure of enrollment (since the primary analysis is from 2000 to 2017).
The FBI data allows for up to five bias motivations per reported hate crime. For the race-based hate crimes, we followed FBI guidance and included reported hate crimes that attributed any bias motivation to anti-Black, anti-White, anti-Arab, anti–American Indian or Native Alaskan, anti-Asian, anti-Hispanic, anti-ethnicity other than Hispanic, anti-multiracial group, and anti–Native Hawaiian or other Pacific Islander. For anti-Black hate crimes, we only included reported hate crimes that included anti-Black as one of the five motivations.
In-state COA is the sum of chg2ay3 (published in-state tuition and fees), chg4ay3 (books and supplies), chg5ay3 (on-campus room and board), and chg6ay3 (on-campus other expenses). The 1999 IPEDS data collection did not collect all of the tuition data that we use, so we supplemented this information with the 2000 IPEDS data collection for the 2019 data (since IPEDS reports tuition for the current data collection year and up to three years prior). We use the percentage of undergraduates receiving federal grants due to IPEDS's consistent collection of this data throughout the analytical time period.
The U.S. Department of Education designated one institution, American Baptist College, as an HBCU in 2013 (Roach 2013). Due to this, it is the only institution with variation in the indicator of HBCU status. We still do not include the HBCU binary in the institution fixed effects models (since only a single institution varies during the analytical time period). We estimate all models not including American Baptist College and the results are qualitatively similar in magnitude, direction, size, and statistical significance.
See these same figures with reports of hate crimes per 500,000 college-eligible residents (at least fifteen years old) in online appendix figures A1 and A2.
Both measures come from the restricted sample of states with at least one HBCU. The average number of first-time students for non-HBCUs in 2017 is 987.79 for total enrollment and 120.87 for Black student enrollment.
These institutions include Spelman College, Howard University, Xavier University, Hampton University, and Morehouse College, according to recent rankings reported in U.S. News and World Report (2022).
We explored the potential for the Black share of tenure-line faculty to be a mediating variable between reports of hate crimes and enrollment at HBCUs. We found no evidence of this relationship (estimates available upon request).
REFERENCES
Author notes
The authors contributed equally to this study and names are listed in alphabetical order. Dr. Baker completed the bulk of this work while at Southern Methodist University.