Differential impacts of college ratings: the case of education deserts
Nicholas W. Hillman
Assistant Professor
Educational Leadership and Policy Analysis
University of Wisconsin-Madison
DRAFT NOT FOR QUOTATION OR CITATION WITHOUT AUTHORS' PERMISSION
Last updated 8/27/2014
Author’s note:
Please email me at [email protected] for update versions, data inquiries, or feedback on this
draft. I would like to thank Zahra Mojtahedi and Taylor Weichman for helpful research
assistance and Matt Lawrence and Gary Orfield for feedback on earlier drafts. Paper presented at
the 2014 Civil Rights Project Research and Policy Briefing at the U.S. Capitol building in
Washington, DC. Views do not necessarily represent those of the Civil Rights Project or event
co-sponsors.
EDUCATION DESERTS
1
Executive Summary
When discussing educational accountability, policymakers rarely consider geography. Yet
geography can be destiny when opportunities richly available for some communities are rare (or
even nonexistent) in others. Many communities where families already face other obstacles have
but one public college option, which is typically an open-access college serving students who
typically have little financial means. These colleges would likely be hit hard by proposed federal
accountability plans, and their communities could lose their only public higher education
opportunity. They would become “education deserts.” In devising accountability policies, it is
important to understand the scope of this risk (which is considerable) and to consider ways to
avoid unintended outcomes.
If federal policymakers assume all students are mobile, that they live in communities with
several public alternatives, or that online education is an adequate alternative to place-based
education, then the findings reported here do not bear on educational equity or opportunity.
However, there is a growing body of evidence to refute these claims. The current study
contributes to this work by exploring the geography of opportunity. It raises important
implications for federal ratings policies, namely that it could disproportionately burden people
who live in minority communities or communities that already have low educational attainment
levels.
A long line of research shows how poor and minority students have unequal chances at pursuing
postsecondary education in the U.S. Factors such as a student’s high school resources, academic
preparations, peer groups, information about college, financial capacity, as well as social and
cultural capital play important roles in shaping whether/where students enroll in college.
However, educational inequalities are also drawn along the lines of geography, especially for
students who are place-bound and have no (or few) public colleges within commuting distance to
home or work. There is limited research on the geography of opportunity, but place certainly
matters since students (particularly low-income and minority students) make choices according
to what colleges are nearby. To understand how community resources and college opportunities
intersect, this study asks:

Research Question 1: To what extent does the number of colleges (public, non-profit,
and for-profit) vary according to a community’s racial/ethnic, socioeconomic, and
local labor market characteristics?

Research Question 2: What are the characteristics of the communities where the
fewest public higher education options exist (i.e., education deserts)?
To answer these questions, the study uses a unique dataset incorporating postsecondary
education data from the U.S. Department of Education with county-level data from various
federal agencies including the Census Bureau, Bureau of Economic Analysis, and Bureau of
Labor Statistics. Using mapping software and multivariate regression analysis, it locates local
commuting zones (i.e., clusters of counties that share a labor force and commuting patterns) that
have the fewest public options. Key findings include:
EDUCATION DESERTS
2

Nearly 1 in 5 public colleges in the U.S. are located in communities where they are
the only public option for prospective students.

Approximately 10% of the U.S. population lives in education deserts: communities
that have few (or no) public alternatives for college.

Communities with rising shares of Hispanic residents are most likely to be classified
as “education deserts.”

Communities that have low educational attainment levels are more likely to have
community colleges rather than four-year colleges.

After accounting for a community’s economic and social resources, those with larger
White populations still have the most educational alternatives.
These findings suggest the geography of educational opportunity is unequal across racial/ethnic
and socioeconomic groups. For place-bound students, some groups will have fewer choices when
considering where to enroll in college. In light of the proposed federal ratings system, these
results show how many communities have unequal opportunities for prospective students to
“shop around” for college. When designing a federal college ratings system, it is important to
consider whether a public college is the only institution in its community. Using the ratings
system to penalize colleges located in “education deserts” may have the unintended consequence
of pushing students into expensive private alternatives. This would disproportionately burden
Hispanic communities and communities that already have low educational attainment levels;
ultimately, a poorly designed rating system could exacerbate educational inequalities.
EDUCATION DESERTS
3
Differential impacts of college ratings: the case of education deserts
Nicholas W. Hillman
University of Wisconsin-Madison
How do students choose where to go to college? This simple question has no easy
answers. In fact, it has been the subject of research studies for several decades, where we are still
learning new insights into the way students make educational choices. One of the most
promising new areas of inquiry into this topic relates to the geography of opportunity, where
researchers study the role of community assets in shaping educational destinations. Despite
advances in technology and distance learning, place still matters when choosing a college since
only about one in ten undergraduates enroll in college exclusively online (U.S. Department of
Education, 2013a). In the public sector of higher education, where students can find the lowest
net price and where the majority (72%) of all college students enroll, a college’s location can
determine whether and where one attends (U.S. Department of Education, 2013b; 2014).
For many students interested in pursuing a public college education, choices are
constrained according to whatever college is located within commuting distance from home and
work (Turley, 2009). Considering that many students work full-time, have dependents to care
for, and commute to campus, the simple question about “how students choose college” becomes
very complicated (D. Kim & Rury, 2011; Perna, 2010; Ziskin, Fischer, Torres, Pellicciotti, &
Player-Sanders, 2014). Instead, a more relevant question becomes “do place-bound students have
alternatives college options near home or work?” Answering this question is made even more
complicated because it is entangled in the deeply rooted inequalities of our postsecondary
marketplace, where communities with the greatest economic advantages are likely to be the same
communities with the most robust educational infrastructure. Alternatively, communities
EDUCATION DESERTS
4
suffering from the worst forms of economic and racial inequality may have the fewest options.
This paper wrestles with these issues and brings data to bear on debates about the geography of
opportunity in postsecondary education.
This matters to ongoing federal higher education policy debates because by the 2015
academic year, the White House expects to roll out its Postsecondary Institution Ratings System
(PIRS). This rating system is built on the market model of higher education, where policymakers
assume students are mobile and they shop around for colleges. By providing information about
each college’s “value,” measured by such factors as the percent of students receiving Pell Grants,
average tuition and student loan debt, or graduation rates and loan repayment rates, students are
expected to make better educational choices. The Obama Administration hopes that by 2018 (and
pending Congressional approval) all federal student aid will be tied to PIRS ratings where
highly-rated colleges would receive larger Pell Grants and more affordable student loans and
poorly-rated colleges could lose access to federal aid altogether (The White House, 2013).
For the ratings system to be effective and equitable, it must be sensitive to the fact that
not all students have the luxury of “shopping around” for college. Interestingly, the research
literature paints an inconsistent picture of the extent of shopping around that occurs. Tracking
where students submit their college applications is one of the few ways researchers measure
college choices, and these measures are often limited to surveys that sample small segments of
the student population. For example, we learn the most about college choices through surveys
that only sample first-time, full-time students (Eagan, Lozano, Hurtado, & Case, 2013) or those
who are immediately out of high school and applying to selective colleges (Smith, 2014). These
surveys are only capturing the experiences of a fraction of the nation’s students, overlooking
community college students, part-time students, transfer students, and many other “non-
EDUCATION DESERTS
5
traditional” students who are increasingly pursuing postsecondary education at non-selective
institutions.
To overcome these limitations, and to include less-selective colleges in the analysis, a
recent study uses administrative data from Texas only to find that most students (2 in 3) apply to
only one state university (Cunha & Miller, 2014). Naturally, each research approach has its
limitations: national surveys select narrow samples and state case studies may not represent
national trends. Nevertheless, it is important to see that a student might apply to but never enroll
in a college, so it is often more important to see where students enroll instead of where they
apply. Table 1 uses the 2012 National Postsecondary Student Aid Survey to document the mean
and median distance between home and the place where undergraduate students enrolled.1 Here,
we see that community college students enroll approximately 30 miles from home and public
four-year college students are only about 80 miles from home. Regardless of “how many”
colleges a student sends their applications to, students tend to stay close to home when attending
public colleges and travel when attending private nonprofit schools.
[Insert Table 1 about here]
When studying college choices, it is important to consider how place matters in college
choices, especially in the public sector. In the community and regional college sectors, where
institutions have the distinct mission of serving their local/regional communities, many students
enroll part-time and select these institutions due to their proximity to home or work. These
colleges disproportionately serve low-income and minority students who have been traditionally
underserved in higher education. Where would a student go if their community college or
regional comprehensive college lost access to federal Pell Grants due to a poor rating? Do all
communities provide multiple educational institutions from which students can “shop around?”
EDUCATION DESERTS
6
Considering the degree of stratification and inequality that exists in higher education, it is
likely that communities with fewer economic resources and higher degrees of racial segregation
have fewer educational options than more privileged communities. A ratings system that is not
sensitive to the structural inequalities of place will do little to reverse them. Accordingly, this
paper extends the concept of “food deserts” into education, where (like food deserts) there are
geographic areas where access to affordable and well-resourced educational opportunities is
scarce (Block & Kouba, 2006; Walker, Keane, & Burke, 2010). Like food deserts, “education
deserts” are expected to be drawn along lines of race and class, where communities with large
minority populations and those with low educational attainment levels may have fewer public
options for pursuing higher education. The purpose of this study is to identify communities
across the U.S. (measured by commuting zones) where citizens have limited public higher
education options. It asks the following research questions:

Research Question 1: To what extent does the number of colleges (public, non-profit,
and for-profit) vary according to a community’s racial/ethnic, socioeconomic, and
local labor market characteristics?

Research Question 2: What are the characteristics of the communities where the
fewest public higher education options exist (i.e., education deserts)?
Conceptual Underpinnings
Geography of opportunity. Geography of opportunity is a key concept for understanding the
nature and extent of educational inequality in America (Miller, 2012; Tate, 2008). As we will
explore later, there are several communities that have limited (and in some cases no) public
options for pursuing postsecondary education. For people who live in these communities,
educational opportunities are constrained not by their own preferences and dispositions, but by
EDUCATION DESERTS
7
their community’s own educational infrastructure. Conceptualizing educational opportunity
through the lens of geography and place, we quickly see how a community’s structural
environment can shape and constrain educational destinations.
To illustrate the importance of place, we can draw lessons from the phenomenon of “food
deserts.” The 2008 Farm Bill defines a food desert as, “an area in the United States with limited
access to affordable and nutritious food, particularly such an area composed of predominantly
lower-income neighborhoods and communities” (P.L. 110-246, 2008). Food deserts do not occur
at random, there are systematic patterns where these communities are located and in how they
have evolved. For example, predominantly black (and low income) neighborhoods have
comparatively fewer supermarkets than white (and higher income) neighborhoods (Walker et al.,
2010; Whelan, Wrigley, Warm, & Cannings, 2002). Similarly, residents of food deserts have
higher probabilities of chronic illnesses related to their diets (e.g., diabetes, heart disease, and
obesity) because their communities have more unhealthy or cost-prohibitive food options (Block
& Kouba, 2006; Lamichhane et al., 2013; Widener, Farber, Neutens, & Horner, 2013). Structural
barriers prevent people living in these communities from having access to healthy and nutritious
food; instead, there is greater access to empty calorie foods (e.g., fast food, convenience stores,
etc.) that make it difficult for residents to maintain a healthy diet (Walker et al., 2010). This
creates an illusion of choice because even the most well informed and savvy consumer in a food
desert will be constrained from making healthy food decisions and will have greater odds of poor
health outcomes.
The structural inequality that exists in food deserts is but one example of how geography of
opportunity shapes life chances. Chetty et al. (2014) draw on geography of opportunity theories
to describe how the unequal resources available in communities contributes to social immobility.
EDUCATION DESERTS
8
Briggs and Wilson (2005) examine how affordable housing is vastly unequal depending on
where one lives, Kennedy (2004) finds unequal access to transportation that is largely defined by
geographic boundaries, and Smedley, et al. (2009) find that high quality and affordable health
care is clustered along spatial domains. In each example, it is clear that place matters and that
individual choices are affected by the quantity and quality of goods and services available near
where they work and live. This is relevant to higher education because not all students are
mobile; in fact, many are place-bound and choose college based on what is nearby and
convenient (Turley, 2009).
Education Deserts. In this paper, the term “education desert” describes communities where
there are limited public alternatives for pursuing a college education. It focuses on public
alternatives because public colleges enroll the majority (72%) of all undergraduate students in
the U.S., and they do so at a lowest net price than the private sector (U.S. Department of
Education, 2013b). When a local community has only one public college, it is impossible for a
prospective student to “choose” an alternative local public option. The majority of “education
deserts” consist of areas where a community college is the only public option available for
students. Someone who is place-bound and living in one of these communities would only have
access to sub-baccalaureate degree programs (e.g., certificates or associate’s degrees) and the
only alternatives would be more expensive private options.
Because of these limited options, the college choice-making process in an education desert
becomes highly constrained. This is why Lopez Turley (2009, p. 126) urges researchers to “stop
treating the college choice process as though it were independent of location and start situating
this process within the geographic context in which it occurs.” Traditional college choice models
focus on the process through which students make educational decisions: students first develop
EDUCATION DESERTS
9
predispositions about college at very early ages, then they search for colleges based on their
educational preferences and expectations, and they eventually choose where to enroll after
weighing the perceived and real costs/benefits (Cabrera & La Nasa, 2000; Hossler & Gallagher,
1987). Over the years, scholars have expanded this three-stage model to incorporate a wider
range of social, cultural, and political factors that shape the way students make educational
choices (Bell, Rowan-Kenyon, & Perna, 2009; J. Kim & Gasman, 2011; Kurlaender, 2006;
McDonough, 1997; Morgan, 2005; Perna, 2006). Among these factors, geography has emerged
as a significant factor in shaping students’ college choice process (Turley, 2009). By theorizing
the college choice process in terms of the geography of opportunity, we can expand the way we
conceptualize college choices and see that not all communities have equal opportunities to public
education.
Literature Review
In the geography of opportunity literature, there are three primary themes explaining how
distance relates to postsecondary destinations. First, the distance elasticity literature suggests that
proximity matters since the further (closer) a one lives to a college the less (more) likely they are
to enroll. Second, the spillover effects literature finds that simply having a college nearby is
strongly associated with enrolling and attaining a degree. Third, community ties are strong forces
that keep students close to home for college. While traditionally overlooked as a lens for
interpreting college choice, there is a growing recognition that place matters in shaping
educational opportunities. This review covers a wide range of interdisciplinary perspectives
including sociology, education, economics, and geography to gain a fuller picture of how and
why place matters.
EDUCATION DESERTS
10
Distance elasticity. Borrowing from a concept commonly used in economics, distance
elasticity examines how sensitive enrollment patterns are relative to proximity from college.
Studies have consistently identified a “deterrent effect” between distance and enrollment, where
the further one lives from a college the less likely they are to enroll (Alm & Winters, 2009). For
example, case studies from Georgia (Alm & Winters, 2009), Washington (Ullis & Knowles,
1975), and West Virginia (K. Ali, 2003) find a one-percent change in distance from a nearby
college reduces the likelihood of enrolling anywhere between 0.4 and 2.2 percent. Similarly,
Leppel’s (1993) study from Pennsylvania found students living within 10 miles of the university
had a 0472 probability of enrolling and their probability dropped to 0.388 (for students living 10
to 50 miles away) and 0.276 (for students more than 50 miles away).
More recently, Niu and Tienda’s (2008) statewide analysis of Texas colleges found similar
results, but took into account the possibility of non-linear relationships (i.e., “u” shaped)
relationship between distance and enrollment. They found the odds of enrolling drop as distance
rises, but after a certain point the odds of choosing a college rises the further they are from that
college. Long (2004) found a similar relationship in her study of 1972, 1982, and 1992 high
school graduates. What this “u-shape” pattern means is that distance plays a less significant role
for students who plan to attend out-of-state or pursue more prestigious colleges (Alm & Winters,
2009). This is consistent with what Hoxby (1997) argues when she examined enrollment patterns
among four-year colleges. She claims the market structure (for four-year colleges) is converging
in ways that makes distance a less relevant factor in college choices. While this may be true for
students pursuing elite universities and migrating out of state, Long (2004) finds that distance is
a significant predictor of college attendance. She finds that the relative impact of distance is
EDUCATION DESERTS
11
diminishing over time, but could be playing an increasingly important for “marginal students”
who come from lower-income families.
Spillover effects. The majority of distance studies focus on “how far” students are from
colleges, but we can also think of distance in terms of “how many” colleges are within
commuting distance of where students live. The number of colleges differs considerably
depending on the state and region where one lives. Eastern states have the most four-year
colleges in proximity to where students live, where only four percent of students living in the
East have no four-year colleges in proximity to where they live; this figure jumps to 19 and 20
percent in the West and South, respectively (Turley, 2009).
Simply having a college nearby can shape people’s enrollment decisions because the more
convenient a college is, the more likely people will participate (Turley, 2009). This relationship
especially true for community colleges, as Rouse (1995) finds living closer to a community
college is associated with being more likely to start college there. This is similar to what Card
(1993) found, where young men living near a four-year college appear to achieve more years of
education and that these effects are “concentrated among men with poorly-educated parents –
men who would otherwise stop schooling at relatively low levels” (p. 25).
These spillover effects likely occur out of convenience since having a college nearby “makes
the transition to college logistically, financially, and emotionally easier” (Turley, 2009, p. 138).
In a case study of the University of San Antonio, De Oliver (1998) finds that commuter students
pay two times more in transportation costs than commuters who live closer to the university.
Chung (2012) finds that students who live in communities with more for-profit colleges nearby
tend to be more likely to enroll in for-profit, likely because of the convenience of enrolling in
these institutions. She also notes growing up near a college lowers the transit costs, which in turn
EDUCATION DESERTS
12
can shape enrollment decisions. In addition to convenience, it is also possible that spillovers exist
because communities with colleges have a raised collective consciousness regarding the
importance of education, or because colleges are civically engaged in these communities. For
example, Griffith and Rothstein (2009) find that having a nearby college can provide positive
spillover effects by raising college-going expectation for youths in the community. Do (2004)
argues the presence of local colleges can lead to greater exposure to role models and greater
awareness of (and access to) information on the benefits of attending college.
Whether spillovers occur out of convenience or out of a raised collective consciousness of a
community, the literature consistently finds that communities with more college options have
greater educational participation and attainment rates. These results are noteworthy because they
suggest having public options can build an environment conducive to college attendance,
positively influencing local young adults and particularly those from lower-SES backgrounds.
Community ties. Proximity to a college can become a significant barrier to participation for
students who are “place-bound” and tied to their community for familial or cultural reasons.
Many community college students, for example, work full-time and seek educational
opportunities that are convenient to their work or commuting patterns (Somers et al., 2006). The
expansion of community colleges in the 1970s was in part a strategy to expand opportunities to
under-served (rural) areas. Before this expansion, Sewell (Sewell, 1963) found opportunities
were divided by geography, but even after the expansion of community colleges in the 1970s,
rural students are less likely to pursue higher education (Karen, 2002). Researchers suspect this
is due to familial expectations and the shaping of educational aspirations that take different forms
in rural versus non-rural communities (S. Ali & Saunders, 2008; James et al., 1999; Kirkpatrick
Johnson, Elder, & Stern, 2005; Roscigno, Tomaskovic-Devey, & Crowley, 2006).
EDUCATION DESERTS
13
Additionally, there are strong communal ties and commitments that ethnic minority groups particularly Hispanic youths – must navigate when deciding whether or where to attend college
(Baum & Flores, 2011). This becomes especially important in relation to whether a campus
provides a safe and nurturing place for students to maintain ties to their home communities while
also making new ones (Nora, 2004). Similarly, tight social networks can shape where students
decide to apply and ultimately enroll since familial commitments and obligations (e.g., care for
siblings and elders) can shape the choice preferences for Black and Hispanic youths
(McDonough, Antonio, & Trent, 1997; Perez & McDonough, 2008). As a result, Hispanic
youths tend to apply to fewer colleges than other racial/ethnic groups and are likely to remain
close to home when they attend college (Hurtado, Inkelas, Briggs, & Rhee, 1997).
Data
To measure the geographic area in which colleges are located, the study utilizes U.S.
Department of Agriculture commuting zone classifications. A commuting zone (CZ) is a cluster
of counties sharing similar labor markets and economic activity; there are 709 CZs accounting
for all 3,147 counties in the country. Tolbert and Sizer (1996) developed this classification
scheme using journey-to-work data that measures the county-to-county flow of commuters.
Unlike metro/micropolitan statistical areas, CZs are not defined by population size and they
incorporate rural counties. This serves as a helpful proxy for measuring the geographic areas in
which prospective college students live, work, and commute.2
To measure whether counties are rural or urban, the dataset uses USDA’s Rural-Urban
Continuum zones to classify counties according to their degree of urbanization and adjacency to
metropolitan areas (USDA, 2014), allowing us to measure the extent to which CZs are more
urban or rural. The dataset also includes Census Division codes to organize CZs according to the
EDUCATION DESERTS
14
following geographic areas: Northeast (New England and Middle Atlantic), Midwest (East and
West), South (Atlantic, East, and West), and West (Mountain and Pacific).
To measure the population of each CZ, it aggregates the bridged-race resident population
estimates from the U.S. Census Bureau and National Center for Health Statistics. This provides
population estimates for the following non-Hispanic/Latino groups: American Indian or Alaskan
Native, Asian or Pacific Islander, Non- Black or African American, and White and for
Hispanic/Latino groups. To measure the labor market conditions of each CZ, the dataset includes
unemployment data from Bureau of Labor Statistics (BLS) and educational attainment3 data
from the U.S. Census Bureau’s American Community Survey (ACS). Per capita personal income
and the share of the labor force employed in manufacturing are from the Bureau of Economic
Analysis (BEA). Table 2 in the following section provides a summary of these data sources.
Finally, the dataset includes U.S. Department of Education’s Integrated Postsecondary Education
Data System (IPEDS) records to identify the number of colleges located in each CZ and basic
enrollment characteristics and financial resources of those institutions.
While IPEDS provides the most comprehensive list of colleges, it is not without limitations.4
In some cases, a public college will report data to IPEDS only at the system-level, rather than
campus-level, making it difficult to decipher exactly where campuses (rather than system offices)
are located.5 Failing to account for these data reporting limitations would result in a
misclassification of where colleges are located, so I employed two strategies to minimize these
concerns. First, the analytical dataset includes any Title IV institution that reported to IPEDS
between the academic years 2010-11 and 2012-13 as a strategy to capture institutions that may
have reported at the campus-level one year, but at the system-level the next. The second
approach consisted of a state-by-state search for campuses that exist but are unreported in
EDUCATION DESERTS
15
IPEDS, such as branch campuses, completion centers, or satellite campuses located on
community college campuses.6 Together, the dataset yields 7,756 institutions enrolling a total
headcount of 25.2 million students.
Analytical technique
To estimate the relationship between colleges and their commuting zone characteristics,
the study employs two basic regression models. The first research question, which examines the
count of college per CZ, uses a Poisson regression. The second research question, which
identifies the CZ characteristics associated with being an “education desert,” employs a logistic
regression. In both models, commuting zones are the unit of analysis (n=708) and the same
vector of controls are included to account for observable differences between each CZ’s
racial/ethnic, socioeconomic, and local labor market conditions. One CZ is missing from Alaska,
which is why this analysis includes 708 rather than all 709. Since each CZ is nested within its
respective state, the error terms are clustered at the state level.7 In addition, Poisson results are
displayed as incident rate ratios (IRRs) and the logistic results are displayed as odds ratios (ORs)
for ease of interpretation.
Both regression models control for the same demographic, socioeconomic, and labor market
characteristics that are thought to be associated with the number of colleges in a commuting zone
(Research Question 1) and whether a commuting zone is located in an education desert
(Research Question 2). These controls include a dummy variable for each Census Division where
the CZ is located with the “West Midwest” division is the reference group (IA, KS, MN, MO,
NE, ND, and SD). To account for the urbanization of the CZ, it controls for the share of
population living in non-metropolitan (i.e., rural) counties. Importantly, it also measures the
racial/ethnic profile of each county by taking the log number of American Indians, African
EDUCATION DESERTS
16
Americans, Asians, Hispanics, and Whites living in each CZ. To measure the labor market, the
models control for the share of the labor force that is unemployed and the percentage of jobs that
are in manufacturing industries. They also control for percent of each CZ that has less than a
bachelor’s degree to account for variations in educational attainment levels. Finally, per capita
income is used to account for some of the financial resources available in each CZ. Table 2
provides the data sources for each variable included in the analysis.
[Insert Table 2 about here]
Robustness checks. I conducted three robustness checks to test how sensitive the final models
are to alternative specifications. One could argue the Poisson distribution is unnecessary since it
is over-disbursed (i.e., the variance is much larger than the mean), so I first ran the same models
using a Negative Binomial, Zero-Inflated Poisson, and Tobit regression with zero as the lower
limit. In each case, results were not substantially different from what is found in the Poisson
models, so I kept the Poisson model for ease of interpretation. Next, I ran the models using
“colleges per 100,000 population” as an alternative outcome variable (instead of “count of
colleges”) and found similar patterns as the count model. I prefer using the count model since it
is consistent with the underlying policy logic that students “shop around” among alternative
institutions: the raw number of institutions in a CZ matters when deciding where to enroll.8
Finally, I restricted the models to a subsample of CZs that are non-metropolitan with populations
less than 250,000 (n=489) due to concerns about over-dispersion. This eliminates large
commuting zones (e.g., Chicago, Los Angeles, etc.) reducing the variance to be nearly identical
with the mean. Nevertheless, results were similar under this alternative model suggesting the
findings are not driven by larger metropolitan areas.
EDUCATION DESERTS
17
These robustness checks ensure the results are consistent across various designs and
assumptions that I use in the final analysis. Using alternative measurements for the outcome
variable, taking a close look at sub-samples, and employing alternative regression techniques
allow me to rule out the plausibility the results would change if different techniques were
employed.
Key Findings
Half of all commuting zones are classified as “education deserts,” and approximately 34.2
million people (11% of the U.S. population) live in these areas. Education deserts contain 353
public colleges and enroll 1.7 million undergraduate students; most of these colleges (n=309) are
community colleges. See Table 3 for more details. The following discussion will highlight
additional findings, including:

Communities with rising shares of Hispanic residents are most likely to be classified
as “education deserts.”

Communities that have low educational attainment levels are more likely to have
community colleges rather than four-year colleges.

After accounting for a community’s economic and social resources, those with larger
White populations still have the most educational alternatives.
[Insert Table 3 about here]
Table 4 below shows how education desert CZs differ from other CZs on the variables
included in the analysis. Education deserts differ from other CZs in their urbanicity, where
approximately 88% of education deserts’ populations live in “non-metropolitan” counties while
only 38% of non-deserts live in these areas. Education deserts tend to also have smaller
populations than non-deserts, but on other measures (income, unemployment rates, and
manufacturing’s share of the labor force) these differences are not as stark when simply looking
at the summary statistics displayed in Table 4. These statistics help display broad characteristics,
EDUCATION DESERTS
18
but they do not answer the central research questions concerning the number of colleges per CZ
or the odds of classifying a CZ as an education desert. For those answers, we turn to multiple
regression analysis, where we find the following patterns.
[Insert Table 4 about here]
Research Question #1. The first research question asks how a CZ’s racial/ethnic,
socioeconomic, and local labor market characteristics are associated with having public colleges
within commuting distance. The number of public options varies along racial/ethnic lines in
systematic and unequal ways; namely, CZs with growing Hispanic populations are associated
with having fewer public four-year and public two-year colleges. A 1% increase in Hispanic
population is associated with decreasing the rate of public four-year college by a factor of 0.837
and 0.957 (or 83.7 and 95.7 percent). Since these rates are below 1, they suggests a negative
relationship were Hispanic communities are associated with having fewer public colleges. This
relationship is statistically significant for public four-year colleges, suggesting there are
systematic inequalities with regard to the number of public options Hispanic communities have,
relative to other groups. While Hispanic communities have the fewest public options, Table 5
below shows that each 1% increase in population increases the number of for-profit college by a
rate of 15.4%. These percentages can be unintuitive and difficult to interpret, so Figure 1
displays how discrete changes in population are associated with the number of colleges in a CZ
(i.e., marginal effects).
[Insert Table 5 about here]
[Insert Figure 1 about here]
In addition to these patterns in Hispanic communities, CZs with growing Black
populations tend to have more private colleges: a 1% increase in Black population is associated
EDUCATION DESERTS
19
increasing the number of private for-profit and non-profit colleges by a rate of 18.6% and 16.8%,
respectively. Many of these institutions may be Historically Black College and Universities,
which would require further research to confirm. However, larger Hispanic communities tend to
have fewer public options (and more for-profit options), larger Black communities tend to have
both more private colleges and more community colleges (but not public four-year). These
patterns are unique to these two racial/ethnic minority groups, as CZs with large Asian
populations have a large number of public four-year colleges and American Indian CZs tend to
have more public options and fewer (but statistically insignificant) private options. Many of the
public colleges near communities with large share of American Indian populations are likely to
be Tribal Colleges, but this needs to be confirmed. Unlike Blacks and American Indians,
however, there are very few colleges founded with the distinct mission of serving Hispanic
students. The only group that consistently has more colleges per CZ (public or private) is White:
a 1% increase in White population, after controlling for other factors, is associated with
increasing the rate of colleges operating in the CZ by 29.0% to 153%.
In addition to these racial/ethnic differences, we observe inequalities related to
educational attainment rates and the number of public colleges per CZ. Communities where a
larger share of the population has less than a bachelor’s degree (i.e., low educational attainment
levels) tend to have larger numbers of community colleges and more for-profit colleges from
which to choose to enroll. These same communities have fewer (though not statistically
significant) public four-year colleges. These implications are discussed later, but it is important
to note the structural inequalities that perpetuate in CZs with low educational attainment levels
(i.e., they have greater access to sub-baccalaureate, rather than baccalaureate, degrees).
[Insert Figure 2 about here]
EDUCATION DESERTS
20
Research Question #2: Now we shift attention to the communities where public options
are the most constrained: education deserts. This employs two slightly different definitions of
education deserts, displayed in Table 5, to test the robustness of the findings. The first definition
(column V) accounts for CZs where community colleges are the only public option (i.e., no
public four-year colleges) or where there are no public colleges whatsoever. The second
definition (column VI) accounts for the same CZs as in column V, but adds any CZ where public
four-year colleges are the only option and there are no community colleges in the CZ. There are
only 44 additional CZs in this latter definition, but they are included because these represent
communities with no public alternatives. Similar to the previous Research Question, I find
unequal patterns that cut along lines of race/ethnicity, socioeconomics, and local labor market
conditions. Again, even after controlling for a number of important variables, education deserts
are drawn along lines of race/ethnicity, socioeconomic, and labor market conditions. Still, as the
share of a CZ’s Hispanic population rises, so too does its odds of being an education desert. For
each additional 1% increase in Hispanic population, the odds of being a desert increases by
approximately 1.4 times.9
Consistent with findings from the count models (Research Question #1), the odds of
living in an education desert decline as the number of American Indians and Asians rises. This is
not surprising since Asians are very strongly concentrated in a few large metropolitan areas that
likely have more colleges nearby and many Tribal colleges are public. Interestingly, a CZ’s
White population is no longer a systematic factor associated with being an education desert:
changes in White population are not associated with having constrained public alternatives. As
the number of Blacks increase for a CZ, the odds of being an education desert declines. These
patterns are discussed more below, but this provides additional evidence that public college
EDUCATION DESERTS
21
options are not equal across lines of race/ethnicity. Also consistent with the count models, CZs
with lower educational attainment levels are more likely to be located in education deserts. As
the share of a CZ with low educational attainment levels rises by 1%, their odds of being a desert
increase by 1.1 times. Relatedly, CZs with high shares of the labor force working in
manufacturing industries are also more likely to be in an education desert and have the most
constrained public options. A 1% increase in manufacturing’s share of the labor force is
associated with having 1.1greater odds of being an education desert, as shown in Figure 3.
[Insert Figure 3 about here]
Summary of findings. When using commuting zones to represent local communities
where people live, work, and pursue formal education, this study finds inequalities in terms of
the number of public colleges and universities available to their residents. Communities with
rising shares of Hispanics tend to have fewer public options (and more private options) while
Whiter communities tend to have more alternatives (public or private). Communities with
growing Hispanic populations not only have fewer options, but they are the most likely to be
classified as “education deserts.” These deserts are characterized by having the least public
choice, they have no public college, or a community college is the only options available nearby.
When expanding this definition to include communities that have no community college but
have a public four-year college as the only option, these results do not change.
Communities with larger share of Asian American populations tend to have more public
four-year college options, while communities with larger African American populations tend to
have more community colleges. White and American Indians are the only two groups where
population increases are associated with larger numbers of public four-year and public two-year
colleges nearby. In addition to these racial/ethnic trends, a community’s educational attainment
EDUCATION DESERTS
22
rate is systematically related to the number of nearby colleges: places with low attainment rates
are characterized as having more public two-year and for-profit colleges that often do not offer
bachelor’s degrees. Similarly, when the share of the labor force employed in manufacturing rises,
the odds of being an education desert also rises.
These results are not entirely surprising and they are inter-related in complex ways. The
purpose of this study is to offer an exploration into the geographical patterns of opportunity;
namely, to identify how the number of alternative public options varies according to several
community characteristics. Communities with growing Hispanic populations, lower educational
attainment rates, and larger share of the workforce employed in manufacturing industries are the
least likely to have public college options. In fact, they are likely to have the most constrained
public alternatives and be “education deserts.”
Policy Implications
Geography of opportunity is largely missing from college access and college choice research.
Not surprisingly, the importance of place has not played a significant role in ongoing federal
policy debates concerning the potential consequences of rationing Title IV financial aid to a
college’s rating. Instead, the underlying logic of the federal policy framework is that college
choices are made independent of place and that a federal rating system “can empower students
and families to make good choices” about where to attend college (Obama, 2013). This is a
market-based logic that assumes students are either mobile or that they live in communities
where there are several public options from which to choose. This study shows the limitations of
this line of thinking, where we see that geography and place matters in the college choice
process.
EDUCATION DESERTS
23
These results contribute to the limited but growing body of research on the geography of
opportunity in postsecondary education, which finds that lower-income and minority students
tend to stay closer to home than middle and upper-income or White students (e.g., Turley, 2009).
There are some studies showing that “high achieving” low-income students are in high demand
among elite college, but recruiting efforts tend to concentrate on only a few large metropolitan
areas rather than places like those described in this study (C. Hoxby & Avery, 2013; C. M.
Hoxby & Turner, 2013). Contrary to the prevailing policy logic, many students stay close to
home for college. By analyzing commuting zones across the U.S., this study finds communities
with growing Hispanic populations, lower educational attainment levels, and higher shares of
manufacturing jobs are the same communities with the fewest public options. Based on the
geography of opportunity literature, it is plausible that most people living in these communities
are place-bound; therefore, they will have the fewest public alternatives.
In the higher education marketplace, access to public options matters because they are often
the lowest-priced option where students tend to borrow at lower rates (and accumulate less debt)
than students in private non-profit or for-profit sectors (U.S. Department of Education, 2014).
However, this study shows there are several communities across the country with either no
public colleges or only community colleges as the sole public option. Community college are the
workhorse of public higher education since they serve the majority of public sector students and
do so with far fewer resources than four-year colleges, so the implications should not discredit
their work. Rather, this study illustrates that not all students are mobile and may be unwilling or
unable to travel long distances to attend an alternative colleges; believing so is inconsistent with
the research on student mobility. It is also inconsistent with the social and economic factors
EDUCATION DESERTS
24
researchers have found to matter with regard to “why” people prefer to stay close to home when
pursuing a college education.
If federal policymakers assume all students are mobile, that they live in communities with
several public alternatives, or that online education is an adequate alternative to place-based
education, then the findings reported here do not bear on educational equity or opportunity.
However, there is a growing body of evidence to refute these claims. The current study
contributes to this body of evidence by exploring the geography of opportunity. It raises
important implications for federal ratings policies, namely that it could disproportionately burden
people who live in minority communities or communities that already have low educational
attainment levels.
An example illustrates the implications of these findings. The cities of Uvalde and Eagle
Pass, Texas, are in the same four-county commuting zone located between San Antonio and the
border town of Piedras Negras, Mexico. The Uvalde-Eagle Pass commuting zone has
approximately 100,000 residents and most (87%) are Hispanic. The commuting zone’s
unemployment rate is 13% and approximately 87% of its residents have less than a bachelor’s
degree. This community is serviced by Southwest Texas Junior College (STJC), which enrolls
approximately 6,000 undergraduate students. According to the College Navigator, the federal
government’s consumer information tool that the federal ratings system will likely be combined
with, shows the only other college within a 100 mile radius is Southwest School of Business and
Technical Careers (a small for-profit college specializing in cosmetology certificates).10 Sul Ross
State College, a public four-year institution, offers upper-level courses at satellite locations in the
commuting zone; however, these are not full academic degree programs nor are they reported in
IPEDS or on the College Navigator page.
EDUCATION DESERTS
25
One of the accountability measures proposed for the rating system is a college’s Cohort
Default Rate, the percent of borrowers who default on their federal loans within three years of
entering repayment. Another possible measure is the college’s net price, which is the cost of
attendance (tuition, books, supplies, room, and board) minus all grant aid. At STJC, the CDR is
high (23.7%), its graduation rate low (21%) and its net price is $6,805 is moderate. Depending
on what is measured in the ratings system, this college could easily receive a poor federal rating.
If this occurs, and if we assume ratings will affect college choices, then prospective students may
see STJC as a low quality option due to its poor rating. By this logic, their only other “choice”
would be to attend a for-profit college (with a net price of $10,811) or forego college altogether.
This is not unique to the Uvalde-Eagle Pass community; this can be multiplied hundreds of times
over all across the U.S. One in ten citizens live in an “education desert” not too dissimilar from
Uvalde-Eagle Pass, Texas.
This brief example illustrates how a poorly designed ratings system could disproportionately
burden communities of color, working class communities, and those that already have low
educational attainment levels. If federal officials are going to pursue a college rating system, then
they should account for geographic factors such as whether the college is the only public option
within commuting distance. It could also identify communities where there are
disproportionately large number of private colleges relative to public colleges to locate areas
where prospective students will likely “choose” more expensive alternatives if their public
institution receives a poor rating. Similarly, communities that have a long history of low
educational attainment levels could receive waivers from the ratings system since many of these
communities may not have the capacity to deliver upper-level courses that lead to a bachelor’s
degree.
EDUCATION DESERTS
26
That federal policymakers have not fully engaged in debates about the equity or fairness
implications, particularly those that deal with geographic inequality, of a ratings system is
concerning. Instead, federal officials may operate with the belief that a ratings system is a tool
for improving college “choices” and to help students seek out “better” institutions. In many
communities, these can be false alternatives when working class and people of color tend to have
the fewest public alternatives. A rating system that does not account for the local context of an
institution will disproportionately burden minority and working class communities that have a
long history of educational inequality.
EDUCATION DESERTS
27
Table 1:
Distance from student's home to college (in miles)
Public 2-year
Public 4-year
Private nonprofit 4-year
Total
Mean
31
82
258
Median
8
18
46
107
13
EDUCATION DESERTS
28
Table 2:
Variables and data sources constructed for analysis
Variable
Commuting zone
Census regions
Rural continuum
Population
Unemployment
Educational attainment
Income
Manufacturing labor force
Number of colleges
Source
U.S. Department of Agriculture
Census Bureau
U.S. Department of Agriculture
Census Bureau and National Center for Health Statistics
Bureau of Labor Statistics – Local Area Unemployment
Census Bureau – American Community Survey
Bureau of Economic Analysis
Bureau of Economic Analysis
Integrated Postsecondary Education Data System
EDUCATION DESERTS
29
Table 3:
Population, enrollment, and number of public colleges located in education deserts
Total
Education Desert
Share of total
Number of
CZs
708
405
(57%)
Population
Public
Number of
(mil.)
enrollment (mil.) public colleges
313.8
18.7
2,076
34.2
1.7
353
(11%)
(9%)
(17%)
EDUCATION DESERTS
30
Table 4:
Descriptive statistics used in analysis
Non-Desert
Mean
S.D.
Northeast (New England)
4%
19%
Northeast (Middle Atlantic)
7%
26%
Midwest (East North Central)
14%
35%
Midwest (West North Central)^
12%
33%
South (South Atlantic)
21%
41%
South (East South Central)
9%
28%
South (West South Central)
16%
36%
West (Mountain)
10%
30%
West (Pacific)
8%
27%
Non-metro share of population
38%
39%
Asian population (logged)
9.0
1.8
American Indian population (logged)
7.8
1.4
Black population (logged)
10.3
1.9
Hispanic population (logged)
10.2
1.8
White population (logged)
12.6
1.1
Unemployment rate
9.1%
2.6%
Share of population with less than BA
75.7%
7.2%
Per capita income (in $1000s)
38.1
7.8
Manufacturing share of labor force
8.2%
4.5%
Note: ^ denotes reference group in regression analysis
Education Desert
Mean
S.D.
1%
9%
1%
9%
8%
27%
30%
46%
9%
28%
11%
31%
15%
36%
17%
38%
8%
28%
88%
28%
5.8
1.6
6.0
1.6
6.9
2.2
7.7
1.7
10.4
1.3
8.5%
3.2%
81.5%
6.1%
38.5
10.1
8.3%
7.3%
EDUCATION DESERTS
31
Table 5:
Poisson and logistic regression estimates for number of colleges and education deserts
Northeast (New England)
Northeast (Middle Atlantic)
Midwest (East North Central)
South (South Atlantic)
South (East South Central)
South (West South Central)
West (Mountain)
West (Pacific)
Non-metro share of population
Asian population (logged)
American Indian population
(logged)
Black population (logged)
Hispanic population (logged)
White population (logged)
Unemployment rate
Share of population with less
than BA
Per capita income (in $1000s)
Manufacturing share of labor
force
Observations
BIC
AIC
Wald χ2
RQ1: number of colleges per CZ
(Incident Rate Ratios)
Public
Public
Non-profit For-profit
4-yr
2-yr
(I)
(II)
(III)
(IV)
2.132*** 1.710**
1.239
0.705***
(0.392)
(0.456)
(0.218)
(0.076)
1.987***
1.006
1.116
0.546***
(0.385)
(0.179)
(0.179)
(0.085)
1.192
0.833
0.772**
0.835
(0.219)
(0.182)
(0.092)
(0.112)
1.222
0.858
0.546***
0.629***
(0.200)
(0.190)
(0.075)
(0.067)
1.069
0.673*
0.688**
0.821
(0.204)
(0.150)
(0.112)
(0.126)
1.437***
0.911
0.378***
0.862
(0.191)
(0.231)
(0.059)
(0.175)
1.187
1.101
0.332***
1.184
(0.276)
(0.223)
(0.071)
(0.241)
1.097
1.570*
0.951
0.873
(0.277)
(0.392)
(0.228)
(0.159)
0.998
1.007***
1.003*
0.992***
(0.002)
(0.002)
(0.002)
(0.002)
1.321***
0.982
0.980
0.969
(0.108)
(0.087)
(0.085)
(0.070)
1.088**
1.124*
0.944
0.933
(0.047)
(0.067)
(0.051)
(0.059)
1.100
1.254***
1.168***
1.186***
(0.068)
(0.061)
(0.063)
(0.078)
0.837***
0.957
1.088
1.154*
(0.050)
(0.051)
(0.065)
(0.095)
1.290*
1.870***
2.499***
2.536***
(0.182)
(0.194)
(0.292)
(0.307)
0.970
0.962*
0.938**
0.966**
(0.023)
(0.020)
(0.030)
(0.016)
0.994
1.042***
0.990
1.014**
(0.011)
(0.010)
(0.008)
(0.007)
0.829
1.128
0.661***
0.771
(0.187)
(0.184)
(0.053)
(0.224)
0.983
0.992
1.008
0.984*
(0.011)
(0.009)
(0.012)
(0.009)
708
708
708
708
1398.4
2002.7
1934.6
2176.7
1311.7
1916.0
1847.9
2090.0
2183.0
4342.2
217.8
194.8
Pseudo-R
0.422
0.469
0.746
0.844
Note: Clustered standard errors in parentheses, * p<0.1, ** p<0.05, *** p<0.01
0.489
0.545
2
2061.5
5098.7
RQ2: education deserts
(Odds Ratios)
Pub. 2-yr
Pub. 2-yr or
only
Pub. 4-yr
(V)
(VI)
0.364
0.224
(0.421)
(0.236)
0.121
0.273
(0.182)
(0.254)
0.877
0.951
(0.354)
(0.351)
1.070
1.186
(0.620)
(0.672)
1.266
2.229**
(0.538)
(0.910)
0.580
0.794
(0.231)
(0.306)
1.250
1.335
(0.601)
(0.717)
3.212*
3.234
(1.977)
(2.312)
1.005
1.001
(0.004)
(0.005)
0.393***
0.364***
(0.113)
(0.112)
0.819**
0.779***
(0.080)
(0.072)
0.779*
0.724**
(0.117)
(0.114)
1.450**
1.468*
(0.237)
(0.320)
1.041
0.670
(0.290)
(0.264)
1.051
0.997
(0.079)
(0.073)
1.108***
1.015
(0.036)
(0.037)
5.419
1.491
(9.291)
(1.472)
1.066***
1.066**
(0.024)
(0.028)
708
708
626.0
564.4
539.3
477.7
EDUCATION DESERTS
32
Figure 1:
Estimated number of colleges per CZ (by sector and race/ethnicity)
Public 4-yr
Public 2-yr
Non-profit
Hispanic
Black
White
Population (logged)
For-profit
EDUCATION DESERTS
33
Figure 2:
Estimated number of colleges per CZ (by educational attainment rates)
Public 4-yr
Public 2-yr
Non-profit
For-profit
Percent of CZ with bachelor’s degree or less
EDUCATION DESERTS
34
Figure 3:
Predicted probability of being classified an “education desert” (by race/ethnicity))
Public 2-yr only
Public 2-yr or Public 4-yr
Hispanic
Black
White
Population (logged)
EDUCATION DESERTS
35
References
Ali, K. (2003). Analysis of Enrollment: A Spatial-interaction Model. The Journal of Economics,
29(2), 67–86.
Ali, S., & Saunders, J. (2008). The career aspirations of rural Appalachian high school students.
Journal of Career Assessment.
Alm, J., & Winters, J. V. (2009). Distance and intrastate college student migration. Economics of
Education Review, 28(6), 728–738.
Autor, D. H., & Dorn, D. (2013). The Growth of Low-Skill Service Jobs and the Polarization of
the US Labor Market. American Economic Review, 103(5), 1553–1597.
doi:10.1257/aer.103.5.1553
Baum, S., & Flores, S. M. (2011). Higher education and children in immigrant families. The
Future of Children, 21(1), 171–193.
Bell, A. D., Rowan-Kenyon, H. T., & Perna, L. W. (2009). College knowledge of 9th and 11th
grade students: Variation by school and state context. The Journal of Higher Education,
80(6), 663–685.
Block, D., & Kouba, J. (2006). A comparison of the availability and affordability of a market
basket in two communities in the Chicago area. Public Health Nutrition, 9(07), 837–845.
Briggs, X. N. D. S., & Wilson, W. J. (2005). The geography of opportunity: Race and housing
choice in metropolitan America. Brookings Institution Press.
Cabrera, A. F., & La Nasa, S. M. (2000). Understanding the College-Choice Process. New
Directions for Institutional Research, 2000(107), 5–22.
Card, D. (1993). Using geographic variation in college proximity to estimate the return to
schooling (No. 4483). Cambridge, MA: National Bureau of Economic Research.
Retrieved from http://www.nber.org/papers/w4483
Chetty, R., Hendren, N., Kline, P., & Saez, E. (2014). Where is the Land of Opportunity? The
Geography of Intergenerational Mobility in the United States. Cambridge, MA: National
Bureau of Economic Research. Retrieved from http://www.nber.org/papers/w19843
Chung, A. S. (2012). Choice of for-profit college. Economics of Education Review, 31(6), 1084–
1101.
Cohen, A. M., & Brawer, F. B. (2008). The American Community College (5th ed.). San
Francisco: Jossey-Bass.
Cunha, J. M., & Miller, T. (2014). Measuring value-added in higher education: Possibilities and
limitations in the use of administrative data. Economics of Education Review, 42, 64–77.
doi:10.1016/j.econedurev.2014.06.001
De Oliver, M. (1998). Geography, race, and class: A case study of the role of geography at an
urban public university. American Journal of Education, 273–301.
Do, C. (2004). The effects of local colleges on the quality of college attended. Economics of
Education Review, 23(3), 249–257.
EDUCATION DESERTS
36
Eagan, K., Lozano, J. B., Hurtado, S., & Case, M. H. (2013). The American Freshman: National
Norms Fall 2013. Cooperative Institutional Research Program at the Higher Education
Research Institute at UCLA. Retrieved from
http://heri.ucla.edu/monographs/TheAmericanFreshman2013-Expanded.pdf
Griffith, A. L., & Rothstein, D. S. (2009). Can’t get there from here: The decision to apply to a
selective college. Economics of Education Review, 28(5), 620–628.
Hossler, D., & Gallagher, K. S. (1987). Studying Student College Choice: A Three-Phase Model
and the Implications for Policymakers. College and University, 62(3), 207–21.
Hoxby, C., & Avery, C. (2013). The Missing “One-Offs”: The Hidden Supply of HighAchieving, Low-Income Students. Washington, D.C.: Brookings Institution. Retrieved
from
http://www.brookings.edu/~/media/Projects/BPEA/Spring%202013/2013a_hoxby.pdf
Hoxby, C. M. (1997). How the Changing Market Structure of U.S. Higher Education Explains
College Tuition (Working Paper No. 6323). Cambridge, MA: National Bureau of
Economic Research. Retrieved from http://www.nber.org/papers/w6323
Hoxby, C. M., & Turner, S. (2013). Informing students about their college options: a proposal
for broadening the Expanding College Opportunities Project (No. 2013-03). Washington,
DC: Brookings Institution. Retrieved from
http://www.brookings.edu/~/media/research/files/papers/2013/06/26-expanding-collegeopportunity-hoxby-turner/thp_hoxbyturner_final.pdf
Hurtado, S., Inkelas, K. K., Briggs, C., & Rhee, B.-S. (1997). Differences in college access and
choice among racial/ethnic groups: Identifying continuing barriers. Research in Higher
Education, 38(1), 43–75.
James, R., Wyn, J., Baldwin, G., Hepworth, G., McInnis, C., & Stephanou, A. (1999). Rural and
Isolated School Students and Their Higher Education Choices: A Re-Examination of
Student Location, Socioeconomic Background, and Educational Advantage and
Disadvantage. ERIC. Retrieved from http://eric.ed.gov/?id=ED449924
Jaquette, O., & Parra, E. E. (2014). Using IPEDS for panel analyses: Core concepts, data
challenges, and empirical applications. In M. Paulsen (Ed.), Higher education: Handbook
of theory and research (Vol. 29, pp. 467–533). Dordrecht, The Netherlands: Springer.
Retrieved from http://link.springer.com/chapter/10.1007/978-94-017-8005-6_11
Karen, D. (2002). Changes in access to higher education in the United States: 1980-1992.
Sociology of Education, 191–210.
Kennedy, L. G. (2004). Transport and Environmental Justice. Retrieved from
http://trid.trb.org/view.aspx?id=790035
Kim, D., & Rury, J. L. (2011). The Rise of the Commuter Student: Changing Patterns of College
Attendance for Students Living at Home in the United States, 1960-1980. Teachers
College Record, 113(5).
Kim, J., & Gasman, M. (2011). In Search of a “Good” School: First and Second Generation
Asian American Students Describe their College Choice Process. Journal of College
EDUCATION DESERTS
37
Student Development. Retrieved from
http://works.bepress.com/cgi/viewcontent.cgi?article=1055&context=marybeth_gasman
Kirkpatrick Johnson, M., Elder, G. H., & Stern, M. (2005). Attachments to family and
community and the young adult transition of rural youth. Journal of Research on
Adolescence, 15(1), 99–125.
Kurlaender, M. (2006). Choosing community college: Factors affecting Latino college choice.
New Directions for Community Colleges, 2006(133), 7–16.
Lamichhane, A. P., Warren, J., Puett, R., Porter, D. E., Bottai, M., Mayer-Davis, E. J., & Liese,
A. D. (2013). Spatial patterning of supermarkets and fast food outlets with respect to
neighborhood characteristics. Health & Place, 23(Vol. 23 – 2013), 157 – 164.
doi:10.1016/j.healthplace.2013.07.002
Leppel, K. (1993). Logit estimation of a gravity model of the college enrollment decision.
Research in Higher Education, 34(3), 387–398. doi:10.1007/BF00991851
Long, B. (2004). How have college decisions changed over time? An application of the
conditional logistic choice model. Journal of Econometrics, 121(1), 271–296.
McDonough, P. M. (1997). Choosing Colleges: How Social Class and Schools Structure
Opportunity. Albany, NY: Suny Press.
McDonough, P. M., Antonio, A. L., & Trent, J. W. (1997). Black Students, Black Colleges: An
African American College Choice Model. Journal for a Just and Caring Education, 3(1),
9–36.
Miller, P. M. (2012). Mapping Educational Opportunity Zones: A Geospatial Analysis of
Neighborhood Block Groups. The Urban Review, 44(2), 189–218. doi:10.1007/s11256011-0189-7
Morgan, S. L. (2005). On the Edge of Commitment: Educational Attainment and Race in the
United States. Stanford, CA: Stanford University Press.
Niu, S. X., & Tienda, M. (2008). Choosing colleges: Identifying and modeling choice sets.
Social Science Research, 37(2), 416–433.
Nora, A. (2004). The role of habitus and cultural capital in choosing a college, transitioning from
high school to higher education, and persisting in college among minority and
nonminority students. Journal of Hispanic Higher Education, 3(2), 180–208.
Obama, B. (2013, August 22). Remarks by the President on College Affordability. Retrieved
May 28, 2014, from http://www.whitehouse.gov/the-press-office/2013/08/22/remarkspresident-college-affordability-buffalo-ny
P.L. 110-246. (2008). 2008 Farm Bill. Retrieved August 10, 2014, from
http://www.ag.senate.gov/issues/2008-farm-bill
Perez, P. A., & McDonough, P. M. (2008). Understanding Latina and Latino college choice: A
social capital and chain migration analysis. Journal of Hispanic Higher Education.
Retrieved from
http://jhh.sagepub.com/content/early/2008/05/19/1538192708317620.short
EDUCATION DESERTS
38
Perna, L. (2006). Studying college access and choice: a proposed conceptual model. In Higher
Education: Handbook of Theory and Research (Vol. XXI, pp. 99–157). Dordrecht, The
Netherlands: Springer. Retrieved from http://link.springer.com/chapter/10.1007/1-40204512-3_3
Perna, L. (2010). Understanding the working college student. Academe, 96(4), 30–32.
Roscigno, V. J., Tomaskovic-Devey, D., & Crowley, M. (2006). Education and the inequalities
of place. Social Forces, 84(4), 2121–2145.
Rouse, C. E. (1995). Democratization or diversion? The effect of community colleges on
educational attainment. Journal of Business & Economic Statistics, 13(2), 217–224.
Sewell, W. H. (1963). The educational and occupational perspectives of rural youth.
Washington, DC: National Committee for Children and Youth. Retrieved from
http://eric.ed.gov/?id=ED019169
Smedley, B. D., Stith, A. Y., & Nelson, A. R. (2009). Unequal treatment: confronting racial and
ethnic disparities in health care (with CD). National Academies Press.
Smith, J. (2014). The effect of college applications on enrollment. The BE Journal of Economic
Analysis & Policy, 14(1), 151–188.
Somers, P., Haines, K., Keene, B., Bauer, J., Pfeiffer, M., McCluskey, J., … Sparks, B. (2006).
Towards a theory of choice for community college students. Community College Journal
of Research and Practice, 30(1), 53–67.
Tate, W. F. (2008). “Geography of opportunity”: Poverty, place, and educational outcomes.
Educational Researcher, 37(7), 397–411.
The White House. (2013, August 22). Fact Sheet on the President’s Plan to Make College More
Affordable: A Better Bargain for the Middle Class. Retrieved February 21, 2014, from
http://www.whitehouse.gov/the-press-office/2013/08/22/fact-sheet-president-s-planmake-college-more-affordable-better-bargainThelin, J. R. (2011). A history of American higher education (2nd ed.). Baltimore, MD: Johns
Hopkins University Press. Retrieved from
Tolbert, C. M., & Sizer, M. (1996). US commuting zones and labor market areas: A 1990
update. Washington, DC: US Department of Agriculture. Retrieved from
http://trid.trb.org/view.aspx?id=471923
Turley, R. N. L. (2009). College proximity: Mapping access to opportunity. Sociology of
Education, 82(2), 126–146.
U.S. Department of Education. (2013a). Digest of Education Statistics. Retrieved August 12,
2014, from http://nces.ed.gov/programs/digest/d13/tables/dt13_311.15.asp
U.S. Department of Education. (2013b). Total 12-month enrollment in degree-granting
postsecondary institutions. Retrieved April 29, 2014, from
http://nces.ed.gov/programs/digest/d13/tables/dt13_308.10.asp
U.S. Department of Education. (2014). Out-of-Pocket Net Price for College. Retrieved August
13, 2014, from http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2014902
EDUCATION DESERTS
39
Ullis, J. J., & Knowles, P. L. (1975). A study of the intrastate migration of Washington college
freshmen: A further test of the gravity model. The Annals of Regional Science, 9(1), 112–
121. doi:10.1007/BF01284992
USDA. (2014). Rural-Urban Continuum Codes. Retrieved August 1, 2014, from
http://www.ers.usda.gov/data-products/rural-urban-continuumcodes/documentation.aspx#.U9uaNGNZbZc
Walker, R. E., Keane, C. R., & Burke, J. G. (2010). Disparities and access to healthy food in the
United States: a review of food deserts literature. Health & Place, 16(5), 876–884.
Whelan, A., Wrigley, N., Warm, D., & Cannings, E. (2002). Life in a’food desert’. Urban
Studies, 39(11), 2083–2100.
Widener, M. J., Farber, S., Neutens, T., & Horner, M. W. (2013). Using urban commuting data
to calculate a spatiotemporal accessibility measure for food environment studies. Health
& Place, 21, 1–9. doi:10.1016/j.healthplace.2013.01.004
Ziskin, M., Fischer, M. A., Torres, V., Pellicciotti, B., & Player-Sanders, J. (2014). Working
Students’ Perceptions of Paying for College: Understanding the Connections between
Financial Aid and Work. The Review of Higher Education, 37(4), 429–467.
EDUCATION DESERTS
Appendix A:
Number of public colleges by commuting zone and county 11
40
End Notes
1
This excludes all students who enroll exclusively online. The variable names from PowerStats include:
DISTANCE, SECTOR4 and ALTONLN2. WTA000 weight is used. Source: U.S. Department of Education,
National Center for Education Statistics, 2011-12 National Postsecondary Student Aid Study (NPSAS:12).
2
For examples of how researchers are now using CZs as a unit of analysis, see Autor & Dorn’s (2013) work on local
labor market wages and Chetty, Hendren, Kline, & Saez’s (2014) work on intergenerational mobility.
3
This is measured as a mean of 2008 to 2012 educational attainment levels.
4
See (Jaquette & Parra, 2014).
5
In Indiana, for example, the state’s community college system (Ivy Tech Community College) has 31 campuses
across the state and one central administrative office in Indianapolis. Until 2011, the system reported IPEDS data for
the branch campuses, but now only reports at the system-level. In IPEDS, it appears that there is only one Ivy Tech
campus when in fact there are 31. Several other colleges are grouped as systems, rather than individual campuses.
6
This search included going to each state’s higher education office to identify locations of branch and satellite
campuses. When information was unavailable, the research team searched for the terms “college” and “university”
in counties that are classified as education deserts. It is possible that we are undercounting the number of colleges in
counties that are outside of education deserts since the main objective was to ensure that the places we identified as
deserts indeed have no public four-year branch campuses. For example, the University of Texas System has its
flagship campus in Austin and several regional campuses (e.g., University of Texas at Tyler) throughout the state
that individually report to IPEDS. In IPEDS, UT-Tyler is located in Tyler, TX; however, UT-Tyler has two
additional branch campuses in other cities (Longview and Palestine) that are not reported in IPEDS. Fortunately,
many of these unreported campuses are located in the same commuting zone as the reporting campus, minimizing
the effects of under-reporting.
7
In the event that a CZ spans multiple states, the CZ is attributed to the state in which the largest population share of
that CZ reside. For example, the Memphis commuting zone spans three states: Tennessee, Mississippi, and
Arkansas. Since the majority of this CZ’s population resides in Tennessee, it is clustered around Tennessee.
8
In the interest of space, these results are not provided here. These results are available upon request.
9
This is calculating by taking the exponent of the log odds: (exp(0.378) and exp(0.346)).
10
See here, using ZIP Code 78852: http://nces.ed.gov/collegenavigator/?s=all&zc=78852&zd=100&of=3
11
Map created in Tableau ®, public access to the interactive map here:
https://public.tableausoftware.com/views/Hillman-CRP/Sheet1?:embed=y&:display_count=no