Policy Analysis Educational Evaluation and

http://eepa.aera.net

Policy Analysis

Educational Evaluation and

http://epa.sagepub.com/content/early/2011/04/13/0162373711399092

The online version of this article can be found at:

DOI: 10.3102/0162373711399092

published online 13 April 2011EDUCATIONAL EVALUATION AND POLICY ANALYSIS

Thomas J. Cooke and Paul Boyle

The Migration of High School Graduates to College

Published on behalf of

American Educational Research Association

and

http://www.sagepublications.com

can be found at:Educational Evaluation and Policy AnalysisAdditional services and information for

http://eepa.aera.net/alertsEmail Alerts:

http://eepa.aera.net/subscriptionsSubscriptions:

http://www.aera.net/reprintsReprints:

http://www.aera.net/permissionsPermissions:

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

The Migration of High School Graduates to College

Thomas J. Cooke

University of Connecticut

Paul Boyle

University of St. Andrews

The National Center for Education Statistics reports that over 250,000 high school graduates moved

across state lines to enroll in college in 2008. The choices made by these high ability individuals may

have long-lasting implications for state economies; not only do they contribute to state and local econo-

mies through their tuition and daily living costs while studying, but many of them will be retained in the

workforce in the state in which they graduate. This paper argues that inadequate attention has been paid

to the spatial processes that underpin such migrations. Specifically, models are required that simultane-

ously consider the characteristics of both migration origins and destinations and their relative spatial

arrangements. Thus, the purpose of this research is to present an alternative, explicitly spatial, approach

to modeling the migration of high school graduates to college. Our results provide new insights into the

factors that determine such flows and have direct relevance to policy-making in this sphere.

Keywords: migration, college enrollment

geographic concentration of highly skilled

workers is an important determinant of regional

economic growth (Storper & Scott, 2009). However,

the factors that attract skilled workers into certain

areas are relatively underresearched (Hansen &

Niedomysl, 2009). Particularly important are the

flows of students to college; the National Center

for Education Statistics (2009) reported that over

250,000 high school graduates moved across state

lines to enroll in college in 2008. The choices made

by these high-ability individuals may have long-

lasting implications for state economies; not only

do they contribute to state and local economies

through their tuition and daily living costs while

studying, but at least some them will be retained

in the workforce in the states in which they gradu-

ate (Groen, 2004; Parsad & Gray, 2005).

Previous research on student migration has

generally focused on either in-migration (the

percentage nonresident enrollment at a university)

or out-migration (the percentage of a state’s high

school graduates attending school in another state).

These studies have found that the out-migration

of students is positively related to public university

tuition and negatively related to the quality of both

public and private universities, the number of

enrollment opportunities, and the availability of

a broad-based public university merit scholarship

program (see Mak & Moncur, 2003; Orsuwan &

Heck, 2009; Zhang & Ness, 2010), while the in-

migration of students is positively related to the

quality of the university and the size of the univer-

sity and negatively related to tuition (see Adkisson

& Peach, 2008). Results regarding the effect of

tuition on nonresident enrollment are mixed,

with strong evidence that elite national universities

enjoy a high degree of pricing power, so that for

them, nonresident enrollment is positively related

Educational Evaluation and Policy Analysis

Month XXXX, Vol. XX, No. X, pp. xx–xx

DOI: 10.3102/0162373711399092

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

Cooke and Boyle

to tuition (Baryla & Dotterweich, 2001, 2006;

Dotterweich & Baryla, 2005).

However, the movement of students from one

state to another state is a joint function of both the

characteristics of the state of origin and potential

destination states. Models that focus on just out-

migration or in-migration cannot fully capture how

the joint characteristics of origins and destinations

influence migration behavior. For example, a

model of out-migration rates can include only the

characteristics of the origin state and cannot

include those of the chosen destination state, even

though those destination-specific characteristics

clearly have an influence on out-migration. A par-

allel argument can be made for in-migration. Thus,

models of in- and out-migration rates may not be

correctly specified, because they exclude important

determinants of the migration decision. The pur-

pose of this research is to introduce to the student

migration literature a well-established technique

used in migration research that addresses this issue.

Our results provide new insights into the factors

that determine student migration behavior and have

direct relevance to policy making in this sphere.

Background

The decision to attend college is clearly spatial;

student enrollment choices are inevitably based

to some degree on the spatial distribution of enroll-

ment opportunities relative to their place of high

school residence. Students who live in close prox-

imity to a diverse range of enrollment opportunities

are more likely to apply to college and to attend

colleges closer to home, while students who live

in areas with few enrollment opportunities are less

likely to apply to college and more likely to attend

colleges far from home (Frenette, 2006; Leppel,

1993; Lopez Turley, 2009; Mulder & Clark,

2002; Sa, Florax, & Rietveld, 2006). Thus, the out-

migration of students from a state is not just a

function of the state’s characteristics but is also

influenced by the opportunities in every other

state and their spatial arrangement relative to

the origin state. The same can be said for the in-

migration of students into colleges within states.

The importance of considering simultaneously

how the characteristics of origins, destinations,

and their relative spatial arrangement influence

migrant flows is well established in the geography

literature. Failure to incorporate these factors into

migration models may introduce specification bias

and likely ignores significant factors that ulti-

mately determine net migration patterns (Rogers,

1990). To resolve these issues, geographers have

developed spatial interaction models, a special case

of which is the seminal gravity model formulation

(Stewart, 1948; Stouffer, 1940, 1960; Zipf, 1949),

from which the determinants of out-, in-, and net

migration patterns can be appropriately identified

(Haynes & Fotheringham, 1984; Wilson, 1971).

The spatial interaction framework conceptualizes

the gross volume of migration, M, between an

origin (i) and a destination (j) as a function of the

attributes of the origin, O, the attributes of the

destination, D, and attributes describing the spatial

arrangement of origins and destinations, S:

= f(O

, D

, S

So, for example, the dependent variable in a

spatial interaction model of migration between each

of the 50 states would be each of the 2,450 observed

intrastate migration flows (50

– 50 or n

– n pos-

sible flows after excluding intrastate migration

flows). As the flows are counts of migrants, and

there will be many zero, several low, and few high

values, a Poisson regression model is an appro-

priate functional form for a spatial interaction

model (Congdon, 1991, 1993; Flowerdew, 1991;

Flowerdew & Aitkin, 1982).

Although any number of independent variables

can be incorporated into a spatial interaction

model, the key variables relating to a traditional

gravity model specification would include an

origin variable reflecting the size of the popula-

tion at risk for migrating, a destination variable

reflecting the number of opportunities available

to migrants, and the distance between each origin

and the destination. Commonly, a second spatial

variable would also be used to reflect the spatial

structure of the various origins and destinations,

such as Stouffer’s (1940, 1960) intervening oppor-

tunities variable, whereby the amount of migration

between an origin and a destination will be reduced

by the degree to which there are alternative, inter-

vening migration destinations that lie between the

two. Failure to include a spatial structure variable

such as this may subject the analysis to both speci-

fication and autocorrelation bias.

Although nearly all models of student migra-

tion have incorporated some sort of variable that

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

The Migration of High School Graduates to College

measures spatial opportunities surrounding an

origin for models of out-migration, or a destination

for models of in-migration, very few studies have

modeled migration flows from origins to destina-

tions within a spatial interaction approach. Slater

(1976) created migration regions from a matrix of

student migration flows, Johns & Viehland (1989)

described the visual patterns observed in a matrix

of student migration flows, and Abbott and Schmid

(1975), Fryman (1988), and Kyung (1996) observed

a negative effect of distance on interstate student

migration. More substantively, Alm and Winters

(2009) estimated models of intrastate student

migration flows between Georgia’s counties and

institutions of higher education and found that dis-

tance mediates the effects of the other independent

variables. Sa, Florax, and Rietveld (2004) reached

a similar conclusion in models of student migra-

tion flows in the Netherlands.

These studies notwithstanding, McHugh and

Morgan (1984) performed the only study to have

estimated models of the flows between the 48

contiguous U.S. states for a sample of students

enrolled in public universities. They found that

the size of the migration flow increased with the

number of students in the origin state, the per

capita income in the origin and the destination,

and private college costs of attendance in the

origin, and decreased with measures of educa-

tional quality in the destination. With respect to

spatial factors, the size of migration flows declined

with the distance between the origin and destina-

tion and increased with the average distance of

the origin state from all other states. The analysis

indicates that both origin- and destination-specific

characteristics as well as their spatial arrangement

jointly, but not necessarily symmetrically, influ-

ence college-bound students’ decisions. However,

the study was limited by the fact that it was

restricted only to public university students, and

the results were not interpreted with respect to

the net gain or loss of college students by state.

Finally, the specification of the “intervening

opportunities” variable (the average distance of

all states from the origin state) was not defined

consistently with the variables commonly used

in the migration literature, and its positive value

is counterintuitive: The expected flow between

any two randomly selected states should be lower,

rather than higher, if there are many opportunities

between the two states.

The spatial interaction perspective models

migration flows between a set of origins and

destinations as a function of the characteristics of

the origins, the destinations, and their relative

spatial arrangement. Models that focus on just out-

migration or in-migration cannot fully capture how

the joint characteristics of origins and destinations,

and their spatial arrangement, influence migration

behavior and may be improperly specified. The

purpose of this research is to introduce this approach

to the student migration literature through the esti-

mation of a spatial interaction model of the inter-

state migration of high school students to college.

Data and Methods

The primary source of data is the 2007 Integrated

Postsecondary Education Data System provided by

the U.S. Department of Education’s National Center

for Education Statistics, which reports annual data

on enrollments, program completions, graduation

rates, faculty and staff, finances, institutional prices,

and student financial aid for all institutions that

participate in federal student aid programs (National

Center for Education Statistics, 2009). The analysis

uses data from the 2006–2007 academic year and

focuses on public and private colleges and universi-

ties that offer accredited bachelor’s degrees in a

primarily traditional residential setting.

Importantly,

these data report, for each institution, the number

of students by state when they applied to college.

It is from these data that gross interstate migration

flows are calculated.

The analysis is limited to 47 states and the

District of Columbia (which for simplicity is treated

and referred to as a state throughout the analysis).

Hawaii and Alaska are not included because their

separation from the contiguous United States

would introduce modeling issues that would not

appreciably contribute to the overall quality of the

model because of the small amount of migration

to and from those two states. Wyoming is also

deleted, because the model specification requires

independent variables on the characteristics of each

state’s private universities. However, Wyoming

is a small state with only one large public university

and no private universities.

The resulting sample of 2,256 observations rep-

resents the number of college-bound students migrat-

ing from each of the 48 selected states to every

other state (48

– 48 interstate migration flows after

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

TABLE 1

Variable Definitions

Variable Definition

Spatial variables

Distance between origin and

destination

Distance between origin and destination states (logged)

Adjacency 1 if states are adjacent, 0 if not

Intervening opportunities See text

Geographic variables

High school graduates, origin Number of in-sample college-bound high school graduates

(logged)

High school graduates, destination Number of in-sample college-bound high school graduates

(logged)

Change in unemployment, origin Change in unemployment rate from 2006 to 2007

Change in unemployment,

destination

Change in unemployment rate from 2006 to 2007

Per capita income, origin 2006 per capita income (logged)

Per capita income, destination 2006 per capita income (logged)

Amenities, origin A measure of the quality of natural amenities based on

topography and climatic conditions (Peters, 2000

Amenities, destination A measure of the quality of natural amenities based on

topography and climatic conditions (Peters, 2000)

% Urban, origin Percentage of the state population living in urban areas

% Urban, destination Percentage of the state population living in urban areas

% Population 18–24, origin Percentage of the state population between the ages of 18 and

24 years

% Population 18–24, destination Percentage of the state population between the ages of 18 and

24 years

Public university variables

Public enrollment, origin Full-time equivalent in-sample enrollment (logged)

Public enrollment, destination Full-time equivalent in-sample enrollment (logged)

Public ACT 75th percentile, origin Enrollment-weighted average 75th-percentile score on the

ACT

Public ACT 75th percentile,

destination

Enrollment-weighted average 75th-percentile score on the

ACT

Public cost of enrollment, origin Enrollment-weighted average in-state public cost of

attendance (logged)

Public cost of enrollment, destination Enrollment-weighted out-of-state public tuition cost of

attendance (logged)

Public admissions rate, origin Enrollment-weighted university admissions rate

Public admissions rate, destination Enrollment-weighted university admissions rate

Merit scholarship program, origin Presence of a broad-based merit scholarship program (1 = yes,

0 = no) (Orsuwan & Heck, 2009)

Merit scholarship program,

destination

Presence of a broad-based merit scholarship program (1 = yes,

0 = no) (Orsuwan & Heck, 2009)

Private university variables

Private enrollment, origin Full-time equivalent in-sample enrollment (logged)

Private enrollment, destination Full-time equivalent in-sample enrollment (logged)

Private ACT 75th percentile, origin Enrollment-weighted average 75th-percentile score on the ACT

Private ACT 75th percentile,

destination

Enrollment-weighted average 75th-percentile score on the ACT

Private cost of enrollment, origin Enrollment-weighted average cost of attendance (logged)

Private cost of enrollment,

destination

Enrollment-weighted average cost of attendance (logged)

Private admissions rate, origin Enrollment-weighted university admissions rate

Private admissions rate, destination Enrollment-weighted university admissions rate

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

The Migration of High School Graduates to College

deleting intrastate moves). These gross migration

flows represent the number of students moving from

their states of residence (defined as their states of

residence when they applied to college in the spring

of 2006) to their states of college matriculation in

the fall of 2006. These flows are then estimated in

a Poisson regression framework as a function of

four broad sets of variables (see Table 1).

The first set of independent variables measure

the effect of the spatial arrangement of origins and

destinations on gross migration flows between

states. Distance is included because one of the

most consistent findings in the migration literature

is that flows generally decline with distance, reflect-

ing the cost of migration, the quality of informa-

tion, and, for students, the separation from friends

and family (Frenette, 2006; Leppel, 1993; Lopez

Turley, 2009; Mulder & Clark, 2002; Sa et al.,

2006). In this case, distance between states is mea-

sured by calculating the spherical distance between

the population-weighted geographic centroids of

each state (U.S. Census Bureau, 2009).

However,

Euclidean distance is a crude measure, especially

for contiguous states. These may be poorly esti-

mated because the distance calculated between

contiguous zones is commonly an overestimate of

the average distance moved between such places.

Consequently, a dummy variable is often used to

identify those pairs of places that are adjacent

(Zhang & Ness, 2010), and in this case, a dummy

variable is included reflecting whether a pair of

states are adjacent to each other. This may also

capture regional agreements that encourage stu-

dents to attend universities in neighboring states

(see Zhang & Ness, 2010). Additionally, migration

from the origin to the destination is likely affected

by intervening opportunities: We expect that migra-

tion from one state to another will be reduced

if the origin state is closely surrounded by a large

number of higher education opportunities, and

likewise, migration from one state to another will

increase if the origin state is located in an area with

very few nearby higher education opportunities.

The variable is defined as

log( )

FTEj

i j i

























≠

∑

where FTEj is the in-sample full-time enrollment

in both public and private universities at the

destination, and d

is the distance from state i to

state j.

The second set of independent variables

reflects the geographic characteristics of each

state. College students migrate as single individu-

als with low costs of living and a lifetime to

recoup the cost of the migration decision (Perna,

2006). Thus, they are likely to be drawn to high-

amenity destinations, like many other young

populations with more expendable income (Black,

Gates, Sanders, & Taylor, 2000; Chen & Rosenthal,

2008; Plane, Henrie, & Perry, 2005; Whisler,

Waldorf, Mulligan, & Plane, 2008). Variables

such as natural amenities in the origin and des-

tination, the age characteristics of the population

in the origin and destination, urbanization in the

origin and destination, income levels in the ori-

gin and destination, and the number of potential

migrants in the origin and destination are therefore

included.

The third and fourth sets of independent vari-

ables are included to measure the effects of public

and private university characteristics on migration

flows. The decision to migrate to college is also a

human capital investment (Parsad & Gray, 2005;

Sjaastad, 1962). Quite literally, students are mak-

ing an investment in time and money with the direct

goal of increasing their lifetime utility and, prob-

ably, their lifetime earnings. Thus, the quality of

the enrollment opportunities relative to the cost of

those opportunities is also a factor in the migration

decision. These factors are measured by variables

indicating enrollment in the origin and destination,

admissions rates in the origin and destination, total

cost of attendance in the origin and destination,

the presence of a merit scholarship program in the

origin and destination, and measures of university

quality in the origin and destination.

The model parameters give clear indications as

to how specific variables affect the flow of students

from origins to destinations. However, interpreting

the parameters of the model is not clear cut, because

many of the key variables are log transformed, and

the model itself is not linear. A more insightful,

alternative approach is to make predictions with

respect to these key variables. First, the combined

impact of the spatial variables on state net migration

is made by predicting every migration flow while

holding the spatial variables at their sample means.

These predicted flows are then summed by origins

and destinations to calculate the net migration rate

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

for each state while holding these spatial variables

constant. The difference in this value with the pre-

dicted net migration rate is a measure of the impact

of spatial structure on net migration rates. Similarly,

the effect of implementing policy changes to improve

net migration rates by lowering resident and non-

resident costs of attendance and establishing a

broad-based merit scholarship program is calculated

in the same way.

Results

As described above, the ultimate aim of this

analysis is to gain a better understanding of net

migration, and Figure 1 shows the observed

patterns. First, the states with the largest nega-

tive net migration are either small, densely settled

states of the East Coast (Maryland, Delaware,

and New Jersey) or large, generally populous

states (Minnesota, Illinois, and Georgia), Nevada

being the exception, although it is similar to this

last group in terms of size. The second pattern

relates to the first, as the states with the largest

positive net migration are generally adjacent, or

proximate, to the previous set of negative net migra-

tion states (e.g., Iowa, Indiana, South Carolina,

Alabama, Utah, Pennsylvania, and West Virginia).

These states are also less densely settled and/or

more rural than the negative net migration states.

In these cases, the spatial proximity to large num-

bers of high school graduates may provide their

colleges with an advantage in attracting out-of-

state students.

Table 2 reports the model estimates along with

the likelihood ratio test and the Wald test for the

inclusion of the three spatial variables. These

tests are all highly significant, indicating that the

spatial variables make a significant contribution

to the estimation of the migration of students

between states. Clearly, a spatial perspective

improves the explanation of the interstate migra-

tion of college-bound high school graduates, and

the first set of parameter estimates demonstrate

that migration flows decrease with the distance

between the origin and destination, increase if

states are adjacent to one another, and increase

(decrease) if the origin is surrounded by states

with few (many) higher education opportunities.

These are not surprising results, but their sig-

nificance is important to emphasize, as previous

research on student migration has neglected to

include them.

This point is emphasized in Figure 2, which

shows the contribution of each state’s spatial

FIGURE 1. Net migration of college-bound high school graduates.

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

variables to its overall net migration rate. Positive

values indicate that the combined effect of the

spatial variables (distance, intervening opportu-

nities, and adjacency) act to improve the net

migration of a state, while negative values indi-

cate that spatial structure acts to harm the net

migration of a state. Thus, the large negative value

for New Jersey indicates that it is harmed by its

relative spatial location. In this case, New Jersey

high school graduates are in close proximity to

out-of-state enrollment opportunities as measured

by distance, adjacency, and intervening opportuni-

ties, all of which increases out-migration. Second,

residents in adjacent states also have many nearby

out-of-state opportunities other than New Jersey

that decrease in-migration to New Jersey. Finally,

TABLE 2

Parameter Estimates

Variable Parameter

Spatial variables

Distance between origin and destination –1.040 .000

States are adjacent 0.807 .000

Intervening opportunities –8.475 .000

Geographic variables

High school graduates, origin 0.023 .098

High school graduates, destination 0.073 .000

Change in unemployment, origin 0.130 .000

Change in unemployment, destination –0.144 .000

Per capita income, origin 2.871 .000

Per capita income, destination –0.387 .000

Amenities, origin –0.154 .000

Amenities, destination 0.164 .000

% Urban, origin 0.004 .000

% Urban, destination –0.027 .000

% Population 18–24, origin –0.262 .000

% Population 18–24, destination 0.169 .000

Public university variables

Public enrollment, origin 0.630 .000

Public enrollment, destination 0.250 .000

Public ACT 75th percentile, origin 0.055 .000

Public ACT 75th percentile, destination 0.113 .000

Public tuition, origin 1.199 .000

Public tuition, destination –0.111 .000

Public cost of enrollment, origin 0.001 .040

Public cost of enrollment, destination 0.003 .000

Merit scholarship program, origin –0.132 .000

Merit scholarship program, destination –0.011 .155

Private university variables

Private enrollment, origin –0.108 .000

Private enrollment, destination 0.447 .000

Private ACT 75th percentile, origin 0.066 .000

Private ACT 75th percentile, destination 0.027 .000

Private cost of enrollment, origin –0.359 .000

Private cost of enrollment, destination 0.659 .000

Private admissions rate, origin –0.017 .000

Private admissions rate, destination 0.020 .000

Constant –30.557 .000

Test of model fit

Pseudo-R

.8310 .000

Tests of adding spatial variables

Wald test 280,000 .000

Likelihood ratio test 264,512 .000

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

not only is New Jersey far from potential in-

migrants because of its position on the East Coast,

but it is in the center of the country’s greatest

concentration of enrollment opportunities, and

hence, from the perspective of students in other

states, there are many other intervening oppor-

tunities between any potential origin state and

New Jersey. Although these likely affect all East

Coast states, this combination of spatial factors is

most difficult for New Jersey.

In contrast, the large negative effect of spatial

structure on net migration on the western states

of Montana, Washington, Nevada, and California

is due to a different set of circumstances. Montana,

Washington, and California are spatially isolated

by distance and intervening opportunities from

potential high school students and surrounded

by less populous states. Nevada appears to be an

anomaly, but in fact the location of its two major

universities in Reno and Las Vegas effectively puts

the center of the state’s higher education sector on

the border with California, meaning that it suffers

from the same sort of spatial issues as Montana,

Washington, and California. With respect to those

states in Figure 2 with a positive spatial effect,

almost without exception these are less densely

settled, more rural states surrounded by more

densely settled, more urban states. As a result,

they benefit from their spatial proximity to a

large number of potential in-migrants relative

to the small number of potential out-migrants

within their borders.

The second set of parameter estimates in Table 2

relates to the geographic characteristics of states

as origins and destinations. One way to interpret

these results is in terms of the direction of migration

flows. For example, the parameter for the number

of high school graduates in the origin is 0.023,

while the parameter for the number of high school

graduates in the destination is 0.073. This implies

that the number of high school graduates has a

greater positive effect on in-migration than on out-

migration. Thus, student migration flows are inef-

ficient in the sense that states with many high

school graduates experience positive net migration

as a result. This may be due to multicollinearity with

population size or even the spatial variables, but

alternative model specifications including the popu-

lation of each origin state did not alter these results.

The other geographic variables are more directly

interpreted and consistent with expectations regard-

ing a young, upwardly mobile population: Student

migration flows from states with increasing unem-

ployment toward states with decreasing unemploy-

ment, from states with higher incomes toward states

with lower incomes, from states with fewer natural

FIGURE 2. Spatial effects on net migration rate.

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

The Migration of High School Graduates to College

amenities toward states with more natural ameni-

ties, from highly urbanized states to less urbanized

states, and from states with low proportions of 18-

to 24-year-olds to states with high proportions of

18- to 24-year-olds.

With respect to the characteristics of public

universities, the enrollment parameters indicate

the same sort of inefficiency as with the high

school graduate parameters: The parameter for

public enrollment in the origin is 0.630, while the

parameter for public enrollment in the destination

is 0.250. This implies that states with large public

enrollments have net out-migration of students.

This too may be a spurious correlation with popu-

lation size, but alternative model specifications

including the population of each origin state did

not alter these results. The other public university

variables behave more to expectations: Student

migration flows toward states with higher quality

public universities (as measured by standardized

test scores), from states with higher in-state costs

of attendance toward states with lower out-of-

state costs of attendance, and toward states with

higher admissions rates. The parameters associ-

ated with the presence of merit scholarship pro-

grams only indicate that they reduce out-migration

and have no effect on in-migration. This is expected

because merit scholarships are not available to

out-of-state students.

Finally, the private university parameters indi-

cate that student migration flows from states with

lower private enrollment toward states with higher

private enrollment, from states with higher quality

private universities toward states with higher pri-

vate costs of attendance, and from states with

lower private admissions rates toward states with

higher private admissions rates. These results

confirm Dotterweich and Baryla’s (2005) conclu-

sion that “it appears that the very selective, expen-

sive, private [institutions of higher education] are

perceived differently by non-resident students

and may have a special cache in the education

marketplace” (p. 381).

The effect of the size of the public higher edu-

cation sector notwithstanding and within the limi-

tations of a cross-sectional analysis, the policy

prescriptions are quite clear. Out-migration can

be reduced by developing a merit scholarship

program and lowering in-state costs of attendance,

while in-migration can be increased by improving

quality (as measured by standardized test scores),

reducing out-of-state costs of attendance, and

increasing admissions rates. The impact of these

policy prescriptions is evaluated in the same

manner that the spatial effects were estimated.

However, lowering admissions rates and raising

standardized tests scores are not directly achieved

and have to be addressed within the context of a

host of other changes at each university, most

notably a likely increase in enrollment. Therefore,

this exercise focuses on the two most easily imple-

mented of these policies: decreasing both in-state

and out-of-state public tuition and the adoption

of in-state merit scholarship programs.

The effects of implementing such policies are

estimated on a sample of six states, all of which

have large negative net migration rates and none

of which have merit scholarship programs (Orsuwan

& Heck, 2009): Connecticut, Delaware, Illinois,

Maryland, Minnesota, and New Jersey. For refer-

ence, total costs of enrollment range from a low of

$17,798 for resident students and $25,326 for non-

resident students in Minnesota to a high of $23,263

for resident students and $30,732 for nonresident

students in New Jersey. Figure 3 shows that all of

the states would benefit from reductions in non-

resident and resident costs of enrollment. Most

prominently, New Jersey, with a net migration rate

of –369 per 1,000 and with the highest resident and

nonresident costs of enrollment of the six states,

would see an increase in net migration to –248 per

1,000 with a 20% drop in costs of attendance. The

other five states would see relatively similar improve-

ments in net migration. Indeed, Maryland could

nearly eliminate its predicted net migration of

–244 per 1,000 to –43 per 1,000 by reducing costs

of attendance by 20%. The effects of establishing

a broad-based merit scholarship program such as

that implemented in Georgia are less meaningful.

Indeed, the largest effect is for Delaware, which

would see its predicted net migration fall from –452

per 1,000 to –391 per 1,000. Thus, student migra-

tion is not completely determined by spatial and

geographic differences between places but is also

responsive to policy prescriptions.

Discussion

This analysis makes several contributions to

the study of the migration of high school students

to college. First, this analysis introduced a spatial

interaction approach, based on migration theory,

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

to the student migration literature. Although many

previous studies have included regional variables

and some spatial variables such as distance, this

is the first to explicitly model the interstate migra-

tion of high school graduates to college within the

spatial interaction framework. Our results are

broadly consistent with previous findings: States

with higher quality, competitively priced public

universities and with higher priced public universi-

ties have a positive net flow of college freshmen.

Students also are attracted toward more rural, high-

amenity states with younger populations. How-

ever, the spatial perspective demonstrates the

importance of the unequal distribution and arrange-

ment of both high school students and colleges

across the United States. For example, even though

New Jersey is in the population center of the coun-

try, its position relative to surrounding states and

on the edge of the continent places a severe struc-

tural impediment to achieving a positive flow of

students into that state. On a more positive note,

many less densely settled states proximate to

larger, more densely settled states, especially in

the eastern half of the United States, experience a

positive externality in the form of increased net

migration because of their spatial location alone.

We have also shown that the net flow of college

students could be responsive to policy intervention:

The model estimates indicate that net student

migration can be improved by lowering in-state

costs of attendance, improving quality (as mea-

sured by standardized test scores), reducing out-

of-state costs of attendance, and increasing

admissions rates. And indeed, simulations of the

impact of a change in costs of attendance and estab-

lishing a broad-based merit scholarship program

significantly improve net migration among a set

of states that currently experience net out-migra-

tion, even for those such as New Jersey that are at

a distinct spatial disadvantage. However, this

analysis does not take the cost of these programs

into account and additional analyses are needed

to more precisely estimate the specific impacts

of specific policy changes on student migration

flows and the costs of those policies. In particular,

future research should address the endogeneity

between migration flows and many of the inde-

pendent variables (e.g., tuition) through the use

of panel rather than cross-sectional data and more

precisely measure the role of proximity between

states in shaping interstate migration by incorporat-

ing the effect of interstate reciprocity

agreements.

Notes

1. The full list of selected institutions is available

from the authors. The selection criteria exclude

FIGURE 3. Effects of policy variables.

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

The Migration of High School Graduates to College

institutions that (a) are not 4-year colleges, (b) are not

degree-granting institutions, (c) do not offer bachelor’s

degrees, (d) are primarily associate’s degree–granting

institutions that also offer bachelor’s degrees, (e) are

primarily or exclusively graduate level, (f) are primarily

or exclusively associate level, (g) are tribal colleges,

and (h) are for profit.

2. Spherical distances were calculated using the

SPHDIST Stata Module (Rising, 1999).

3. The University of the District of Columbia is the

only public university in the district, but the cost of

attendance data were not reported in this data set. The

cost was estimated from the College Board (2009).

References

Abbott, W. F., & Schmid, C. F. (1975). University

prestige and first-time undergraduate migration in

the United States. Sociology of Education, 48(2),

168–185.

Adkisson, R. V., & Peach, J. T. (2008). Non-resident

enrollment and non-resident tuition at land grant

colleges and universities. Education Economics,

16(1), 75–88.

Alm, J., & Winters, J. V. (2009). Distance and intrastate

college student migration. Economics of Education

Review, 28(6), 728–738.

Baryla, E. A., Jr., & Dotterweich, D. (2001). Student

migration: Do significant factors vary by region?

Education Economics, 9(3), 269–280.

Baryla, E. A., Jr., & Dotterweich, D. (2006). Institutional

focus and non-resident enrollment. International

Journal of Educational Management, 13, 239–248.

Black, D., Gates, G., Sanders, S., & Taylor, L. (2000).

Demographics of the gay and lesbian population in

the United States: Evidence from available system-

atic data sources. Demography, 37(2), 139–154.

Chen, Y., & Rosenthal, S. S. (2008). Local amenities

and life-cycle migration: Do people move for jobs or

fun? Journal of Urban Economics, 64(3), 519–537.

College Board. (2009). College search. Retrieved

July 9, 2009, from http://collegesearch.collegeboard

.com/search/index.jsp

Congdon, P. (1991). General linear modelling: Migration

in London and south east England. In J. Stillwell &

P. Congdon (Eds.), Migration models: Macro and micro

approaches (pp. 113–136). London, UK: Belhaven.

Congdon, P. (1993). Approaches to modelling overdis-

persion in the analysis of migration. Environment

and Planning A, 25(10), 1481–1510.

Dotterweich, D., & Baryla, E. A., Jr. (2005). Non-res-

ident tuition and enrollment in higher education:

Implications for tuition pricing. Education Economics,

13(4), 375—385.

Flowerdew, R. (1991). Poisson regression modeling

of migration. In J. Stillwell & P. Congdon (Eds.),

Migration models: Macro and micro approaches

(pp. 92–112). London, UK: Belhaven.

Flowerdew, R., & Aitkin, M. (1982). A method of fitting

the gravity model based on the Poisson distribution.

Journal of Regional Science, 22(2), 191.

Frenette, M. (2006). Too far to go on? Distance to

school and university participation. Education

Economics, 14(1), 31–58.

Fryman, J. F. (1988). Factors in interstate migration

of college students. College and University, 63(3),

234–247.

Groen, J. A. (2004). The effect of college location on

migration of college-educated labor. Journal of

Econometrics, 121(1–2), 125–142.

Hansen, H. K., & Niedomysl, T. (2009). Migration of

the creative class: Evidence from Sweden. Journal of

Economic Geography, 9(2), 191–206.

Haynes, K. E., & Fotheringham, A. S. (1984). Gravity

and spatial interaction models. Beverly Hills, CA:

Sage.

Johns, V., & Viehland, D. W. (1989). Migration pat-

terns of first-time freshmen in the United States.

Research in Higher Education, 30(5), 485–502.

Kyung, W. (1996). In-migration of college students to

the state of New York. Journal of Higher Education,

67(3), 349–358.

Leppel, K. (1993). Logit estimation of a gravity model

of the college enrollment decision. Research in

Higher Education, 34(3), 387–398.

Lopez Turley, R. N. (2009). College proximity: Map-

ping access to opportunity. Sociology of Education,

82(2), 126–146.

Mak, J., & Moncur, J. E. T. (2003). Interstate migration

of college freshmen. Annals of Regional Science,

37(4), 603–612.

McHugh, R., & Morgan, J. N. (1984). The determinants

of interstate student migration: A place-to-place

analysis. Economics of Education Review, 3(4),

269–278.

Mulder, C. H., & Clark, W.A.V. (2002). Leaving home

for college and gaining independence. Environment

and Planning A, 34(6), 981–999.

National Center for Education Statistics. (2009).

IPEDS data center. Retrieved June 3, 2009, from

http://nces.ed.gov/ipeds/datacenter/login.aspx

Orsuwan, M., & Heck, R. (2009). Merit-based student

aid and freshman interstate college migration: Test-

ing a dynamic model of policy change. Research in

Higher Education, 50(1), 24–51.

Parsad, B., & Gray, L. (2005). Interstate migration pat-

terns of recent recipients of bachelor’s and master’s

degrees in science and engineering. Arlington, VA:

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from

Cooke and Boyle

National Science Foundation, Division of Science

Resource Statistics.

Perna, L. (2006). Studying college access and choice:

A proposed conceptual model. In J. C. Smart (Ed.),

Higher education: Handbook of theory and research

(Vol. 21, pp. 99–157). Amsterdam, the Netherlands:

Springer.

Peters, D. J. (2000). Understanding natural ameni-

ties: Impacts on population and employment in

Missouri. Jefferson City, MO, Missouri Depart-

ment of Economic Development: Research and

Planning.

Plane, D. A., Henrie, C. J., & Perry, M. J. (2005).

Migration up and down the urban hierarchy and

across the life course. Proceedings of the National

Academy of Sciences, 102(43), 15313–15318.

Rising, B. (1999). SPHDIST: Stata module to com-

pute spherical distances. Boston: Boston College,

Department of Economics.

Rogers, A. (1990). Requiem for the net migrant. Geo-

graphical Analysis, 22(4), 283–300.

Sa, C., Florax, R., & Rietveld, P. (2004). Determinants

of the regional demand for higher education in the

Netherlands: A gravity model approach. Regional

Studies, 38(4), 375–392.

Sa, C., Florax, R.J.G.M., & Rietveld, P. (2006). Does

accessibility to higher education matter? Choice

behaviour of high school graduates in the Netherlands.

Spatial Economic Analysis, 1(2), 155–174.

Sjaastad, L. A. (1962). The costs and returns of human

migration. Journal of Political Economy, 70(5),

80–93.

Slater, P. B. (1976). The use of state-to-state college

migration data in developing a hierarchy of higher

educational regions. Research in Higher Education,

4(4), 305–315.

Stewart, J. W. (1948). Demographic gravitation: Evi-

dence and applications. Socimetry, 11, 31–47.

Storper, M., & Scott, A. J. (2009). Rethinking human

capital, creativity and urban growth. Journal of

Economic Geography, 9(2), 147–167.

Stouffer, S. A. (1940). Intervening opportunities:

A theory relating mobility and distance. American

Sociological Review, 5, 845–867.

Stouffer, S. A. (1960). Intervening opportunities and

competing migrants. Journal of Regional Science,

2(1), 2–26.

U.S. Census Bureau. (2009). Population and population

centers by state: 2000. Retrieved June 19, 2009, from

http://www.census.gov/geo/www/cenpop/cntpop2k

.html

Whisler, R. L., Waldorf, B. S., Mulligan, G. F., &

Plane, D. A. (2008). Quality of life and the migration

of the college-educated: A life-course approach.

Growth and Change, 39(1), 58–94.

Wilson, A. G. (1971). A family of spatial interaction

models, and associated developments. Environment

and Planning A, 3(1), 1–32.

Zhang, L., & Ness, E. (2010). Does state merit-based

aid SETM brain drain? Educational Evaluation and

Policy Analysis, 32, 143–165.

Zipf, G. K. (1949). Human behavior and the principle

of least effort: An introduction to human ecology.

Cambridge, MA: Addison-Wesley.

Authors

THOMAS J. COOKE is a Professor in the Depart-

ment of Geography and Director of the Center for

Population Research at the University of Connecticut,

Storrs, CT 06269. E-mail: thomas.cooke@uconn.

edu<mailto:[email protected]>. His research

interests are in the geographic and spatial operation of

labor markets: How people search for, commute, and

migrate for employment and how these processes

impact both communities and individuals.

PAUL BOYLE is Chief Executive of the Economic

and Social Research Council and a Professor of Geog-

raphy at the School of Geography and Geosciences,

University of St Andrews, St Andrews, KY16 9AL

Scotland; P.Boyle@st-andrews.ac.uk<mailto:P.

Boyle@st-andrews.ac.uk>. His research falls within

the fields of Demography and Epidemiology, having

published mainly on migration, fertility, mortality, and

health inequalities.

at UNIV OF CONNECTICUT on May 4, 2011http://eepa.aera.netDownloaded from