Who goes early?:
A multi-level analysis of enrollment via early action and early decision admissions
DRAFT COPY
Please do not reproduce or circulate without the permission of the authors.
Julie J. Park and M. Kevin Eagan
The authors would like to thank the USC Center for Research on Enrollment, Policy, and Practice for
their generous support of our research.
1
Introduction
“I think there are lots of very talented students out there from poor and moderate-income
backgrounds who have been discouraged by this whole hocus-pocus of early admissions by many
of the nation’s top colleges,” -William R. Fitzsimmons, Harvard College’s dean of admissions
and financial aid upon Harvard’s announcement to end its early action program (qtd. in Finder &
Arenson, 2006).
“Penn plans to continue with its successful early-decision plan…It's worked well for us, and the
quality of the students we're getting is exceptional.” -Lee Stetson, University of Pennsylvania
dean of admissions (qtd. in Yale Alumni Magazine, 2002).
As these two quotations show, top institutions continue to disagree on the value of early
admissions programs. In recent years, early action and early decision programs in college
admissions have received greater scrutiny as some institutions, including Stanford and Yale,
have moved from early decision to early action and others such as Harvard have ended all early
admissions. Early decision and early action make up the two types of early admissions programs
offered by a number of higher education institutions. Deadlines for both types of programs come
earlier than regular decision deadlines for applications. In early decision, students must sign an
agreement that they will attend the institution if accepted. Because of the binding nature of early
decision programs, students are urged to only apply to one school under early decision. Early
action programs are non-binding; students are not required to attend if accepted, but they receive
earlier notification of acceptance, denial, or deferral (Avery, Fairbanks, & Zeckhauser, 2003).
Although research has investigated the types of institutions that tend to offer both early
action and early decision programs, the types of students who apply to these programs, and the
types of high schools that they come from (Avery, et al., 2003), no study has examined these
2
three contexts in unison. Drawing from frameworks of cultural capital and socialization theory,
this research employs cross-classified hierarchical generalized linear modeling (CCHGLM) to
examine how high school, college, and student-level characteristics affect a student’s decision to
apply to college via early action or early decision programs.
Literature Review
Research on student enrollment via early decision or early action is critical due to the
policy’s growing role in the college admissions process. The percentage of students who reported
that being admitted through early action or early decision was an important factor in them
choosing a college has increased steadily from 6.9% in 1999 to 10.9% in 2006 (Pryor, Hurtado,
Sáenz, Santos & Korn, 2007). Although these figures indicate a recent increase in the proportion
of students enrolling in higher education via early action and early decision programs,
institutions have used such practices for decades. Beginning in the mid-1950s, institutions such
as Harvard, Princeton, and Yale introduced versions of early action and early decision programs
to gain advantages over their competitors for attracting the best students (Avery et al., 2003).
The primary reason for the controversy over early admissions programs is whether they
give an unfair advantage to students who tend to be from higher income families or schools with
greater resources -- students who already have a leg up in the selective college admissions race.
Because students have to agree to attend an institution if accepted through early decision,
students who can afford to attend a college regardless of the financial aid package have a greater
incentive to apply early (Lucido, 2002). Some institutions have responded to this criticism by
replacing early decision programs with non-binding early action programs (Farrell, 2006, Flores,
2002). Still, due to having to submit college applications earlier in the year, early action
3
programs may also attract students who tend to have higher levels of college-going knowledge or
financial resources.
While colleges often tout that early admissions programs allow them to secure a group of
highly motivated students who are committed to attending their institution as a key motivator for
using early admissions program, another key reason why institutions use early admissions
policies is for the purpose of enrollment management. Prior to 2003, the U.S. News and World
Report college rankings included yield, and early decision was criticized because colleges were
thought to have an incentive to use such programs in order to increase yield. However, in 2003
the magazine dropped yield in its calculations of rankings (Steinberg, 2003). Still, early
admissions programs generally assist institutions in their enrollment projections for the incoming
class of students by pushing up the timetable for knowing which students will attend or are
accepted in a time when students are applying to more colleges.
Early admissions programs are also highly attractive to students who are able to take
advantage of them. Besides the comfort of knowing whether one is accepted into an institution
earlier in the year, the cohort of students that applies early often has a higher acceptance rate than
students who apply at the regular deadline. Fallows (2001) notes that for students who applied to
college during the 1999-2000 school year that Yale admitted 37% of early applicants but only
16% of regular applicants; Amherst 35% of early applicants and 19% of regular applicants, and
the University of Pennsylvania 47% of early applicants and 26% of regular applicants. While
universities may argue that the acceptance rates are skewed due to a stronger, more motivated
pool of students who apply early, does applying early actually offer an advantage?
Avery et al. (2003) answered this question by conducting the most extensive study on
early admissions policies to date. Using student records from applicants to 14 highly selective
4
institutions, they found that students who applied to the early action institutions had SAT scores
about 10 to 20 points greater than students who applied regular, while there was no notable
difference between students who applied through early decision and those who applied regular.
In fact, early decision applicants had SAT scores that were slightly lower than regular applicants.
Thus Avery et al. (2003) did not find evidence to support the claim that the pool of students who
applied via early admissions was notably stronger than the regular pool, although the acceptance
rate was notably higher for students who applied early. They concluded that students who
applied to college under early decision or early action programs had a substantial advantage in
the admissions process equivalent to approximately a 100-point increase in a student’s
Standardized Aptitude Test (SAT). Using multiple regression, they found that applying through
early action increased an average applicant’s chances of being accepted by at least 15% at 11 of
the 14 institutions and that applying early decision increased chances by at least 25% at the
institutions that offered early decision.
Research has also demonstrated that more affluent students have both the financial
resources to apply early as well as an understanding of the possible advantages of applying early
(Avery, Fairbanks, & Zeckhauser, 2000, 2001; Ehrenberg, 2003). Given that students from
higher socioeconomic backgrounds typically have a number of advantages already, such as the
ability to pay for college and stronger academic preparation, further privileging these students
via early admissions programs raises questions regarding the equity of such policies. The Journal
of Blacks in Higher Education (1999) found that few Black students apply through early
decisions programs, citing the inability to compare financial aid packages as the primary reason.
Avery et al. (2005) found similar results, finding that early applicant pools tend to be
disproportionately White and from higher income families.
5
Significance
Although the institutions that offer early admissions programs serve a relatively small
sector of students in the entire postsecondary education system, the high value and prestige
attached to a college diploma from an elite institution cannot be understated. Graduates from
these institutions make up a disproportionate number of leaders in society. The entire industry
that has sprung up around the college applications process reflects the high amount of
competition for these limited spots. While some institutions have dropped early admissions
programs because of equity concerns, or have switched over to early action from early decision,
the programs continue at a number of selective and highly selective institutions. When early
admissions offer an advantage in the college applications process to a group that is already
substantially advantaged, further research on the dynamics of early admissions programs using
national, comprehensive datasets is critical.
Prior research on early decision programs examined factors affecting student behavior
almost solely from an individual student perspective, identifying patterns such as a student’s
income or race/ethnicity being related to patterns of applying early. This approach tends to
ignore how larger contextual effects from both students’ high schools and their prospective
higher education institutions may affect a student’s ability and eventual decision whether to
apply to an institution via early action or early decision. High school resources, such as the
number of counselors and Advanced Placement (AP) courses available to students, have
significant weight in the college application process (McDonough, 2005). Similarly, students’
high school peers may affect their college choice process, as research consistently has
demonstrated the strong influence of peer effects (Weidman, 1989). At the same time, college
characteristics likely play a role in a students’ decision to apply early. Students aspiring to enroll
6
at more selective colleges, which often feature lower acceptance rates, may apply early to college
to increase their chances of gaining admission to their first-choice institution (Ehrenberg, 2003).
Our study takes all three of these contexts into account while drawing on an unprecedented
dataset that provides information about students’ individual characteristics, as well as high
school and postsecondary institutional characteristics.
Conceptual Framework
Drawing from the research and its associated limitations noted above, this study
incorporates the frameworks of socialization theory and cultural capital to understand the
individual and institutional forces shaping students’ decisions to enroll in college through being
accepted via early admissions. Weidman (1989) suggests that students experience normative
effects from a variety of forces, including their peers and teachers. These different groups apply
normative pressures that prompt students to reconsider their beliefs and values, as students
compare themselves to their peers. Although Weidman bases his theory on undergraduate
college students, his framework may apply to high school students who are influenced by their
classmates, parents, or counselors not just to apply to colleges but to apply to colleges that
maintain a certain level of prestige.
We are also influenced by work on cultural capital that might explain why certain
populations of students are more likely to apply and enroll via early action or decision. Cultural
capital refers to how privilege tends to accumulate within certain cultures or subgroups due to
the generational transference of resources, attitudes, or knowledge (Bourdieu & Passeron, 1977).
We look at applying early as a form of cultural capital in three ways. First, students from higher
socioeconomic backgrounds may have the material resources to invest in their college
application processes, purchasing services such as private college counselors who may offer
7
advice on the advantages of applying early (McDonough, Korn, & Yamasaki, 1997). They are
more likely to attend well-resourced public or private high schools that provide college
preparatory tracks and counseling that make them attuned to the advantages of applying early. A
second, closely related way that applying early may correspond with an elevated level of cultural
capital is that, to apply early, students must have the necessary college-going knowledge, which
is circulated more widely in certain environments or high school settings. Coupled with this form
of college-going knowledge may be information on other strategies that help students navigate
the selective institution application process that are linked to applying early, such as the
advantages of taking standardized tests earlier or multiple times. This knowledge may come from
sources such as their peer environments, academic settings, or families. Finally, students’
cultural capital relates to an individual’s decision to apply via early action or early decision in
that students with higher cultural capital generally have an overall advantage in a highly
competitive college admissions environment. Early admissions policies can perpetuate the
advantages that higher income students already have in the college admissions process, giving a
privilege to the already privileged.
Using these conceptual frameworks and prior research, this study seeks to understand the
individual, high school, and college characteristics that influence students’ decisions to enroll in
college via early action or early decision programs. Specifically, this study intends to address the
following research questions:
1. Controlling for students’ background characteristics, do student-level characteristics such
as their perceptions of their academic abilities and receipt of private college counseling
affect their decision to enroll in college via early action or early decision programs?
8
2. Controlling for student-level characteristics, does the rate at which students enroll via
early admission programs vary across high schools? If so, do the number of AP courses,
the number of counselors, and the college-going rate of students from the high school
account for this variation?
3. Controlling for student-level characteristics, does the rate at which students enroll via
early admission programs vary across higher education institutions? If so, do factors such
as region, control, institutional expenditures, and selectivity account for some of this
variation?
Methods
Data and Sample
This study utilizes student, high school, and college data collected by the Cooperative
Institutional Research Program (CIRP) within the Higher Education Research Institute (HERI).
In 2005, more than 300,000 first-year college students completed the Freshman Survey, which
asked students questions regarding high school involvement, aspirations, and attitudes, among
other things. Data from the Freshmen Survey were merged with the 2000 College Board High
School Survey, which includes information on the high schools that the survey respondents had
attended. We also merged college and university data collected from the Integrated
Postsecondary Education Data System (IPEDS) into the same dataset. The final analytic sample
for this study features more than 175,163 students originating from 5,607 high schools and
situated within 543 colleges and universities across the U.S.
Variables
We derived the dependent variable from an item on 2005 Freshman Survey that asked
students how important being admitted through an early admission program was in their ultimate
9
decision to attend their particular college or university. Students had the option of choosing “not
important,” “somewhat important,” or “very important.” We coded students responding
“somewhat important” or “very important” as having enrolled via an early admission program;
conversely, we coded students who responded to this item by selecting “not important” as not
having enrolled via an early admission program. This coding scheme resulted in a dichotomous
dependent variable that served as a proxy as to whether a student enrolled via early admission
program.
We included a number of student-level variables in the model. We controlled for
demographic characteristics such as race, gender, parental income, and parental education. We
used dichotomous race variables representing Black, American Indian, Asian American, Latino,
and “other race” students, respectively, with White students representing the reference group.
The dichotomous gender variable had men as the reference group. Income represented a selfreported variable from the survey and ranged from 1 (less than $10,000) to 14 ($250,000 or
more). The parental education variable controlled for the highest level of education earned by
either parent and ranged from 1 (grammar school or less) to 8 (graduate degree). Finally, we
controlled for the extent to which students expressed concerns about their ability to finance their
college education.
In addition to the demographic characteristics, we controlled for key experiences that
students had while enrolled in high school as well as students’ reasons for pursuing a college
degree. We included in the model the extent to which students’ parents and counselors
influenced their decision to enroll in college. We also examined whether a desire to prepare for
graduate school had a significant association with students’ decision to enroll via early action or
early decision programs. Additionally, we created two factors that refer to students’ academic
10
self-confidence as well as students’ commitment to academic and professional success. The
academic self-confidence factor included five items from the Freshman Survey: academic ability,
drive to achieve, mathematical ability, intellectual self-confidence, and writing ability. The
academic self-confidence factor had a Cronbach alpha of 0.62, which falls slightly below the
recommended reliability threshold for the social sciences (Pehauzar & Schmelkin, 1991). We
included four items related to students’ goals in the construction of students’ commitment to
future success. These four items included: becoming an authority in my field, obtaining
recognition from my colleagues for contributions to my special field, being well off financially,
and becoming successful in a business of my own. The Cronbach alpha for this factor was 0.66.
The final block of student-level variables in the model controlled for students’ intended
major in college. We had dichotomous variables referring to the following groups of majors:
science, technology, engineering, and math; arts and humanities; social sciences; and the
professions, which include business and education. Students who were undecided in their
intended major comprised the reference group.
In addition to student-level variables, we examined several high school characteristics.
We controlled for the type of high school a student attended: private Catholic, private other
religious affiliation, and private no religious affiliation compared to public high schools. The
data included the counselor-to-student ratio, calculated by dividing the number of counselors in a
high school by the total enrollment of the high school Additionally, the model included a
variable measuring the total number of AP courses available to students within a particular high
school. These variables offer insight into the type of resources and opportunities available to
students for advanced academic preparation and likely, a more informed college search process.
Finally, we controlled for the region of the country in which the high school was located.
11
Our last grouping of variables examined the influence of specific college and university
characteristics that might have prompted students to seek enrollment via early action or early
decision programs. We controlled for public institutions compared to private and also included a
control for the geographic region of the institution. Key variables of interest included
institutional expenditures on research, instruction, and financial aid per full-time-equivalent
student. Additionally, we examined the association between students’ decision to enroll via early
action or early decision programs and institutional selectivity, as measured by average
Standardized Aptitude Test (SAT) scores. The expenditure and selectivity data serve as proxies
for institutional prestige, which prior research suggests serve as a factor in student rationale for
applying to and enrolling in colleges via early decision programs (Ehrenberg, 2003).
Analyses
The complex dataset used in this study provided an opportunity to consider the unique
effects of college, high school, and individual factors affecting students’ decision to enroll via
early action or early decision programs. To account for these unique effects, we employed crossclassified hierarchical generalized linear modeling (CCHGLM). This advanced statistical
technique accounts for the multiple clustering effects of both students within high schools as well
as students within colleges and universities (Raudenbush & Bryk, 2002). This type of analysis
enables us to partition group-level variance between students’ high schools and their colleges
and universities. Additionally, this technique represents an appropriate method to examine the
effects of independent variables on a dichotomous dependent variable. Using CCHGLM enables
us to simultaneously consider how student-level traits and experiences, college variables, and
high school characteristics interact together to facilitate or discourage students' decision to enroll
in a four-year higher education institution via early action or early decision. Most current
12
research draws only from student and college data or from student and high school variables, or,
in the most limited of studies, from only student-level data. Additionally, other studies that have
examined students’ decisions related to early action or early decision programs have tended to
use single-level statistical techniques, which ignore the clustered nature of the data, to estimate
the multi-level effects influencing students' decisions to enroll under early action and early
decision programs. By using the advanced techniques offered through CCHGLM, we link the
environmental contexts of students' high schools and higher education institutions with students'
individual characteristics to provide the most comprehensive depiction of the factors affecting
students' decisions to enroll in higher education via early decision and early action programs.
When constructing a hierarchical generalized linear model, it is important to begin with
the null model, also known as the fully unconditional model since it has no predictors. Because
of the dichotomous nature of the dependent variable, the level-1 sampling model is Bernoulli, as
represented in Equation 1:
Prob (Yij = | βij) = Φij
(1)
Generally, use of hierarchical linear modeling techniques requires consideration of the extent to
which the outcome measure varies across groups (Raudenbush & Bryk, 2002). In models with a
continuous outcome, the variation attributed to groups can be found through the intra-class
correlation (ICC). Because of the hetereoscedasticity of the variance in the outcome at level-1,
the ICC is not instructive for our model. Instead, we examined box-plots of Empirical Bayes
(EB) residuals to determine the extent of variation across high schools and across colleges in
students’ average likelihood to enroll via early decision or early action. Seeing substantial
variation across both types of institutions (colleges and high schools), we proceeded to build the
level-1 and level-2 models in CCHGLM. Equation 2 represents the student-level model:
13
 ij 
Log 
 = π0jk + π 1jk * (Female)ijk + π 2kj * (Race)ijk
1  ij 
(2)
+ π 3jk * (High School GPA)ijk + π4jk * (Income) + π5jk * (Parental Support) ijk
+ π6jk * (Prepare for grad school) ijk + π7jk * (Time spent talking with teachers) ijk
+ π8jk * (Financial concerns) ijk + π9jk * (Private counselor advice) ijk
+ π10jk * (Parental education) ijk + π11jk * (Intended major) ijk + π12jk * (SAT) ijk
+ π13jk * (Academic self-confidence) ijk + π14jk * (Commitment to academic success) ijk
+ eijk
where π0jk corresponds to the average likelihood of a student enrolling in college via early action
or early decision at college j from high school k. The level-2 model is given by Equation 3:
π0jk = θ0 + γ01 * (Instructional Expenditures)j + γ02 * (Research expenditures)j (3)
+ γ03 * (Region)j + γ04 * (Public)j + γ05 * (Selectivity)j
+ γ06 * (Cost of attendance)j + γ07 * (Average aid per student)j
+ β01 * (Total enrollment)k + β02 * (Number of AP courses)k
+ β03 * (High School Type)k + β04 * (Counselor-Student Ratio)k
+ β05 * (Region)k + b00j + c00k
where γ corresponds to the fixed effect for any college-level variable at institution j, β represents
the fixed effect of any high school variable at high school k, and b00j and c00j represent the
residual random effects of colleges and high schools, respectively, on π0jk after taking into
account the predictors included in the model (Raudenbush & Bryk, 2002). We constrained all
level-2 parameters to be fixed across all colleges and schools. Similarly, we did not allow level-1
parameters to vary across high schools or across colleges.
14
Finally, utilization of CCHGLM necessitates consideration of centering. We left all
dichotomous variables in the model un-centered. We centered the ordinal and continuous
variables in our model around their grand mean (Raudenbush & Bryk, 2002).
Limitations
This study has several limitations. First, because we relied on secondary data in our
analyses, we were limited by the variables and definitions included in the dataset. The survey did
not include some data that may be helpful to understanding factors that might influence a
student’s decision to apply early. Additionally, we relied on a derived survey item to construct
our dependent variable. Because we relied on this self-reported item that asked students about
the importance of early action or early decision in their choice to enroll in this institution, we
may have missed students who indeed enrolled in college via early action or early decision but
did not rate that program as important. We also were not able to account for the much broader of
phenomena of students who applied via early admissions, but did not end up attending a school
that they had applied to early. Second, our data do not constitute a nationally representative
sample, which may limit the generalizability of our findings. Finally, although CCHGLM has a
number of advantages over single-level logistic regression, it requires a substantial amount of
data. Because of this limitation, we had to eliminate high schools and their associated students
from our sample if fewer than five students from a particular high school had completed the
Freshman Survey.
Results
Descriptive statistics
Table 1 presents descriptive statistics for the individual, college, and high school
variables included in the study. Descriptive analyses of the data indicated that 25% of the sample
15
had enrolled via early action or early decision programs. Analyses show that 80% of our sample
identified as White, 9% as Asian American, 7% as Latino, and 7% as Black. Approximately 57%
of students in our sample identified as female, and the average self-reported high school GPA
was a B+. Students appeared modestly concerned about their ability to finance their college
education, and the average student came from a well-educated (at least one parent with a college
degree) household.
The high schools in the study had an average of 1 college counselor for every 100
students; however, some high schools appeared to have no college counselors on staff. On
average, high schools reported offering just fewer than 9 AP courses. Approximately 80% of
high schools in the sample were public. Average enrollment among the high schools hovered just
below 1,200 students. Finally, the highest concentration of feeder high schools in our sample was
located in the Midwest, and the lowest concentration of high schools was in the Northeast.
Among the colleges and universities, institutions spent an average of $9,178 per FTE
student on instruction and $1,883 per FTE on research. Only 23% of colleges and universities in
the sample were publicly controlled. The colleges and universities in our sample had a moderate
level of selectivity, as the mean SAT score of entering students was 1,110 for the average
institution.
Cross-classified hierarchical generalized linear modeling results
We report the results from the analyses in delta-p statistics for easier interpretation in
Table 2, and we use the method recommended by Petersen (1985) to calculate the delta-p
statistics. The student-level predictor with the strongest association with students’ decision to
enroll via early action or early decision programs was having received advice from a private
college counselor. Students who reported choosing their particular higher education institution
16
based on the advice of a private college counselor appeared to be approximately 19% more likely
to enroll via early action or early decision programs compared to their peers who did not receive
such advice.
The results in Table 2 also suggest that Black (delta-p = -3.05%), Asian American (deltap= -2.53%), and Latino (delta-p = -1.65%) students had significantly lower probabilities of
enrolling via early action or early decision programs compared to their White classmates. Aside
from controls for race, the results suggest that many demographic characteristics had significant
yet marginal associations with students’ decision to take advantage of early action or early
decision programs. Women appeared to have a slightly higher probability (delta-p = 0.95%) than
their male counterparts to enroll in college through early action or early decision programs.
Parental income (delta-p = 0.38%) and parental education (delta-p = 0.38%) had significant,
positive, albeit modest, associations with students’ likelihood of enrolling via early action or
early decision programs.
Considering students’ academics, the results indicate that as students’ high school grade
point average increases, their probability of enrolling in college via early action or early decision
significantly increases (delta-p = 1.14%). We found a significant negative association between
students’ composite SAT score and their likelihood to enroll via early action or early decision.
For every 100-point increase in composite SAT scores, students experienced a 1.65% reduction
in their probability of enrolling in college via early action or early decision.
Students who reported receiving encouragement from parents to pursue a college
education (delta-p = 1.53%) and students who wanted to go to college to prepare for graduate
school (delta-p = 2.32%) had significantly higher probabilities of enrolling in college via early
action or early decision compared to their peers who received less support from their parents or
17
who did not feel drawn to college to prepare for graduate school. We found a significant and
positive association between students’ academic self-confidence and their decision to enroll via
early action or early decision (delta-p = 0.38%); however, this effect appears to be marginal
compared to other predictors in the model. Likewise, we see a significant yet marginal
association between students’ commitment to academic and professional success and their
decision to enroll via early action. Finally, the results indicate that, for many students, knowing
their intended major prior to starting college significantly improved their chances of enrolling via
early action or early decision. Students who planned to major in professional fields (delta-p =
2.12%) or in the science, technology, engineering, and math fields (delta-p = 1.34%) were
significantly more likely to enroll via early action or early decision programs compared to their
classmates who were unsure of their major at the start of college.
Among high school variables, the strongest association emerges between a school’s
counselor-student ratio and students’ decision to enroll in college via early action or early
decision. For every 1% increase in a high school’s counselor-student ratio (in other words,
adding one counselor for every 100 students in a school), students become 1.14% more likely to
take advantage of early action or early decision programs. Also, each additional Advanced
Placement course offered by a high school increases students’ probability of participating in an
early action or early decision program by 0.38%. In other words, students at an average high
school (in this case, a high school offering approximately 9 Advanced Placement courses) have a
3.42% higher probability of participating in early action or early decision programs compared to
their peers at high schools without any Advanced Placement courses. Students who attended a
private, non-religious high school had a 0.56% lower probability of enrolling in college via early
action or early decision compared to their counterparts in public high schools. Individuals
18
enrolled in high schools located in the mid-Atlantic region were 3.32% more likely to enroll in
college via early decision or early action than their peers in the Northeast. Conversely, students
who attended high school in the western U.S. were 2.88% less likely to participate in early action
or early decision than their peers in the northeast.
Considering college-level variables, we find a significant and positive association
between institutional selectivity and students’ probability of enrolling via early action or early
decision. A 100-point increase in institutional selectivity, measured by the average SAT scores
of entering students, corresponded to a 3.12% percent increase in students’ probability of
participating in early action or early decision. Thus, students in the most selective institution in
this sample (SAT composite average = 1,510) were 22.9% more likely to take part in early action
or early decision programs than their peers in the least-selective institution in our sample (SAT
composite average = 775). Results in Table 2 suggest that as average institutional financial aid
per FTE student increases, students’ likelihood of taking advantage of early action or early
decision also increase. For every $1,000 increase in financial aid per FTE student, individuals
see an increase of 0.57% in their probability of participating in an early action or early decision
program. Finally, similar to the high school variables, we found a significant association between
regional affiliation of colleges and students’ likelihood to take advantage of early action or early
decision. Students enrolling in colleges and universities in the Great Lakes (delta-p = -3.39%)
and the Plains (delta-p = -6.74%) regions were significantly less likely than their peers attending
college in the Northeast to enroll via an early action or early decision program.
Looking at the model statistics, we find that the college-level variables explained
approximately 30.19% of the variance in students’ decision to participate in early action or early
decision programs across colleges. Additionally, we find that our high school variables explained
19
83.56% of the variance in early decision or early action enrollment occurring across high schools
in the sample. Overall, our level-2 model explained approximately 33.23% of the variance
occurring across colleges and high schools. Because of the heteroscedasticity of the variance at
level-1, we are unable to provide an accurate estimation of the variance or explained variance
occurring at the student level.
Discussion
By utilizing a unique database that combines high school institutional characteristics, student
attitudes and behaviors, and characteristics of the colleges attended by students, we have an
unprecedented opportunity to better understand the phenomena of early decision and early admission
in college admissions. Perhaps our most significant finding is the role of private college counseling;
students who received advice from a private college counselor were 19% more likely to enroll via
early action or early decision compared to their peers who did not report having access to this
resource. While previous research has posited that students from wealthier families are more likely to
apply early, we suggest that students from more affluent backgrounds have the resources to purchase
services like private college counseling. With this caveat in mind, even after controlling for factors
related to socioeconomic status, such as parental education, family income, and students’ concerns
about financing their college education, we continue to see a strong, positive association between
receipt of private college counseling and students’ decision to enroll via early action or early
decision. The strong influence of private college counseling becomes even more important when
considering two other individual predictors that may serve as proxies for students’ cultural capital:
parental encouragement for college and students’ desire to enroll in college to prepare for graduate
school. Aside from racial characteristics, these three variables, which loosely relate to students’
20
cultural capital, represent three of the strongest influences on students’ decision to enroll via early
action or early decision programs.
Earlier we suggested that early admissions functions as a form of cultural capital that works
to reproduce privilege in three closely related ways. First, students from higher socioeconomic
backgrounds have parents who can purchase resources such as private college counseling who may
alert them to the benefits of applying early. Second, knowledge around early admissions is a form of
college-going knowledge that may be transmitted through peer networks or knowledge that families
acquire about how to navigate the college applications process. Third, early admissions offer a very
tangible advantage to applicants, akin to a 100 point score increase on the SAT. These three means
may overlap substantially, for instance, a student may acquire college-going knowledge about the
way that early admissions may offer an advantage to applicants. In our study, while parental
education and income had modest, significantly positive effects on student enrollment through early
admissions, the effect of these variables may be mediated through two forms of cultural capital that
serve to advantage students in the college applications process: the resources that wealthier parents
can buy for their children such as private college counseling and knowledge that more educated
parents tend to have about navigating the college application process. The relationship between a
student’s desire to enroll in college to prepare for graduate school also suggests that such a student is
attuned to the advantage that early admissions offers, akin to approximately a 100-point increase on
the SAT. Students who already thinking ahead to graduate school may have acquired college-going
knowledge on how they can have a hand up in the selective college admissions game that will
strategically position them to meet their goals around graduate education.
We also see this link between cultural capital and students’ decision to enroll via early action
or early decision among our high school variables. Students attending more resource-rich high
21
schools, as measured by the counselor-student ratio and number of Advanced Placement courses
available, appear to have a significant advantage over their peers attending less-affluent schools.
High schools with more college counselors per student provide their students with increased
opportunities to obtain guidance from college counselors, and college counselors are able to focus
their attention on assisting fewer students. Similarly, high schools that offer more Advanced
Placement courses provide their students with several advantages: students have the potential to earn
more college credits while enrolled in high school; students in these courses likely experience smaller
classes and have greater access to their teachers, and students in these courses generally have an
opportunity to augment their grade point average, as Advanced Placement courses generally award
weighted grade points because of the rigor found in the curriculum.
Thus at the high school level, it appears that attending a better-resourced high school
increases the probability that a student will enroll through early admissions. Once again, privilege
and the cultural capital that comes through privilege are mutually reinforcing. We know that students
who attend such high schools tend to come from higher socioeconomic backgrounds. The privileges
of attending a well-resourced high school are reinforced when such resources are channeled into
college-going knowledge that serves students in the overall college application process or through
strategies like applying early that can advantage students in the admissions process.
However, resources and organizational structures are not the only way that high schools
facilitate the college-going process; high schools also cultivate normative practices and expectations
in this area (Falsey & Heyns, 1984; Hill, 2008; McDonough, 1997). The case of early admissions is
an example of how structure and norms interact to inform students’ college-going knowledge and
behavior. Structures such as low counselor-student ratios and greater offerings of Advanced
Placement courses point to a greater amount of resources devoted to assisting students in the
22
admissions process and to positioning students to be academically prepared for college. Furthermore,
such resources may create a culture of college-going norms where practices such as applying early
are seen as part of one’s college application strategy and reinforced through information students
receive from counselors and peer networks that are cultivated in environments such as Advanced
Placement classes. As Hill (2008) asserts, few studies have systematically examined the dynamics
that exist between and within high schools in the area of college-linking strategies that reflect both
resources and normative practices, and the way that awareness about early admissions practices is
cultivated within high schools is an area that needs additional examination.
The role of parents, private college counselors, and high school college counselors also
highlight the effect of the different forces that influence students to reconsider their behaviors
and decisions (Weidman, 1989). Unfortunately we were not able to test variables that tap into the
role of peer influence on the college decision making, though we suspect that peers play an
important role in alerting one another to means of navigating the college admissions process.
Future research in this area is much needed.
Recommendations and Implications for Policy
In the end, do our findings do more to support the comments of Fitzsimmons or Stetson?
In a sense, both are true. Early admissions policies have worked to attract a pool of applicants
who tend to be better financially resourced, and early admissions has also worked to attract
talented, high quality students. Like previous research, we found that those who enroll due to
early admissions tend to be White, with higher family incomes and parents with greater levels of
education. We make a unique contribution to the literature by simultaneously examining three
levels of context: the student, the high school, and the higher education institution. Ultimately we
identified that enrollment via early admissions policies is linked to access to resources that
23
individuals with greater financial means tend to have access to: private college counseling, high
schools with low college counselor-to-student ratios, and academically rigorous curriculum.
Given the underrepresentation of low-income students at selective colleges and
universities and the many barriers that these students face to accessing higher education, we find
it troubling that many universities employ a policy that tends to work as another sorting
mechanism in a higher education system that is already stratified by race and class. We join the
colleges and universities that have already ended such policies to recommend that higher
education act to level the playing field by ending early admissions. We realize that in a time
period where schools are increasingly worried about how to project yield and allocate financial
resources that not all institutions have the luxury that a place like Harvard has in being able to
end early action. Avery et al. (2003) point out that even if early admissions was eradicated that
certain students would still face advantages in the admissions race, be it through ways that
universities identify students with high interest or communication with college counselors at elite
high schools, and they identified other possible recommendations to reform admissions policies
such as a “Gold-Star-Only” system where applicants would rank their top preferences in order to
indicate interest and commitment. At a minimum, institutions need to look inward and ask
serious questions about the patterns of who applies and is accepted early, and the implications of
offering advantages to students who generally already are advantaged in the admissions process.
There also needs to be more systematic research conducted on trends of applicants, accepted, and
attending students for institutions that ended early admissions policies or switched from early
decision to early action to assess the impact of the policy on applicant and enrollment trends.
Ending or reforming early admissions policies will have little effect on making the
overall higher education system more equitable without greater change in the K-12 educational
24
system, especially at the high school level. In her study of college-linking strategies used by high
schools, Hill (2008) identified how schools that she labeled “brokering schools” were
characterized by dedicating a high amount of resources to facilitate college attendance, but also
by engaging parents in the college-going process and acting as a broker of its resources to
families and students. Such schools see themselves as playing an active role in getting students to
college, rather than simply providing students with information on the college-going process that
students can choose to take or ignore. Other intervention strategies include college outreach
programs and broadening access to rigorous high school curriculum. Clearly, there are many
changes that we need to be made, but we recommend that institutions begin by seriously
examining the equity of the policies that they already have in place.
References
Avery, C., Fairbanks, A., & Zeckhauser, R. (2000). What worms for the early bird: Early
admissions at elite colleges. Cambridge, MA: John F. Kennedy School of Public Policy.
Avery, C., Fairbanks, A., & Zeckhauser, R. (2001). Joining the elite: The early admissions game.
Cambridge, MA: John F. Kennedy School of Public Policy.
Avery, C., Fairbanks, A., Zeckhauser, R. (2003). The early admissions game: Joining the elite.
Cambridge: Harvard University Press.
Bordieu, P., & Passerson, J. (1977). Reproduction in education, society, and culture. London:
Sage.
Ehrenberg, R. G. (2003). Reaching for the brass ring: The U.S. News and World Report rankings
and competition. The Review of Higher Education, 26(2), 145-162.
25
Fallows, J. (2001, September). The early-decision racket. The Atlantic Monthly. Available at
http://www.theatlantic.com/doc/print/200109/fallows
Falsey, B. & Hayns, B. (1984). The college channel: Private and public schools reconsidered.
Sociology of Education, 57, 111-22.
Farrell, E. (2006, September). Princeton drops its early–admission option. The Chronicle of
Higher Education. Available at http:/chronicle.come/daily/2006/09/2006091903n.html
Retrieved April 25, 2008.
Finder, A. & Arenson, K. (2006, September 2006). Harvard ends early admission. The New
York Times. Available at http://www.nytimes.com/2006/09/12/education/12harvard.html
Retrieved September 10, 2008.
Flores, C. (2002, May 3). U. of North Carolina at Chapel Hill drops early-decision admissions.
Chronicle of Higher Education, p. A38.
Hill, D.H. (2008). School strategies and the “college-linking” process: Reconsidering the effects
of high schools on college enrollment. Sociology of Education, 81(1), 53-76.
The Journal of Blacks in Higher Education. (1999). Why few Blacks apply for early admission.
The Journal of Blacks in Higher Education, 24, 66-68.
Lucido, J. (2002). Eliminating early decision: Forming the snowball and rolling it downhill. The
College Board Review, No. 197, 4-29.
McDonough, P. (2005). Counseling matters: Knowledge, assistance, and organizational
commitment in college preparation. In W.G. Tierney, Z.B. Corwin, & J.E. Colyar (Eds.),
Preparing for college: Nine elements of effective outreach. Albany, NY: State University
of New York Press.
26
McDonough, P. (1997). Choosing colleges: How social class and schools structure opportunity.
Albany: State University of New York Press.
McDonough, P., Korn, J., & Yamasaki, E. (1997). Access, equity, and the privatization of
college counseling. The Review of Higher Education, 20(3), 297-317.
n.a. (2002, December). College will end binding early decision. Yale Alumni Magazine.
Available at http://www.yalealumnimagazine.com/issues/02_12/l_v.html#1
Pryor, J.H., Hurtado, S., Saenz, V.B., Santos, J. L., & Korn, W. S. (2007). The American
Freshman: Forty Year Trends. Los Angeles: Higher Education Research Institute.
Steinberg, J. (2003, July 10). College rating by U.S. News drops factor in admissions. The New
York Times. Available at
http://query.nytimes.com/gst/fullpage.html?res=9C03E7D8133DF933A25754C0A9659C
8B63
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods (2nd edition). Thousand Oaks, CA: Sage Publications.
Weidman, J. C. (1989). Undergraduate socialization: A conceptual approach. In J. Smart (ed.),
Higher education: Handbook of theory and research (Vol. 5). New York: Agathon.
27
Table 1
Descriptive statistics of variables included in the analyses
N
Dependent variable
Enrolled via early action or early decision program
Student-level variables
Female
Black
American Indian
Asian American
Latino
Other race
High school GPA
Parental income
Reason for college: parents wanted me to go
Reason for college: to prepare for graduate school
Mean
S.D.
Min.
Max
175163
0.25
0.43
0.00
1.00
175163
175163
175163
175163
175163
175163
175163
175163
175163
175163
0.57
0.07
0.02
0.09
0.07
0.03
6.38
9.35
2.22
2.46
0.50
0.26
0.12
0.28
0.25
0.16
1.42
2.89
0.76
0.71
0.00
0.00
0.00
0.00
0.00
0.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
8.00
14.00
3.00
3.00
Hours per week spent talking with teachers outside
class
175163
2.56
1.03
1.00
8.00
Concerns about being able to finance college
education
175163
1.77
0.65
1.00
3.00
Chose this college because a private college
counselor advised me
175163
1.16
0.43
1.00
3.00
175163
175163
175163
175163
175163
175163
175163
175163
6.11
0.25
0.13
0.12
0.42
11.81
0.01
0.00
1.75
0.43
0.33
0.32
0.49
1.57
1.00
1.00
1.00
0.00
0.00
0.00
0.00
4.00
-5.13
-2.58
8.00
1.00
1.00
1.00
1.00
16.00
2.28
2.04
543
543
543
543
543
543
543
543
543
543
543
9.18
1.88
0.12
0.25
0.15
0.10
0.23
0.05
0.02
0.10
0.23
6.81
6.25
0.32
0.43
0.35
0.30
0.42
0.21
0.13
0.30
0.42
2.60
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
71.85
84.28
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
Highest level of education attained by either parent
Major - professions (education, business)
Major - social sciences
Major - arts and humanities
Major - science, technology, engineering, math
SAT composite score (100)
Academic self-confidence
Commitment to academic/professional success
College variables
Instructional expenditures ($1,000)
Research expenditures ($1,000)
Region - northeast (reference group)
Region - middle east
Region - great lakes
Region - plains
Region - southeast
Region - southwest
Region - rocky mountains
Region - far west
Public
28
Selectivity (100)
Cost of attendance ($1,000)
Average aid per student ($1,000)
High school variables
Total enrollment
Number of Advanced Placement courses
Type - private, non-religious
Type - private, Catholic
Type - private, other religion
Type - public (reference group)
Counselor-student ratio
Region - west
Region - south
Region - mid-Atlantic
Region - mid-west
Region - northeast (reference group)
29
543
543
543
11.10
27.10
8.42
1.16
8.83
5.14
7.75
8.87
0.25
15.10
44.00
24.54
5607
5607
5607
5607
5607
5607
5607
5607
5607
5607
5607
5607
1179.08
8.39
0.05
0.11
0.04
0.80
0.01
0.18
0.22
0.21
0.29
0.10
747.97
5.87
0.22
0.31
0.19
0.20
0.01
0.38
0.41
0.41
0.45
0.30
10.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
5266.00
31.00
1.00
1.00
1.00
1.00
0.10
1.00
1.00
1.00
1.00
1.00
Table 2
Results from cross-classified hierarchical generalized linear modeling analyses
Student-level variables
Female
Black
American Indian
Asian American
Latino
Other race
High school GPA
Parental income
Reason for college: parents wanted me to go
Reason for college: to prepare for graduate school
Hours per week spent talking with teachers outside class
Concerns about being able to finance college education
Chose this college because a private college counselor advised
me
Highest level of education attained by either parent
Major - professions (education, business)
Major - social sciences
Major - arts and humanities
Major - science, technology, engineering, math
SAT composite score
Academic self-confidence
Commitment to academic/professional success
College variables
Intercept
Instructional expenditures
Research expenditures
Region - middle east
Region - great lakes
Region - plains
Region - southeast
Region - southwest
Region - rocky mountains
Region - far west
Public
Selectivity
Cost of attendance
Average aid per student
30
Logodds
S.E. Delta-P
0.05
-0.17
-0.01
-0.14
-0.09
-0.08
0.06
0.02
0.08
0.12
0.04
-0.08
0.01
0.03
0.05
0.02
0.03
0.04
0.01
0.01
0.01
0.01
0.01
0.01
0.95%
-3.05%
0.86
0.02
0.11
0.04
-0.04
0.07
-0.09
0.02
0.04
0.01
0.01
0.02
0.03
0.03
0.02
0.01
0.01
0.01
19.06%
0.38%
2.12%
-1.22
0.01
-0.01
-0.13
-0.19
-0.40
0.05
-0.11
0.24
0.02
-0.17
0.16
-0.01
0.03
0.09
0.01
0.01
0.09
0.10
0.12
0.10
0.14
0.21
0.11
0.11
0.03
0.01
0.01
-2.53%
-1.65%
-1.47%
1.14%
0.38%
1.53%
2.32%
0.76%
-1.47%
1.34%
-1.65%
0.38%
0.76%
16.04%
-3.39%
-6.74%
3.12%
0.57%
High school variables
Total enrollment
Number of Advanced Placement courses
Type - private, non-religious
Type - private, Catholic
Type - private, other religion
Counselor-student ratio
Region - west
Region - south
Region - mid-Atlantic
Region - mid-west
Model Statistics
College-level variance
College-level variance explained
High school-level variance
High school-level variance explained
Overall level-2 variance explained
0.01
0.02
-0.03
-0.07
-0.04
0.06
-0.16
0.01
0.17
-0.05
0.00844
30.19%
0.00012
83.56%
33.23%
31
0.01
0.01
0.01
0.04
0.04
0.03
0.04
0.03
0.03
0.03
0.38%
-0.56%
1.14%
-2.88%
0.19%
3.32%