The link between advanced placement experience and early college success.
Klopfenstein, Kristin ; Thomas, M. Kathleen
1. Introduction
At its inception in 1957, the Advanced Placement (AP) Program was
designed to allow high school students to earn credit, or at least
advanced placement, for college-level coursework, thereby avoiding
needless repetition once these students arrived at college. The Program
primarily served students from elite private high schools. While the
structure of the AP Program has not changed in 50 years, its scope has
broadened dramatically. In 1960, 890 secondary schools participated in
the AP Program. Forty years later, that number had risen to 13,253.
Today, 15,122 U.S. schools offer AP courses, and 12,037 of those schools
are public high schools (College Board 2007). Regardless of the
exam-taking that earns students college credit, AP course-taking has
become a primary signal used to identify motivated, high-achieving
students in the college admissions process (Breland et al. 2002). In
addition, state policy makers have begun mandating the inclusion of AP
courses in their districts and high schools (see Table 1). This
expansion proceeds with very little rigorous empirical evidence
regarding the benefits and costs of AP participation.
The College Board provides plenty of studies showing that passing
AP exam scores are strong predictors of college success (e.g.,
Willingham and Morris 1986; Morgan and Manackshana 2000; Hargrove,
Godin, and Dodd 2007; and Keng and Dodd 2007). However, the public,
including legislators, interpret this to mean that high AP exam scores
cause college success, and subsequently infer that the expansion of the
AP program will improve college outcomes for an expanded set of AP
takers. This faulty reasoning does not come without costs. AP Programs
can be emphasized, or even mandated, at the expense of other proven
curricula or programs in the following ways: i) the best, most
experienced teachers are assigned to small AP classes, while the non-AP
classes necessarily grow larger; ii) schools, often with subsidies from
the state or federal government, pay for AP teachers to attend the AP
summer institute; iii) science labs must be equipped with more expensive
equipment; iv) AP exam fees are subsidized, usually for low income
students, but sometimes for all students regardless of need; and v)
textbooks must be replaced more frequently (Klopfenstein and Thomas
2007).
This paper uses an extensively specified regression analysis to
investigate AP course-taking as a potential cause of early college
success. Because it is impossible to randomly assign students to AP
courses, as an ideal experiment necessitates, we employ regression
analysis using an extensive administrative database of all Texas public
school students who entered Texas public universities directly after
graduating from high school in May 1999. Our data are unique in that we
are able to include a broad range of variables describing the
student's non-AP curricular experience. We show that failing to
control for the student's non-AP curricular experience leads to
positively biased AP coefficients. Math is frequently shown to be a
strong predictor of college success, so the omission of math-taking
information in previous studies is particularly problematic (e.g., Rose
and Betts 2001; Sadler and Tai 2007). We find no evidence that AP
course-taking increases the likelihood of early college success beyond
that predicted by the non-AP curriculum for the average student,
regardless of race or family income. This finding casts serious doubt on
the causal effect of AP experience on college success.
AP course experience matters now more than ever. In 2000, a survey
of 962 four-year public and private colleges and universities showed
that AP experience factors directly or indirectly into five of the top
six criteria in college admissions (see Table 2). Grades on AP exams
rank ninth. Note that the five criteria valued most heavily by colleges
do not depend on AP exam scores, but on course participation. A 2005
survey of 539 public and private four-year and two-year colleges and
universities supports the Breland survey. According to Sathre and Blanco
(2006), 91% of the postsecondary institutions surveyed take AP into
account in their admissions process.
Given that grade point averages (GPA) and class rank, as calculated
by the high school, are the number one criteria in the admissions
process, it is necessary to examine the non-trivial role of AP
experience in these outcomes. Weighting grades in AP courses more
heavily than grades in other courses is common practice. While grade
weighting is not mandated in Texas, as it is in some other states (e.g.,
North Carolina), 98% of all Texas public high schools weight AP grades
more heavily than other course grades when calculating class rank
(College Board 2006, author's survey). (1) The College Board
provides no grade-weighting guidelines, so weighting schemes vary
dramatically across schools. The most common methods are to add one
point on a four-point scale, yielding a 25% weight, or to add 10 points
on a 100-point scale, yielding a 10% weight. In schools with a large
number of AP course offerings, students must take a substantial number
of AP courses to remain competitive in class rank.
The College Board is conspicuously silent on the use of AP in
admissions decisions. Intentionally or not, the College Board supports
the use of AP for admissions purposes by advertising AP classes as
"college prep." Moreover, as revealed by President George W.
Bush's remarks in his 2006 State of the Union Address, government
at all levels is devoting considerable resources to expand the program
further under the pretense that AP courses are college preparatory. In
Section 1702 of the No Child Left Behind Act of 2001, the federal
government supports "state and local efforts to raise academic
standards through advanced placement programs, and thus further increase
the number of students who participate and succeed in advanced placement
programs" (U.S. Department of Education 2004). In 2006, they
granted $5,867,284 to 26 states to help them achieve that goal through
test fee subsidies (U.S. Department of Education 2006). Such efforts are
driven, in large part, by a competitive college admissions process that
places extraordinary emphasis on AP course work. The consequences of the
decision to emphasize AP in admissions reach beyond the university
itself. Fully aware of admissions policies, upper-class parents demand
that schools maximize their AP course offerings. This places school
administrators in the unenviable position of deciding whether to
redirect resources from other areas of need in order to expand their AP
Program.
Admissions officials, parents, and state legislators are not the
only stakeholders in the expansion of AP. The Education Commission of
the States (ECS) is a prominent think-tank that used to manage the
National Assessment of Educational Progress for the federal government.
ECS currently recommends that every state adopt a comprehensive AP
policy (Dounay 2006). Its central recommendation is for every state to
mandate that a minimum number of AP courses be offered at every high
school. At a minimum, they propose that all states provide financial
incentives to encourage schools and districts to offer AP. Implicit
behind policies that encourage AP participation is the belief that AP
experience prepares students for college-level work. However, the
widespread belief that AP experience produces positive outcomes in
college has remained largely untested.
2. Conceptual Framework
There are two theories to explain why AP experience might be a good
predictor of early college success. First, AP experience signals two
important but difficult to measure personal characteristics: ability and
motivation. Second, AP experience might build human capital, in which
case AP participation is good preparation for college. AP exposes
students to college-level material in a supportive high school
environment, where students are, presumably, more likely to receive the
individualized attention they need to develop study skills and habits of
mind that will serve them well in college.
The two models are not mutually exclusive, and colleges are
indifferent with respect to which model is at work, because a
high-quality student is identified either way. However, from a policy
standpoint, distinguishing which avenue is more or less at play is
important. The human capital model provides justification for broadly
expanding AP participation, while the signaling model does not. Before
these theories can be disentangled, however, it must be determined that
AP experience is in fact a good predictor of college success at all.
Prior research on the predictive power of AP course experience on
college success is not compelling. Studies from the College Board, owner
of the AP trademark, and the Educational Testing Service (ETS),
administrator of the yearly AP examinations, are frequently cited by AP
Program proponents (Willingham and Morris 1986; Morgan and Maneckshana
2000). The descriptive nature of these studies, however, is insufficient
for isolating the independent impact of the AP Program, given that the
typical AP student is bright, motivated, and likely to experience
positive college outcomes regardless of AP experience. The public
enthusiasm for AP waxes unabated, however, as shown by Newsweek's
ranking of high schools based exclusively on the ratio of the number of
AP exams taken to the number of graduating seniors. "The idea is
that schools should be recognized for pushing even average students to
take challenging AP courses, the more, the better" (Winerip 2006).
One study frequently cited by AP proponents as evidence of the
program's success is Adelman's Answers in the Tool Box (1999).
In The Tool Box Revisited (2006), a modified replication of his earlier
study, Adelman makes it clear that his results have been repeatedly
misinterpreted. In his original study, Adelman finds that a rigorous
high school curriculum, of which AP is one component, is an important
factor in obtaining a bachelor's degree. He does not find that AP
participation alone contributes to bachelor's degree completion.
Although Adelman never intended to investigate the independent impact of
AP course experience on college success, he explores the issue in his
2006 study to address the misreading of the original Tool Box. He
develops a model that replaces his index measuring the academic
intensity of a student's high school curriculum with proxy
variables measuring science momentum, foreign language study, and AP,
while also controlling for academic performance and various demographic
characteristics. He finds that AP does not explain bachelor's
degree completion, and this is after controlling for a very limited
representation of the student's high school experience.
A recent upsurge of independent research studying the AP Program is
of significantly higher quality than the College Board and ETS studies
and targets AP course experience specifically (Geiser and Santelices
2004; Dougherty, Mellor, and Jian 2005; Sadler and Tai 2007).
Multivariate regression models are used in an attempt to identify the
unique impact of AP after controlling for class rank, test scores,
and/or high school quality. With the notable exception of Sadler and Tai
(2007), these new studies remain problematic in that they omit student
experience in non-AP coursework. Given that AP-taking and other rigorous
course-taking are positively correlated, and that other rigorous
courses, particularly math and science, have an established positive
impact on the likelihood of college success, omitting student experience
in these other courses leads to positive bias on the AP coefficients. In
other words, it is misleading in favor of AP being effectual to consider
the effect of AP on college outcomes without controlling for the body of
the student's non-AP curricular experience. Our research suggests
that, for the average AP student, much of the estimated AP effect found
in previous studies is actually the effect of non-AP coursework in math
and science.
It is worth noting that the effect of AP experience on first
semester GPA may be negatively biased in regression models if students
who pass AP exams enroll in more challenging first semester classes than
non-AP students and consequently earn lower grades. Several studies
provide evidence that such bias is unlikely to be large. For example,
from a random sample of 8594 students in 128 first semester introductory
college science courses at 63 colleges and universities, Sadler and Tai
(2007) find that it is not uncommon for students who earn scores of 3 or
higher on an AP science exam to retake the course at the university.
Among the students in their sample, 283 out of 1029 AP-takers had earned
a score of 3 or higher, yet were enrolled in the comparable introductory
level course. Students reported several reasons for this: Some colleges
require a score higher than a 3 for advanced placement; some colleges do
not accept AP credit at all; some departments require a placement exam
in addition to passing AP scores; and some students voluntarily
re-enrolled in an effort to improve their understanding.
Further evidence comes from a recent National Research Council
(2002) survey showing that, while substantially more than half of
mathematics departments grant credit to students with passing scores on
AP calculus exams, only one third of departments allow placement in
advanced courses without additional testing and/or interviews. Hurdles
such as these reduce the number of students placed directly into more
advanced classes in their freshman year, which might lower first
semester grades as a result. Lichten (2000) finds that only 22% of AP
calculus students earning a 3 on the exam took a more advanced calculus
course at any point in their college career. In his sample, which comes
from ETS, 24% of students who earn 3s took no additional calculus, and
17% took a remedial course. Students who do place into more difficult
courses in the popular subject areas of calculus, English language and
composition, and biology generally do well in these classes (Dodd et al.
2002).
3. Data
We estimate the effect of AP course experience on early success in
college using the Texas Schools Microdata Panel (TSMP). Our sample
consists of over 28,000 Texas high school graduates who attended 31
four-year Texas public universities in the fall of 1999. We measure
early college success via second year retention and first semester GPA.
The vast majority of students who drop out of college do so during, or
immediately following, the freshman year (Pascarella and Terenzini 1980;
Tinto 1993, 1998), and "academic performance was the overwhelmingly
most significant factor affecting a freshman's decision to continue
into the sophomore year" (Braunstein, McGrath, and Pescatrice 2000,
p. 191). If the AP Program is truly college preparatory, AP experience
should improve academic performance in college and increase the
likelihood of returning for the second year. Because the AP curriculum
replicates freshman-level college courses, any preparatory benefits
students derive from the program should be apparent within the first
year of college. (2)
In our study, students who have a GPA of less than 2.0 and do not
return to any four-year institution in Texas for their second year of
study, including those who transfer to two-year postsecondary
institutions are "not retained." While our data do not include
information on students who transfer to private Texas universities or
out of state, measurement error should be minimized by the substantial
difficulty students would face transferring from one four-year
institution to another with a GPA below 2.0. (3)
White AP students retain at the highest rate among the 1999 cohort
of Texas public university students studied and non-AP taking black and
Hispanic students at the lowest rates (see Table 3). The freshmen
retention rates in our data are consistent with national trends given
the range of colleges and universities represented in the sample (U.S.
News and World Report 2003). While just 10% of white AP-takers do not
return for a second year, this represents 870 students and provides
substantial variation with which to estimate the model. Average first
semester GPA is also highest for white AP-takers (2.77) and lowest for
black students with no AP experience (2.01).
A handful of research, most of it quite recent, does account for
individual ability and motivation by including such variables as high
school GPA and test scores (Willingham and Morris 1986; Geiser and
Santelices 2004; Dougherty, Mellor, and Jian 2005; Sadler and Tai 2007).
However, with the exception of Sadler and Tai (2007), they fail to
control for the body of the non-AP curriculum taken by AP students. Our
data allow us to consider the years of science taken, years of foreign
language taken, and the highest level of math completed, as well as
participation in honors courses. Table 4 presents summary statistics of
these variables.
We include a host of additional controls. Student variables include
race, sex, SAT scores, high school GPA (standardized to a 4.0 scale),
whether a student was in the top 10% of their graduating class, and
whether Hispanic students have ever been designated as Limited English
Proficient. Family characteristics include parent education and family
income, as well as whether the student received a Stafford Loan. High
school characteristics include the percentage of students who qualify
for free or reduced price lunch, percentage of students who took college
entrance exams, the student/teacher ratio, percentage of inexperienced
teachers, and school size. We also include fixed effects for the
university attended and a variable indicating whether the student
enrolled part time. (4) Our retention logit and GPA regression are
specified as follows:
Pr(R = 1) = X[[beta].sub.1] + Y[[beta].sub.2] + Z[[beta].sub.3] +
U[[beta].sub.4] + C[[beta].sub.5] + AP[[beta].sub.6] + [epsilon],
GPA = X[[alpha].sub.1] + Y[[alpha].sub.2] + Z[[alpha].sub.3] +
U[[alpha].sub.4] + C[[alpha].sub.5] + AP[[alpha].sub.6] + [mu],
where R is a dummy variable equal to one if the student returns for
the second year; GPA is the student's first semester college GPA on
a four-point scale; the [beta]s and [alpha]s are vectors of regression
coefficients; X, Y, and Z are matrices of student, family, and high
school characteristics, respectively; U is a matrix of university
dummies; C is a matrix of non-AP curriculum variables; and AP is a
matrix of AP participation variables.
4. Results
We consider the impact of the total number of AP credits taken in
core subject areas on college retention and GPA, as well as the effect
of experience in specific AP subject areas on the same outcomes. The
appropriate modeling technique for persistence, which is a dichotomous
variable equal to one if a student returns for a second year, is
different from that for GPA, which is a continuous variable between zero
and four. We model persistence using a logit model and GPA using
ordinary least squares (OLS). (5) In every model, we include the
student, family, and high school characteristics previously described,
as well as college fixed effects. We estimate each model two ways. We
first simulate previous studies by excluding the non-AP curriculum
variables, and then we show how the addition of a host of non-AP course
controls reduces the magnitude of the AP variable coefficients, in most
cases to the point of eliminating statistical significance at
conventional levels.
Retention Models
First consider the impact of the number of AP credits taken in high
school on college retention. (6) In this model, the effect of AP credits
on retention is allowed to follow a quadratic path, since diminishing
returns are likely. Figure 1 summarizes the effect of the total number
of AP credits in core courses on the probability a student persists to
the second year of college when non-AP courses are included and excluded
from the analysis. (7) Note that the marginal impact of AP courses for
each specification is represented by the slope of the curve rather than
the intercept. Differences in the predicted probabilities of retention
between white and black students are generated solely by differences in
mean characteristics because the coefficient estimates for white and
black students are statistically indistinguishable. (8) However,
marginal effects for Hispanic students differ based on both coefficient
estimates and mean characteristics.
[FIGURE 1 OMITTED]
Consistent with prior research that also omits non-AP course-taking
variables, we find a statistically significant positive, albeit small,
effect of AP experience on the likelihood of persistence when non-AP
course experience is omitted from the analysis. Each additional AP
course taken (up to five credits) has a constant positive impact on
white students, while black and Hispanic student retention increases
only for the first two or three AP courses. Using these same data but
including non-AP course-taking variables in the model, we find that
these positive and significant findings vanish for all but Hispanic
students. While there is generally an upward trend in the college
retention rate for students with AP experience, the effects are small
and insignificantly different from zero, indicating that any upward
trend is likely due to random chance. The positive bias displayed in
Figure 1 is theoretically predictable, given that difficult non-AP
courses have a positive expected impact on college retention and are
positively correlated with AP course-taking. It is important to
recognize that omitted variable error leads to estimates that are biased
and inconsistent, and the bias will not diminish as the sample size
increases.
The AP effect on retention may be biased downward in Figure 1 under
both specifications if AP experience increases the likelihood of college
attendance for first generation students, but colleges and universities
do not support traditionally underrepresented students once they arrive
on campus. Many high school administrators and AP teachers in schools
serving a large proportion of low income and minority students believe
that AP experience increases college awareness and helps traditionally
underrepresented students identify themselves as college material, but
this hypothesis has yet to be tested (Spencer 2005). Similarly, AP
experience may increase the likelihood of a student enrolling at a more
selective institution. This might introduce a negative bias on the AP
coefficients in both the GPA and retention models if such students are
in over their heads at these institutions. (9)
Interestingly, Hispanic AP-taking continues to increase the
likelihood of retention even after the inclusion of non-AP control
variables. We investigate this robust result further in order to discern
which core AP subject area(s) facilitate Hispanic retention: English,
math, science, economics, government, history, and/or psychology. Once
again, we provide results for two logit models: one with AP
course-taking only and one with additional course-taking information.
Table 5 provides estimates of the marginal effects of AP course-taking
for the "average student," the characteristics of whom are
determined by the mean values of all continuous variables in the model
(with each race calculated separately). Dummy variables are turned on or
off to describe the "average student," based on which category
contains the largest percentage of observations within each race.
In the model of Hispanic students with controls for a broad measure
of curricular experience, we see that the entire AP effect on retention
for the average Hispanic student is driven by AP science. Note, however,
that AP science is not a significant predictor of retention for either
white or black students, while the marginal effect of AP science is
quite large and significant for Hispanic students: AP science increases
the probability of retention to the second year of college by 2.9
percentage points, or 3.6%, for the average Hispanic student. This
estimate, however, is 25% smaller than the estimated effect of AP
science in the model without controls for the non-AP curriculum. (10)
Given the unexpected magnitude of the AP science effect for
Hispanic students, we considered possible sources of omitted variable
bias. Candidates included variables that were correlated with AP
science-taking, being Hispanic, and staying in college. Interventions
that target first generation students in largely Hispanic regions of the
state may be responsible for the observed results. One such program, the
Texas Prefreshmen Engineering Program (TexPREP), began in 1979 in San
Antonio. Currently, eight San Antonio universities host "a
three-year mathematics-based summer program of approximately eight
weeks' duration ... [where the] curriculum is made up of
interdisciplinary applied subject matter, with an emphasis on math-based
logic and preparation for Advanced Placement classes" (emphasis
added). (11) UT San Antonio, the original home of TexPREP, is a largely
Hispanic institution, having graduated the third largest number of
Hispanic students (46% of its graduating class) of any university in the
nation in 2004-2005 (Rodriguez 2006). Although the program does not
maintain detailed longitudinal data on its participants, 70% of Texas
TexPREP sites are located at universities in the predominantly Hispanic
Rio Grande Valley and in south Texas. Participants in the program have
enjoyed high rates of college matriculation and graduation: In a 2002
general survey of 5380 former TexPREP participants over the age of 18,
88% reported attending college. Of those, 87% stayed at colleges in
Texas, making them potential members of our sample, and 90% earned a
postsecondary degree. (12) Although we can not empirically test the
hypothesis that TexPREP is responsible for the Hispanic AP science
effect, the facts that TexPREP university sites are located at campuses
serving large proportions of Hispanic students, that TexPREP emphasizes
AP science-taking, and that program participants have high rates of
college success lead us to believe that TexPREP (or a similar
intervention) is one possible cause.
The significant effect of AP economics on retention stands out in
the white/black pooled sample, as does the positive bias when non-AP
curriculum is excluded. Few high schools offer AP economics, and the
significant coefficients may be driven by unobserved characteristics of
schools and/or teachers who offer the course; however, the marginal
effect of AP Economics on retention drops by as much as 20% when
additional coursework is included in the model. The AP English effect
shrinks by an even more impressive 70-76% when other courses are
included.
As evidenced by both course-taking and exam-taking patterns, the
vast majority of high schools involved in the AP Program offer, at a
minimum, calculus AB, English, and history. These courses will likely be
the first offered by schools under state mandates requiring an AP
curriculum where none currently exists. These are the very
courses--those central to the AP Program--that have no predictive power
after accounting for a student's other rigorous high school
courses.
GPA models
Unlike in the retention models, F tests confirm that the
coefficients are statistically different for white and black students,
as well as for Hispanic students in the GPA models. Figure 2 describes
the relationship between AP course experience and changes in GPA with
and without controlling for non-AP courses. The bias in much of the
research on AP can be clearly seen. The coefficient estimate on the
number of AP credits taken by white students is 1.6 times larger when
only AP courses are considered, and only the AP effect for white
students remains statistically different from zero. While the effect of
the number of AP courses on GPA is insignificant for black and Hispanic
students in both cases, bias is evident nonetheless: The coefficient on
the number of AP courses is nearly eight-fold larger for Hispanic
students when the non-AP curriculum is excluded, and the black
coefficient behaves similarly. Thus, the omission of non-AP curriculum
in previous studies can lead to erroneous conclusions regarding the
effectiveness of the AP program for improving college outcomes,
particularly for traditionally underserved students.
Given the significant effect of AP experience on first semester
college GPA for white students, we again disaggregate AP courses by
subject to identify the source of the positive result (see Table 6).
This time, it appears that AP government is the driving force once the
appropriate non-AP courses are included in the model. As expected, none
of the individual courses emerge as significant for black students, but
for Hispanic students, AP science is once again a positive factor, as
are AP economics and AP psychology. AP government, economics, and
psychology are not flagship courses of the AP program. The positive
coefficients on these courses are most likely capturing some unobserved
characteristics of the high schools that can offer an AP curriculum of
such breadth and the students who choose to take AP courses outside the
core. The most striking result of our analysis is that, just as in the
retention model, the three most popular AP courses--calculus, English,
and history--have no effect on first semester GPA for any group.
Participation in the core AP courses has no effect on early college
success. Our large sample sizes, coupled with low correlation
coefficients among and between honors courses and AP courses, make it
unlikely that these results are driven by collinearity. (13)
[FIGURE 2 OMITTED]
AP math, which theoretically includes both calculus and statistics
classes, but in reality is heavily dominated by calculus classes, has a
statistically insignificant impact on both retention and GPA. On the
surface, this result appears to contradict the finding that rigorous
math prepares students for success in college. However, calculus (with
or without an AP designation) is included among the math curriculum
variables and has the expected positive and large impact on both
retention and GPA (see Appendix). The inclusion of the AP math dummy
captures the additional effect of converting a non-AP calculus class
into an AP calculus class; the insignificant but negative coefficient
reveals that converting to an AP class confers no additional benefit in
terms of college preparation. (14)
5. Conclusions
The exam-based structure of the AP Program was designed in 1957 to
provide a mechanism by which students might engage in accelerated
learning in high school and then bypass previously mastered material
once in college. The use of AP course experience as a criterion in
college admissions is a relatively recent phenomenon and an application
of the AP Program that was, we believe, unanticipated. Despite this,
policy makers at all levels of government and many members of the public
do not recognize the distinction between these two very different,
though not necessarily mutually exclusive, applications of the program.
Well-intentioned education advocates have come to believe that AP is an
appropriate, and even necessary, component in the portfolio of the
well-prepared college student. Our research finds no conclusive evidence
that, for the average student, AP experience has a causal impact on
early college success.
Our findings support a clear distinction between courses that are
"college preparatory" and those that are "college
level." The former type of course emphasizes the development of
skills needed to succeed in college, such as note taking, study skills,
and intellectual discipline; the latter type assumes that such skills
are already in place. At-risk high school students particularly benefit
from skills-based instruction, including "how to study, how to
approach academic tasks, what criteria will be applied, and how to
evaluate their own and others' work," where writing and
revising are ongoing (Darling-Hammond, Ancess, and Ort 2002, p. 658).
AVID, Gear Up, and TexPREP are three programs that provide explicit
training in these skills, and implementing such a program in conjunction
with a limited, aligned, high-quality AP Program is a promising way to
improve college outcomes (Watt, Yanez, and Cossio 2003; Dougherty,
Mellor, and Jian 2006). (15) Future research should synthesize the
existing data on these and similar programs and disentangle the most
effective aspects of the programs.
It is important to recognize that prediction and causality are not
the same, and that the practice of placing extraordinary weight on AP
participation in the college admissions process absent evidence of human
capital gains from program participation distorts incentives. Our
research finds that AP course-taking alone may be predictive of college
success, a finding that is consistent with College Board research by
Dodd et al. (2007) but casts doubt on the notion that AP participation
imparts a positive causal impact on college performance for the typical
student. (16) Our research indicates that the predictive power of
AP-taking is likely the result of signaling: high ability, motivated
students take more AP classes to differentiate themselves from other
students in the college application process. Once other rigorous high
school courses and demographic and school characteristics are
considered, however, students typically do well in college regardless of
their AP experience. The power of the AP Program as a signal will be at
least partially diminished if the AP Program continues to expand its
enrollment to students with less ability and/or motivation under the
guise of human capital benefits.
Stakeholders in the AP Program, as well as members of state and
federal governments, mistakenly interpret the predictive power of AP as
a causal impact. The belief that AP enhances human capital leads to the
policy prescription of assisting more, perhaps many more, students to
enroll in AP courses with the goal of improving their chances of college
success (beyond the admissions process). Unfortunately, under the
signaling model, this policy is suboptimal, as high-achieving students
take more and more AP classes to differentiate themselves from the
less-qualified college candidates taking fewer. Inefficiency arises when
the high-ability students continue to enroll in high numbers of AP
classes even when the costs (stress; reduction in sports, music, or
other extracurricular activities; reduction in social interactions)
exceed the benefits (greater chance of college admission, possibly a
semester or more of college credit) (Klopfenstein and Thomas 2007).
Moreover, the marginal student who responds to changes in legislative
policy and enrolls in an AP class overestimates the benefits of AP
participation in believing that AP-taking improves their college
readiness. AP courses are not explicitly designed to develop the study
skills and discipline necessary to succeed in college, and the benefit
of having students who have not mastered high-school-level material take
college-level classes in high school is unproven. Under the signaling
scenario, college admissions officers will ultimately have to find an
alternative signal as the pool of AP-taking applicants becomes of lower
average ability.
It would be more efficient for postsecondary institutions to focus
on the years of high school science and math studied as math and science
experience consistently emerge as strong predictors of college and labor
force success, and there is much stronger evidence of a causal link
between math and science training and life success (e.g., Rose and Betts
2001). Given the sophistication of current data systems and the increase
in whole file review for universities engaging in affirmative action in
admissions, the costs of changing from an emphasis on AP to math- and
science-taking should be relatively small. If universities change the
intensity of their admissions focus from AP courses to math and science,
either with or without the AP designation, parents of college-bound
students and state policy makers will have the incentive to shift their
own emphasis toward math and science learning. Currently, the trend
among policy makers is to legislate quite strongly in favor of AP,
necessarily at the expense of other areas of need. Much more study is
needed about the benefits and costs of the AP Program for all students,
including those who do not participate in AP, before such policies are
further expanded.
Appendix
Coefficient Estimates of Non-AP Curriculum Variables
Disaggregated GPA model Disaggregated Retention Model (a)
Variable White (b) Black (b)
Science = 3 years
(relative to <3) 0.19 *** (0.07) 0.19 *** (0.07)
Science > 3 years
(relative to <3) 0.16 ** (0.07) 0.16 ** (0.07)
Foreign language = 2 years
(relative to <2) 0.02 (0.08) 0.02 (0.08)
Foreign language > 2 years
(relative to <2) 0.03 (0.09) 0.03 (0.09)
High math algebra -0.59 *** (0.18) -0.35 ** (0.17)
High math geometry -0.45 *** (0.11) -0.21 ** (0.11)
High math algebra 2 -0.24 *** (0.05) omitted
High math trigonometry -0.05 (0.08) 0.20 *** (0.08)
High math pre-calculus omitted 0.24 *** (0.05)
High math calculus 0.19 ** (0.08) *** (0.09)
Honors English 0.20 *** (0.06) 0.20 *** (0.06)
Honors science -0.17 (0.06) -0.17 (0.06)
Honors social science 0.01 (0.06) 0.01 (0.06)
N ([double dagger]) 23,127 23,127
Disaggregated
Retention Disaggregated
Disaggregated GPA model Model (a) GPA model
Variable Hispanic White
Science = 3 years
(relative to <3) *** (0.13) 0.05 ** (0.03)
Science > 3 years
(relative to <3) 0.39 *** (0.14) 0.02 (0.03)
Foreign language = 2 years
(relative to <2) 0.02 (0.16) -0.01 (0.03)
Foreign language > 2 years
(relative to <2) -0.02 (0.17) 0.01 (0.04)
High math algebra -0.21 (0.38) -0.31 *** (0.10)
High math geometry -0.26 (0.26) -0.21 *** (0.06)
High math algebra 2 -0.21 ** (0.10) -0.09 *** (0.02)
High math trigonometry 0.06 (0.15) -0.01 (0.02)
High math pre-calculus omitted omitted
High math calculus 0.07 (0.15) 0.05 ** (0.02)
Honors English 0.12 (0.11) 0.02 (0.02)
Honors science -0.10 (0.11) -0.07 (0.02)
Honors social science -0.07 (0.11) 0.04 ** (0.02)
N ([double dagger]) 5194 19,281
Disaggregated GPA model Disaggregated GPA model
Variable Black Hispanic
Science = 3 years
(relative to <3) 0.02 (0.05) 0.06 (0.06)
Science > 3 years
(relative to <3) 0.07 (0.06) 0.02 (0.06)
Foreign language = 2 years
(relative to <2) 0.10 ** (0.06) 0.07 (0.07)
Foreign language > 2 years
(relative to <2) 0.10 * (0.07) 0.08 (0.07)
High math algebra -0.11 (0.10) 0.03 (0.18)
High math geometry 0.07 (0.07) -0.04 (0.10)
High math algebra 2 omitted -0.08 ** (0.04)
High math trigonometry 0.07 (0.08) -0.002 (0.05)
High math pre-calculus 0.08 ** (0.04) Omitted
High math calculus 0.10 (0.08) 0.12 *** (0.05)
Honors English 0.04 (0.05) 0.02 (0.04)
Honors science -0.10 (0.05) -0.04 (0.04)
Honors social science 0.09 (0.05) -0.02 (0.04)
N ([double dagger]) 3017 5037
(a) As with all logit estimates, the coefficients presented for the
retention model are not equal to the marginal effects.
(b) "Disaggregated models" are those including the seven categories
of AP courses.
(c) White and black retention results based on a pooled sample, and
the sample consists of 20,260 white and 2867 black students
*** p [less than or equal to] 0.01; ** p [less than or equal to]
0.05; * p [less than or equal to] 0.10 based on one-tailed
hypothesis tests. Standard errors are in parentheses.
We express our deepest gratitude to the late John F. Kain and the
staff at the UTD Texas Schools Project. Thanks to Trevor Packer at the
College Board and Bob Leal and Rudy Reyna at TexPREP. Thanks also to
seminar participants at the meetings of the American Economic
Association, Georgia State University, and The University of
Mississippi, and very helpful referees. Any errors are our own.
Received February 2007; accepted April 2008.
References
Adelman, Clifford. 1999. Answers in the tool box: Academic
intensity, attendance patterns, and bachelor's degree attainment.
Washington, DC: U.S. Department of Education Office of Educational
Research and Improvement.
Adelman, Clifford. 2006. The toolbox revisited: Paths to degree
completion from high school through college. Washington, DC: U.S.
Department of Education.
Betts, Julian R., and Darlene Morell. 1999. The determinants of
undergraduate GPA: The relative importance of family background, high
school resources, and peer group effects. Journal of Human Resources
34(2):268-93.
Braunstein, Andrew, Michael McGrath, and Donn Pescatrice. 2000.
Measuring the impact of financial factors on college persistence.
Journal of College Student Retention 2(3):191-203.
Breland, Hunter, James Maxey, Renee Gernand, Tammie Cummins, and
Catherine Trapani. 2002. Trends in College Admission 2000: A report of a
national survey of undergraduate admission policies, practices, and
procedures. Accessed 29 June 2005. Available at
http://airweb.org/trends.html.
College Board. 2006. Federal and state support for AP. Accessed 26
May 2006. Available at
http://www.collegeboard.com/student/testing/ap/cal_fed.html.
College Board. 2007. Advanced Placement Report to the Nation 2007.
Accessed 5 June 2007. Available at
http://www.collegeboard.com/prod_downloads/about/news_info/ap/2007
/2007_ap-report-nation.pdf.
Darling-Hammond, Linda, Jacqueline Ancess, and Susanna W. Ort.
2002. Reinventing high school: Outcomes of the coalition campus schools
project. American Educational Research Journal 39(3):639-73.
Dodd, Barbara G., Steven J. Fitzpatrick, R. J. De Ayala, and Judith
A. Jennings. 2002. An investigation of the validity of AP grades of 3
and a comparison of AP and non-AP student groups. College Board Research
Report No. 2002-9.
Dougherty, Chrys, Lynn Mellor, and Shuling Jian. 2005. The
relationship between Advanced Placement and college graduation. National
Center for Educational Accountability Study Series Report 1.
Dougherty, Chrys, Lynn Mellor, and Shuling Jian. 2006. Orange juice
or orange drink? Ensuring that "advanced courses" live up to
their labels. National Center for Educational Accountability Policy
Brief No. 1.
Dounay, Jennifer. 2006. "Advanced Placement." Education
commission of the states. Accessed 4 June 2007. Available at
http://www.ecs.org/clearinghouse/67/44/6744.htm.
Geiser, Saul, and Veronica Santelices. 2004. The Role of Advanced
Placement and Honors Courses in College Admissions. Accessed 1 April
2005. Available at http://repositories.cdlib.org/cshe/CSHE-4-04/.
Hargrove, Linda, Donn Godin, and Barbara Dodd. 2007. College
outcomes comparisons by AP and non-AP high school experiences. Working
paper, The University of Texas at Austin.
Keng, Leslie, and Barbara Dodd. 2007. An investigation of college
performance of AP and non-AP student groups. Working paper, The
University of Texas at Austin.
Klopfenstein, Kristin, and M. Kathleen Thomas. 2007. Advanced
placement participation: Evaluating the policies of states and colleges.
Manuscript presented at Conference on Advanced High School Coursework in
Science and Mathematics, Cambridge, MA, 10 May 2007.
Lichten, William. 2000. Whither advanced placement? Education
Policy Analysis Archives 8(29), Available at
http://epaa.asu.edu/epaa/v8n29.html.
Morgan, Rick, and Behroz Maneckshana. 2000. AP students in college:
An investigation of their course-taking patterns and college majors.
Statistical Report 2000-09. Princeton, NJ: Educational Testing Service.
Accessed 15 September 2003. Available at
http://www.collegeboard.com/ap/pdf/validity2.pdf.
National Research Council. 2002. Learning and understanding:
Improving advanced study of mathematics and science in U.S. high
schools. Center for Education, Division of Behavioral and Social
Sciences and Education. Washington, DC: National Academy Press.
Pascarella, Ernest T., and Patrick T. Terenzini. 1980. Predicting
freshman persistence and voluntary dropout decisions from a theoretical
model. Journal of Higher Education 51(1):60-75.
Rodriguez, Kris. 2006. UTSA Continues Leading in Degrees to
Hispanics. UTSA Today, 7 June. Accessed 4 June 2007. Available at
http://www.utsa.edu/today/2006/06/rank.cfm?header=-1.
Rose, Heather, and Julian R. Betts. 2001. Math matters: The links
between high school curriculum, college graduation, and earnings. San
Francisco: Public Policy Institute of California.
Sadler, Philip M., and Robert H. Tai. 2007. Advanced Placement exam
scores as a predictor of performance in introductory college biology,
chemistry, and physics courses. Science Educator 16(2):1-19.
Sathre, Colleen O., and Cheryl D. Blanco. 2006. Policies and
practices at postsecondary institutions. In Accelerated learning
options: Moving the needle on access and success. Boulder, CO: Western
Interstate Commission for Higher Education, pp. 25-36.
Spencer, Jason. 2005. Advanced Placement: Passing the test? Houston
Chronicle, 13 November, B1.
Stinebrickner, Ralph, and Todd R. Stinebrickner. 2003.
Understanding educational outcomes of students from low-income families:
Evidence from a liberal arts college with a full tuition subsidy
program. Journal of Human Resources 38(3):591-617.
Tinto, Vincent. 1998. Colleges as communities: Taking research on
student persistence seriously. Study of Higher Education 21(2):167-77.
Tinto, Vincent. 1993. Leaving college: Rethinking the causes and
cures of student attrition. 2nd edition. Chicago: University of Chicago
Press.
U.S. Department of Education. 2004. No Child Left Behind Act of
2001. Section 1702. Accessed 16 December 2007. Available at
http://www.ed.gov/policy/elsec/leg/esea02/pg14.html#sec1702.
U.S. Department of Education. 2006. "Advanced Placement test
fee program." Guide to U.S. Department of Education Programs.
Accessed 17 December 2007. Available at
http://www.ed.gov/programs/apfee/index.html.
U.S. News and World Report. 2003. America's Best Colleges
2003. Available at http://colleges.usnews.rankingsandreviews.com/college/ national-freshmen-least-most-likely-return.
Watt, Karen M., Darlene Yanez, and Griselda Cossio. 2003. AVID: A
comprehensive school reform model for Texas. National Forum of
Educational Administration and Supervision Journal 19(3):43-59.
Willingham, Warren W., and Margaret Morris. 1986. Four years later:
A longitudinal study of Advanced Placement students in college. College
Board Report No. 86-2; ETS RR No. 85-86. Princeton, NJ: Educational
Testing Service.
Winerip, Michael. 2006. Odd math for 'Best High Schools'
list. New York Times, 17 May. Available at
http://www.nytimes.com/2006/05/17/education/17education.html.
Kristin Klopfenstein, Department of Economics, Texas Christian
University, Box 298510, Fort Worth, TX 76129, USA; E-mail
k.klopfenstein@tcu.edu.
M. Kathleen Thomas, Department of Finance and Economics,
Mississippi State University, Box 9580, 326 McCool Hall, Mississippi
State, MS 39762, USA; E-mail kthomas@cobilan.msstate.edu; corresponding
author.
(1) Klopfenstein conducted a telephone survey of counselors at all
Texas public high schools during the 2003-2004 academic year regarding
AP-related practices, including grade weighting. Surveys were completed
by approximately 80% of high schools. Among respondents, 15% did not
offer AP Programs, which resulted in a sample size of approximately 725
AP-offering high schools.
(2) While passing AP exam scores should reduce overall time to
graduation by earning students credit, this outcome is fully consistent
with the original purpose of the AP Program and not the outcome of
interest in this study.
(3) This approach is modeled after that taken by Geiser and
Santelices (2004). Students who leave the data set with a GPA above 2.0
are assumed to have transferred to a private or out-of-state
institution. Regression results are qualitatively identical with and
without the GPA restriction.
(4) Students in our sample attended 1066 different high schools. If
we use high school fixed effects in addition to college fixed effects,
we lose all significance in our regressions. Consequently, we control
for measurable differences across high schools as described.
(5) Although GPA is restricted to between zero and four, it is
commonly modeled using OLS. See Betts and Morell (1999) and
Stinebrickner and Stinebrickner (2003). Moreover, in our samples,
truncation is a minor issue: 6% of the white sample earned a 4.0 GPA, as
did just 1.3% of the black sample and 3% of the Hispanic sample.
(6) Psychology, Microeconomics, Macroeconomics, U.S. Government,
and Comparative Government count for half of a credit each, while
English, Science, Math, and History courses count as a full credit.
(7) Complete regression results are available from the authors upon
request.
(8) In addition, coefficient estimates for students with family
income below the median are statistically indistinguishable from those
with family income greater than or equal to the median.
(9) Because our data do not include college information for
students attending private universities or those outside the state of
Texas, it is not possible to test these hypotheses here.
(10) Changes in the marginal effects from one model to another are
calculated using the percent differences from the baseline (the numbers
in parentheses). Changes in the point differences are not comparable
across models because the baselines differ.
(11) http://www.prep-usa.org/portal/texprep/default.asp, accessed
February 1, 2008. There are also four TexPREP sites at universities in
the Rio Grande Valley (Brownsville, Harlingen, Edinburg, and Laredo),
three sites in south Texas (Corpus Christi, Victoria, and Houston), one
site in west Texas (Lubbock), and five sites in north Texas (Fort Worth,
Arlington, Austin, and two in Dallas).
(12) http://www.prep-usa.org/portal/generaldetail.asp?ID=107,
retrieved February 1, 2008.
(13) Because the average AP student takes courses in just two of
the seven AP subjects we examine, the correlations among the AP courses
are low. Furthermore, the AP course coefficients do not change in sign
or significance when the honors courses are jointly removed from the
model. While the absence of high pairwise correlations between AP
courses and other independent variables does not eliminate the
possibility of collinearity involving more than two variables, the
robustness of our results is further supported by the general math and
science curriculum variables, which are of the expected sign and
significance (see Appendix).
(14) Sample sizes are large enough to facilitate the division of
calculus into non-AP and AP sections. In the white sample, 18% of
students took AP calculus, and 18% took non-AP calculus, and the two
groups are essentially mutually exclusive. In the black sample, 9% of
students took AP calculus, and 6% took non-AP calculus; in the Hispanic
sample, 13% took AP calculus, and 11% took non-AP calculus. These
numbers do not align with those presented in Table 4 because the
variable AP math includes AP statistics.
(15) The AVID program started in Texas in 1999, after the cohort we
study graduated from high school. For information on AVID, see
http://www.avidonline.org/. For information on Gear Up, see
http://www.ed.gov/programs/gearup/index.html. For information on
TexPREP, see http://www.prep-usa.org/portal/texprep/default.asp.
(16) Passing scores on AP exams are likely to be better predictors
of college success than AP course-taking alone, particularly if a
student's general pattern of high school course-taking is not
considered (Geiser and Santelices 2004). Because we do not have access
to AP exam score data, we cannot test this hypothesis.
Table 1. State Mandated AP Course Offerings
All Districts school All Districts
State Must Offer AP Must Offer AP
Arkansas [check]
(2008-09)
Idaho
Indiana [check]
Kentucky
Mississippi (a) [check]
Ohio
Oregon
South Carolina (b) [check]
Vermont
Virginia
West Virginia [check]
All High Schools All Districts Must
Must Offer Offer Advanced
Advanced Classes Classes That
State That May Included Ap May Include AP
Arkansas
Idaho [check]
Indiana
Kentucky [check]
Mississippi (a)
Ohio [check]
Oregon [check]
South Carolina (b)
Vermont [check]
Virginia [check]
West Virginia [check]
(2008-09)
Source: Education Commission of the States. Accessed 29 January 2008.
Available at http://mb2.ecs.org/reports/Report.aspx?id=996.
(a) Mississippi accepts online delivery as an acceptable alternative.
(b) South Carolina's mandate is contingent upon school size.
Table 2. AP and College Admissions
Rank Factors in College Admissions
1 High school GPA or class rank
2 SAT/ACT score
3 Pattern of high school coursework
4 College level work in HS
5 AP course enrollments (a)
6 AP course grades (b)
7 Letters of recommendation
8 Essays
9 AP Exam Grades (c)
Source: Survey of 962 four-year public and private colleges and
universities in 2000 (Breland et al. 2002).
(a) Moves up to number 4/tied for number 3 for private/public
universities if exclude "not considered" in the average importance
computation.
(b) Moves up to number 5/tied for number 3 for private/public
universities if exclude "not considered" in the average importance
computation.
(c) Jumps above both letters of recommendation and essays but below
SAT II (otherwise ranked number 12) for public universities if
exclude "not considered" in the average importance computation.
Table 3. Descriptive Statistics of Dependent Variables
White Black
No AP AP Taker No AP AP Taker
Percent
retained 83.7 90.6 78.7 84.9
Average fall 2.43 2.77 2.01 2.33
GPA (1.02) (0.96) (0.98) (1.02)
N 10,112 9240 2093 939
Hispanic
No AP AP Taker
Percent
retained 78.0 86.0
Average fall 2.11 2.39
GPA (1.06) (1.05)
N 2883 2154
Source: Texas Schools Microdata Panel.
Table 4. Descriptive Statistics of Curriculum Variables (a)
Variable White Black Hispanic
Science = 3 years 0.39 0.44 0.38
Science > 3 years 0.47 0.32 0.48
Foreign language = 2
years 0.41 0.51 0.45
Foreign language > 2
years 0.48 0.29 0.44
High math geometry 0.02 0.07 0.02
High math algebra 2 0.23 0.40 0.27
High math trigonometry 0.09 0.06 0.09
High math pre-calculus 0.38 0.30 0.36
High math calculus 0.29 0.14 0.25
Honors English 0.58 0.42 0.54
Honors science 0.49 0.30 0.44
Honors social science 0.46 0.31 0.41
AP math 0.19 0.09 0.14
AP science 0.15 0.10 0.13
AP English 0.29 0.18 0.25
AP economics 0.13 0.07 0.10
AP government 0.16 0.09 0.14
AP history 0.13 0.07 0.08
AP psychology 0.03 0.03 0.01
AP-taker 0.47 0.31 0.42
Number of AP courses
taken take one 2.3 (1.5) 2.0 (1.3) 2.1 (1.4)
N 19,801 3126 5240
Source: Texas Schools Microdata Panel.
(a) The means for dummy variables represent the proportion of the
sample reporting a one. Standard deviations are reported in
parentheses for continuous variables.
Table 5. Disaggregated Marginal Effects (a) of AP Experience on
Student Retention
White
AP Only Broad Curriculum (b)
Baseline Pr(retain) 85.20 88.23
AP math 1.91 ** (2.25) -0.02 (-0.02)
AP science 0.23 (0.27) 0.22 (0.25)
AP English 1.20 ** (1.40) 0.29 (0.33)
AP economics 2.68 *** (3.15) 2.21 *** (2.51)
AP government 0.90 (1.06) 0.66 (0.75)
AP history 0.92 (1.08) 0.72 (0.82)
AP psychology 0.21 (0.25) 0.67 (0.76)
N (b) 23,127 23,127
Black
AP Only Broad Curriculum (b)
Baseline Pr(retain) 73.58 73.25
AP math 3.00 ** (4.08) -0.04 (-0.05)
AP science 0.35 (0.48) 0.41 (0.56)
AP English 1.86 ** (2.53) 0.55 (0.75)
AP economics 4.24 *** (5.76) 4.31 *** (5.89)
AP government 1.40 (1.91) 1.27 (1.73)
AP history 1.44 (1.95) 1.37 (1.87)
AP psychology 0.33 (0.45) 1.28 (1.74)
N (b) 23,127 23,127
Hispanic
AP Only Broad Curriculum (b)
Baseline Pr(retain) 75.33 82.29
AP math 3.17 * (4.21) -1.25 (-1.52)
AP science 3.58 ** (4.75) 2.93 * (3.57)
AP English 0.04 (0.05) -1.09 (-1.32)
AP economics -0.03 (-0.05) -1.86 (-2.26)
AP government 1.86 (2.46) 1.13 (1.37)
AP history 0.21 (0.28) -0.56 (-0.68)
AP psychology -3.06 (-4.07) -4.59 (-5.58)
N (b) 5194 5194
(a) Marginal effects are presented as point differences from the
baseline with the percent differences from the baseline in
parentheses.
(b) Broad curriculum includes the following: the highest level of
math achieved (six categories); years of science (three categories);
years of foreign language (three categories); and a dummy variable
each for honors English, natural science, and social science.
(c) Black and white students are pooled in the retention model, and
the sample consists of 20,260 white and 2867 black students.
*** p [less than or equal to] 0.01; ** p [less than or equal to]
0.05; * p [less than or equal to] 0.10 based on one-tailed hypothesis
tests.
Table 6. Disaggregated Marginal Effects (a) of AP Experience
on First Semester Grade Point Average
White
AP Only Broad Curriculum (b)
AP math 0.07 *** (0.02) 0.03 (0.02)
AP science -0.01 (0.02) -0.01 (0.02)
AP English 0.03 ** (0.02) 0.02 (0.02)
AP economics 0.02 (0.03) 0.02 (0.03)
AP government 0.04 * (0.03) 0.04 * (0.02)
AP history 0.02 (0.02) 0.02 (0.02)
AP psychology -0.01 (0.04) 0.02 (0.04)
N 19,281 19,281
Black
AP Only Broad Curriculum (b)
AP math 0.11 ** (0.07) 0.07 (0.09)
AP science 0.00 (0.06) 0.01 (0.06)
AP English -0.02 (0.05) -0.05 (0.05)
AP economics -0.02 (0.09) -0.05 (0.09)
AP government 0.10 (0.08) 0.09 (0.08)
AP history 0.02 (0.07) -0.01 (0.07)
AP psychology 0.10 (0.10) 0.10 (0.10)
N 3017 3017
Hispanic
AP Only Broad Curriculum (b)
AP math 0.01 (0.04) -0.09 (0.06)
AP science 0.08 ** (0.04) 0.10 ** (0.05)
AP English -0.01 (0.04) -0.01 (0.04)
AP economics 0.08 * (0.06) 0.08 * (0.06)
AP government 0.02 (0.06) 0.02 (0.06)
AP history -0.09 (0.05) -0.07 (0.05)
AP psychology 0.16 * (0.11) 0.15 * (0.11)
N 5037 5037
(a) Standard errors in parentheses.
(b) Broad curriculum includes the following: the highest level of
math achieved (six categories); years of science (three
categories); years of foreign language (three categories); and a
dummy variable each for honors English, natural science, and social
science.
*** p [less than or equal to] 0.01; ** p [less than or equal to]
0.05; * p [less than or equal to] 0.10 based on one-tailed
hypothesis tests.
