Content uploaded by Daniel L. Bennett
Author content
All content in this area was uploaded by Daniel L. Bennett on Oct 20, 2017
Content may be subject to copyright.
HOMESCHOOL BACKGROUND, TIME USE, & ACADEMIC PERFORMANCE
AT A PRIVATE RELIGIOUS COLLEGE
Daniel L. Bennett*
Department of Entrepreneurship, Hankamer School of Business, Baylor University
Economic & Business Analytics, Patrick Henry College
Elyssa Edwards
Economic & Business Analytics, Patrick Henry College
Courtney Ngai
Economic & Business Analytics, Patrick Henry College
Abstract
We study the effects of homeschool background and time use on academic performance among
students at Patrick Henry College, a private religious institution with a 63-credit core classical
liberal arts curriculum. Using OLS regression analysis, we examine four research questions: (1)
Does time use influence academic performance? (2) Do homeschooled students perform
differently than traditionally schooled students? (3) Does parental education moderate the impact
of homeschooling on academic performance? (4) Does homeschooling moderate the impact of
ACT scores on academic performance?
Keywords: Academic Performance; Homeschool Background; Parental Education; Personality;
Student Time Use
*Corresponding author. Email: bennettecon@gmail.com Website: www.BennettEcon.com
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 1
1. Introduction
Homeschooling, or parent-led home-based education, has a long and distinguished history in the
United States, with many prominent historical figures (e.g., George Washington, Thomas
Jefferson, Benjamin Franklin, Andrew Carnegie, and Franklin Roosevelt) receiving at least part
of their education at home (Coulson, 1999). With the rise of compulsory school attendance laws
in the mid-nineteenth and early twentieth centuries, the practice of homeschooling diminished
significantly (Lips and Feinberg, 2008), with the number of homeschooled students dropping to
around 13,000 by the early 1970s (Lines, 1991). By this time, homeschooling had become “an
unacceptable practice for satisfying compulsory education requirements in most states” (Lines
2000, p. 77).
Homeschooling, however, began to regain popularity beginning in the 1970s among so-
called pedagogues, who believed that the growing bureaucratization and professionalization of
public schools were undermining instruction of their children. The so-called ideologues, seeking
to impart religious values on their children, joined the homeschool movement by the 1980s (Van
Galen, 1991). Despite these emerging trends, the right to homeschool was largely not protected
by law, as only three states permitted homeschooling in 1980 (Coulson, 1999). Over the next two
decades, state legislatures gradually changed their laws. By the start of the twenty-first century,
homeschooling was legal in all 50 states (Lines, 2000), facilitating rapid growth in the number of
homeschooled students (Lips and Feinberg, 2008). The Department of Education estimated that
the number of homeschooled students increased from 850,000 (1.7 percent of the 5 to 17 year-
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 2
old population) in 1999 to 1.8 million (3.4 percent) in 2012 (Redford et al., 2017). Ray (2016)
estimated that the number of homeschooled students reached 2.3 million by 2016.
1
Although homeschooling is now legal across the country and is widely accepted by the
public as a viable education option (Lines, 2000), the degree to which it is regulated differs
considerably across states (Lips and Feinber, 2008; Ray and Eagleson, 2008) and there remains
considerable debate over the degree to which homeschooling should be regulated (see, for
example, Ray and Eagleson (2008) and references therein). A lack of reliable evidence about the
academic outcomes of homeschooled students is one of the main concerns of proponents for
stricter governmental regulation of homeschooling (e.g., Reich, 2005).
2
Indeed, there is mixed
evidence on the collegiate academic performance of homeschooled students relative to
traditionally schooled students (Cogan, 2010; Jones and Gloeckner, 2004; Snyder, 2014).
The current study contributes to the literature on the academic performance of
homeschooled students by examining the factors affecting the academic performance of students
at Patrick Henry College (PHC). PHC provides a well-controlled environment for such a study
for several reasons. As discussed further in the methods section, the relative homogeneity in
terms of religious beliefs and educational background of the student population, as well as
relative homogeneity of the coursework and residency status, provide an excellent environment
for a case study of factors that contribute to differences in academic performance among a
population of primarily homeschooled students. We conducted an anonymous and voluntary
survey of PHC students to gather data about their grade point average (GPA) and time use during
the Fall 2014 semester. Additionally, we collected information about the students’ homeschool
1
Collom (2005) provides a review of research on parental motivations to homeschool. Among the most
common reasons are: dissatisfaction with public schools; academic and pedagogical concerns; religious
values; and family needs.
2
Romanowski (2006) discussed several other common criticisms of homeschooling.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 3
background, parental education level, personality type, major, and class rank. Using Ordinary
Least Squares (OLS) regression analysis, we test the following four research questions:
1. Does time use influence academic performance?
2. Do homeschooled students perform differently than traditionally schooled students?
3. Does parental education moderate the impact of homeschooling on academic
performance?
4. Does homeschooling moderate the impact of ACT scores on academic performance?
The remainder of the paper is organized as follows. Section 2 discusses the methods. A
review of related literature is provided in Section 3. The results are presented in Section 4.
Section 5 offers concluding remarks.
2. Methods
2a. Methodological Approach
The current study examines the determinants of academic performance at PHC, a private
selective religious liberal arts college that predominantly serves homeschooled students. We
employ a quantitative research approach using OLS regression analysis to address the research
questions posed in the introduction.
2b. Hypotheses
This study attempts to address four research questions related to the influence of homeschooling
and time use on academic performance among students at PHC. Specifically, the following null
hypotheses were tested using a quantitative research approach:
Students who spend more time on non-academic activities (e.g., extra-curricular
activities, paid work, dating, sleep) do not perform differently academically than students
who spend less time on non-academic activities;
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 4
There is no difference in academic performance between homeschooled students and
non-homeschooled students;
There is no difference in academic performance between homeschooled students with
college-educated parents and homeschooled students whose parents do not have a college
education; and
The impact of pre-college academic preparation on academic performance is not
conditional on whether a student was homeschooled.
2c. Study Design
This paper contributes to the literature by examining the factors affecting academic performance
among students at PHC, a selective private Christian residential liberal arts college in northern
Virginia. PHC provides a well-controlled environment for such a study for several reasons.
First, PHC was founded in 2000 with the objective of providing a high-quality classical
liberal arts education for Christian homeschooled students. The majority (79%) of PHC students
have at least some homeschool background (Patrick Henry College Fact Book, 2014-2015).
Second, PHC has a classical liberal arts curriculum, and of the 123 credits needed to graduate, all
students are required to take the same 63-credit core curriculum; further, the recommended
course sequence is very similar across majors. This enhances the comparability of grades
significantly, particularly once we control for class and major.
Third, the vast majority of students (90% of all students, 99% of freshman) reside on
campus (Patrick Henry College Fact Book, 2014-2015), so the potential effect of residency status
on academic performance is not a serious factor. Lastly, PHC is a religious institution that
voluntarily chooses not to accept government funding, and all students are required to sign a
statement of faith prior to enrolling. These institutional characteristics likely exert a strong
student selection effect, minimizing the potential impact of unobserved cultural differences
attributable to religious and social upbringing.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 5
2d. Sample Selection
As discussed in the study design section, PHC is a relatively well-controlled environment to
examine the impact of homeschool background and other factors such as time use and
personality type on academic performance. The data used for this study was collected through a
voluntary and anonymous survey of the PHC student population during the Spring 2015
semester.
The survey, which is provided in the Appendix, consisted of 14 questions. Voluntary
responses were solicited through an email sent through the campus email system to all students,
as well as two posts to the college-wide Facebook page.
3
Students were asked to report their
grade point average (GPA) for Fall 2014 and to answer questions related to their time use that
semester. In addition, the survey contained questions related to homeschooling background,
standardized test scores, major, class, and their parents’ educational attainment. The survey was
designed to extract individual-level characteristics that have been previously found to influence
collegiate academic performance (see literature review section for additional information).
The electronic survey was set up to ensure student responses were anonymous and
limited students to completing only one survey. Of the 326 students enrolled at PHC at the time
of survey, 109 responded to the survey, so our sample accounts for one-third of the student
population. The vast majority of the 109 students in our sample (94%) completed at least some
of their pre-college education at home and more than three-fourths (76%) were homeschooled
during elementary, middle, and high school.
To incentivize participation, students were offered the chance to be entered in a drawing
for a $5 campus gift card for completing the survey. Students choosing to enter the drawing were
3
The email solicitation generated 38 responses, while 71 students responded to the Facebook posts.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 6
provided with a link to a separate survey after completion of the initial survey and were asked to
provide their name. Two-thirds of respondents opted to enter the drawing, suggesting that the
incentive was at least somewhat effective in eliciting responses from the student population.
2e. Variables
Responses from the survey were used to construct the variables used in the empirical analysis of
this study. Table 1 describes and provides summary statistics for all of the variables used in this
study.
Dependent variable. Survey respondents indicated their overall GPA (to 2 decimals) for the
Fall 2014 semester. GPA is a continuous variable and is used as the measure of academic
performance in the current study. GPA is the dependent variable in the regressions.
Independent variables. The main variables of interest in this study are homeschool
background, time use, personality type, and parental education.
Survey respondents were asked to indicate if they were homeschooled during: (i) elementary
school; (ii) middle school; and (iii) high school. We used this information to create a dummy
variable equal to one if a student responded affirmatively to all three questions, and zero
otherwise. The resulting variable, Home, is our measure of homeschool background. More than
three-fourths of the sample (76%) were homeschooled during elementary, middle, and high
school.
Additionally, survey respondents were asked several questions related to their time use
during the Fall 2014 semester, including: (i) number of credit hours enrolled for the semester
(Credit); (ii) typical number of hours spent on extracurricular activities per week (Extra); (iii)
typical number of hours spent working a job per week (Work); (iv) typical number of hours spent
sleeping per night (Sleep); and (v) whether the student was single, married, or in a relationship
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 7
(Single). Responses to (i) – (iv) were used to construct continuous variables, while Single was
constructed as a dummy variable equal to one if the respondent indicated that they were single,
and zero if either married or in a relationship.
Survey respondents were also asked to select the highest level of education achieved by
both their father and mother from the following list: (i) High school; (ii) Some college; (iii)
Bachelor’s degree; (iv) Master’s degree; (v) Ph.D. We used this information to construct
educational attainment dummy variables for each parent equal to one if the parent earned a
bachelor’s degree or higher, and zero otherwise. The correlation between the two parental
education variables, Dad_college and Mom_college, is only 0.3, so we include both variables
simultaneously in the regressions.
4
Control variables. The survey asked a number of additional questions that were used to
construct control variables that potentially impact academic performance. These variables
include:
• ACT: ACT-equivalent standardized admissions test score;
5
• Judge: dummy variable equal to one if Judging personality type;
6
4
The results are both qualitatively and quantitatively similar when including only one parental education
variable in the baseline interaction model. Results omitted for space.
5
Respondents were asked to indicate their ACT or SAT score. 37 of the 109 respondents reported ACT scores
and the remaining 76 students reported SAT scores. Of the latter group, 55 reported a score above 1600, so
these scores are unambiguously on the 2400-point scale. The remaining SAT scores ranged from 1150-1600,
leaving some ambiguity as to whether these scores were on the 1600 or 2400-point scale. Through
discussions with the admissions department at PHC, we learned that the school rarely admits students with
an SAT score below 1000 (on the 1600 scale) and enrolls numerous students with SAT scores above 1500
every year. It was therefore assumed that the SAT scores less than or equal to 1600 were on the 1600-point
scale, although we flagged observations of 1500 and above to test the sensitivity of our results to this
assumption. SAT scores were converted to ACT-equivalent scores using the College Board’s concordance
table.
6
The Myers-Briggs Type Indicator (MBTI) instrument, which uses self-reported answers of easily recognized
reactions to identify people’s basic preferences regarding Perception and Judgment, is commonly used as a
barometer of personality. As of Fall 2014, all PHC students were required to take the MBTI assessment during
their first year as part of the college’s “Wisdom and Eloquence Portfolio” course. The MBTI assessment is
administered by the college each year for incoming students and the results included in each student’s
portfolio. Survey respondents were asked to indicate their personality type from the PHC-administered MBTI
assessment. This data was used to construct a dummy variable, Judge, equal to one if a student indicated that
they were Judging type, and zero if they indicated that they were Perceiving type.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 8
• Campus: dummy variable equal to one if lived on campus during Fall 2014;
• Class_fre: dummy variable equal to one for freshmen;
• Class_soph: dummy variable equal to one for sophomores;
• Class_jun: dummy variable equal to one for juniors;
• Class_sen: dummy variable equal to one for seniors;
• Major_eba: dummy variable equal to one if an Economics and Business Analytics
major;
• Major_gov: dummy variable equal to one if an American Politics & Policy,
International Politics & Policy, Political Theory, Strategic Intelligence or General
Government major;
• Major_lib: dummy variable equal to one if a classical liberal arts, history, or literature
major; and
• Major_jou: dummy variable equal to one if a journalism major.
7
2f. Data Analysis
We used the program Stata to perform the statistical analysis. OLS regression was selected as the
most appropriate multivariate method to analyze the determinants of academic performance for
several reasons:
1. The dataset is cross-sectional in nature;
2. The dependent variable is continuous;
3. We assume that the population model is linear in its parameters;
4. The data was collected using a random sampling technique, as discussed in the Sample
Selection sub-section above;
5. There is no multi-collinearity between any of the independent variables; and
6. The independent variables are assumed to be exogenous.
The satisfaction of conditions 3-6 ensures that the OLS estimator is unbiased (Woolridge,
2013). Because we are testing whether the impact of homeschool background on college
academic performance is conditional on parental education and whether the impact of pre-
college admission test scores on college academic performance is moderated by homeschooling
background, it is assumed that the variance in the error term of the model is conditional on the
7
As of the time of the survey, PHC only offered 10 different majors. Because of the limited sample size and
concern over degrees of freedom in the regression analysis, we grouped the majors into the four categories
described above.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 9
independent variables. We correct for this heteroskedasticity by using robust standard errors
(White, 1980).
Equation 1 describes the baseline model used to estimate the determinants of academic
performance, measured by Fall 2014 GPA, where:
• E is a matrix of dummy variables indicating educational attainment of parents;
• T is a matrix of the time use variables;
• C is a matrix of dummy variables for class status;
• M is a matrix of dummy variables for academic major;
• , , , , and are scalar partial effects;
• , and are vectors of partial effects; and
• e is an idiosyncratic error.
(1)
The two interaction terms, and , are included to test null
hypotheses and , respectively. A failure to reject suggests that there is no difference in
academic performance between homeschooled students with college-educated parent(s) and
homeschooled students whose parents do not have a college education, whereas a rejection of
suggests the opposite, that homeschooled students with college-educated parent(s) perform
better academically in college than their homeschooled peers whose parent(s) do not have a
college education. Similarly, suggests that there is no difference in the effect of standardized
tests scores on academic performance between homeschooled and traditionally schooled
students, whereas a rejection ofsuggests the opposite, that the effect of standardized test
scores on academic performance is moderated (or enhanced) for students who were
homeschooled.
8
8
The null hypothesis requires a two-sided test, so the effect could plausibly be enhanced by homeschool
background as well.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 10
As discussed by Brambor et al. (2006), we cannot draw meaningful statistical inference about
null hypotheses and from the standard errors associated with the respective interaction
terms in isolation. This is because we are interested in the partial effects for Home, or
, and ACT, or
. As such, we test the joint significance of the
partial effects (Woolridge, 2013), as further discussed in the results section below.
2g. Limitations
As with all empirical studies, ours has several limitations. First, the results should be viewed as
correlational and not causal because of the potential endogeneity of some of the independent and
control variables. Second, the results may suffer from omitted variable bias as we are unable to
directly control for ability or intelligence. We believe that parental education and admissions test
results serve as decent proxies, so this bias is likely small. Third, our study only examines time
use and academic performance for a single semester so the results may differ across time periods.
Lastly, given the uniqueness of the institution and its student body, the results of the sample are
likely not generalizable to the U.S. student population.
3. Literature Review
3a. Student Time Use and Academic Performance
How a student chooses to allocate her time while in college may impact her academic
performance. Time is a scarce resource so the more time that a student spends on non-academic
activities such as dating, extracurricular activities, sleep, and work, the less time that she will
have to devote to homework and studying, which could lead to poorer academic performance.
The concept of diminishing marginal returns suggests that rather than academic performance
being a linear function of time spent on coursework, the marginal impact of studying declines the
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 11
more one does it. The standard microeconomic assumption of convex utility suggests that
individuals prefer a mixture of goods and services. Because time is a binding constraint that a
student must allocate across activities, it follows that students should prefer to expend their time
“consuming” a mixture of activities. To achieve higher levels of satisfaction, a student is faced
with a smaller supply of time to allocate towards studying and may therefore either face a
tradeoff between schoolwork and other activities, hindering academic performance, or develop
more efficient study habits and/or better time management skills, resulting in better academic
performance. Study time may even exert a negative burnout effect for students who consistently
burn the symbolic midnight oil, sacrificing sleep and other activities that provide them with
satisfaction. Previous empirical research on student time use and academic performance has
produced mixed results.
Academic success in college is linked to student retention, for it requires staying enrolled.
Two theories from the education literature guide research here (Danbert et. al., 2014). First is the
theory of departure, which suggests that student retention is closely tied to student integration
with an institution and its members (Tinto, 1975). Next is the student involvement theory, which
posits that the more involved a college student, the greater his learning and personal
development (Astin, 1999). Based on these theories, the level of student involvement—
“investment of physical and psychological energy in various objects”—affects academic success,
usually measured by college GPA (Astin, 1984: 519), which is the best predictor of degree
completion (Kuh et al., 2006; Pascarella and Terenzini, 2005).
Zacherman and Foubert (2014) found that low and medium levels of activities per week
are both associated with higher GPA, with the former exhibiting a larger effect, while high levels
are associated with lower GPA. Zehner (2011) found that among students with similar SAT
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 12
scores, highly engaged students out-performed the less-engaged regardless of class level,
although upperclassmen tended to have higher GPAs regardless of engagement. More physically
active students tend to have better alertness, time management skills, and less stress (Zehner,
2014; Neubert, 2013), so GPA may also be related to physical fitness and wellness. Neubert
(2014) provided evidence that academic performance is higher among students frequenting the
fitness center at Purdue University, and Danbert et al. (2014) documented that freshmen and
sophomores at Michigan State University with fitness center memberships exhibit higher GPAs
and retention rates. In the current study, we estimate the impact of extracurricular activities on
GPA among PHC students during the Fall 2014 semester.
Sleep habits also affect one’s wellness, and sleep-related problems are prominent among
college students (Gilbert and Weaver, 2010). Trockel et. al. (2000) found that sleep habits,
especially weekday and weekend wake-up times, affected first-year GPA more than other health-
related factors. Kelly et al. (2001) found that short sleepers (6 or fewer hours) averaged
significantly lower GPAs than long sleepers (9+ hours), while the GPA of average sleepers (7-8
hours) did not significantly differ from that of long or short sleepers. Taylor et al. (2010) found
that later bedtimes and wakeup times, longer time awake after rising, and inconsistent sleep
habits are all related to lower cumulative GPA. Gilbert and Weaver (2010) found a negative
correlation between the GPA of non-depressed college students and self-reported sleep quality
(2010). In the current study, we estimate the impact of average number of hours slept on GPA
among PHC students during the Fall 2014 semester.
Having a job may also affect a student’s academic performance. Working reduces the
amount of time available to study, so work may be associated with lower grades. Although some
studies have found that hours worked are negatively associated with academic performance
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 13
(Stinebrickner and Stinebrickner, 2003; Brennan et al., 2005; DeSimone, 2008; Kalenkoski and
Pabilonia, 2010), some have found that only working more than 20 hours a week has a
detrimental effect, and others that students working a limited number of hours actually perform
better than students who do not work at all (NCES, 1994; Light, 2001; Dundes and Marx, 2006).
Ehrenberg and Sherman (1987) meanwhile found no relationship between student work and GPA
among male college students, and Astin (1975) provides evidence that on-campus jobs are
positively associated with GPA, possibly because they entail greater involvement with the
institution (Furr and Elling, 2000). In the current study, we estimate the impact of average work
hours per week on GPA among PHC students during the Fall 2014 semester.
The number of credits that a student takes can also affect their grades. Students taking
larger course loads have less time to spend studying for each course, so they may receive lower
average grades than a similar student taking a lighter course load. Two studies have found,
however, that students taking more credits tended to earn higher GPAs, regardless of academic
department or major (Szafran, 2001; Khouj et. al., 1982). In the current study, we estimate the
impact of the number of credit hours taken on GPA among PHC students during the Fall 2014
semester.
A student’s relationship status could influence their academic performance. There is
substantial empirical evidence that being married or in an intimate relationship is associated with
better physical and psychological well-being (e.g., Coombs, 1991; Dolan et al., 2008).
Braithwaite et al. (2010) provide evidence of a link between romantic relationships and well-
being among college students. Two theories of why people in relationships experience better
well-being, social support (Coombs, 1991) and behavioral regulation (Liktwak et al., 1989),
could also serve as mechanisms linking student relationship status to academic performance.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 14
However, students in serious romantic relationships may have greater external demands on their
time than single students, leading to less time available to study, hindering academic
performance. In the current study, we estimate the impact of relationship status on GPA among
PHC students during the Fall 2014 semester.
3b. Homeschool Background and Academic Performance
Cogan (2010) provided evidence that homeschooled students score higher on standardized tests
and earn higher GPAs than traditionally schooled students. Jones and Gloeckner (2004),
however, found no statistical difference in first-year GPA between homeschooled and
traditionally schooled students. Snyder (2013), meanwhile, observed a significant difference in
overall GPA between homeschooled and public-schooled students and between homeschooled
and Catholic-schooled students. However, neither major GPA nor core GPA significantly
differed between homeschooled and public schooled or Catholic schooled students. In the current
study, we estimate the impact of the number of credit hours taken on GPA among PHC students
during the Fall 2014 semester.
First-generation college students have lower persistence and graduation rates due to
factors like poorer high school records, lower educational aspirations, different priorities and
goals, and less college-related knowledge and experience (Pike and Kuh, 2005; Kuh et. al., 2006;
High School Survey of Student Engagement, 2005). Warburton et al. (2001) found them less
likely to be as academically prepared than second-generation students, meaning generally lower
GPAs. However, first-generation students with high school educations comparable to that of
second-generation students did not significantly differ in first-year GPAs. Studies by Ray (2000),
Rudner (1999), and Collom (2005) found that students homeschooled by more educated parents
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 15
perform better academically. In the current study, we also test if the impact of homeschooling
background on academic performance is conditional on parental educational attainment.
Pre-college academic preparation is a good predictor of academic performance in college.
For instance, SAT scores are positively correlated with college GPA (Betts and Morell, 1999;
Zehner, 2011), although some researchers have found that the positive effect of aptitude tests on
GPA diminishes after accounting for involvement in campus activities (Kuh et al., 2008; Sackett
et al., 2009). In the current study, we control for the impact of ACT score on GPA among PHC
students during the Fall 2014 semester. Additionally, we test if homeschool background has a
moderating impact on the effect of standardized tests on academic performance.
3c. Other Determinants of Academic Performance
Academic performance may also be affected by non-cognitive factors such as personality type.
Research suggests that Judging types have, on average, higher GPAs and retention rates than
Perceiving types (Schurr and Ruble, 1988; Barrineau, 2005; DiRienzo et. al., 2010; Sanborn,
2013). Differences in academic performance between personality types have been linked to
differences in learning styles, responses to teaching styles, and project time management styles
(Schurr and Ruble, 1988; Beckham, 2012). In the current study, we control for the impact of
Judging vis-à-vis Perceiving personality type on GPA among PHC students during the Fall 2014
semester.
Academic performance differs by major at many colleges and universities. For instance,
Rask (2010) found that students with the lowest average GPAs majored in subjects such as
chemistry, math, economics, psychology, and biology; students majoring in subjects such as
education, language, English, music, and religion exhibited the highest average GPAs. Grades
can provide a powerful incentive regarding course choices. Rask (2010) also found that some
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 16
students majoring in the STEM fields become discouraged by their grades—which were
significantly lower than those of their peers majoring in other areas of study—to the point of
changing majors. Because many colleges and universities have high- and low-grading
departments, this can hinder students from knowing their relative strengths and weaknesses. It
also makes grades and various proxies of ability (e.g., SAT score or parental education) less
accurate or poor predictors of grades in high-grading departments (Sabot, 1991). In the current
study, we control for the potential impact of major on academic performance. We also control
for the potential impact of a student’s class on academic performance because most students
follow a very similar course sequence at PHC during their first two years of college, but their
field of study determines their course sequence during their final two years.
Additionally, we control for whether a student lived on or off-campus during the Fall
2014 semester because on-campus living has been linked to better academic performance
(Thompson et al., 1993; Araujo and Murray, 2005).
4. Results
Table 2 reports the OLS regressions estimates of Equation 1 for the full sample, with
heteroskedastic-robust standard errors in parentheses. Model 1 omits the interaction terms
depicted in Equation 1 and serves as the baseline model. Model 2 introduces the ACT
interaction term to the baseline model. Model 3 adds the Dad_college and
Mom_college interaction terms to the baseline model. Model 4 simultaneously includes
all three interaction terms. Models 5 and 6 re-estimate Model 4 for the samples of underclassmen
and upperclassmen, respectively. Table 3 re-estimates the same models as Table 2, excluding
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 17
observations for which the student reported an SAT score between 1500 and 1600 (see Footnote
5 for details).
In the below analysis, statistical significance is tested at the 10 percent level. We find a
positive and statistically significant association between GPA and the control variables ACT and
Judge throughout Table 2. These results are consistent with previous research that has found
entrance exams (Betts and Morell, 1999; Zehner, 2011) and judging personality types (Schurr
and Ruble, 1988; Barrineau, 2005; DiRienzo et. al., 2010; Sanborn, 2013) to be a strong
predictor of academic performance. We also find that journalism majors have higher GPAs, all
else equal, but this result only holds for the full sample. We do not find statistically significant
results for the remaining control variables. Findings with respect to the null hypotheses outlined
above are discussed in the following sub-sections.
4a. Some Uses of Time Hinder Academic Performance
Among the time use variables, only Extra and Single are statistically significant in the baseline
estimates (Model 1). The coefficient for Extra is negative, suggesting that, inconsistent with null
hypothesis, PHC students who spend more time on extracurricular activities perform worse
academically, all else equal. This result holds and the coefficient is stable when the interaction
terms are introduced in Models 2-4, although the magnitude of the effect is slightly higher for
underclassmen (Model 5) than upperclassmen (Model 6) in both Tables 2 and 3. This result
contributes to the literature on extracurricular activities and academic performance, which has
produced somewhat mixed results (Zacherman and Foubert, 2014; Zechner, 2011).
The coefficient for Single is negative and relatively stable in Models 1-4 of both Tables 2
and 3. The coefficient remains negative when the samples are restricted to underclassmen and
upperclassmen in Models 5 and 6, respectively, but it is no longer statistically significant. The
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 18
negative effect of being single on academic performance suggests that, all else equal, students at
PHC in serious romantic relationships performed better academically than their single
counterparts. This provides some corroboration for the social support (Coombs, 1991) and
behavioral regulation (Liktwak et al., 1989) hypotheses discussed in the literature review section.
Meanwhile, the other time use variables (Credit, Work, and Sleep) are not statistically
significant in any of the models in Table 2, providing some support for null hypothesis that
time spent on non-academic activities has no effect on academic performance.
4b. Homeschooling Does Not Influence Academic Performance
Home is positive but is not statistically significant in the baseline estimates (Model 1) of either
Table 2 or 3, suggesting that, consistent with null hypothesis, there is no difference in
academic performance between homeschooled and traditionally schooled students at PHC.
Model 2 introduces the interaction term, so we must test the joint significance
of the interaction term and the constituent term, Home, as discussed in Section 2f. The p-value
from this test is reported as p(Home). The results suggest that the partial effect of homeschooling
background is not statistically significant in Model 2.
Similarly, Model 3 introduces the and
interaction terms to the baseline model. We test the joint significance of Home and the two
interaction terms, reporting the p-value from this test as p(Home). The results indicate that the
partial effect of Home is not statistically significant. Meanwhile, Model 4 simultaneously
includes all three interaction terms, so we test the joint significant of Home and all three
interaction terms. Once again, the partial effect of Home is not statistically significant, as
indicated by p(Home). Finally, Models 5 and 6 restrict the sample to underclassmen and
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 19
upperclassmen, respectively. Once again, Home is not statistically significant in either model, as
indicated by p(Home).
Overall, the results from Tables 2 and 3 suggest that, consistent with , there is no
difference in academic performance between homeschooled students and traditionally schooled
students at PHC. These findings are consistent with those of Jones and Gloeckner (2004), but
inconsistent with those reported by Cogan (2010) and Snyder (2013).
4c. Homeschooling by College-Educated Parents (Largely) Does Not Influence Academic
Performance
The null hypothesis suggests that there is no difference in academic performance between
homeschooled students with college-educated parents and homeschooled students whose parents
do not have a college education. As mentioned above, Models 3-6 include two interaction
terms, and . The inclusion of these two
interaction terms allows us to testby examining the partial effect of Home. As discussed in
Section 4b, the partial effect of Home is not statistically significant in any of these models,
suggesting that parental education does not influence the effect of homeschooled students’
academic performance at PHC.
Additionally, we separately examine the partial effects of Dad_college and Mom_college by
testing the joint significance of each parental education constituent term and its interaction with
Home. These results are reported as p(Dad) and p(Mom), respectively, but neither is statistically
significant in Models 3-5 in either Table 2 or 3. The partial effect of Dad_college is statistically
significant when the sample is restricted to upperclassmen only (Model 6), suggesting that
homeschooled upperclassmen whose father completed a college education performed better
academically than homeschooled students whose father did not complete a college education.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 20
Overall, the results discussed in this section are largely consistent with. Homeschool
background does not impact academic performance of PHC students, irrespective of whether a
student’s parents completed a college education. The findings also largely indicate that parental
education is not a good predictor of academic performance, regardless of whether a student was
homeschooled or not, although we do find that homeschooled upperclassmen whose father has a
college education perform better academically, all else equal.
4d. Entrance Exams Better Predict Academic Success for Traditionally Schooled Students
The null hypothesis suggests that the impact of pre-college entrance exams on academic
performance is not conditional on whether a student was homeschooled. As mentioned above,
Model 2 and Models 4-6 include the interaction term, allowing us to testby
examining the partial effect of ACT. As before, this involves testing the joint significance of the
constituent term, ACT, and interaction term, . The former term enters positively
and the latter negatively, and the p-value for the joint test of their significance suggests that the
partial effect of ACT is statistically significant.
These results suggest that we reject and conclude that the impact of ACT scores on GPA
is lower for homeschooled than traditionally schooled students at PHC. The coefficient for the
constitutive term, ACT, represents the effect for traditionally schooled students, while the sum of
the coefficients for the constitutive and interaction terms represents the effect for homeschooled
students. The latter is reported as ACT(Home).
5. Conclusion
Previous studies have produced mixed results regarding the impact of homeschooling on
collegiate academic performance (Cogan, 2010; Jones and Gloeckner, 2004; Snyder, 2013). This
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 21
study contributes to this line of literature by examining the determinants of academic
performance during Fall 2014 at PHC, a private selective Christian liberal arts college that
predominantly serves homeschooled students. As discussed in the methods section, PHC
provides a well-controlled environment for such a study because of the relative homogeneity in
terms of religious beliefs and educational background of the student population, as well as
relative homogeneity of the coursework and residency status of students.
We conducted an anonymous and voluntary survey of PHC students to gather data about
their GPA and time use during the Fall 2014 semester, as well as information about their
homeschool background, parental education level, personality type, class rank, and major. Our
sample consists of one-third of the student population at PHC. Using OLS regression analysis,
we tested four research questions.
First, we examined whether student time use impacts academic performance. We tested
the effect of numerous factors, including time spent on extracurricular activities, paid work, and
sleeping, as well as relationship status and the number of credit hours enrolled. The results
suggest that students who spend more time on extracurricular achieve lower GPAs. Additionally,
students not involved in a romantic relationship perform worse than students in a relationship.
Second, we tested whether there is a difference in academic performance between
homeschooled and traditionally schooled students at PHC. The results suggest that, all else equal,
there is no difference in the GPAs of homeschooled and traditionally schooled students. This
result holds when we condition the effect of homeschooling on parental education, which is the
third hypothesis tested. The latter result was particularly surprising, given that previous research
has found that students homeschooled by more educated parents perform better academically
(Ray, 2000; Rudner, 1999; Collom, 2005). The finding that homeschooled students do not
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 22
perform differently academically than traditionally schooled students at PHC should not be
viewed as a detrimental finding regarding the academic achievement of homeschooled students,
as PHC is a selective institution. Rather, it suggests that homeschooled students perform as well
as their traditionally schooled peers who also entered college well-prepared academically.
Finally, we test if homeschooling background moderates the effect of college entrance
exam scores on academic performance. This question is important because previous research has
found that homeschooled students perform as well as or better than traditionally schooled
students on academic achievement tests (Rudner, 1999; Ray, 1997, 2016). ACT and SAT test
scores are routinely used by college admissions officers as a means to assess the likelihood for
academic success of a prospective student. Our results indicate that the partial effect of ACT-
equivalent scores on GPA is higher for traditionally schooled than homeschooled students at
PHC, potentially suggesting that college entrance exam scores may not be as accurate a predictor
of collegiate academic success for homeschooled students.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 23
References
The ACT. Accessed April 3, 2015. http://www.act.org/aap/concordance/estimate.html.
Araujo, P. D. and Murray, J. (2010). Estimating the Effects of Dormitory Living on Student
Performance. Center for Applied Economics and Policy Research, Department Economics at
Indiana University Bloomington.
Astin, A. W. (1999). Student involvement: A developmental theory for higher education.
Journal of College Student Development, 40(5), 518-529.
Barrineau, P. (2005). Personality Types Among Undergraduates Who Withdraw from Liberal
Arts Colleges. Journal of Psychological Type, 64(4), 27-32.
Beckham, M. H. (2012). Building Momentum: The Unconventional Strengths of Perceiving
College Students, Journal of Psychological Type, 72(2), 27-40.
Betts, J. R., and Morell, A. (1999). The Determinants of Undergraduate Grade Point Average:
The Relative Importance of Family Background, High School Resources, and Peer Group
Effects. The Journal of Human Resources, 34(2), 268-293.
Brambor, T., Clark, W.R., and Golder, M. (2006). Understanding Interaction Models: Improving
Empirical Analyses. Political Analysis, 14(1): 63-82.
Braithwaite, S. R., Delevi, R., and Fincham, F.D. (2010). Romantic Relationships and the
Physical and Mental Health of College Students. Personal Relationships, 17(1): 1-12.
Brennan, J., Callender, C., Duaso, A., Little, B. & Van Dyke, R. (2005). Survey of Higher
Education Students’ Attitudes to Debt and Term-Time Working and Their Impact on
Attainment. Centre of Higher Education Research and Information and London South Bank
University.
Center for Applications of Psychological Type. “About the MBTI Instrument.”
Cogan, M. F. (2010). Exploring Academic Outcomes of Homeschooled Students. Journal of
College Admission. National Association for College Admission Counseling. Summer, 18-
25.
Collom, E. (2005). The Ins and Outs of Homeschooling: The Determinants of Parental
Motivations and Student Achievement. Education and Urban Society, 37(3): 307-335.
Coombs, R. H. (1991). Marital Status and Personal Well-Being: A Literature Review. Family
Relations, 40(1): 97-102.
Coulson, A. J. (1999). Market Education: The Unknown History, New Brunswick: Transaction
Publishers.
Danbert, S.J., Pivarnik, J. M., McNeil, R.N., and Washington, I. J. (2014). Academic Success
and Retention: The Role of Recreational Sports Fitness Facilities. Recreational Sports
Journal, 38(1), 14-22.
DeSimone, J. S. (2008). The Impact of Employment during School on College Student Academic
Performance. National Bureau of Economic Research, NBER Working Paper No. 14006.
DiRienzo, C., Das, J., Synn, W., Kitts, J., and McGrath, K. (2010). The Relationship between
MBTI and Academic Performance: A Study across Academic Disciplines, Journal of
Psychological Type, 70(5), 53-67.
Dolan, P., Peasgood, T., and White, M. (2008). Do We Really Know What Makes Us Happy? A
Review of the Economic Literature on the Factors Associated with Subjective Well-Being.
Journal of Economic Psychology, 29(1): 94-122.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 24
Dundes, L. and Marx, J. (2006/2007). Balancing Work and Academics in College: Why do
Students Working 10 to 19 Hours Per Week Excel? Journal of College Student Retention,
8(1), 107-120.
Ehrenberg, R. G. and Sherman, D. R. (1987). Employment While in College, Academic
Achievement and Post-College Outcomes: A Summary of Results. Journal of Human
Resources, 22(1), 1-23.
Gilbert, S. P., and Weaver, C. C. (2010). Sleep Quality and Academic Performance in University
Students: A Wake-Up Call for College Psychologists. Journal of College Student
Psychotherapy, 24(4), 295-306.
High School Survey of Student Engagement (HSSSE) (2005). Getting Students Ready for
College: What Student Engagement Data Can Tell Us.
Jones, P. and Gloeckner, G. (2004). First-Year College Performance: A Study of Home School
Graduates and Traditional School Graduates. Journal of College Admission, 183 (March),
17-20.
Kalenkoski, C. M. and Pabilonia, S. W. (2010). Parental transfers, student achievement, and the
labor supply of college students. Journal of Population Economics, 23(2), 469-496.
Kelly, W. E., Kelly, K. E., and Clanton, R. C. (2001). The Relationship between Sleep Length
and Grade-Point Average Among College Students. College Student Journal, 35(1), 84-86.
Khouj, Z., Mohammad, A., Shami, A., and Ansar, M. (1982). Relationship of Students’ Course
Loads with their Grade Point Average Scores, Study II: An Occasional Paper. Umm Al-Qura
University, Mecca, Saudia Arabia. College of Education.
Kuh, G. D., Cruce, T. M., Shoup, R., Kinzie, J., and Gonyea, R. M. (2008). Unmasking the
Effects of Student Engagement on First-Year College Grades and Persistence. The Journal of
Higher Education, 79(5): 540-563.
Kuh, G. D., Kinzie, J., Buckley, J. A., Bridges, B. K., and Hayek, J. C. (July 2006). What
Matters to Student Success: A Review of the Literature. Commissioned Report for the
National Symposium on Postsecondary Student Success: Spearheading a Dialog on Student
Success.
Lines, P. M. (1991). Estimating the Home Schooled Population. Working Paper OR 91-537.
Office of Research, U.S. Department of Education, Washington, D.C.
Lines, P. M. (2000). Homeschooling Comes of Age. The Public Interest, 140: 74-85.
Lips, D. and Feinberg, E. (2008). Homeschooling: A Growing Option in American Education.
Backgrounder No. 2122, Washington, DC: The Heritage Foundation.
Litwak, E., Messeri, P., Wolfe, S., Gorman, S., Silverstein, M., and Guilarte,
M. (1989). Organizational Theory, Social Supports, and Mortality Rates: A Theoretical
Convergence. American Sociological Review, 54(1): 49–66.
Michigan State University (July 10, 2014). “Want a Higher GPA in College? Join a Gym.”
Neubert, A. P. (April 15, 2013). College students working out at campus gyms get better grades.
Purdue Today, Purdue University.
No author given. September 3, 2014. New college students who visit campus gym regularly see
boost in their first semesters. Purdue Research News, Purdue University.
Pascarella, E.T. and Terenzini, P.T. (2005). How College Affects Students: A Third Decade of
Research. San Francisco: Jossey-Bass.
Patrick Henry College Catalog. 2014-2015. Patrick Henry College.
Patrick Henry College Fact Book. 2014-2015. Patrick Henry College.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 25
Pike, G. R. and Kuh, G. D. (2005). First- and Second-Generation College Students: A
Comparison of Their Engagement and Intellectual Development. The Journal of Higher
Education, 76(3), 276-300.
Rask, K. (2010). Attrition in STEM Fields at a Liberal Arts College: The Importance of Grades
and Pre-Collegiate Preferences. Cornell University, School of Industrial and Labor Relations.
Working Paper 3-2010.
Ray, B. (1997). Strengths of Their Own: Home Schoolers Across America. Salem, OR: National
Home Education Research Institute.
Ray, B. D. (2000). Homeschooling: The Ameliorator of Negative Influences on Learning?
Peabody Journal of Education, 75(1/2): 71-106.
Ray, B. (2016). Research Facts on Homeschooling, National Home Education Research Institute
(March 23).
Ray, B. and Eagleson, B. (2008). State Regulation of Homeschooling and Homeschoolers’ SAT
Scores. Academic Leadership Journal, 6(3), Article 17.
Redford, J., Battle, D., and Bielick, S. (2017). Homeschooling in the United States: 2012 (NCES
2016-096.REV). National Center for Education Statistics, Institute for Education Sciences,
U.S. Department of Education. Washington, D.C.
Reich, R. (2005). Why Homeschooling Should be Regulated. In Homeschooling in Full View: A
Reader. Bruce S. Cooper, ed., Greenwich, CT: Information Age Publishing.
Reynolds, C. L. (2012). Estimating the Causal Effects of Dormitory Residence During College
on Student Outcomes. Kent State University, Department of Economics. Working paper.
Romanowski, M. H. (2006). Revisiting the Common Myths about Homeschooling. The Clearing
House: A Journal of Educational Strategies, Issues and Ideas, 79(3): 125-129.
Rudner, L. M. (1999). Scholastic Achievement and Demographic Characteristics of Homeschool
Students in 1998. Educational Policy Analysis Archives, 7(8): 1-38.
Sabot, R. and Wakeman-Linn, J. (1991). Grade Inflation and Course Choice. Journal of
Perspectives, 5(1), 159-170.
Sackett, P. R., Kuncel, N. R., Arneson, J. J., Cooper, S. J., and Waters, S.D. (2009).
Socioeconomic Status and the Relationship Between the SAT and Freshman GPA: An
Analysis of Data from 41 Colleges and Universities. The College Board Research Report No.
2009-1.
Sanborn, D. K (2013). Myers-Briggs preferences and academic success in the first college
semester. Iowa State University, Graduate Theses and Dissertations. Paper 13095.
Schurr, K. T. and Ruble, V. (1986). The Myers-Brigg Type Indicator and First-Year College
Achievement: A Look Beyond Aptitude Test Results. Journal of Psychological Type, 12(5),
25-37.
——— (1988). Psychological Type and the Second Year of College Achievement: Survival and
the Gravitation Toward Appropriate and Manageable Major Fields. Journal of Psychological
Type, 14(9), 57-59.
Snyder, M. (2013). An Evaluative Study of the Academic Achievement of Home-schooled
Versus Traditionally Schooled Students Attending a Catholic University. Catholic
Education: A Journal of Inquiry and Practice, 16(2), 288-308.
Stinebrickner, R. and Stinebrickner, T.R. (2003). Working during School and Academic
Performance. Journal of Labor Economics, 21(2), 473-492.
Szafran, R. (2001). The Effect of Academic Load on Success for New College Students: Is
Lighter Better? Research in Higher Education, 42(1), 27-50.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 26
Taylor, D. J., Vatthauer, K. E., Bramoweth, A. D., Ruggero, C. and Brandy, M. (2013). The Role
of Sleep in Predicting College Academic Performance: Is it a Unique Predictor? Behavioral
Sleep Medicine, 11(3), 159-172.
Thompson, J., Samiratedu, V., and Rafter, J. (1993). The Effects of On-Campus Residence on
First-Time College Students. NASPA Journal, 31(1), 41-47.
Tinto, V. (1975). Dropout from Higher Education: A Theoretical Synthesis of Recent Research.
Review of Educational Research, 45(1), 89-125.
——— (2006/2007). Research and Practice of Student Retention: What Next? Journal of
College Student Retention, 8(1), 1-19.
Trockel, M. T., Barnes, M. D., and Egget, D. L. (2000). Health-Related Variables and Academic
Performance Among First-Year College Students: Implications for Sleep Other Behaviors.
Journal of American College Health, 49(3), 125-131.
Van Galen, J. A. (1991). Ideologues and Pedagogues: Parents who Teach their Children at
Home. In Homeschooling: Political, Historical, and Pedagogical Perspectives, Jane Van
Galen and Mary Anne Pitman, eds., Norwod, NJ: Ablex.
Warburton, E. C., Bugarin, R., and Nuñez, A. (2001). Bridging the Gap: Academic Preparation
and Postsecondary Success of First-Generation Students. Statistical Analysis Report (May):
National Center for Education Statistics 2001-153. U.S. Department of Education.
White, H. (1980). A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct
Test for Heteroskedasticity. Econometrica, 48 (4): 817–838.
Woolridge, J. M. (2013). Introductory Econometrics: A Modern Approach 5th Edition. Mason,
OH: South-Western, Cengage Learning.
Zacherman, A. and Foubert, J. (2014). The Relationship Between Engagement in Cocurricular
Activities and Academic Performance: Exploring gender differences. Journal of Student
Affairs Research and Practice, 51(2), 157-169.
Zehner, A. (2011). Co-Curricular Activities and Student Learning Outcomes. Purdue University,
Office of Institutional Assessment.
Zehner, A. and Zelaya, T. (2014). Physically Active Students Get Better Grades. Purdue
University, Student Development and Assessment Department.
Zimmerman, D. J. (2003). Peer Effects in Academic Outcomes: Evidence from a Natural
Experiment. The Review of Economics and Statistics, 85(1), 9-23. Accessed April 3,
2015.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 27
TABLE 1: VARIABLE DESCRIPTIONS & SUMMARY STATISTICS
Variable
Mean
Stdev
Min
Max
N
Description
GPA
3.54
0.46
1.92
4.00
109
Fall 2014 grade point average.
Home
0.76
0.43
0.00
1.00
109
Dummy variable =1 if homeschooled in elementary,
middle, and high school.
Time Use Variables
Extra
8.26
6.74
0.00
30.00
109
Average number of hours spent per week on
extracurricular activities during Fall 2014.
Work
10.63
10.94
0.00
60.00
109
Average number of hours worked per week in a job
during Fall 2014.
Sleep
6.76
1.09
4.00
11.00
109
Average number of hours slept per night during Fall
2014.
Credit
15.72
2.01
7.00
21.00
109
Number of credit hours enrolled for during Fall 2014.
Single
0.68
0.47
0.00
1.00
109
Dummy variable =1 if single (not married and not in a
romantic relationship) during Fall 2014.
Parental Education Variables
Dad_college
0.84
0.36
0.00
1.00
109
Dummy variable =1 if father completed a bachelor’s
degree or more.
Mom_college
0.72
0.45
0.00
1.00
109
Dummy variable =1 if mother completed a bachelor’s
degree or more.
Control Variables
ACT
30.29
3.01
22.00
36.00
109
Score on ACT test. Some students reported SAT score.
Converted to ACT using College Board concordance
table.
Judge
0.85
0.36
0.00
1.00
109
Dummy variable =1 if scored as a Judging type on the
Myers-Briggs personality test.
Campus
0.91
0.29
0.00
1.00
109
Dummy variable =1 if living on campus during Fall
2014.
Class_fre
0.28
0.45
0.00
1.00
109
Dummy variable =1 if student a freshman during Fall
2014.
Class_soph
0.25
0.43
0.00
1.00
109
Dummy variable =1 if student a sophomore during Fall
2014.
Class_jun
0.24
0.43
0.00
1.00
109
Dummy variable =1 if student a junior during Fall 2014.
Class_sen
0.24
0.43
0.00
1.00
109
Dummy variable =1 if student a senior during Fall 2014.
Major_eba
0.09
0.29
0.00
1.00
109
Dummy variable =1 if an Economics & Business
Analytics major.
Major_gov
0.56
0.50
0.00
1.00
109
Dummy variable =1 if an American Politics & Policy,
International Politics & Policy, Political Theory,
Strategic Intelligence, or General Government major.
Major_lib
0.30
0.46
0.00
1.00
109
Dummy variable =1 if a classical liberal arts, history, or
literature major.
Major_jou
0.05
0.21
0.00
1.00
109
Dummy variable =1 if a journalism major.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 28
TABLE 2: OLS REGRESSION RESULTS
GPA is dependent variable in all models
(1)
(2)
(3)
(4)
(5)
(6)
Home
0.083
1.380
-0.047
1.545
0.818
3.120*
(0.110)
(1.027)
(0.303)
(1.129)
(1.253)
(1.807)
Parental Education Variables
Dad_College
0.185
0.177
0.149
0.095
0.198
-0.605**
(0.143)
(0.145)
(0.309)
(0.298)
(0.399)
(0.259)
Mom_College
0.041
0.041
-0.058
-0.132
-0.291
-0.425
(0.105)
(0.106)
(0.271)
(0.304)
(0.359)
(0.360)
Time Use Variables
Work
-0.004
-0.003
-0.004
-0.004
-0.018*
0.000
(0.004)
(0.004)
(0.004)
(0.004)
(0.010)
(0.005)
Extra
-0.021***
-0.021***
-0.021***
-0.021***
-0.029**
-0.018*
(0.008)
(0.008)
(0.008)
(0.008)
(0.013)
(0.010)
Sleep
-0.030
-0.020
-0.035
-0.025
0.018
-0.032
(0.037)
(0.037)
(0.038)
(0.038)
(0.057)
(0.066)
Credit
0.017
0.017
0.017
0.017
0.083
0.011
(0.016)
(0.016)
(0.016)
(0.016)
(0.050)
(0.023)
Single
-0.159*
-0.151*
-0.155*
-0.139*
-0.122
-0.108
(0.085)
(0.085)
(0.079)
(0.080)
(0.134)
(0.113)
Control Variables
ACT
0.069***
0.103***
0.069***
0.114***
0.093**
0.187***
(0.013)
(0.032)
(0.013)
(0.038)
(0.039)
(0.062)
Judge
0.306**
0.332**
0.308**
0.343**
0.724***
0.113
(0.141)
(0.142)
(0.139)
(0.136)
(0.207)
(0.137)
Campus
0.154
0.171
0.154
0.177
0.311
0.200
(0.115)
(0.117)
(0.115)
(0.116)
(0.297)
(0.172)
Interaction Terms
HomeACT
-0.043
-0.057
-0.035
-0.138**
(0.033)
(0.040)
(0.043)
(0.067)
HomeDad_college
0.047
0.106
0.044
0.914**
(0.326)
(0.318)
(0.400)
(0.389)
HomeMom_college
0.132
0.230
0.276
0.569
(0.268)
(0.300)
(0.353)
(0.362)
Other Statistics
N
109
109
109
109
57
52
F
4.468
3.857
4.360
3.813
4.370
.
Adj.R-squared
0.339
0.346
0.328
0.342
0.336
0.412
p(ACT)
0.000
0.000
0.004
0.001
p(Home)
0.393
0.684
0.508
0.779
0.112
p(Dad)
0.287
0.325
0.250
0.050
p(Mom)
0.636
0.393
0.720
0.235
ACT(Home)
0.060
0.057
0.058
0.049
Fall 2014 GPA is dependent variable in all models. Heteroskedastic-robust standard errors reported in parentheses. All models
include a constant term and a set of major and class dummy variables as controls, but these results omitted for space. Models 5 and 6
restrict sample of underclassmen and upperclassmen, respectively. p(ACT), p(Home), p(Dad) and p(Mom) are p-values from joint
tests of significance of the partial effects of ACT, Home, Dad_college, and Mom_college, respectively. ACT(Home) is the partial
effect of ACT for homeschooled students. Statistical significance levels: *** p<0.01, ** p<0.05, * p<0.1.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 29
TABLE 3: OLS REGRESSION RESULTS – RESTRICTED SAMPLE
GPA is dependent variable in all models
(1)
(2)
(3)
(4)
(5)
(6)
Home
0.083
1.380
-0.047
1.545
0.818
3.120*
(0.110)
(1.027)
(0.303)
(1.129)
(1.253)
(1.807)
Parental Education Variables
Dad_College
0.185
0.177
0.149
0.095
0.198
-0.605**
(0.143)
(0.145)
(0.309)
(0.298)
(0.399)
(0.259)
Mom_College
0.041
0.041
-0.058
-0.132
-0.291
-0.425
(0.105)
(0.106)
(0.271)
(0.304)
(0.359)
(0.360)
Time Use Variables
Work
-0.004
-0.003
-0.004
-0.004
-0.018*
0.000
(0.004)
(0.004)
(0.004)
(0.004)
(0.010)
(0.005)
Extra
-0.021***
-0.021***
-0.021***
-0.021***
-0.029**
-0.018*
(0.008)
(0.008)
(0.008)
(0.008)
(0.013)
(0.010)
Sleep
-0.030
-0.020
-0.035
-0.025
0.018
-0.032
(0.037)
(0.037)
(0.038)
(0.038)
(0.057)
(0.066)
Credit
0.017
0.017
0.017
0.017
0.083
0.011
(0.016)
(0.016)
(0.016)
(0.016)
(0.050)
(0.023)
Single
-0.159*
-0.151*
-0.155*
-0.139*
-0.122
-0.108
(0.085)
(0.085)
(0.079)
(0.080)
(0.134)
(0.113)
Control Variables
ACT
0.069***
0.103***
0.069***
0.114***
0.093**
0.187***
(0.013)
(0.032)
(0.013)
(0.038)
(0.039)
(0.062)
Judge
0.306**
0.332**
0.308**
0.343**
0.724***
0.113
(0.141)
(0.142)
(0.139)
(0.136)
(0.207)
(0.137)
Campus
0.154
0.171
0.154
0.177
0.311
0.200
(0.115)
(0.117)
(0.115)
(0.116)
(0.297)
(0.172)
Interaction Terms
HomeACT
-0.035
-0.049
-0.022
-0.137*
(0.033)
(0.039)
(0.041)
(0.069)
HomeDad_college
0.070
0.118
0.075
0.913**
(0.330)
(0.323)
(0.398)
(0.396)
HomeMom_college
0.150
0.235
0.292
0.584
(0.269)
(0.305)
(0.354)
(0.375)
Other Statistics
N
109
109
109
109
57
52
F
4.468
3.857
4.360
3.813
4.370
.
Adj.R-squared
0.339
0.346
0.328
0.342
0.336
0.412
p(ACT)
0.000
0.000
0.004
0.001
p(Home)
0.393
0.684
0.508
0.779
0.112
p(Dad)
0.287
0.325
0.250
0.050
p(Mom)
0.636
0.393
0.720
0.235
ACT(Home)
0.060
0.057
0.058
0.049
Fall 2014 GPA is dependent variable in all models. Heteroskedastic-robust standard errors reported in parentheses. Sample excludes
the 4 students who reported an SAT score less than 1500 - see Footnote 5 for details. All models include a constant term and a set of
major and class dummy variables as controls, but these results omitted for space. Models 5 and 6 restrict sample of underclassmen
and upperclassmen, respectively. p(ACT), p(Home), p(Dad) and p(Mom) are p-values from joint tests of significance of the partial
effects of ACT, Home, Dad_college, and Mom_college, respectively. ACT(Home) is the partial effect of ACT for homeschooled
students. Statistical significance levels: *** p<0.01, ** p<0.05, * p<0.1.
HOMESCHOOL BACKGROUND, TIME USE & ACADEMIC PERFORMANCE 30
Appendix: Student Survey Questions
1. What was your SAT or ACT score?
2. Were you homeschooled?
o Yes, during elementary school
o Yes, during middle school
o Yes, during high school
o No
3. What is your father’s highest level of education?
o High school
o Some college
o Bachelor’s degree
o Master’s degree
o Ph.D.
4. What is your mother’s highest level of education?
o High school
o Some college
o Bachelor’s degree
o Master’s degree
o Ph.D.
5. What is your current class standing?
o Freshman
o Sophomore
o Junior
o Senior
6. What is your Myers-Briggs personality type?
7. What was your GPA for the Fall 2014 semester (2 decimal places)?
8. How many credit hours did you take in the Fall 2014 semester?
9. What was your living situation in the Fall 2014 semester?
o On-campus
o Off-campus
10. What was your relationship status during the Fall 2014 semester?
o Single
o In a relationship
o Married
11. How many hours/week did you typically work during the Fall 2014 semester?
12. How many hours/week did you spend in extracurricular activities during the Fall 2014 semester?
(e.g. hobbies, working out, etc. - do not include activities such as debate that you received credit
for)
13. How many hours did you typically sleep (during a 24-hour period) during the Fall 2014 semester?
14. What is your intended major?
o Literature
o History
o Political Theory
o Strategic Intelligence
o American Politics and Policy
o International Politics and Policy
o General Government
o Classical Liberal Arts
o Economics and Business Analytics
o Journalism
