Can school sports reduce racial gaps in truancy and achievement?
Cuffe, Harold E. ; Waddell, Glen R. ; Bignell, Wesley 等
Can school sports reduce racial gaps in truancy and achievement?
But here's the thing: most American principals I spoke with
expressed no outrage over the primacy of sports in school. In fact, they
fiercely defended it. "If I could wave a magic wand, I'd have
more athletic opportunities for students, not less, " Bigham, the
former Tennessee principal, told me. His argument is a familiar one:
sports can be bait for students who otherwise might not care about
school. "I've seen truancy issues completely turned around
once students begin playing sports," he says. "When students
have a sense of belonging, when they feel tied to the school, they feel
more part of the process. "
(Amanda Ripley, "The Case Against High-School Sports,"
The Atlantic, October 2013)
I. INTRODUCTION
In the 2012-2013 academic year, more than 12% of boys in grades 9
through 12 participated in high-school football. Overall, it marks the
24th consecutive year of increase in high-school athletic participation,
with track and field, basketball, soccer, and baseball/softball each
attracting more than one million student athletes. (1) At the same time,
however, there is growing concern among parents and policy makers, who
question the efficacy of participation and the potential imbalance
represented in the escalating pursuit of nonacademic activities within
American educational institutions. Clearly, a better understanding of
the implicit tradeoffs associated with athletic participation is
fundamental to navigating any future provisions of athletic
opportunities. Our contribution will be to move us in that direction.
The related academic literature suggests that high levels of
participation are not without justification. In fact, the meta-narrative
is quite clearly that participation in high-school athletics correlates
with or contributes to improvements in outcomes. For example, existing
research broadly implicates athletic participation in better
labor-market outcomes (Barron, Ewing, and Waddell 2000; Eide and Ronan
2001; Stevenson 2010), higher high-school grade-point average (GPA),
standardized-test performance, and class rank (Barron, Ewing, and
Waddell 2000; Lipscomb 2007; Rees and Sabia 2010), and more favorable in
high-school and college degree attainment (Barron, Ewing, and Waddell
2000; Eide and Ronan 2001; Lipscomb 2007; Pfeifer and CorneliBen 2010;
Stevenson 2010). Persico, Postlewaight, and Silverman (2004) find that
high school athletics participation mediates much of the positive adult
labor market returns to height.
Even with such evidence, it remains that we arrive at better policy
when the underlying mechanisms that contribute to related outcomes from
athletic participation are understood. As athletic funding comes under
increasing scrutiny, any draw down of athletic resources should be
informed by the channels through which benefits might derive. Likewise,
anticipating any gender or racial heterogeneity in the role of athletics
in outcomes allows for more nuance and sophistication in how we navigate
the future of athletics in education.
Among the potential mechanisms of interest in a broad analysis of
the future role of athletics in education, one might suspect that
athletic participation reduces leisure time, may reduce the amount of
nonschool time devoted to studying, or lead students toward taking
less-demanding classes, for example. To the extent such mechanisms are
prominent, athletic participation should adversely affect human-capital
acquisition. However, the rules under which sport is undertaken in
school tend to bundle participation with human-capital investments: to
participate, students must maintain a minimum GPA, for example, and show
up for classes, particularly on days of games. Thus, it would not be
surprising to find that athletic participation has positive effects on
human-capital investments, with such effects potentially large for
students who are otherwise marginal in their academic endeavors.
In this article, we consider the effect of athletic participation
on student-athlete absenteeism. While the consequences of student
absenteeism are not yet well understood in the literature--and seem to
not have been considered at all among athletes--there is evidence
supporting that instructional time may itself be an important factor in
educational achievement and other outcomes. For example, in a
country-level design, Lee and Barro (2001) find that more time in school
improves math and science test scores although reading scores may
suffer. Using international survey data, Lavy (2012) and Rivkin and
Schiman (2013) leverage within-school across-subject variation in weekly
instruction time and find positive effects on standardized-test scores.
Evidence from charter schools also suggests that successful schools tend
to have longer school days or years, and that instructional time
correlates with school effectiveness (Angrist, Pathak, and Walters 2013;
Dobbie and Fryer 2011; Hoxby and Murarka 2009).
Quasi-experimental evidence in Pischke (2007) suggests that student
cohorts exposed to shorter school years exhibit lower academic
achievement and higher rates of grade repetition, holding curriculum
constant. Eren and Millimet (2007) report weak evidence that longer
school years improve math and reading test scores, which they identify
from variation across U.S. states. Using within-state variation in
unscheduled school closures, Marcotte and Hemelt (2008) find that the
percentage of students passing math assessments falls by about one-third
to one-half a percentage point for each day school is closed, with the
effect largest for students in lower grades. Exploiting weather closures
as a source of exogenous variation, Marcotte (2007) suggests that the
share of students testing proficient in winters with average levels of
snowfall (about 17 inches) is about one to two percentage points lower
than in winters with little to no snow. With a similar identification
strategy, Marcotte and Hansen (2010) suggests that a 10-day increase in
instructional days yields a 0.2 SD increase in state math assessments,
or the equivalent of a 3- to 5-percentage-point increase in the number
of students passing math assessments. Using state-mandated changes in
test-date administration in Minnesota, which moved five times in 5
years, Hansen (2011) also shows that more instructional time prior to
test administration increases student performance. Fitzpatrick,
Grissmer, and Hastedt (2011), Carlsson, Dahl, and Rooth (2012), and
Aucejo and Romano (2014) also identify off of variation in the timing of
testing. (2) Perhaps as important, Card and Krueger (1992) and Betts and
Johnson (1997) support that subsequent earnings can increase with
school-year length.
Overall, we are inclined to interpret "instructional
time" estimates as lower bounds on the true effect of absenteeism
on educational achievement, since absenteeism arguably represents both a
reduction in instructional days plus, for example, the misalignment of a
student's progression in coursework and that of the class. While
the literature's use of "number of instructional days" is
a district- or school-specific variable, absenteeism is typically
student specific. As such, falling behind the pace of instruction
partially forecloses on their ability to absorb new material upon
returning. It is in this way that we anticipate that the causal effect
of absenteeism may well be larger than that of instructional days. In
fact, Goodman (2014) reconsiders weather-induced absences alongside
student-specific absences and finds that student absences force teachers
to spend time getting students on the same page as their classmates. In
the end, teachers appear to deal well with coordinated disruptions
(e.g., snow days) but poorly with student-specific absences. (3)
Among the students contributing to our analysis, 97% report having
an absence of some kind every year, while more than 85% report having an
unexcused absence in the average academic year. If absenteeism differs
systematically with athletic participation--we will end up arguing that
it does--it clearly has the potential to contribute to explaining the
observed heterogeneity in outcomes. (4)
Using daily administrative data from Seattle Public Schools (SPS),
we examine the effect of active participation in athletics on school
attendance. Previous literature (Barron, Ewing, and Waddell 2000)
suggests that the main challenge to identifying the effect of athletic
participation on attendance is positive selection into athletic
participation--those who tend to select into athletics are likely the
same as those who tend to select into higher attendance. As such,
comparisons of absenteeism rates among athletes and those of the general
student body are likely to misidentify the causal effect of athletic
participation. In our baseline specifications, then, for years in which
we observe a student participating in one or more high-school sports, we
leverage only the time-series variation in sport-specific seasons to
retrieve an estimate of the causal effect of active participation on
attendance including student-by-year fixed effects and identifying off
of the sport-specific variation in seasons and season length, where each
athlete spends some of the school year in season and some not. We
thereby avoid a fundamental confounder in identifying the effect of
athletic participation on absenteeism. We also absorb unobserved shocks
to absenteeism in school-weeks with school-by-year-by-week fixed
effects. (5)
In this setting, we show that active participation reduces male
athletes' overall absenteeism by 5.2%, while female attendance
appears relatively unresponsive. Notably, the effect is primarily
explained by a reduction in unexcused absences--arguably, the more
malleable sort of absence--which we interpret as consistent with
athletes optimizing around policy-induced incentives. Unexcused absences
fall by 10.3% at the mean for boys. The relationship is strongest for
Black male athletes (12.8% declines at the mean), among boys in homes
with only one parent (11.9%), and in lower grades (14.6% in grades 9 and
10). This heterogeneous response by race serves to reduce racial gaps in
truancy and achievement by approximately 20%.
While we anticipate having been able to retrieve an estimate of the
causal effect of active participation in athletics on attendance, doing
so off of comparisons of in-season athletes to out-of-season athletes
may underestimate important channels through which athletics leads to
attendance gains. We have in mind relationships with coaches,
connections with teammates, school connectedness, for example, or
eligibility criteria that may continue the duration of the school year.
Clearly, we are identifying the marginal contribution of active
participation, over and above whatever else athletic affiliations may
afford but are independent of season. To the extent coaches play a
mentoring role with athletes out of season, for example, our control
group may well attend at higher rates than they would in the absence of
athletic affiliation. To the extent such mechanisms spill into
out-of-season athletes, our identification will under-represent the
extent to which active participation influences attendance.
With sport-specific schedules available, we also relax the
constraint that absenteeism among active athletes is constant across
game days and nongame days. Doing so, we see even stronger evidence that
students are responding to incentives, as absences fall on game days and
rise the day after. This flexibility also reveals a similar pattern
among female athletes, with attendance gains on game days offset by
declines thereafter of a magnitude that leaves the net effect of
athletic participation approximately zero (and therefore insignificant
in specifications that ignore the timing of absences around game days).
While the focus of our analysis is on the effect of athletic
participation on truancy and absenteeism, before concluding we briefly
consider participation's effects on grades. While grades appear
unresponsive to the fraction of class days spent in active competition,
there is an important heterogeneity in apparent responsiveness. That is,
grades fall slightly with active competition in White and Asian boys
while rising with active competition in Black and Hispanic boys. The
response by White and Asian girls mirrors that of boys, while Black and
Hispanic girls' academic performance appears unaffected. Consistent
with the results on attendance, boys from households without two parents
show the largest gains from time spent in active athletic participation.
In Section II we provide additional context for the empirical
exercise to follow, discussing some of the institutional information and
the relevant incentives faced by student athletes. In Section III we
describe our data and in Section IV we set up the empirical problem more
formally and present results. We offer concluding remarks in Section V.
II. BACKGROUND
Reductions in school funding put additional pressure on resources,
and parents and others are seeming to be increasingly nervous about
potential imbalances in athletic versus academic focus in education. For
example, fundamental to the concern of many is that any increase in time
allocated to athletics implies offsetting reductions elsewhere. If not a
direct substitution, the hype and acclaim surrounding athletics may
likewise encourage athletes to focus less on academic preparedness, so
much so that we could anticipate student athletes sub-optimally
investing in nonathletic human capital in favor of sport. Of course,
injury and recovery times may well introduce mechanistic relationships
between athletic participation and absenteeism.
Alone, these concerns make it tempting to anticipate that athletes
will exhibit higher levels of truancy and that, if anything, that
truancy increases in periods of active participation. Yet, to find such
patterns in the data would run counter to a growing literature that
documents positive outcomes associated with athletics, which are more
consistent with lower rates of absenteeism. Moreover, anticipating
tension between academic and athletic pursuits, school districts
typically have safeguards in place to protect academic interests. For
example, in order to be eligible for competition on a game day, athletes
are required to have attended a full day of classes. Athletes can also
jeopardize their eligibility by irregular school attendance, whether or
not the absences occur on game days, or by failing to maintain at least
a 2.0 cumulative-GPA, as well as a 2.0 GPA in the classes they are
currently enrolled in (Seattle Public Schools 2011). While we remain
agnostic, as yet, with respect to the effect of athletics on
absenteeism, such incentives give reason to anticipate that absenteeism
may in fact be lower among athletes. In short, the same "hype"
that worries parents and administrators alike, when appropriately
governed, introduces a currency of sorts in the policy maker's
ability to steward student athletes into classes, as the price of
absenteeism is arguably higher for athletes, and particularly higher
in-season.
While we take the relationships between athletic participation and
longer-run outcomes as given, many of the same challenges to
identification in existing literature will exist as here we attempt to
identify a role for absenteeism as a possible mechanism. For example, if
physical activity itself improves academic performance, then we should
anticipate better average outcomes from athletes than nonathletes, even
without a direct role for organized sport. In this case, the ideal
experiment--hold physical activity constant while varying athletic
participation--is largely unachievable.
Given positive selection into athletics, the role of athletics in
outcomes could also be confounded by third factors that explain both.
For example, Barron, Ewing, and Waddell (2000) document that there is
indeed a large signaling component to high-school athletic
participation--many of the long-run benefits associated with athletics
are also consistent with positive sorting of high-ability students into
athletics (e.g., educational attainment, employment, wages, holding
supervisory positions, receiving piece-rate compensation). (6) With
these concerns in mind, we will identify the effects of athletic
participation on absenteeism by exploiting only the differences in the
timing and lengths of seasons over the school year for given athletes,
and variation in the team-specific dates on which games occur. (7)
III. DATA
In order to speak to the relationship between athletics and
attendance, we acquired restricted-use data from SPS, inclusive of
students' demographics, sport-specific indicators of athletic
participation, and daily attendance records. The data span all 10
traditional SPS high schools over academic years 2008-2009 through 2011
-2012. In order to identify the effect of athletic participation
separately from that due to the effects of other athlete attributes that
themselves might directly contribute to attendance, we discard
nonathletes and exploit only the exogenous variation in athletic seasons
and game days to identify the causal effect of being an athlete on
absenteeism. (8)
From among all SPS athletes, we discard all observations for an
academic year in which a student participates in multiple sports within
the same season (e.g., participation in two SPS-defined "winter
sports") as such students perhaps face a different treatment
intensity or are otherwise different in unobservables. As only 0.79% of
all SPS athletes participate in two sports within the same season, the
cleaner identification purchased here does not come at much of a cost.
We also discard all cheerleaders from the analysis, as their
participation spans the entirety of all other sport seasons and
contribute nothing to our estimate of the effect of being in active
competition on attendance. (9) We also drop any athlete appearing to
transfer from one SPS high school to another midway through a school
year.
In the remaining SPS data, daily records of attendance for all
student athletes yields a sample of more than 2.5 million student-day
observations, or more than 7,000 student athletes. Overall,
approximately 35% of the boys in SPS and 31% of girls in SPS participate
in at least one sport during the average academic year in our sample. In
Table 1, we stratify demographic and attendance data by gender and,
although they will not contribute to the econometric analysis, include
statistics for nonathletes for additional context. Within SPS, Whites
are over-represented in athletics--they account for approximately half
of all athletes despite being about one-third of the nonathlete
population. In terms of grade level, athletes generally reflect the
distribution of the nonathlete population.
As indicated in Table 1, athletes and nonathletes differ
significantly in their absenteeism rates. Both male and female athletes
are recorded absent for 0.38 school periods per day, on average. With
six class periods in the typical school day, this implies that athletes
are missing an average of 1.9 class periods per week, or roughly the
equivalent of one school day every 3.2 weeks. Nonathletes are absent
roughly 0.64 periods per day, or the equivalent of 3.2 class periods per
week, or 1 day every 1.9 weeks.
In addition to a record of the number of periods absent, our data
include whether the absences were "excused" or
"unexcused." According to SPS, absences may be excused for
reasons pertaining to the health of the student or a family member, as
well as for religious holidays, educational activities, a late bus, or a
school-imposed suspension. District policy explicitly prohibits other
reasons from justifying an absence as "excused," and parents
are given 48 hours following an absence to contact the school and
petition to excuse an absence. Table 1 also reveals two notable
differences when absences are tabulated by type. First, despite boys and
girls being absent at similar rates overall, the distribution of absence
types is markedly different by gender, as girls record approximately 22%
more excused absences than are recorded for boys. Second, the proportion
of excused to unexcused absences is approximately 40% higher for
athletes than for nonathletes. Also in Table 1, we see the known
tendency for athletes' GPAs to be higher, on average.
SPS control when athletes are considered active, and the types of
activities that can occur in and out of official seasons. This includes
the maximum number of matches athletes may participate in (between 10
and 21 per year, depending upon the sport). As shown in Table 1, nearly
half of all student-day observations are "in-season," which we
refer to as "active" days.
In Table 2, we summarize attendance and participation, by sport,
across race and gender. Firstly, panel A illustrates the large
attendance gaps that exist across racial groups, particularly with
respect to unexcused absences. Whites and Asian boys miss, on average,
roughly 0.32 periods per day, while Blacks and Hispanics miss 0.52. In
unexcused absences, though, a gap emerges between Whites and Asians, and
widens between Whites and Blacks--White athletes miss only 0.09
unexcused periods per day, while Asian athletes miss about 1.6 times
this amount, and Black athletes 3.3 times the rate of White athletes. In
panel B we see racial compositions varying across sport. Student counts
indicate very little participation by some groups within a few sports
(e.g., female wrestlers). It is apparent that the distribution of
athletes varies both within and across sports. For example, Black
students make up the largest racial group in basketball, while swimming
(another winter sport) largely attracts White and Asian students.
Looking within, rather than across columns paints much the same
picture--Black boys concentrate most heavily in football, basketball
and, to a lesser extent, track and field, while the most popular sports
for White boys appear to be football, soccer, and baseball. The final
panels in Table 2 show that time spent in-season is approximately equal
across race and gender, and that the proportion of athletes
participating in one, two, or three sports per year is similar across
columns.
In Table 3, we provide a simple comparison of average attendance
rates across "active" and "inactive" athletes by
sport. The results show no clear picture of the effect of being in
active competition on attendance as average attendance rates among
athletes in some sports appears to go up during the sport's season
while declining for others. One possible explanation for this is the
fact that these attendance statistics fail to account for general
patterns of attendance over the school year, leaving spring sports
appearing to have a negative affect on attendance, since attendance in
the spring is relatively poor, generally, regardless of sports
participation or in-season status.
To summarize the general patterns in attendance, we plot the
average daily periods absent for active and inactive athletes, and
nonathletes over the academic year in Figure 1. Note that different
students are contributing to the athlete plots in different time
periods, while the nonathlete plot is the same set of students
throughout the school year. It is clear that for unexcused absences in
particular, there is an upward trend through the academic year. What is
interesting, however, is that while the unexcused absences of inactive
athletes follow the general upward trend in the general student
population in the latter part of the year, the unexcused absences of
active athletes fail to rise, until much later in the year when many
athletes are still "active" yet are no longer competing as the
state championship tournament concludes.
IV. EMPIRICS
A. The Effect of In-Season Status on Absenteeism
To begin, we simply consider how being in active competition
changes daily average attendance of athletes. Using ordinary least
squares on a panel of daily attendance records for all SPS students in
the sample years (2008-2009 through 2011-2012) in which they
participated in at least one high-school sport, we estimate the model,
(1) Periods[Absent.sub.asd] = [alpha] + [beta][Active.sub.asd] +
[[epsilon].sub.asd]
where Periods[Absent.sub.asd] is the number of school periods
student athlete a is absent from school s on day d, and is determined by
a's participation status, [Active.sub.asd], which equals one if
athlete a is in-season on day d. In all specifications, standard-error
estimates correct for possible clustering at the school level. (10)
Given differences in absenteeism and possible differences in the
intensity of treatment across genders, we estimate Equation (1)
separately for boys and girls. Of course, if active athletics
participation decreases absenteeism, [beta]^ will be negative.
The results in column 1 of Table 4 imply that active athletes
exhibit lower rates of absenteeism on average. The number of periods
boys are absent declines with active participation by approximately .027
per day, or 7% of the absence rate exhibited by inactive athletes (.39
classes daily). Active girls are absent .022 fewer classes per day, or
5.7% of the mean out-of-season absence rate (also .39 classes). The size
and statistical significance of the estimates are modest, indicating
with the confidence interval around the point estimate on the
relationship between girls' sports participation and their school
attendance spanning zero.
In several dimensions, even though the simple model of Equation (1)
cannot be explained by athlete selection, it fails to account for
athletes selecting systematically into seasons (i.e., sports), which may
correlate with other unobservable student attributes that drive
attendance. For example, if students who have more active days in a
given year are more industrious or more competitive, and also have lower
rates of absenteeism through the year, the point estimate in column 1
will be biased downward. If school attendance is lower in times when
sports participation is higher (as is the case in the spring term), this
variation would also be potentially problematic. Likewise, school-level
factors, more broadly, may correlate with attendance and average season
lengths at the school.
To account for such potential confounders, we re-estimate Equation
(1) as,
(2) Periods [Absent.sub.asdwy] = [alpha] + [beta][Active.sub.asd] +
[[lambda].sub.ay] + [[tau].sub.swy] + [u.sub.asdwy]
with the addition of athlete-by-year fixed effects,
[[lambda].sub.ay], and school-by-year-by-week fixed effects,
[[tau].sub.swy]. Thus, in column 2 of Table 4, we identify the effect of
being in active competition on absenteeism by leveraging the variation
in student-by-year variation in in-season status within a given
school-week. In column 3--our preferred specification--we also absorb
student-specific heterogeneity into the error structure, exploiting the
day-by-day variation in the timing of students' active seasons.
While point estimates do move across columns, the overall story is
insensitive to specification--attendance improves with active
competition. From our fully-specified model, estimates imply that
boys' absences fall with active athletic participation--being
in-season reduces the number of periods absent by 5.2% relative to that
seen out-of-season, on average. Although point estimates suggest that
absences also decline among girls (3.3%), one should hesitate drawing
such a conclusion as the confidence interval also includes zero. Of
course, we should note again that in our setting we are identifying the
marginal contribution of active, in-season participation over and above
any general increase in attendance that athletic affiliation itself may
induce.
B. Are Patterns of Unexcused and Excused Absences Different?
Before we consider the potential for heterogeneous effects, it will
be instructive to consider the distinctions between absences designated
"excused" versus those designated "unexcused." In
particular, by separately identifying the effect of active competition
on the type of absence has the potential to separate competing patterns
of behavior. For example, a decline in unexcused absences would be
consistent with athletes substituting away from oversleeping or leisure
activities crowding out classes--examples of the sort of behavior that
leads to unexcused absences--while excused absences should move
differently, if not at all. We hesitate to think of this as a proper
falsification exercise, as we can easily think of mechanisms that would
move excused absences systematically with sport activity. For example,
offsetting increases in excused absences would be consistent with
students or parents investing additional resources in having given
absences excused while the student is actively participating. As
absences may be "excused" for reasons pertaining to the health
of the student, we likewise might anticipate increases in excused
absences among sports particularly susceptible to injury. Similarly,
sanctioned tournaments that directly conflict with classes would, we
imagine, increase excused absences. However, we see no evidence that
scheduled tournaments conflict with classes.
In Table 5, we report the estimated effect of being active on
excused and unexcused absences, adopting the fully-specific model in
Equation (2). Doing so clearly points to the patterns of absenteeism
among boy athletes being driven by their "unexcused"
behavior--unexcused absences decline 10.3% when athletes are active,
though excused absences decline only insignificantly (1.3%). The small
negative effect of being active on excused absences should not be
ignored, however, as it is by this that we learn that the decline in
unexcused absences cannot be explained by parents investing in having
given absences excused when their children are actively competing. Were
this the case, which we could not rule out a priori, that the point
estimate in column 2 would be positive. In supplemental analysis, we
also find support for the conjecture that declines in full-day absences
are somewhat larger, suggesting that much of the change in behavior
represents deliberate declines in truancy.
As is the case in the restricted model reported in Table 4, the
patterns of absences around girls' active competitions again appear
modest when separated by whether the absence was excused or not.
However, we can still rule out that parents are investing additional
resources in having given absences excused when their girls are actively
competing. Although insignificant, interpreting point estimates directly
implies that unexcused and excused absences decline among girls by 3.3%.
C. Heterogeneity
Race. Assuming that the effect of active participation on
absenteeism is constant across race seems overly restrictive in light of
significant race differences in both school attendance and sports
participation. We therefore stratify the earlier estimates by race,
which we present in Table 6. Among boys, doing so reveals that the
effect of being in active competition on unexcused absences is strongest
among Black and Hispanic boys, who exhibit a 12.8% decline in unexcused
absences relative to their average out-of-season attendance rates (which
are high relative to White and Asian boys). The decline in White and
Asian absenteeism is less than half this size (relative to their mean)
and statistically insignificant at conventional levels. Since
out-of-season Black and Hispanic boys are truant approximately 0.2
periods more, per day, than Whites and Asians, the effect represents a
20% reduction in the racial truancy gap. As in earlier specifications,
excused absences do not increase in-season among any race/gender group,
which we continue to interpret as ruling out that parents are more
inclined to have given absences excused when their children are active.
Indeed, any declines in excused absences would be consistent with
improved health of the student or family, fewer conflicting educational
activities or late buses, or a decline in school-imposed suspensions
during active seasons.
Family Structure. In Table 7 we stratify by a measure of family
structure--whether the student is recorded as living with both parents,
or "not with both parents," which includes those in foster
care, living with grandparents, or, as is predominantly the case, with
one parent. There is a large body of work highlighting the association
between divorce and children's long-run outcomes. As such, we think
that considering the role of family structure in how athletics explains
outcomes is an important aspect for policy makers to consider. (11)
While power is somewhat limited, there is a noteworthy robustness
to the Black and Hispanic result above; Table 7 suggests that this is
capturing a single-parent effect, of a sort. Note, first, that the
stratification by family structure reveals that average absenteeism is
higher among single-parent families in all race/gender cells. Unexcused
absences are highest among Black and Hispanic boys from single-parent
homes (.39), while excused absences are highest among Black and Hispanic
girls from single-parent homes (.32).
Among boys (Panel A), the effect of being in active competition on
unexcused absences is seemingly strongest among Black and Hispanic boys
from single-parent homes--declines in unexcused absences are 12.4%
relative to their mean out-of-season attendance rates. Considering the
racial difference between out-of-season truancy among single-parent
athletes (0.21 periods per day), in-season sports participation serves
to reduce this gap by 23%. The effect among Black and Hispanic boys of
two-parent households is also large--9% relative to the mean--but fails
to exclude zero from the confidence interval. In Panel B we see that
girls' unexcused absences vary with family structure, similar to
boys. Different, however, is that the precision and percentage-impact
appear greater among girls in two-parent households, with sports
participation again reducing the racial gap in truancy. In no case does
stratifying by family structure reveal significant relationships between
athletic participation and excused absences, although the impact for
Black and Hispanic students generally exceeds that of Whites and Asians.
Grade Level. The incentives to please coaches or compete for
playing time may change throughout athletes' high-school careers,
making any grade-level heterogeneity an interesting distinction. In
Table 8, we estimate the effect of being in active competition on
excused and unexcused absences stratifying by upperclassman and
underclassman status, and again by gender. Across both boys and girls,
active engagement in athletics predicts larger declines in absenteeism
in earlier grade levels (i.e., grades 9 and 10). This is particularly so
among boys, where absenteeism rates are some 14.6% lower at the mean, a
reduction of .7 class periods in a 5-day school week. Among
upperclassmen, the estimated effect remains statistically significant,
but drops in size to 6.1% of the mean absenteeism rate. Again, in boys,
excused absences do not appear to be responsive to active sports
participation, and the negative point estimates again suggest that there
is no measurable substitution across unexcused and excused absences.
Excused absences do fall among girls in lower grade levels, although
there is also less heterogeneity across active and inactive athletes. In
unreported analysis, a triple difference reveals somewhat larger
negatives among underclassmen who eventually dropout, suggesting that
the larger effect for upperclassmen may derive from (earlier) attrition
by those students most responsive to active participation. Importantly,
however, data limitations leave us unable to distinguish between
dropouts and transfers (out of SPS). As such, one should be measured in
making related inference.
D. Are There Game-Day Effects on Absenteeism?
According to the Washington Interscholastic Activities Association
(WIAA), athlete attendance is monitored and eligibility is jeopardized
by poor attendance. However, it is notable that athletes are required to
attend a full day of school on the day of any sport competition in order
to be eligible for competition. As such, to the extent students are
responding to incentives, we might expect more than simple level shifts
in attendance for periods of active participation.
To our data, we add nearly 8,500 tournament
events--school-by-sport-specific dates on which competition occurred
within SPS. Schedule data for all sports were collected from daily
online historical records in the Seattle Times, internet records
databases, and from high-school coaches. (12) We believe we have nearly
full coverage over the 4 years, though we suspect some missing dates for
the wrestling tournaments and gymnastics meets occurring in the first 2
years.
In the models of Table 9, we allow athlete a's absenteeism to
vary differently on game days, and on the calendar days immediately
before and after a game. We adopt our preferred specification from
above, but, as sports can tend to follow somewhat regular schedules
(e.g., football tending to play games on Fridays), we report
specifications with and without day-of-week fixed effects.
Allowing for this flexibility proves important to fitting the data,
and reveals an interesting pattern that is consistent with students
optimizing around the incentives they face. For boys, the pattern in
unexcused absences is clearly systematic around tournament dates, with
the overall decline in absences documented above seemingly driven by
general declines in absenteeism, but particularly large declines in
absenteeism on game days, with an offsetting increase on days following
tournaments, where active athletes are apparently not different from
inactive (i.e., out-of-season) athletes. This pattern is consistent with
a strong behavioral response to the policy, but with an offsetting
effect--athletes make up for their increased attendance leading up to
game days with post-game-day retreats, of a sort. This pattern is even
more evident in excused absences. While the earlier results (see Table
5) reveal no net difference in excused absences between in-season and
out-of-season boys, allowing for game-day effects in Table 9 uncovers an
underlying pattern of declining absences leading into game days
(although point estimates are insignificant) and a large and significant
increase in excused absences on days that follow game days. Using the
point estimates from the models without day-of-week fixed effects,
relative to the average active boy on days not surrounding a tournament,
active boys experience only slightly lower absenteeism rates the day
before a game. However, the effect is more than doubled on game days,
reducing truancy by 21% relative to inactive athletes. This is followed
by a rise in truancy on the day after a game, even accounting for the
overall reduction while in-season. Excused absences also increase on the
day after a game--the day after a game, excused absences are 13.6%
higher than we see in active boys on days not surrounding a game. While
this is consistent with parents investing additional resources in having
absences excused, our prior would be for such practice to be
demonstrated leading up to game days, or on game days, where marginal
absences are more likely to trigger ineligibility. Thus, we anticipate
that part of this pattern may be explained by injury and or recovery
times, which we consider below.
Among girls, recall that we found no significant differences in
either unexcused or excused absences overall. However, allowing for
game-day effects in the patterns of absences likewise points to
significant reductions in absences on game days followed by significant
increases in absences on days that follow game days. Relative to the
average inactive girl, active girls exhibit similar absenteeism rates
the day before a game, but 9.1% lower truancy rates on game days. On
days following games, they exhibit 11.6% higher unexcused-absenteeism
rates when compared to active girls on game days or the day prior.
Excused absences decrease 13.7% on game days, but again rise the day
after a game.
While somewhat cumbersome given that tournaments for some sports
fall disproportionately on particular days of the week (e.g., a majority
of football games occur on Fridays, and cross-country invitational meets
on Saturdays), sport-specific estimates are provided in Tables 10 and
11. Across all sports, the model suggests that football players are
potentially the most sensitive to game-day effects. On days prior to
games they are 38.8% less likely to be absent than are inactive male
athletes, 44.1% less likely on game days, but 54.1% more likely to be
absent on days following a game. (13) Other sports in which absenteeism
is seemingly quite responsive to game days include boys' basketball
(8.8% decrease, 43.5% decrease, and 7.7% increase) and girls'
basketball (21.7% decrease, 61.7% decrease, and 10.8% increase). Two
sports--cross country and golf, which both attract mainly White and
Asian students--seemingly having unexcused absences increase with active
participation. (14)
E. Academic Performance
We next turn to an analysis of the effects of active participation
on student achievement, leveraging student-by-class-level administrative
transcript data. With the potential to consider within-class variation
in performance, we generally have six observations per semester for each
athlete--for each we have a course title, subject code, and final letter
grades. For our purpose, we transform letter grades into a standard 4.0
scale, and create a semester-level measure of active participation for
each athlete using the sport-specific schedules considered above. That
is, we regress athlete a's performance in class c on the proportion
of semester t that athlete spent in season,
(3) [Grade.sub.acstg] = [alpha] + [gamma]
Prop.Semester[Active.sub.at] + [[lambda].sub.a] + [[eta].sub.g] +
[[sigma].sub.sct] + [u.sub.acstg],
where [[sigma].sub.sct] is a set of
school-by-course-by-term-by-school-year fixed effects. While attendance
is observed daily, grades are relatively infrequent outcomes; observed
only twice per year. We therefore abandon the individual-by-year fixed
effects, and adopt athlete and grade-level fixed effects
([[lambda].sub.a] and [[eta].sub.g], respectively). Estimated standard
errors again allow clustering at the school level.
While initial specifications in Table 12 suggest some systematic
variation in academic performances with differences in the intensity of
active participation, accounting for unobserved heterogeneity specific
to students and to semester-specific courses within schools (in Column
3) yields point estimates that are small and statistically
insignificant. (15)
In Table 13, we stratify by both gender and race, which reveals an
important difference in how grades move with active participation. In
particular, in boys, point estimates suggest that performance among
Black and Hispanic athletes improves with active participation and the
effect is statistically significant. White and Asian athletes appear to
suffer slight declines, though the estimate is not statistically
significant. The same asymmetry is true among female athletes, with
White and Asian athletes' grades declining and Black and
Hispanics' rising, but both effects are small relative to the
effect for Black and Hispanic boys, and statistically significant at the
10% level only for White and Asian girls.
In Table 14 we again stratify by a measure of family
structure--whether the student is recorded as living with both parents,
or "not with both parents," which includes those in foster
care, living with grandparents, or, as is predominantly the case, with
one parent. Doing so suggests that it is Black and Hispanic boys who do
not live with both parents that drive the relationship between active
participation and improvements in academic outcomes. Moreover, the
magnitude of the point estimate is not inconsequential, suggesting that
Black and Hispanic boys spending most of the semester active could
improve by nearly 0.3 in two classes (e.g., the difference between a B
and a B+). Interestingly, this effect represents an approximately 20%
reduction in the racial gap in grades for boys from this family
structure, which is on par with the gap reduction seen in truancy.
Stratifying by family structure also reveals that there is declining
performance with active participation among White boys living with two
parents. The magnitude of this decline is small, however, relative to
the gains by Black and Hispanic athletes, and marginally significant.
Among girls, the academic performance of those living with both
parents appears unresponsive to the athletic calendar, whether pooled or
stratified by race. The net effect of greater active sports
participation for girls not living with both parents is negative, which
appears to be driven by declines by White and Asian girls.
V. CONCLUSION
Whereas parents and administrators may be concerned that sports
draw students' priorities away from academics, we find little
evidence to support that active participation either decreases school
attendance or student achievement. Indeed, with daily student-level
records of attendance, we find that active athletic participation in
high school reduces absenteeism, with truancy reductions as the primary
mechanism. Moreover, we find particularly strong responses among Black
and Hispanic students, and athletes living in single-parent households,
suggesting that the incentives introduced with high-school athletics may
be of particular benefit to demographic groups that are often in
relative need. This disproportionate response serves to reduce the
racial gap in truancy by more than 20%.
Despite the overall reductions, there are partially offsetting
increases in absenteeism following game days--when we allow for
sport-specific effects, we find as large as a 28% increase in the
propensity for football players to record unexcused absences on days
following a game (relative to inactive athletes). While such truancy
patterns increase confidence in having retrieved estimates of the causal
effect of participation on absenteeism, at the same time we see the
partial undoing of the attendance gains leading up to game days as an
area of immediate concern. Students' transcripts also suggest that
the truancy response we document around active participation shows up in
academic gains--the longer is an athlete's length of season, the
higher is academic performance.
ABBREVIATIONS
GPA: Grade-Point Average
NFHS: National Federation of State High School Associations
SPS: Seattle Public Schools
WIAA: Washington Interscholastic Activities Association
doi: 10.1111.ecin.12452
APPENDIX
TABLE A1
Standard-Error Sensitivity around the Effect of Being in
Active Competition on Unexcused and Excused Absences
Unexcused Excused
Panel A: Boys (n = 1,382,011) (1) (2)
Active athlete -0.018 -0.003
Standard-error estimates: IID 0.001 *** 0.002
Individual 0.003 *** 0.004
School-by-week 0.002 *** 0.003
School-by-year 0.004 *** 0.005
School 0.005 ** 0.006
Ibragimov-Muller (95% CI) [-0.035, [-0.022,
-0.009] 0.012]
Unexcused Excused
Panel B: Girls (n = 1,132,704) (1) (2)
Active athlete -0.004 -0.009
Standard-error estimates: IID 0.001 *** 0.003 ***
Individual 0.002 * 0.004 **
School-by-week 0.002 ** 0.004 **
School-by-year 0.002 0.006
School 0.003 0.008
Ibragimov-Muller (95% CI) [-0.021, [-0.038,
0.005] 0.012]
Notes: All specifications include student-by-year and
school-by-week fixed effects. Ibragimov-Miiller
confidence intervals follow Ibragimov-Miiller
(2010, 2013).
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
REFERENCES
Angrist, J. D., P. A. Pathak, and C. R. Walters. "Explaining
Charter School Effectiveness." American Economic Journal: Applied
Economics, 5(4), 2013, 1-27.
Antecol, H., and K. Bedard. "Does Single Parenthood Increase
the Probability of Teenage Promiscuity, Substance Use, and Crime?"
Journal of Population Economics, 20(1), 2007, 55-71.
Arulampalam, W., R. A. Naylor, and J. Smith. "Am I Missing
Something? The Effects of Absence from Class on Student
Performance." Economics of Education Review, 31(4), 2012, 363-75.
Aucejo, E. M., and T. F. Romano. "Assessing the Effect of
School Days and Absences on Test Score Performance." Mimeo, 2014.
Baker, M. "Industrial Actions in Schools: Strikes and Student
Achievement." Canadian Journal of Economics, 46(3), 2013, 1014-36.
Barron, J. M., B. T. Ewing, and G. R. Waddell. "The Effects of
High School Athletic Participation on Education and Labor Market
Outcomes." Review of Economics and Statistics, 82(3), 2000, 409-21.
Baum, C. F., A. Nichols, and M. E. Schaffer. "Evaluating
One-Way and Two-Way Cluster-Robust Covariance Matrix Estimates."
German Stata Users Group Meeting, July 2011.
Betts, J. R., and E. Johnson. "A Test of Diminishing Returns
to School Spending." Mimeo, University of California San Diego,
1997.
Caceres-Delpiano, J., and E. Giolito. "The Impact of
Unilateral Divorce on Crime." Journal of Labor Economics, 30(1),
2012,215-48.
Cameron, A. C., and D. L. Miller. "A Practitioner's Guide
to Cluster-Robust Inference." Journal of Human Resources, 50(2),
2015, 317-72.
Card, D., and A. Krueger. "Does School Quality Matter? Returns
to Education and the Characteristics of Public Schools in the United
States." Journal of Political Economy, 100(1), 1992, 1-40.
Carlsson, M., G. B. Dahl, and D.-O. Rooth. "The Effect of
Schooling on Cognitive Skills." Working Paper No. 18484, National
Bureau of Economic Research, 2012.
Chen, J., and T.-F. Lin. "Class Attendance and Exam
Performance: A Randomized Experiment." Journal of Economic
Education, 39(3), 2008, 213--27.
Cortes, K. E., J. Bricker, and C. Rohlfs. "The Role of
Specific Subjects in Education Production Functions: Evidence from
Morning Classes in Chicago Public High Schools." The B.E. Journal
of Economic Analysis and Policy, 12, 2012, Article 27 (Contributions).
Dobbie, W., and R. G. Fryer. "Getting Beneath the Veil of
Effective Schools: Evidence from New York City." Working Paper No.
17632, National Bureau of Economic Research, 2011.
Dobkin, C., R. Gil, and J. Marion. "Skipping Class in College
and Exam Performance: Evidence from a Regression Discontinuity Classroom
Experiment." Economics of Education Review, 29(4), 2010, 566-75.
Durden, G. C., and L. V. Ellis. "The Effects of Attendance on
Student Learning in Principles of Economics." American Economic
Review, 85(2), 1995, 343-46.
Eide, E. R., and N. Ronan. "Is Participation in High School
Athletics an Investment or a Consumption Good? Evidence from High School
and Beyond." Economics of Education Review, 20(5), 2001, 431-42.
Eren, O., and D. Millimet. "Time to Learn? The Organizational
Structure of Schools and Student Achievement." Empirical Economics,
32(2), 2007, 301-32.
Fitzpatrick, M. D., D. Grissmer, and S. Hastedt. "What a
Difference a Day Makes: Estimating Daily Learning Gains during
Kindergarten and First Grade Using a Natural Experiment." Economics
of Education Review, 30(2), 2011,269-79.
Goodman, J. "Flaking Out: Snowfall, Disruptions of
Instructional Time, and Student Achievement." Mimeo, 2014.
Gruber, J. "Is Making Divorce Easier Bad for Children? The
Long-Run Implications of Unilateral Divorce." Journal of Labor
Economics, 22(4), 2004, 799-833.
Hansen, B. "School Year Length and Student Performance: Quasi
Experimental Evidence." 2011. Accessed April 10, 2017.
https://ssm.com/abstract=2269846
Hoxby, C., and S. Murarka. "Charter Schools in New York City:
Who Enrolls and How They Affect Their Students' Achievement."
Working Paper No. 14852, National Bureau of Economic Research, 2009.
Ibragimov, R., and U. K. Muller. "r-Statistic Based
Correlation and Heterogeneity Robust Inference." Journal of
Business & Economic Statistics, 28(4), 2010, 453-68.
--. "Inference with Few Heterogeneous Clusters." July
2013. Accessed April 10, 2017.
http://www.webmeets.com/files/papers/lacea-lames/2013/731/BehrensFisher.pdf
Kuhn, P, and C. Weinberger. "Leadership Skills and
Wages." Journal of Labor Economics, 23(3), 2005, 395-436.
Laughlin, N. T. "Athletic Participation and the Grade Point
Average, Absences, Cuts, and Disciplinary Referrals of High School
Athletes." International Journal of Sport Psychology, 9(2), 1978,
79-89.
Lavy, V. "Do Differences in Schools' Instruction Time
Explain International Achievement Gaps? Evidence from Developed and
Developing Countries." Working Paper No. 16227, National Bureau of
Economic Research, 2012.
Lee, J.-W., and R. J. Barro. "Schooling Quality in a
Cross-Section of Countries." Economica, 68(272), 2001, 465-88.
Lipscomb, S. "Secondary School Extracurricular Involvement and
Academic Achievement: A Fixed Effects Approach." Economics of
Education Review, 26(4), 2007,463-72.
Marburger, D. R. "Absenteeism and Undergraduate Exam
Performance." Journal of Economic Education, 32(2), 2001,99-109.
--. "Does Mandatory Attendance Improve Student
Performance?" Journal of Economic Education, 37(2), 2006, 148-55.
Marcotte, D. E. "Schooling and Test Scores: A MotherNatural
Experiment." Economics of Education Review, 26(3), 2007, 629-40.
Marcotte, D. E., and B. Hansen. "Time for School?"
Education Next, 10(1), 2010,53-59.
Marcotte, D. E., and S. W. Hemelt. "Unscheduled School
Closings and Student Performance." Education Finance and Policy,
3(3), 2008, 316-38.
Persico, N., A. Postlewaight, and D. Silverman. "The Effect of
Adolescent Experience on Labor Market Outcomes: The Case of
Height." Journal of Political Economy, 112(5), 2004, 1019-53.
Pfeifer, C., and T. Cornelissen. "The Impact of Participation
in Sports on Educational Attainment--New Evidence from Germany."
Economics of Education Review, 29(1), 2010,94-103.
Pischke, J.-S. "The Impact of Length of the School Year on
Student Performance and Earnings: Evidence from the German Short School
Years." The Economic Journal, 117(523), 2007, 1216-42.
Rees, D. I., and J. J. Sabia. "Sports Participation and
Academic Performance: Evidence from the National Longitudinal Study of
Adolescent Health." Economics of Education Review, 29(5), 2010,
751-59.
Rivkin, S. G., and J. C. Schiman. "Instruction Time, Classroom
Quality, and Academic Achievement." Working Paper No. 19464,
National Bureau of Economic Research, 2013.
Romer, D. "Do Students Go to Class? Should They?" Journal
of Economic Perspectives, 7(3), 1993, 167-74.
Seattle Public Schools. "Student Athletic Handbook."
2011. Accessed April 10, 2017.
http://roosevelths.seattleschools.org/UserFiles/Servers/Server_7419/
File/Academics/Athletics%20Info/full_student_handbook_august2011_web_version.pdf
Silliker, S. A., and J. T. Quirk. "The Effect of
Extracurricular Activity Participation on the Academic Performance of
Male and Female High School Students." School Counselor, 44(4),
1997, 288-93.
Stanca, L. "The Effects of Attendance on Academic Performance:
Panel Data Evidence for Introductory Microeconomics." Journal of
Economic Education, 37(3), 2006, 251-66.
Stevenson, B. "Beyond the Classroom: Using Title IX to Measure
the Return to High School Sports." Review of Economics and
Statistics, 92(2), 2010, 284-301.
HAROLD E. CUFFE, GLEN R. WADDELL and WESLEY BIGNELL*
* The views expressed in this paper represent those of the authors
and should not be interpreted as representing those held by Seattle
Public Schools or any other representative or employee of Seattle Public
Schools. While retaining responsibility for any shortcomings, the
authors thank Erica Birk, Luke Chu, Sue Dynarski, Giacomo De Giorgi,
David Figlio, Caroline Hoxby, Dean Hyslop, David Johnson, Jason Lindo,
Dave Marcotte, Christian Pfeifer, Je Smith, and seminar participants at
Virginia (Batton), Victoria University of Wellington, the 24th New
Zealand Econometric Study Group Meeting, Colorado-Denver, Otago, and the
NBER Economics of Education meetings.
Cuffe: Lecturer, Victoria University of Wellington, School of
Economics and Finance, Wellington, 6011, New Zealand. Phone +64 4 463
5380, Fax +64 4 463 5014, E-mail
[email protected]
Waddell: Professor, Department of Economics, University of Oregon,
Eugene, OR 97403-1285. Phone (541) 3461259, Fax (541) 346-1243, E-mail
[email protected]
Bignell: Graduate Student, Department of Sociology, University of
Washington, Seattle, WA 98195211. Phone (404) 513-3104, Fax (206)
543-2516, E-mail
[email protected]
(1.) National Federation of State High School Associations (NFHS),
www.nfhs.org, and the National Center for Education Statistics,
http://nces.ed.gov.
(2.) Representing still others from this large literature, Baker
(2013) exploits disruptions to instruction associated with labor action,
and Cortes, Bricker, and Rohlfs (2012) exploit the quasi-random
variation between students in the ordering of classes, where the higher
rate of absenteeism in early class creates a wedge in instructional time
that significantly reduces grades in those courses.
(3.) Aucejo and Romano (2014) also consider the relative
effectiveness of reducing absences in extending instructional time,
identifying the effect of absences on performance using within-student
variation in the number of absences in grade three and grade five on
associated standardized-test scores. Test-score gains are much larger
than those deriving from instructional time, with 10-day reductions in
absenteeism implying 5.8% and 3% of a standard deviation increases in
math and reading, respectively.
(4.) Absenteeism has also been directly implicated in determining
academic success among university students. For example, Dobkin, Gil,
and Marion (2010) exploit a discrete change in mandatory attendance for
students scoring below the median on midterm exams, and find that a 10%
increase in attendance increased subsequent exam scores by 0.17 standard
deviations. Using variation from randomly determined class times to
instrument for attendance, Arulampalam, Naylor, and Smith (2012) also
find that missing class leads to poorer performance, a relationship that
is particularly strong among high-performing students. For additional
considerations of absenteeism in post-secondary education see Romer
(1993), Durden and Ellis (1995), Marburger (2001, 2006), Stanca (2006),
and Chen and Lin (2008).
(5.) Given the within-student design, potential imbalance in
unobservables across the "treatment" and "control"
observations is not of great concern. By definition, no studentathlete
contributes to treatment observations without also contributing to
control observations. Thus, any difference is only driven by the
possible unequal contributions particular student athletes may make to
treatment and control groups. The traditional balance tests raise no
concerning patterns, however.
(6.) As not all of the variation in outcomes is explained by
selection, there is room for additional human capital having been
acquired through participation, which is also consistent with Kuhn and
Weinberger (2005), who find that higher wages are partly attributable to
the leadership skills developed through athletic participation.
(7.) There are examples in the existing literature that consider
the effect of sports seasons on attendance by comparing average
attendance during the season to average attendance out of season. For
example, Laughlin (1978) finds that among 243 high-school wrestlers,
attendance is higher during the season. Silliker and Quirk (1997)
conclude that among 123 high-school soccer players, attendance at school
seemed to be better in- season, but the difference was not significant.
In no case that we know of is class-level or day-level variation in
attendance exploited, or unobserved student heterogeneity considered.
(8.) Nonathletes can be retained in order to better identify other
parameters in the model but, without variation in their status (i.e.,
they are never "active"), they contribute nothing to the
estimate of interest. The inclusion of nonathletes would be justified if
their patterns of attendance provided at least as good a counterfactual
for active athletes' attendance as did inactive athletes. However,
there are reasons to believe that this may not be the case (e.g.,
athletes tending to be better students in other dimensions)
(9.) As is the case for the inclusion of nonathletes, retaining
cheerleaders (e.g., to better identify school effects) does not change
the reported results.
(10.) With few clusters (e.g., we have ten schools), the asymptotic
approximation of the autocorrelation within clusters may not be valid.
We have conducted sensitivity analysis with respect to estimated
standard errors, correcting for possible clustering at the school-year,
and at the school-week level. Of the options, correcting for clustering
at the school-level yields the largest standard-error estimates, which
we choose to present. This is consistent with Baum, Nichols, and
Schaffer (2011) who note that "with nested levels of clustering,
clusters should be chosen at the most aggregate level ... to allow for
correlations among individuals at that level," and with Cameron and
Miller (2015), who argue that "the consensus is to be conservative
and avoid bias and use bigger and more aggregate clusters when possible,
up to and including the point at which there is concern about having too
few clusters." T-tests are conducted using G - 1 = 9 degrees of
freedom when correcting for clustering at the school level. In Table A1
we demonstrate the sensitivity of standard-error estimates to various
assumptions on the pattern of clustering. We also provide confidence
intervals for the main results following the "few clusters"
approach of Ibragimov and Muller (2010, 2013).
(11.) Using the National Longitudinal Survey of Youth, Antecol and
Bedard (2007) "find that an additional five years with the
biological father decreases the probability of smoking, drinking,
engaging in sexual activity, marijuana use, and conviction."
Exploiting variation in the divorce rate generated by changes in
states' unilateral divorce laws, Gruber (2004) demonstrates that
those adults exposed to unilateral divorce as children in the decade
after the law change obtained less total education. This same
identifying variation leads Caceres-Delpiano and Giolito (2012) to
conclude that individuals who were children at the time of the reforms
exhibited higher rates of violent crime arrests as adults. Both Antecol
and Bedard (2007) and Caceres-Delpiano and Giolito (2012) cite reduced
supervision and adult interaction as possible mechanisms. We believe
this explanation to be particularly relevant for athletes as coaches
have the potential to serve as de facto surrogates to active athletes,
providing adult supervision and role modeling otherwise absent.
(12.) See www.athletic.net,www.maxpreps.com, and
www.nwcaonline.com. The records we collect are for all varsity
tournaments. For many sports, there is either no distinction between
varsity- and nonvarsity-level athletes, or tournaments include varsity
and nonvarsity heats. Likewise, inspecting more recent event calendars
for sports in which seniors and nonseniors do not compete against one
another, there exists a high degree of symmetry in the tournament
calendars, with both varsity and junior varsity teams competing on the
same day at different times or locations. As we do not have information
on the varsity-status of athletes, we treat all athletes as following
the varsity schedule. We anticipate that any remaining measurement error
would bias us towards finding no effect.
(13.) Relative to the off-season rates of absenteeism specifically
among football players (.34), these numbers are also quite large, at
-19.4, -22.1, and 27.1.
(14.) The racial heterogeneity apparent in Table 6 may instead be
driven by a combination of sport-specific effects and differences in
participation rates by race and sport. In an unreported table, we
re-estimate the equation used to produce Table 10, but interact the
variables associated with the two sports with the most significant
participation by minority boys (i.e., football and basketball) with an
indicator for Black/Hispanic status. In short, while the game-day
dynamics we report do not appear to be race-specific, the magnitudes are
significantly larger for Black and Hispanic boys.
(15.) Given the discrete nature of letter grades, a point estimate
of 0.095 cannot represent a 0.095 average increase across all courses in
a semester which is spent entirely in-season. Rather, accounting for
such an effect arising from across six classes, the effect is equivalent
to a student experiencing roughly a 0.3 point increase in two classes
(e.g., going from B to B+) or a 0.6 point increase in one class (e.g.,
one grade being changed from B to A-).
Caption: FIGURE 1 Mean Daily Absences by Week, by Gender: (A) All
Absences, (B) Excused Absences, and (C) Unexcused Absences
TABLE 1
Summary Statistics for Athletes and Nonathletes
Athletes Boys Nonathletes All
Students 3,941 7,360 11,301
Student-years 7,907 13,344 21,251
Student-days 1,382,011 2,258,556 3,640,567
White .49 .36 .41
Asian .20 .26 .24
Black .21 .22 .22
Other .11 .17 .15
Grade 9 .28 .32 .31
Grade 10 .26 .24 .25
Grade 11 .24 .21 .22
Grade 12 .22 .23 .23
Both parents .69 .54 .60
Periods absent per day .38 .65 .55
Periods excused .22 .28 .26
Periods unexcused .16 .38 .29
Semester GPA 3.01 2.40 2.63
Athletes Girls Nonathletes All
Students 3,217 7,290 10,507
Student-years 6,453 13,395 19,848
Student-days 1,132,704 2,278,796 3,411,500
White .52 .31 .38
Asian .24 .26 .25
Black .14 .28 .23
Other .11 .16 .14
Grade 9 .28 .29 .29
Grade 10 .28 .24 .26
Grade 11 .24 .22 .23
Grade 12 .20 .24 .23
Both parents .70 .53 .58
Periods absent per day .38 .64 .56
Periods excused .27 .33 .31
Periods unexcused .11 .31 .25
Semester GPA 3.37 2.73 2.94
Notes: Statistics are calculated from administrative
records spanning 10 SPS high schools over academic years
2008-2009 through 2011 -2012. A student is considered an
"athlete" in a given year if (s)he participates in at least
one of the following sports on a school-organized varsity,
junior varsity, or freshman team: baseball, basketball, cross
country, football, golf, gymnastics, soccer, softball, swimming,
tennis, track and field, volleyball, and wrestling. Six periods
make up the standard school day. However, in the two most recent
years, one school implemented a schedule with only five periods
per day. Additionally, three schools added an additional study
hall period twice a week in a number of the observed academic
years, bringing the total possible periods absent to seven on
these days. Observations from days with either one fewer or
one additional period make up 2.4% and 5.3% of the data,
respectively. Robustness checks demonstrate that our regression
results are not sensitive to the particular class scheduling in
these school years. (For example, our results will be robust to
the inclusion of fixed effects for school-by-year-by-day-of-week,
which cleans up much of the natural variation in absenteeism than
may not be attributable to active sport.) GPA is measured on a
standard 4.0 point scale. Students are "active" or "in-season"
during district-defined periods when tournaments and regular
training sessions may be held. Dropped from the analysis are
students who transfer to another SPS school midway though an
academic year (<1% of all athletes), and students who participate
in multiple sports, contemporaneously (<1% of all athletes).
TABLE 2
Summary Statistics for Athletes by Race, Gender, and Sport
Boys
White Asian Black Hispanic
Panel A: Absenteeism statistics across race and gender
Periods absent .32 .30 .52 .51
Periods excused .23 .17 .24 .25
Periods unexcused .09 .14 .29 .25
Panel B: Athlete-by-year counts by race, gender, and sport
Baseball 693 150 156 98
Basketball 378 130 622 83
Cross country 658 163 71 76
Football 847 289 647 258
Golf 323 64 -- --
Gymnastics -- -- -- --
Soccer 724 170 268 260
Softball
Swimming 495 221 30 49
Tennis 333 456 29 59
Track 634 258 389 121
Volleyball
Wrestling 326 203 107 100
Panel C: Percent of school year spent in-season
In-season .50 .48 .51 .48
Panel D: Percent participating in 1, 2, or 3 sports per year
1 sport .64 .71 .66 .70
2 sports .31 .23 .27 .25
3 sports .05 .06 .08 .05
Girls
White Asian Black Hispanic
Panel A: Absenteeism statistics across race and gender
Periods absent .35 .32 .53 .49
Periods excused .28 .21 .30 .31
Periods unexcused .07 .11 .23 .18
Panel B: Athlete-by-year counts by race, gender, and sport
Baseball
Basketball 288 90 325 119
Cross country 488 136 63 48
Football
Golf 89 59 -- --
Gymnastics 287 110 36 42
Soccer 805 171 114 125
Softball 385 180 124 115
Swimming 615 263 19 87
Tennis 482 511 89 96
Track 510 137 268 100
Volleyball 505 297 170 142
Wrestling 34 36 -- --
Panel C: Percent of school year spent in-season
In-season .48 .46 .48 .48
Panel D: Percent participating in 1, 2, or 3 sports per year
1 sport .70 .73 .71 .72
2 sports .25 .24 .23 .23
3 sports .05 .04 .05 .06
Notes: Refer to the notes found in Table 1 for a
description of the data. To protect the privacy of
students, results calculated from data on 10 or
fewer students are suppressed (indicated with a "-")
TABLE 3
Athletes' In-Season and Out-of-Season
Absences, by Sport
Boys Girls
Out-of- In- Out-of- In-
Season Season Season Season
Panel A: Absent Periods--Unexcused
Fall sports
Cross country 0.09 0.04 0.10 0.04
Football 0.34 0.16
Golf 0.10 0.05 0.09 0.11
Volleyball 0.17 0.08
Winter sports
Basketball 0.25 0.19 0.24 0.18
Gymnastics 0.08 0.07
Swimming 0.08 0.06 0.08 0.05
Wrestling 0.21 0.17 0.21 0.20
Spring sports
Baseball 0.11 0.15
Soccer 0.14 0.18 0.12 0.06
Softball 0.14 0.17
Tennis 0.08 0.08 0.06 0.09
Track 0.11 0.16 0.12 0.16
Panel B: Absent Periods--Excused
Fall sports
Cross country 0.21 0.16 0.28 0.21
Football 0.28 0.24
Golf 0.25 0.21 0.22 0.28
Volleyball 0.33 0.23
Winter sports
Basketball 0.24 0.24 0.30 0.33
Gymnastics 0.28 0.29
Swimming 0.19 0.21 0.28 0.24
Wrestling 0.22 0.24 0.24 0.25
Fall sports
Baseball 0.23 0.27
Soccer 0.18 0.24 0.31 0.23
Softball 0.26 0.32
Tennis 0.16 0.18 0.19 0.28
Track 0.17 0.25 0.24 0.32
Notes: Refer to the notes found in Table 1 for a description
of the data. To protect the privacy of students, results
calculated from data on 10 or fewer students are
suppressed.
TABLE 4
The Effect of Being in Active Competition
on Class Absences
Panel A: Boys (n = 1,382,011) (1) (2) (3)
Active athlete -0.027 ** -0.040 *** -0.020 **
(0.011) (0.010) (0.008)
Mean out-of-season absences .39 .39 .39
% Impact -7 -10.3 -5.2
School-by-week FE No Yes Yes
Student-by-year FE No No Yes
Panel B: Girls (n = 1,132,704) (1) (2) (3)
Active athlete -0.022 -0.027 ** -0.013
(0.013) (0.012) (0.009)
Mean out-of-season absences .39 .39 .39
% Impact -5.7 -6.9 -3.3
School-by-week FE No Yes Yes
Student-by-year FE No No Yes
Notes: Estimated standard errors are reported in parentheses,
adjusted for any clustering at the school level. Percent
impacts are relative to the mean number of periods
absent among out-of-season athletes.
*** Significant at 1%; ** significant at 5%; * significant at 10%.
TABLE 5
The Effect of Being in Active Competition
on Unexcused and Excused Absences
Unexcused Excused
Panel A: Boys (n = 1,382,011) (1) (2)
Active athlete -0.018 ** -0.003
(0.005) (0.006)
Mean out-of-season absences .17 .22
% Impact -10.3 -1.3
Unexcused Excused
Panel B: Girls (n = 1,132,704) (1) (2)
Active athlete -0.004 -0.009
(0.003) (0.008)
Mean out-of-season absences .12 .27
% Impact -3.3 -3.3
Notes: All specifications include student-by-year and
school-by-week fixed effects. Estimated standard errors are
reported in parentheses, adjusted for any clustering at the
school level. Percent impacts are relative to the mean number
of periods absent among out-of-season athletes.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 6
Racial Heterogeneity in the Effect of
Active Competition on Absences
Panel A: Boys Unexcused
White/Asian Black/Hispanic
(1) (2)
Active athlete -0.005 -0.040 ***
(0.005) (0.007)
Mean out-of-season absences .11 .31
% Impact -4.9 -12.8
Observations 951,020 430,991
Students 5,416 2,491
Panel B: Girls Unexcused
White/Asian Black/Hispanic
(1) (2)
Active athlete -0.001 -0.016
(0.001) (0.010)
Mean out-of-season absences .09 .22
% Impact -1.3 -7.4
Observations 851,045 281,659
Students 4,840 1,613
Panel A: Boys Excused
White/Asian Black/Hispanic
(3) (4)
Active athlete -0.003 -0.003
(0.005) (0.009)
Mean out-of-season absences .21 .24
% Impact -1.4 -1.2
Observations 951,020 430,991
Students 5,416 2,491
Panel B: Girls Excused
White/Asian Black/Hispanic
(3) (4)
Active athlete -0.006 -0.017
(0.007) (0.013)
Mean out-of-season absences .26 .3
% Impact -2.2 -5.6
Observations 851,045 281,659
Students 4,840 1,613
Note: See notes for Table 5.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 7
Heterogeneity across Family Structure in the
Effect of Active Participation on Absences
Living with Two Parents
All White/Asian Black/Hispanic
(1) (2) (3)
Panel A: Boys
Unexcused absences
Active athlete -0.006 -0.004 -0.019
(0.005) (0.004) (0.010)
Mean out-of-season absences .12 .09 .21
% Impact -5.3 -4.1 -9
Excused absences
Active athlete -0.003 -0.003 -0.004
(0.006) (0.006) (0.012)
Mean out-of-season absences .21 .2 .23
% Impact -1.6 -1.5 -1.7
Observations 954,449 761,856 192,593
Students 5,434 4,327 1,107
Panel B: Girls
Unexcused absences
Active athlete -0.003 * -0.001 -0.017 ***
(0.001) (0.001) (0.004)
Mean out-of-season absences .09 .07 .15
%-Impact -3.2 -.9 -11.6
Excused absences
Active athlete -0.010 -0.007 -0.021
(0.007) (0.007) (0.014)
Mean out-of-season absences .26 .25 .28
%-Impact -3.7 -2.8 -7.5
Observations 799,453 658,688 140,765
Students 4,538 3,737 801
Not Living with Two Parents
All White/Asian Black/Hispanic
(4) (5) (6)
Panel A: Boys
Unexcused absences
Active athlete -0.034 *** -0.009 -0.048 ***
(0.007) (0.011) (0.007)
Mean out-of-season absences .29 .18 .39
% Impact -11.9 -4.8 -12.4
Excused absences
Active athlete 0.000 0.001 -0.001
(0.008) (0.007) (0.011)
Mean out-of-season absences .25 .25 .25
% Impact 0 .2 -.4
Observations 427,562 189,164 238,398
Students 2,473 1,089 1,384
Panel B: Girls
Unexcused absences
Active athlete -0.008 -0.002 -0.019
(0.009) (0.006) (0.018)
Mean out-of-season absences .19 .13 .28
%-Impact -4.3 -1.3 -6.9
Excused absences
Active athlete -0.010 -0.004 -0.017
(0.013) (0.013) (0.022)
Mean out-of-season absences .31 .3 .32
%-Impact -3.2 -1.2 -5.2
Observations 333,251 192,357 140,894
Students 1,915 1,103 812
Note: See notes for Table 5.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 8
Active Competition on Class Absences, by Grade Level
Panel A: Boys Grades 9-10
Unexcused (1) Excused (2)
Active athlete -0.022 *** -0.005
(0.006) (0.008)
Mean out-of-season absences. .15 .21
% Impact -14.6 -2.4
Observations 752,462 752,462
Students 4,298 4,298
Panel A: Girls Grades 9-10
Unexcused (1) Excused (2)
Active athlete -0.005* -0.017 *
(0.003) (0.009)
Mean out-of-season absences .09 .25
% Impact -5.9 -6.9
Observations 637,870 637,870
Students 3,625 3,625
Panel A: Boys Grades 11-12
Unexcused (3) Excused (4)
Active athlete -0.012 * -0.002
(0.006) (0.007)
Mean out-of-season absences. .2 .23
% Impact -6.1 -.8
Observations 629,549 629,549
Students 3,609 3,609
Panel A: Girls Grades 11-12
Unexcused (3) Excused (4)
Active athlete -0.003 -0.000
(0.005) (0.007)
Mean out-of-season absences .15 .3
% Impact -1.9 -.1
Observations 494,834 494,834
Students 2,828 2,828
Note: See notes for Table 5.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 9
Game-Day Effects of Active Sports Participation
on Class Absences
Panel A: Boys Unexcused
(n = 1,382,011)
(1) (2)
Active athlete -0.016 *** -0.013 **
(0.005) (0.004)
Active x GameDay -1 -0.004 * -0.006 **
(0.002) (0.002)
Active x GameDay -0.019 *** -0.026 ***
(0.005) (0.006)
Active x GameDay +1 0.021 *** 0.014 ***
(0.005) (0.003)
Mean out-of-season absences .17 .17
Day-of-week FE No Yes
Panel B: Girls Unexcused
(n = 1,132,704)
(1) (2)
Active athlete -0.005 -0.005
(0.003) (0.003)
Active x GameDay -1 0.002 0.005
(0.003) (0.003)
Active x GameDay -0.006 *** -0.007 **
(0.003) (0.003)
Active x GameDay +1 0.014 *** 0.008 **
(0.003) (0.003)
Mean out-of-season absences .12 .12
Day-of-week FE No Yes
Panel A: Boys Excused
(n = 1,382,011)
(3) (4)
Active athlete -0.001 -0.003
(0.005) (0.005)
Active x GameDay -1 -0.007 -0.006
(0.004) (0.004)
Active x GameDay -0.020 -0.017
(0.026) (0.026)
Active x GameDay +1 0.030 *** 0.034 ***
(0.005) (0.004)
Mean out-of-season absences .22 .22
Day-of-week FE No Yes
Panel B: Girls Excused
(n = 1,132,704)
(3) (4)
Active athlete -0.005 -0.007
(0.005) (0.005)
Active x GameDay -1 0.003 0.006
(0.005) (0.005)
Active x GameDay -0.037 * -0.033 *
(0.017) (0.017)
Active x GameDay +1 0.018 ** 0.020 **
(0.008) (0.007)
Mean out-of-season absences .27 .27
Day-of-week FE No Yes
Note: See notes for Table 5.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 10
Game-Day Effects by Sport, Boys
Unexcused Excused
Boys (n = 1,382,011) (1) (2)
Active x Baseball -0.027 *** -0.020 *
(0.008) (0.009)
Baseball x GameDay -1 -0.007 0.001
(0.005) (0.010)
Baseball x GameDay -0.029 ** -0.026
(0.010) (0.031)
Baseball x GameDay +1 0.001 0.012
(0.003) (0.008)
Active x Basketball -0.012 ** -0.019
(0.005) (0.013)
Basketball x GameDay -1 -0.003 0.015
(0.007) (0.029)
Basketball x GameDay -0.062 *** -0.021
(0.011) (0.053)
Basketball x GameDay +1 0.025 ** 0.079 **
(0.011) (0.029)
Active x Cross Country 0.027 *** -0.011
(0.008) (0.007)
Cross Country x GameDay -1 0.011 * 0.002
(0.005) (0.009)
Cross Country x GameDay 0.034 *** 0.036
(0.006) (0.032)
Cross Country x GameDay +1 0.019 ** 0.036
(0.006) (0.020)
Active x Football -0.037 ** 0.027 **
(0.012) (0.009)
Football x GameDay -1 -0.029 ** -0.040 ***
(0.010) (0.012)
Football x GameDay -0.038 *** -0.072 ***
(0.011) (0.020)
Football x GameDay +1 0 129 *** 0.183 ***
(0.038) (0.047)
Active x Golf 0.023* 0.010
(0.012) (0.012)
Golf x GameDay -1 0.007 -0.016
(0.005) (0.027)
Golf x GameDay 0.014 -0.008
(0.009) (0.030)
Golf x GameDay +1 -0.002 -0.021
(0.006) (0.015)
Active x Soccer -0.021 * -0.021
(0.010) (0.012)
Soccer x GameDay -1 -0.008 0.017 *
(0.007) (0.009)
Soccer x GameDay -0.017 -0.001
(0.010) (0.033)
Soccer x GameDay +1 0.012 0.017
(0.012) (0.012)
Active x Swimming 0.010 -0.006
(0.006) (0.008)
Swimming x GameDay -1 0.003 -0.052 ***
(0.007) (0.010)
Swimming x GameDay 0.011 0.008
(0.007) (0.021)
Swimming x GameDay +1 0.006 0.025
(0.005) (0.047)
Active x Tennis -0.018 0.000
(0.017) (0.012)
Tennis x GameDay -1 0.007 -0.026
(0.005) (0.018)
Tennis x GameDay 0.018 0.012
(0.011) (0.028)
Tennis x GameDay +1 0.002 -0.006
(0.006) (0.016)
Active x Track -0.025 * 0.011
(0.011) (0.015)
Track x GameDay -1 -0.021 ** -0.025 *
(0.009) (0.013)
Track x GameDay 0.004 -0.010
(0.011) (0.022)
Track x GameDay +1 0.037 ** 0.036 **
(0.013) (0.013)
Active x Wrestling 0.008 -0.002
(0.009) (0.009)
Wrestling x GameDay -1 0.006 -0.009
(0.015) (0.011)
Wrestling x GameDay -0.005 -0.016
(0.016) (0.016)
Wrestling x GameDay +1 0.023 0.024 **
(0.018) (0.010)
Note: See notes for Table 5.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 11
Game-Day Effects by Sport, Girls
Unexcused Excused
Girls (n = 1,132,704) (1) (2)
Active x Basketball -0.014 0.001
(0.012) (0.008)
Basketball x GameDay -1 -0.012 0.013
(0.013) (0.023)
Basketball x GameDay -0.043 *** -0.057 **
(0.010) (0.023)
Basketball x GameDay +1 0.029 ** 0.087 ***
(0.010) (0.019)
Active x Cross Country 0.009 -0.007
(0.008) (0.006)
Cross Country x GameDay -1 -0.001 -0.036 *
(0.007) (0.017)
Cross Country x GameDay 0.019 *** -0.019
(0.005) (0.032)
Cross Country x GameDay +1 0.002 0.029
(0.004) (0.017)
Active x Golf 0.012 0.033
(0.013) (0.019)
Golf x GameDay -1 -0.022 -0.040
(0.013) (0.031)
Golf x GameDay -0.005 -0.044
(0.020) (0.032)
Golf x GameDay +1 0.039 ** 0.011
(0.016) (0.026)
Active x Gymnastics 0.011 0.015
(0.007) (0.013)
Gymnastics x GameDay -1 0.027 -0.028
(0.017) (0.024)
Gymnastics x GameDay 0.005 -0.023
(0.012) (0.036)
Gymnastics x GameDay +1 0.000 0.061 **
(0.010) (0.021)
Active x Soccer 0.017 ** -0.003
(0.007) (0.010)
Soccer x GameDay -1 0.002 0.012
(0.004) (0.009)
Soccer x GameDay 0.009 -0.028
(0.005) (0.015)
Soccer x GameDay +1 0.006 ** -0.004
(0.002) (0.010)
Active x Softball -0.026 ** -0.015
(0.009) (0.015)
Softball x GameDay -1 0.000 -0.015
(0.011) (0.009)
Softball x GameDay -0.000 -0.048
(0.011) (0.029)
Softball x GameDay +1 0.005 -0.017
(0.007) (0.013)
Active x Swimming 0.013 *** -0.012
(0.004) (0.007)
Swimming x GameDay -1 -0.006 0.019
(0.005) (0.015)
Swimming x GameDay 0.014 * 0.004
(0.007) (0.020)
Swimming x GameDay +1 0.009 0.011
(0.011) (0.030)
Active x Tennis -0.030 *** -0.007
(0.009) (0.007)
Tennis x GameDay -1 0.011 * -0.000
(0.006) (0.017)
Tennis x GameDay 0.006 0.009
(0.006) (0.026)
Tennis x GameDay +1 0.007 -0.003
(0.013) (0.013)
Active x Track -0.006 -0.012
(0.011) (0.015)
Track x GameDay -1 -0.017 -0.034 *
(0.010) (0.018)
Track x GameDay -0.001 -0.017
(0.009) (0.016)
Track x GameDay +1 0.045 *** 0.059 **
(0.013) (0.023)
ActiveX Volleyball 0.004 -0.017
(0.004) (0.010)
Volleyball x GameDay -1 -0.004 -0.012
(0.004) (0.008)
Volleyball x GameDay -0.016 *** -0.063 ***
(0.005) (0.015)
Volleyball x GameDay +1 0.013 0.017
(0.008) (0.011)
Active x Wrestling 0.032 * -0.023
(0.015) (0.029)
Wrestling x GameDay -1 0.022 0.048
(0.039) (0.035)
Wrestling x GameDay -0.029 0.042
(0.029) (0.032)
Wrestling x GameDay +1 0.024 0.082 **
(0.033) (0.029)
Note: See notes for Table 5.
*** Significant at 1%; ** significant at 5%;
* significant at 10%.
TABLE 12
Does the Degree of Active Participation in
a Semester Change Final Course Grades?
Panel A: Boys (n = 89,807) (1) (2) (3)
Prop, semester active 0.095 ** 0.077 ** 0.007
(0.037) (0.028) (0.010)
Mean out-of-season grade 2.98 2.98 2.98
Course-by-semester-by-school FE No Yes Yes
Student FE No No Yes
Panel B: Girls (n = 73,841) (1) (2) (3)
Prop, semester active 0.055 0.078 *** -0.006
(0.042) (0.024) (0.010)
Mean out-of-season grade 3.38 3.38 3.38
Course-by-semester-by-school FE No Yes Yes
Student FE No No Yes
Notes: All regressions include grade fixed effects.
Estimated standard errors are reported in parentheses,
adjusted for any clustering at the school level. By "course,"
we mean to imply a particular subject and course code (e.g.,
Geometry 101) rather than a particular section of Geometry 101.
Schools may teach concurrent sections of the same course within
a semester.
*** Significant at 1%; ** significant at 5%; *
* significant at 10%.
TABLE 13
Racial Heterogeneity in the Effect of Active
Participation on Grades
White/Asian Black/Hispanic
Panel A: Boys (1) (2)
Prop, semester active -0.012 0.058 ***
(0.012) (0.018)
Mean out-of-season grade 3.18 2.5
Observations 61,666 28,141
White/Asian Black/Hispanic
Panel B: Girls (1) (2)
Prop, semester active -0.017 * 0.008
(0.009) (0.023)
Mean out-of-season grade 3.51 2.93
Observations 55,441 18,400
Notes'. All regressions include student,
school-year-semester-course, and grade fixed effects.
Estimated standard errors are reported in parentheses,
adjusted for any clustering at the school level.
*** Significant at 1%; ** significant at 5%:
* significant at 10%.
TABLE 14
Heterogeneity across Family Structure in the
Effect of Active Participation on Grades
Living with Two Parents
All White/Asian Black/Hispanic
Panel A: Boys (1) (2) (3)
Prop, semester active -0.009 -0.017 * 0.035
(0.010) (0.009) (0.052)
Mean out-of-season grade 3.15 3.27 2.68
Observations 61,984 49,405 12,579
Living with Two Parents
All White/Asian Black/Hispanic
Panel B: Girls (1) (2) (3)
Prop, semester active 0.002 -0.003 -0.000
(0.009) (0.011) (0.040)
Mean out-of-season grade 3.38 3.56 3.1
Observations 52,153 42,949 9,204
Not Living with Two Parents
All White/Asian Black/Hispanic
Panel A: Boys (4) (5) (6)
Prop, semester active 0.046 * 0.022 0.087 ***
(0.022) (0.037) (0.024)
Mean out-of-season grade 2.59 2.86 2.35
Observations 27,823 12,261 15,562
Not Living with Two Parents
All Black/Hispanic All
Panel B: Girls (4) (5) (6)
Prop, semester active -0.023 -0.052 ** 0.003
(0.023) (0.023) (0.074)
Mean out-of-season grade 3.12 3.34 2.75
Observations 21,688 12,492 9,196
Note: See notes for Table 13.
*** Significant at 1%; ** significant at 5%; * significant at 10%.
COPYRIGHT 2017 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.
