Page 1
Genetic and environmental influences on being expelled and suspended
from school
Kevin M. Beavera,b,⁎, Joseph A. Schwartzc, J.C. Barnesd, Joseph L. Nedelecd, John P. Wrightb,d,
Brian B. Boutwelle,f, Eric J. Connollyg
aCollege of Criminology and Criminal Justice, Florida State University, Tallahassee, FL 32306-1127, USA
bCenter for Social and Humanities Research, King Abdulaziz University, Jeddah, Saudi Arabia
cSchool of Criminology and Criminal Justice, University of Nebraska Omaha, Lincoln, NE 68588-0561, USA
dSchool of Criminal Justice, University of Cincinnati, Cincinnati, OH 45221-0389, USA
eCriminology & Criminal Justice, School of Social Work, Saint Louis University, St. Louis, MO 63103, USA
fDepartment of Epidemiology, Saint Louis University, St. Louis, MO 63104, USA
gDepartment of Criminal Justice, Pennsylvania State University, Abington, Abington, PA 19001, USA
a b s t r a c ta r t i c l ei n f o
Article history:
Received 20 October 2015
Accepted 4 November 2015
Available online xxxx
Keywords:
Add Health
Genetics
Environments
Expulsions
Suspensions
Therehas beena significantamountof interestinunderstanding someofthekeyissuesrelatedtoschoolsuspen-
sionsandexpulsions.Oneofthemoreintriguingandstudiedoftheseissueshastodowithfactorsthatcontribute
tovariationinschooldisciplinarysanctions.Todate,however,noresearchhasexaminedthegeneticarchitecture
to either suspensions or expulsions. The current study addresses this gap in the literature by analyzing a sample
oftwinpairsdrawn fromtheNationalLongitudinalStudyof AdolescenttoAdultHealth(AddHealth).Theresults
oftheanalysesrevealedthat sharedand nonshared environmental factorsaccountedfor thevariation insuspen-
sions. Genetic influences, in contrast, were the dominant source of variation for expulsions. We conclude by
discussing the implications of our findings and avenues for future research.
© 2015 Elsevier Ltd. All rights reserved.
There has been a considerable amount of interest in unpacking the
various factors related to school suspensions and expulsions. Research
has, for instance, focused on black–white differences in rates of suspen-
sions and expulsions (Wright, Morgan, Coyne, Beaver, & Barnes, 2014),
whethervarious policies leadto changesin therates of suspensionsand
expulsions (Cornell, Gregory, & Fan, 2011), and the various con-
sequences associated with being suspended or expelled (Rocque &
Paternoster, 2011). Yet, one of the more elusive issues focuses on varia-
tion in the underlying etiological processes that contribute to variation
in suspension and/or expulsion. Numerous explanations have been ad-
vanced, including discrimination by teachers, prejudicial views held by
schooladministrators,anddifferentialinvolvementinbehaviorsthatvi-
olate school policies, to name just a few (Kinsler, 2011; Moore, 2002;
Wrightet al., 2014). What is noticeably absentfrom this lineof research
isinformationonthepotentialinfluenceofgeneticfactorsonexplaining
individual differences in school suspensions and expulsions. This gap in
the existing literature is all the more surprising when juxtaposed
against the fact that previous research has shown that nearly every
human phenotype, including problem behavior, is at least partially in-
fluenced by genetic factors (Polderman et al., 2015).
The results generated from this body of research have revealed con-
sistently that genetic factors explain about 50% of the variance in most
phenotypes. This finding is so well established that it has been dubbed
the first law of behavior genetics (Turkheimer, 2000). The remaining
variance not accounted for by genetic influences is attributable to
non-genetic environmental effects (and error). Two types of environ-
mental influence can account for environmental variance: shared envi-
ronmental influences and nonshared environmental influences. Shared
environmentsareenvironmentsthatarethesamebetweensiblingsand
that make siblings more similar phenotypically. Nonshared environ-
ments, in contrast, are environments and non-genetic factors that
cause siblings to be different from each other. Collectively, the genetic
effect (referred to as heritability) along with shared and nonshared en-
vironmental influences (i.e., non-genetic influences) account for 100%
of the variance in all phenotypes.
Given that these genetic and environmental effects are so well
established, some may question whether it is necessary to conduct ad-
ditional univariate studies on the genetic and environmental basis of
human behaviors. There is a reason to suspect, however, that suspen-
sionsand expulsionsmayhave differentetiologies andthus be differen-
tially affected by genetic and environmental influences. Suspensions,
for example, can be driven by school-specific policies and differential
enforcement by teachers and administrators (e.g., Cornell et al., 2011).
Stated differently, environmental factors somewhat outside the control
Personality and Individual Differences 90 (2016) 214–218
⁎ Corresponding author at: College of Criminology and Criminal Justice, Florida State
University, 145 Convocation Way, Tallahassee, FL 32306-1273, USA.
E-mail address: kbeaver@fsu.edu (K.M. Beaver).
http://dx.doi.org/10.1016/j.paid.2015.11.008
0191-8869/© 2015 Elsevier Ltd. All rights reserved.
Contents lists available at ScienceDirect
Personality and Individual Differences
journal homepage: www.elsevier.com/locate/paid
Page 2
of youth might affect the probability that a child or adolescent is
suspended from school. Moreover, suspensions, when compared to ex-
pulsions, tend to be associated with less serious types of misconduct.
These are often minor forms of misbehavior that are age-normative
and, on average, under less genetic influence when compared against
more serious types of misconduct (Barnes, Beaver, & Boutwell, 2011).
In comparison, there is less discretion when it comes to policies re-
lated to expulsions. Expulsions are driven, at least in part, by more seri-
ous types of behaviors, such as physical fighting, possession of a deadly
weapon, threatening conduct, and the selling of drugs at school. Behav-
iors that lead to expulsions are thus likely to be significantly more seri-
ous when compared to those that result in a suspension. Additionally,
previous research has revealed that genetic influences are more heavily
involved in the development of more serious forms of antisocial behav-
ior relative to less serious forms of antisocial behavior (Barnes et al.,
2011). This suggests that genetic influences may be more likely to un-
derlie the etiology of expulsions compared to suspensions.
To test this possibility, we employ the well-known twin-based re-
search design. With the twin-based research design, the phenotypic
similarity of monozygotic (MZ) twins is compared to the phenotypic
similarity of dizygotic (DZ) twins. Given that MZ twins share twice as
muchgeneticmaterialasDZtwins,andbothtypesoftwinsareassumed
to share relatively equal environments—an assumption that recent re-
search has indicated is typically upheld (Barnes et al., 2014)—then the
only reason that MZ twins should be more similar to one another than
DZ twins is because of genetic influences (Plomin, DeFries, Knopik, &
Neiderhiser, 2013). As the similarity of MZ twins increases relative to
DZtwins,thenthegeneticeffectincreases,too.Thetwin-basedresearch
design has been used in thousands of studies, the assumptions that ac-
companythisdesignhavebeentestedandretested,andoverallpatterns
of findings have been replicated using other types of research designs
(e.g., adoption research designs). Against this backdrop, there is reason
to suspect that the twin-based method is among the most robust re-
search designs employed in the social and behavioral sciences.
1. Methods
1.1. Sample
DataforthisstudyweredrawnfromtheNationalLongitudinalStudy
ofAdolescenttoAdultHealth (AddHealth; Udry,2003).TheAddHealth
is a four-wave prospective study of a nationally representative sample
of American adolescents who were attending middle or high school dur-
ing the 1994–1995 school year. The first wave of questionnaires—known
as the wave 1 in-school surveys—was administered during a regular
school day to nearly 90,000 adolescents. In order to gather additional in-
formation, a subsample of youth along with their primary caregiver was
selected to be re-interviewed in their homes. In total, 20,745 youths
participated in the wave 1 in-home component. Approximately
1.5 years later, the second round of interviews was completed
with 14,738 of the youths. The final two waves of data were collect-
ed in 2001–2002 (n = 15,197) and 2007–2008 (n = 15,701), re-
spectively (Harris et al., 2003). Given that few respondents were
still in high school at wave 3, the current study employs data from
only the first two waves of the Add Health.
Embedded within the Add Health data are a number of subsamples.
One of these subsamples includes twins that can be used for quantita-
tive genetic analysis (Harris, Halpern, Smolen, & Haberstick, 2006).
Twins were oversampled for inclusion in the study. Specifically, during
wave 1 interviews, respondents were asked whether they had a co-
twin. If they responded affirmatively, then their co-twin was also in-
cluded in the sample. Overall, close to 800 twin pairs were included in
the Add Health data. After removing cases with missing data and after
eliminating cases with unknown zygosity, the final analytical sample
consisted of 289 MZ twin pairs (578 individuals) and 248 same-sex
DZ twin pairs (496 individuals).
1.2. Measures
1.2.1. Suspension
Three measures of suspension were used in the current study. First,
during wave 1 interviews, respondents were asked whether they had
ever received an out-of-school suspension. Responses to this question
were coded dichotomously, such that 0 = never suspended and 1 =
suspended. Second, during wave 2 interviews, respondents were again
asked if they had received an out-of-school suspension but this time
they were asked to report only those incidences that occurred during
the current school year (if the interview was conducted in the summer,
theywereaskedabouttheprecedingschoolyear).Responseswereonce
again coded dichotomously, wherein 0 = not suspended and 1 =
suspended. Third, these two dichotomous suspension variables were
summedtogetherandthendichotomizedtoarriveatalifetimemeasure
of suspension. With this coding scheme, 0 = never suspended and 1 =
suspended at least one time.
1.2.2. Expulsion
Three measures of expulsion were analyzed in this study and they
were measured in much the same way as the suspension variables.
First, during wave 1 interviews, respondents were asked whether they
had ever been expelled from school. Responses to this question were
coded dichotomously, such that 0 = not expelled and 1 = expelled. Sec-
ond, during wave 2 interviews, respondents were asked whether they
had been expelled in the current school year (if the interview was con-
ducted in the summer, they were asked about the preceding school
year). Responses were once again coded dichotomously with 0 = not
expelled and 1 = expelled. Third, these two dichotomous expulsion var-
iables were summed together and then dichotomized to create a life-
time measure of expulsion, where 0 = never expelled and 1 = expelled
at least one time. Table 1 presents the prevalence of both suspensions
and expulsions, separated according to twin zygosity type.
1.3. Plan of analysis
The analysis for the current study followed a two-step process.
First, logistic regression models were estimated by zygosity to deter-
mine whether there is an association between one twin having been
suspended (expelled) and the odds that their co-twin had been
suspended (expelled). If there is a genetic effect on being suspended
(expelled), then the association for MZ twins should be significantly
greater than the association for DZ twins. Second, to more formally esti-
mate the genetic, shared environmental, and nonshared environmental
influences on expulsions and suspensions, liability threshold versions of
the ACE model were estimated for each of the examined outcomes
using the statistical software program Mplus (Muthén & Muthén, 2010).
The liability threshold model is a biometric model fitting technique
that is similar to the traditional univariate ACE model, but is acceptable
for both categorical and dichotomous outcome measures (Prescott,
2004). The liability threshold model decomposes the variance in each
of the examined outcome measures into three latent estimates: genetic
influences (symbolized as A), shared environmental influences (sym-
bolized as C), and nonshared environmental influences (symbolized as
Table 1
Prevalence of suspensions and expulsions by zygosity.
MZ twinsDZ twins
Suspension (wave 1)
Suspension (wave 2)
Expulsion (wave 1)
Expulsion (wave 2)
Ever suspended
Ever expelled
N (pairs)
N (individuals)
23.6%
12.9%
4.6%
2.1%
26.5%
6.3%
289
578
30.2%
11.3%
5.5%
2.5%
30.0%
5.7%
248
496
215
K.M. Beaver et al. / Personality and Individual Differences 90 (2016) 214–218
Page 3
E).The A componentof themodel provides a latent estimate of thepro-
portionofthevarianceintheoutcomemeasurethatcanbeexplainedby
additivegenetic influences.TheCcomponentprovidesa latentestimate
of the proportion of the variance in the examined outcome that can be
explained by shared environmental influences that make siblings
more similar (e.g., SES and school policies). The E component provides
a latent estimate of the proportion of the variance in the outcome mea-
surethatisexplainedbyuniqueornonsharedenvironmentalinfluences
that make siblings different from one another (e.g., different peer
groups), and also includes error. All three factors cumulatively explain
100% of the variance in the examined outcome measure.
2. Results
The analysis began by estimating binary logistic regression models
for suspensions at wave 1 and wave 2. The results of these models are
presented as predicted probabilities, conditional on co-twin suspension
statusandzygosity.TheestimatesarepresentedinFig.1.Ascanbeseen,
the odds ratios (ORs)—which are provided in the figure caption—are all
larger for MZ twins compared to DZ twins, suggesting that there could
beageneticeffectontheprobabilityofsuspension.Importantly,howev-
er, the predicted probabilities of being suspended are largely compara-
ble for MZ and DZ twins.
Next, the same models were calculated except that the wave 1 and
wave 2 expulsion variables were used instead of the suspension vari-
ables. The results of the logistic regression models are presented in
Fig. 2. For these variables, the ORs are all larger for MZ twins than for
DZ twins. The predicted probabilities clearly indicated that the effect
of having an MZ co-twin who was expelled increased the predicted
probability of being expelled more when compared to the effect of hav-
ing a DZ co-twin who was expelled. This pattern of results is consistent
with a partial genetic explanation for the odds of expulsion.
The last set of logistic regression models examined the composite
measures of ever being suspended or ever being expelled. The results
of these models are presented in Fig. 3 and show a pattern of results
that is consistent with the wave-specific findings. Specifically, the ORs
and predicted probabilities for the ever suspended variable are compa-
rable for MZ and DZ twins, suggesting genetic influences on being
suspended are minimal. For expulsions, the OR is greater for MZ twins
than for DZ twins, which is reflected in the drastically different predict-
ed probabilities of being expelled between both types of twins. This
latter finding therefore suggests a potential genetic effect on being
expelled.
Last, liability threshold models were conducted to estimate the ge-
netic, shared environmental, and nonshared environmental influences
on the suspension and expulsion measures. Several goodness of fit sta-
tistics were estimated to evaluate model fit. More specifically, the Com-
parative Fit Index (CFI), the Tucker–Lewis Index (TLI), and the Root
Mean Square Error of Approximation (RMSEA) were estimated. Model
fit was assessed using results from all of the estimated statistics using
criteria specified by Hu and Bentler (1999). A model that fits the data
wellshouldreportCFIandTLIvaluesof.95orgreaterandRMSEAvalues
of .06 or smaller. Wald's difference in coefficient test was used to max-
imize model parsimony and compare nested models alongside baseline
models. Chi-square statistics were calculated to examine whether the
prevalence of suspensions and expulsions varied significantly by zygos-
ity (see Table 1). The results revealed that the chi-square values were
non-significant for wave 1 and wave 2 expulsions, wave 2 suspensions,
the ever suspended measure, and the ever expelled measure. The only
chi-square statistic that was significant was for wave 1 suspensions
(chi-square = 5.80, df = 1, p = .02).
The results of the liability threshold models and corresponding model
fit statistics are presented in Table 2. As can be seen, the wave-specific
suspension measures were not influenced by genetic factors, but rather
by a combination of shared environmental influences (between 64%
Fig. 1. Predicted probabilities of being suspended based on zygosity and whether co-twin
had been suspended. Notes: MZ twins (wave 1): b = 2.210, SE = .322, OR = 9.120,
p b .001; MZ twins (wave 2): b = 3.202, SE = .515, OR = 24.571, p b .001; DZ twins
(wave 1): b = 1.765, SE = .308, OR = 5.843, p b .001; DZ twins (wave 2): b = 2.470,
SE = .530, OR = 11.818, p b .001.
Fig. 2. Predicted probabilities of being expelled based on zygosity and whether co-twin
had been expelled. Notes: MZ twins (wave 1): b = 2.750, SE = .715, OR = 15.636,
p b .001; MZ twins (wave 2): b = 4.392, SE = 1.306, OR = 80.800, p b .001; DZ twins
(wave 1): b = 1.277, SE = .829, OR = 3.587, p = .123; DZ twins (wave 2): b = 2.714,
SE = 1.260, OR = 15.083, p b .05.
Fig. 3. Predicted probabilities of ever being suspended/expelled based on zygosity and
whether co-twin had ever been suspended/expelled. Notes: MZ twins (ever suspended):
b = 2.110, SE = .360, OR = 8.246, p b .001; MZ twins (ever expelled): b = 3.489, SE =
.717, OR = 32.743, p b .001; DZ twins (ever suspended): b = 1.980, SE = .358, OR =
7.239, p b .001; DZ twins (ever expelled): b = 1.546, SE = .861, OR = 4.694, p = .073.
216
K.M. Beaver et al. / Personality and Individual Differences 90 (2016) 214–218
Page 4
and 77% of the variance) and nonshared environmental influences
(between 23% and 36% of the variance). A different pattern of findings
emerged for the wave-specific expulsion measures. Specifically, genetic
factors explaining 64% of the variance in the liability for expulsion at
wave 1 and 86% of the variance in the liability at wave 2. The remaining
variance in liability was attributable to nonshared environmental sources
ofvariance.Inthefinalmodels,measuresassessingeverbeingsuspended
and ever being expelled were analyzed and the results revealed that 65%
of the variance in liability of ever being suspended was the result of
shared environmental factors, while the remaining 35% of the variance
was the result of nonshared environmental factors. With respect to ever
being expelled, ACE model parameter estimates from the best-fitting
model suggested that 80% of the variance in liability was due to genetic
factors, while the remaining 20% of the variance in liability was the result
of nonshared environmental factors.
3. Discussion
During the past decade, we have seen a resurgence of research ex-
amining some of the key factors related to youth who receive serious
school sanctions, particularly suspensions and expulsions. Despite the
amount of research devoted to this topic, to date there has not been
any research examining the genetic architecture of these types of disci-
plinary outcomes. The currentstudy addressed this gapin theliterature
by analyzing a sample of twin pairs drawn from the Add Health. The
results of our analyses revealed two broad findings. First, the results
revealed that variation in suspensions was not influenced by genetic
factors, but rather was theresult of shared and nonshared environmen-
tal effects. This is a particularly interesting finding given that virtually
every humanphenotype, at least to some extent, is impacted by genetic
factors (Polderman et al., 2015). We revisit this finding momentarily.
Second, and in direct contrast, were the results from the analysis of ex-
pulsion. These results suggested genetic factors accounted for most of
thevarianceinliabilityforexpulsion,withnonsharedenvironmentalin-
fluences accounting for the remaining variance.
Now to address the question: What accounts for these somewhat
disparate findings in regard to the genetic and environmental contribu-
tors to suspensions and expulsions? As previously mentioned, suspen-
sions are often reserved for less serious offenses and may be guided
more by school-specific policies. Suspensions might also be the result
of subjective assessments and enforced at the discretion of teachers
and school administrators. Expulsions, however, are often reserved for
themostserioustypesof behavior and for students whodisplaydisrup-
tive and serious misconduct repeatedly. Not only are these types of be-
haviors under strong genetic influence (Barnes et al., 2011), but
teachers and administrators likely have less discretionary oversight in
administering punishments in such instances as most of the time they
fall under zero-tolerance policies. If this is the case, then the end result
is that individual behavior may drive expulsions to a greater extent
than it does for suspensions, suggesting that genetic factors are more
likely to underlie the former.
Thesefindingshavepotentialimportantramificationsforcurrentre-
search on suspensions and expulsions. Using the current findings as a
backdrop, it appears as though expulsions are driven by individual be-
havior, not by some type of widespread discriminatory nature of school
administrators. Variation in suspensions, however, is driven more by
environmental influences. Much more research is needed on this topic
in order to unpack the specific shared and nonshared environmental
factors that are involved in creating variation in suspensions. Until
such research is conducted, it is impossible to identify the specific envi-
ronmental factors that are related to school suspensions among youth.
Although our study is the first to estimate genetic and environmen-
talinfluencesonsuspensionsandexpulsions,thereareanumberoflim-
itationsthatshouldbeaddressedinfuturestudies.First,themeasuresof
suspensions and expulsions were based on self-reports, not on school
administration data. As a result, there is the possibility that these esti-
mates arenot entirelyaccurate. Butwehave littlereason tosuspectsys-
tematic bias given that self-reports of arrest (a much more serious
outcome, meaningit is more likely to show biasthan self-reportsofsus-
pensionsorexpulsions)tendtoprovideanaccuratepictureofoffenders'
arrest histories (Pollock, Menard, Elliott, & Huizinga, 2015). Second, the
analyses were based on a sample of twins and thus the generaliz-
ability of these estimates remains in question. With that said, re-
search exists showing that the twin sample of the Add Health is
not significantly different from the nationally representative sample
of the Add Health (Barnes & Boutwell, 2013). Last, the current anal-
ysis was only able to decompose variance in suspensions and ex-
pulsions without providing any information regarding the specific
genetic polymorphisms, the specific shared environments, and the
specific nonshared environments that might be accounting for this
variance. Future research would benefit greatly from identifying
these sources of variance.
Before concluding, it is important to note that just because expul-
sions are under significant genetic influence does not mean that these
behaviors are inevitable outcomes for those who are genetically
predisposed for being expelled. Genetic influences can be exacerbated
or dampened depending on exposure to different types of environ-
ments (Rutter, 2006). Certain school and classroom environments
may suppress or, conversely, may exacerbate antisocial predispositions.
Beginning to identify which environments—particularly those found
within the school—might be able to attenuate genetic effects could go
a long way towards reducing problem behavior at school and, as a
consequence, expulsions. In order for this to be achieved, the possibility
that genetic influences are at play should be given full consideration in
future studies on the causes of suspensions and expulsions.
Table 2
ACE model parameter estimates and fit statistics for suspensions and expulsions.
Parameter estimatesModel fit statistics
ACE
χ2
Δχ2
CFI TLIRMSEA
Suspension (wave 1)
ACE .18
AE .59
CE
.00
E .00
.50⁎⁎
.00
.64⁎⁎
.00
.32⁎⁎
.08
.36⁎⁎
1.00⁎⁎
7.18
15.87
7.70
167.11
.971
.918
.975
.000
.981
.959
.987
.553
.072
.105
.059
.348
8.04⁎⁎
.72
147.17⁎⁎
Suspension (wave 2)
ACE .27
AE
CE
.00
E .00
.54⁎
.00
.77⁎⁎
.00
.19⁎⁎
.14⁎
.23⁎⁎
1.00⁎⁎
1.39
7.38
2.49
1.00
.979
1.000
.000
1.007
.990
1.005
.565
.000
.060
.000
.388
.86⁎⁎
5.39⁎
1.03
164.11⁎⁎
181.11
Expulsion (wave 1)
ACE
AE
CE .00
E .00
.64⁎
.64⁎⁎
.00
.00
.54⁎⁎
.00
.36⁎
.36⁎
.46⁎⁎
1.00⁎⁎
1.05
1.23
2.93
23.37
1.000
1.000
1.000
.048
1.068
1.072
1.028
.619
.000
.000
.000
.117
.00
5.98⁎
21.31⁎⁎
Expulsion (wave 2)
ACE.54
AE
CE.00
E.00
.30
.00
.78⁎⁎
.00
.16
.14
.22
1.00⁎⁎
1.72
2.00
2.52
45.06
1.000
1.000
1.000
.000
1.022
1.026
1.019
.590
.000
.000
.000
.185
.86⁎⁎
.30
.78
41.10⁎⁎
Ever suspended
ACE
AE
CE
E
.04
.78⁎⁎
.00
.00
.62⁎⁎
.00
.65⁎⁎
.00
.34⁎⁎
.22⁎⁎
.35⁎⁎
1.00⁎⁎
2.06
13.91
2.03
135.71
1.000
.918
1.000
.000
1.005
.959
1.008
.569
.000
.103
.000
.335
10.76⁎⁎
.04
123.32⁎⁎
Ever expelled
ACE
AE
CE
E
.80⁎⁎
.80⁎⁎
.00
.00
.00
.00
.73⁎⁎
.00
.20⁎
.20⁎
.27⁎⁎
1.00⁎⁎
.46
.61
3.78
71.40
1.000
1.000
1.000
.000
1.026
1.026
1.002
.591
.000
.000
.000
.239
.00
66.92⁎⁎
66.93⁎⁎
Chi-square difference tests were performed using Wald's test of parameter constraints.
Note: Best-fitting model are bold-typed.
⁎ p b .05.
⁎⁎ p b .01.
217
K.M. Beaver et al. / Personality and Individual Differences 90 (2016) 214–218
Page 5
Acknowledgements
ThisresearchusesdatafromAddHealth,aprogramprojectdesigned
by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and
funded by a grant P01-HD31921 from the Eunice Kennedy Shriver
National Institute of Child Health and Human Development, with coop-
erative funding from 17 other agencies. Special acknowledgment is due
to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original
design. Persons interested in obtaining data files from Add Health
should contact Add Health, Carolina Population Center, 123 W. Franklin
Street, Chapel Hill, NC 27516-2524 (addhealth@unc.edu). No direct
support was received from grant P01-HD31921 for this analysis.
References
Barnes, J. C., & Boutwell, B. B. (2013). A demonstration of the generalizability of twin-based
research on antisocial behavior. Behavior Genetics, 43, 120–131.
Barnes, J. C., Beaver, K. M., & Boutwell, B. B. (2011). Examining the genetic underpinnings
to Moffitt's developmental taxonomy: A behavioral genetic analysis. Criminology, 49,
923–954.
Barnes, J. C., Wright, J. P., Boutwell, B. B., Schwartz, J. A., Connolly, E. J., Nedelec, J. L., &
Beaver, K. M. (2014). Demonstrating the validity of twin research in criminology.
Criminology, 52, 588–626.
Cornell, D. G., Gregory, A., & Fan, X. (2011). Reductions in long-term suspensions
following adoption of the Virginia Student Threat Assessment Guidelines.
NASSP Bulletin, 95, 175–194.
Harris, K. M., Florey, F., Tabor, J., Bearman, P. S., Jones, J., & Udry, J. R. (2003). The National
Longitudinal Study of Adolescent Health: Research design [www document]. URL:
http://www.cpc.unc.edu/projects/addhealth/design
Harris, K. M., Halpern, C. T., Smolen, A., & Haberstick, B. C. (2006). The National Longitudinal
Study of Adolescent Health (Add Health) twin data. Twin Research and Human Genetics,
9, 988–997.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural equation modeling: A
multidisciplinary journal, 6, 1–55.
Kinsler, J. (2011). Understanding the black–white school discipline gap. Economics of
Education Review, 30, 1370–1383.
Moore, A. L. (2002). African-American early childhood teachers' decisions to refer
African-American students. International Journal of Qualitative Studies in Education, 15,
631–652.
Muthén, L. K., & Muthén, B. O. (2010). Mplus User's Guide: Statistical Analysis with Latent
Variables: User's Guide. Muthén & Muthén.
Plomin, R., DeFries, J. C., Knopik, V. S., & Neiderhiser, J. M. (2013). Behavioral genetics (6th
ed.). New York, NY: Worth.
Polderman, T. J.C., Benyamin, B., de Leeuw, C. A., Sullivan, P. F., van Bochoven, A., Visscher,
P. M., & Posthuma, D. (2015). Meta-analysis of the heritability of human traits based
on fifty years of twin studies. Nature Genetics, 47, 702–709.
Pollock, W., Menard, S., Elliott, D. S., & Huizinga, D. H. (2015). It's official: Predictors of
self-reported vs. officially recorded arrests. Journal of Criminal Justice, 43, 69–79.
Prescott, C. A. (2004). Using the Mplus computer program to estimate models for contin-
uous and categorical data from twins. Behavior Genetics, 34, 17–40.
Rocque, M., & Paternoster, R. (2011). Understanding the antecedents of the school-to-jail
link: The relationship between race and school discipline. Journal of Criminal Law &
Criminology, 101, 633–665.
Rutter, M. (2006). Genes and behavior: Nature–nurture interplay explained. Malden, MA:
Blackwell.
Turkheimer, E. (2000). Three laws of behavior genetics and what they mean. Current
Directions in Psychological Science, 9, 160–164.
Udry, J. R. (2003). The National Longitudinal Study of Adolescent Health (Add Health), waves
I and II, 1994–1996; wave III, 2001–2002 [machine-readable data file and documentation].
Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel
Hill.
Wright, J. P., Morgan, M. A., Coyne, M. A., Beaver, K. M., & Barnes, J. C. (2014). Prior
problem behavior accounts for the racial gap in school suspensions. Journal of
Criminal Justice, 42, 257–266.
218
K.M. Beaver et al. / Personality and Individual Differences 90 (2016) 214–218
Download full-text 
