Vous êtes sur la page 1sur 17

Student Discipline and High School Performance

Author(s): David E. Myers, Ann M. Milne, Keith Baker and Alan Ginsburg
Source: Sociology of Education, Vol. 60, No. 1 (Jan., 1987), pp. 18-33
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2112616 .
Accessed: 19/03/2014 23:51
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
Sociology of Education.
http://www.jstor.org
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE
DAVID E. MYERS KEITH BAKER
ANN M. MILNE ALAN GINSBURG
DRC U.S. Department of Education
Sociology of Education 1987, Vol. 60 (January):18-33
This paper examines the relationship between student misbehavior and academic performance and the
effects offamily structure and mother's employment on misbehavior and performance. Using panel data
on high school sophomores from the High School and Beyond (HSB) survey, we estimate a number of
linear panel models. The findings indicate that sophomores with low grades misbehave more as seniors
than those with high grades. Academic achievement in the sophomore year has little effect on changes
in misbehavior. Misbehavior, however, has negative effects on changes in grades and achievement test
scores. Finally, living in a single-parent family and mother's employment negatively affect both
achievement and behavior.
INTRODUCTION
For the last sixteen years, the Gallup education
opinion polls have found that the public's major
concern about schools is the lack of discipline
(Gallup 1984). Recently, A Nation at Risk
(Commission on Excellence in Education 1983)
and other national reports have called attention to
the implications of the decline in test scores for this
nation's youth. It has been hypothesized that poor
academic performance is a function of student
misbehavior and, more generally, lack of disci-
pline in the classroom. While educators have
focused on the consequences of misbehavior,
researchers in the area of juvenile delinquency
have hypothesized that misbehavior is attributable
to poor school performance.
During the same period, there have been a
number of changes in family demographics that
may be related to students' school behavior and
success. Foremost among these trends are in-
creases in the incidence of single-parent families
and working mothers. Between 1970 and 1980, the
proportion of children living in one-parent families
increased from about 11 percent to nearly 19
percent (U.S. Bureau of the Census 1982). The
U.S. Bureau of Labor Statistics (1983) reports that
in the same decade, the labor force participation of
mothers with children under 18 increased from
about 42 percent to more than 56 percent.
REVIEW OF RESEARCH LITERATURE
Each of these issues-the relationship between
academic performance and misbehavior, and the
impact of family structure and mother's employ-
ment on academic performance and misbehavior-
has been addressed in the literature.
The Relationship between Academic Performance
and Misbehavior
A number of theories have been developed in the
juvenile delinquency literature to explain the
causes of misbehavior and its relationship to
achievement. Early theorists held that delinquency
was a problem associated with low socioeconomic
status. In an attempt to account for this relation-
ship, Merton (1968) and, more specifically,
Cloward and Ohlin (1960) proposed that the poor
socialization of lower-class students or the aca-
demic deficiencies they brought to school would
lead to school failure, which would subsequently
lead to discipline problems. Cohen (1955) ex-
panded this notion by theorizing that many
students with these lower-class characteristics,
when confronted with the middle-class values
inherent in the schools, would become frustrated
and fail.
However, Call (1965) and Polk and Halferty
(1966) found little class difference in the back-
grounds of delinquents and nondelinquents, and
Stinchcombe (1964) suggested that it was not
family status but status prospects (i.e., a student's
educational and occupational outlook) that were
important. Stinchcombe showed that misbehavior
at all class levels was lower among college-
oriented or high-achieving students. Thus, he
predicted greater discipline problems among low-
achieving middle-class boys, a hypothesis not
This paper was prepared under contract no. 300-83-021 1
with the U. S. Department of Education. An earlier
version was presented at the 1985 meetings of the
American Educational Research Association, Chicago.
The authors thank Michael Finch and Sandra Hanson for
providing comments on an earlier version, Jane Burnette
for editorial support, and two anonymous reviewers for
helping to clarify a number of points in this paper. The
opinions expressed in this paper do not necessarily reflect
the position or policy of the U.S. Department of
Education or DRC. Address all correspondence to Dr.
David E. Myers, DRC, 1828 L Street, NW, Fifth Floor,
Washington, DC 20036.
18
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 19
borne out by his own data or by Polk's (1967) or
Elliott's (1966).
Given the importance of expectations of achieve-
ment in the development of these theories, a
natural extension was the focus on labeling theory
(e.g., see the summary by Pink [1979]) and the
hypothesis that schools, through various types of
labeling, create the situation for failure. Labeling
theory argues that students who are "successfully
labeled as inferior and treated as inferior [and who
internalize the concept] will in fact perform in an
inferior fashion. Such students lack commitment to
the school, see school (and its rules) as unimpor-
tant to them, and are thus prone to rebellion and
delinquency" (Pink 1979, p. 46). Pink, like those
before him, posits that the direction of causality is
from academic failure to misbehavior.
Unlike criminologists, who have focused on one
direction of the relationship between school
success and misbehavior, sociologists have pointed
out that there is also a causal path from
misbehavior to school performance (Coleman,
Hoffer, and Kilgore 1982; Purkey and Smith 1983;
DiPrete 1981; Baker 1985). These authors argue
that good discipline is a prerequisite for learning,
and their research has consistently shown that a
good discipline climate is one of the few variables
associated with academic achievement.
Thus, two hypotheses concerning the relation-
ship between school success (performance) and
misbehavior can be derived from the literature.
The criminological literature suggests that poor
school performance will result in more misbehav-
ior. The educational literature suggests that
misbehavior will result in poorer performance.
DiPrete (1981) tested these two hypotheses
using cross-sectional data rather than the longitudi-
nal data used here and found that the effect of
misbehavior on grades was weaker than the effect
of grades on misbehavior. He notes that the
restriction to cross-sectional data "creates difficul-
ties in determining the correct causal ordering of
school outcomes" (1981, p. 154). While sugges-
tive, his analyses suffer from the inability to
deduce directionality from cross-sectional data. (In
the analyses reported here, we used the cross-
sectional data used by DiPrete as the first wave of
our panel data.)
Family Situation, Misbehavior, and Academic
Performance
Several studies (e.g., Coleman et al. 1966;
Jencks et al. 1972) have demonstrated the effects
of socioeconomic status on achievement. Less
attention has been paid to those aspects of family
structure that have changed dramatically in recent
years-the number of parents in the home and
mother's employment.
Single Parents. There are a number of reasons
why the absence of a parent in the home affects
school outcomes. Single parents have high rates of
poverty (Bogue 1985), have less time for the
supervision and care of their children (Hetherington,
Camara, and Featherman 1981), and may have
emotional problems stemming from divorce and
separation (Hetherington et al. 1981). Each of
these factors can be expected to have negative
implications for school outcomes. The earliest
studies on single-parent families (e.g., see reviews
by Herzog and Sudia 1973; Hetherington et al.
1981; Shinn 1978) found effects of father absence
on delinquency in boys, although they did not
always control socioeconomic status variables. In a
recent study, DiPrete (1981) found higher rates of
discipline problems among students from single-
parent families. Hetherington et al. (1981) and
Kelly and Wallerstein (1979) have suggested that
children from single-parent families created by
divorce have more discipline problems than
children from single-parent families created by
death. They attribute this difference to the
emotional disturbance caused by separation and
divorce. Milne et al. (1986) found that students
from single-parent families have lower achieve-
ment and that much of the difference in achieve-
ment can be attributed to the lower incomes of
these families.
Working Mothers. The effects of mother's
employment are also related.to the lack of time for
supervision and care, but they may be offset by
increased family income and by mother's function
as a role model, particularly for daughters. Gold
(1963) found more delinquency among the sons of
working mothers with white-collar husbands than
among the sons of working mothers in other social
classes. Etaugh (1974) reviewed a number of
studies relating mother's employment to lack of
behavioral adjustment in school but found that the
effects differed by age and sex of the child, by
social class, and by mother's satisfaction with her
job. Hoffman (1979, 1980), in particular, notes the
importance of using control and intervening
variables when studying the adjustment of children
who have working mothers.
Hoffman (1980) states that while mother's
employment may have no effect on the achieve-
ment of girls (partly because of positive role
modeling) and lower-class boys, it may have a
negative effect on the achievement of middle-class
boys. Milne et al. (1986) show that mother's
employment has negative effects for white students
from two-parent families but positive effects for
black students from single-parent families. The
potential beneficial effects of mother's employ-
ment on achievement in poor and black families is
also noted by Hoffman (1980) and by Heyns
(1982).
THE MODEL
The conceptual model used to guide the
investigation of the relationship between student
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
20 MYERS ET AL.
misbehavior and academic performance is shown
in Figure 1. The panel design permits us to explore
the effects of student misbehavior on changes in
two measures of academic performance: (1)
self-reported grades and (2) reading and mathemat-
ics achievement test scores. Similarly, we can
examine the extent to which academic perfor-
mance influences changes in student misbehavior
between the sophomore and senior years of high
school. With this panel design, we are able to
assess, for example, the impact of misbehavior in
the sophomore year on changes in academic
performance between the sophomore and senior
years.
The simultaneous equation model, which as-
sumes a simultaneous relationship between misbe-
havior and academic performance, is an altemative
to the panel design. That is, we could postulate a
reciprocal relationship between misbehavior and
performance in which both are measured at the
same point in time (e.g., when the students were
seniors in high school). The panel model, on the
other hand, postulates that misbehavior and
academic performance in the sophomore year
affect misbehavior and performance in the senior
year. Thus, the simultaneous equation model and
the panel model represent different formulations of
the process.
The criminologist's theories imply that student
grades but not achievement tests influence changes
in student misbehavior. Students rarely know their
scores on standardized achievement tests; there-
fore, these scores may not affect students'
perceptions of how well they are doing in school.
On the other hand, students know their grades and
use them to measure their success or failure in
school relative to their classmates. Thus, it is
expected that grades will have a direct effect on
student misbehavior and that achievement scores
will have little, if any, effect.
Students who perceive that they are failing in
school can be expected to exhibit nonnormative
behavior. For example, students with low perfor-
mance may resort to alternative forms of behavior
to attain specific goals. In addition, students not
performing well in school may be labeled bad
students by both teachers and classmates and, in
tum, may act in a manner consistent with those
expectations.
On the other hand, misbehavior in the sopho-
more year is expected to influence both grades and
achievement scores. Students who misbehave
(e.g., cut class) are likely to perform at a lower
level than their classmates who regularly attend
class and behave in prescribed ways. Further, a
student's misbehavior may influence a teacher's
subjective rating of that student and result in poor
grades.
In addition to emphasizing the relationship
between misbehavior and changes in academic
performance, the model also states that the number
of parents in the family, mother's employment,
SINGLE PARENT. SENIOR MISBEHAVIOR
MOTHER WORKS SENIOR GRADES
MOTHER'S EDUCATION SENIOR ACHIEVEMENT
SCORES
NUMBER OF SIBLINGS
FAMILY INCOME
HIGH SCHOOL PROGRAM
SOPHOMORE EDUCATONAL
ATTAINMENT EXPECTATIONS
SOPHOMORE MISBEHAVIOR
SOPHOMORE GRADES
SOPHOMORE ACHIEVEMENT
SCORES
Figure 1. The conceptual model. (Models are estimated separately by sex.)
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 21
mother's educational attainment, family income,
number of siblings, and high school curriculum
have both direct and indirect effects on each of the
endogenous variables in the model. We expect
students from single-parent families and students
whose mothers work to misbehave more than
students from two-parent families and students
whose mothers do not work because of the more
limited potential for supervision in these families.
Further, we expect mother's educational attain-
ment and family income to have negative effects
on misbehavior and positive effects on academic
performance. These influences may be attributed to
family socialization (Cloward and Ohlin 1960), to
expectations about educational performance, and
to the availability of material and intellectual
resources in the home (Benson, Medrich, and
Buckley 1980). A large number of siblings is
expected to have a positive influence on misbehav-
ior and a negative influence on academic perfor-
mance. This effect may be attributed to reduced
supervision by adult family members. Enrollment
in a general or vocational program rather than an
academic program is expected to increase student
misbehavior (Schafer and Olexa 1971) and de-
crease academic performance (Alexander and McDill
1976). Finally, the student's educational attainment
expectations are expected to mediate the effects of
the exogenous variables and to directly influence
changes in misbehavior and academic performance.
Those with high aspirations are expected to expe-
rience decreases in misbehavior and increases in
academic performance.
We conducted separate analyses for white
males, white females, black males, and black
females. This allowed us to test for possible
interactions between race and sex and the remain-
ing variables in each of the models. DiPrete (1981)
found that the processes relating family back-
ground and misbehavior differed for males and
females, and Milne et al. (1986) observed strong
interaction effects between race and other variables
on academic achievement.
DATA AND METHODS
Data
Data from the base year and first follow-up of
the HSB survey (National Center for Education
Statistics 1983a, 1983b) were used in the analyses
reported here. The base-year data, collected in the
spring of 1980, were obtained from a multistage,
stratified, cluster sample of about 30,000 sopho-
mores in over 1,100 schools; 36 sophomores were
randomly selected from each school. In the first
follow-up, conducted in the spring of 1982,
information was obtained from nearly all base-year
respondents. For purposes of the analyses reported
here, we excluded students who did not participate
in the base-year survey or the follow-up survey,
those who transferred from one school to another
school, and those who graduated early. In
addition, we confined our analyses to white and
black students. Constraining the sample in the
above manner reduces the overall number of
students from 30,000 to 19,000.
In both the base-year and follow-up surveys,
students were asked about their family back-
ground, socioeconomic status, family composition,
and aspirations concerning postsecondary school-
ing. For the analyses reported here, the back-
ground (exogenous) variables refer to student and
family characterstics in 1980. Achievement tests in
subject areas such as reading and mathematics
were administered to all students in 1980 and to
base-year sophomores again in 1982. Grades
represent students' self-reports of the grades they
had earned in high school so far, measured in the
sophomore and senior years.
The measure of student misbehavior for each
time period is a Guttman scale derived from three
items, each coded 0 or 1. Students were asked (1)
if they cut class, (2) if they were perceived by their
classmates as troublemakers, and (3) if they had
ever been in serious trouble with the law. Thus,
with the exception of the second item, the
misbehavior items capture cumulative behavior
and not behavior at a specific point in time.
Responses to the second question measure behav-
ior in the sophomore and senior years. Initially, we
considered adding other items to the scale but
found that they did not contribute additional
information. A positive value on the scale indicates
misbehavior, and zero indicates no discipline
problems. The following table presents summary
statistics of the performance of the scales. The
coefficients of reproducibility and scalability for
misbehavior in the sophomore year and in the
follow-up year suggest that these items scale
reasonably well, particularly in the follow-up data.
Sophomores Seniors
Coefficient of reproducibility .89 .94
Coefficient of scalability .51 .76
A reviewer noted that if the misbehavior scale is
primarily a function of students cutting class, then
this variable may be a weak indicator of
misbehavior. A detailed examination of the items,
however, shows that 25 percent of all students
reported that they cut class, 25 percent reported
that their classmates considered them troublemak-
ers, and 4 percent reported that they had been in
trouble with the law. Further examination of the
scale shows that 59 percent of the sophomores in
1980 had no behavioral problems (as measured by
the three items), 41 percent had at least some
disciplinary problems (a score of 1 or more), 12
percent had a moderate or high score on the scale
(2 or 3), and slightly less than 2 percent had a high
score of 3 on the misbehavior scale. Thus, the
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
22 MYERS ET AL.
misbehavior scale is not dominated by class
cutting; it captures a wide range of behaviors that
at the extreme can be thought of as delinquent.
The remaining variables used in our analysis are
briefly described in Table 1. Univariate statistics
for each of the four subsamples are presented in
Table 2.
Previous analyses of family background data
provided by students and parents have shown that
students provide less reliable information than
parents (Rosenthal et al. 1983). Thus, we expect
that some of the estimates reported here will be
attenuated.
Like many large social surveys, the HSB has its
share of missing data. In response to missing data,
researchers typically (1) discard observations with
Table 1. Variable Descriptions
Variable Description
SINGLE Coded 1 if only the mother and no extended family is present, 0 otherwise. Derived
from questions BB036B, BB036C, BB036D, BB036E, BB036G, BB036J, and
BB036K on the base-year survey.
MWPT amA MWFT MWFT coied I ii -stuene' mol'ner woikei ThY-time 'belore ithe studenr was m
elementary school, while the student was in elementary school, and while the student
was in high school; coded 0 otherwise. MWPT coded 1 if student's mother worked in
at least one period but less than full-time during each of the three periods, 0 otherwise.
Derived from BB037A, BB037B, and BB037C.
MEDUC Mother's educational attainment, coded 10 if mother did not complete high school, 12
if she graduated from high school, 13 if she attended a vocational school or college for
less than two years, 14 if she attended a vocational school or college for two years or
more but did not graduate from a four-year program, 16 if she graduated from college
or attended graduate school. Derived from BB042.
NUMSIB Number of children in the household. Derived from BB096A, BB096B, BB096C,
BB096D, and BB096E.
LNINC Logarithm of family income. Family income was coded 3500 if student reported family
income of $6,999 or less, 9500 if income reported was $7,000 to $11,999, 14000 if
income reported was $12,000 to $15,999, 18000 if income reported was $16,000 to
$19,999, 22500 if income reported was $20,000 to $24,999, 31500 if income reported
was $25,000 to $37,999, and 44500 if income reported was $38,000 or more. Derived
from BB1O1.
EDUCEXP Plans to attend college, coded 1 if student did not plan to attend, 2 if student did not
know, and 3 if student did plan to attend college. Derived from BBI 15.
YBDISC Guttman scale based on responses to three items:
1. Every once in a while I cut class. (true or false)
2. Do other sophomores in your school see you as a troublemaker? (very, somewhat,
not at all)
3. I have been in serious trouble with the law. (true or false)
Derived from BB059E, YBO53F, and BB061A.
FYDISC Same as YBDISC, but derived from FY66F, FY74F, and FY76A on the follow-up
survey.
YBREAD Sophomore reading achievement test score. Derived from YBREADFS.
FYREAD Derived from FYREADFS.
YBMATH Sum of two math test scores (items YBMTHIFS and YBMTH2FS).
FYMATH Sum of two math test scores in the follow-up survey (items FYMTHIFS and
FYMTH2FS).
ACAD, GEN, and VOC ACAD coded 1 if student was enrolled in an academic or college-preparatory program,
0 otherwise. GEN coded 1 if student was in a general program, 0 otherwise. VOC
coded 1 if student was in a vocational program, 0 otherwise. Derived from BB002.
SEX Coded 0 if male, 1 if female.
YBGRADE Coded 3.75 if student reported mostly A's, 3.5 if about half A's and half B's, 3.0 if
mostly B's, 2.5 if about half B's and half C's, 2.0 if mostly C's, 1.5 if about half C's
and half D's, 1.0 if mostly D's, and .5 if mostly below D. Derived from BB007.
FYGRADE Coded the same as YBGRADE, but derived from FY7 on the follow-up survey.
RACE Coded 1 if black, 0 if white.
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 23
Table 2. Means and Standard Deviations of Variables, by Sex and Race
White Females White Males Black Females Black Males
Variables Mean SD Mean SD Mean SD Mean SD
SINGLE .108 .311 .109 .311 .318 .466 .307 .461
MWPT .671 .470 .644 .479 .475 .500 .473 .499
MWFT .127 .333 .139 .346 .441 .497 .426 .495
MEDUC 12.624 1.795 12.767 1.809 12.235 1.751 12.438 1.797
NUMSIB 2.814 1.936 2.796 2.055 3.846 2.684 3.730 2.857
LNINC 9.809 .555 9.927 .555 9.403 .710 9.493 .714
GEN .449 .497 .462 .499 .389 .488 .400 .490
VOC .156 .363 .179 .383 .305 .461 .297 .457
EDUCEXP 2.551 .705 2.436 .786 2.625 .658 2.549 .701
YBDISC .467 .698 .692 .825 .377 .599 .589 .767
FYDISCa .492 .658 .776 .835 .382 .581 .562 .685
YBREAD 7.668 4.640 7.960 4.792 4.177 3.685 4.697 4.012
FYREADa 9.250 4.851 9.543 4.986 5.174 4.123 5.947 4.378
YBMATH 14.162 9.129 15.602 10.198 6.219 7.039 6.741 7.866
FYMATHa 16.539 9.888 18.723 10.866 7.972 7.945 9.480 9.039
YBGRADE 2.879 .707 2.645 .779 2.604 .688 2.414 .722
FYGRADEa 3.008 .620 2.768 .685 2.717 .635 2.522 .647
N (base year) 8,281 8,019 1,580 1,350
N (in school
at time of
follow-up survey) 7,710 7,416 1,428 1,169
a
Based on the sample of students enrolled in school at the time of the follow-up survey.
missing data (listwise deletion), (2) use pairwise
covariance matrices when applicable, or (3) impute
missing data. We have selected the last altemative.
A description of the imputation procedure is
provided in Wise and McLaughlin (1980). We
chose this method over pairwise deletion because
the iterative maximum likelihood procedures used
here require complete individual-level data. We
did not use listwise deletion because it resulted in a
large reduction in the number of sample observa-
tions and poor estimates of selected univariate
parameters.
Methods
To obtain estimates of the relationships depicted
in Figure 1, we estimated a series of linear
regression equations. These equations are assumed
to represent the behavioral process in the popula-
tion. The equations were specified for each
individual, i, as follows:
Ytjii=XalBlj +Elji
(=1,
. 5;i=1, . .
,N) (1)
and
Y2ki takYlki
+
X2iB2k
+
E2ki
(k= I, . . . 4; i= 1, . . ., N), (2)
where
Yljf
and Y2k, are the endogenous variables
measured while a student was a sophomore and a
senior, respectively; Xli is a vector of exogenous
variables (including a constant);
Blj
is a conform-
able vector of parameters to be estimated;
Elji
is a
random disturbance;
X2j
is a vector of predeter-
mined variables (not including Ylki); B2k is a
conformable vector of parameters to be estimated;
Otk is a parameter that relates the lagged endogenous
variable Ylk, to the current endogenous variable
Y2ki; and E2ki is a random disturbance term.
Unfortunately, panel data for the sophomore
cohort are incomplete on two outcome variables;
that is, only those who were still in school during
the first follow-up survey were asked about their
in-school misbehavior and the grades they had
earned so far in high school. Since leaving high
school is not a random process, we are left with a
nonrandom sample of students who were high
school sophomores in 1980 and who were still
enrolled in high school in 1982. All analyses
involving dependent variables measured two years
after the base-year survey are based on the sample
of students who remained in the same school that
they attended when they were sophomores. The
sophomore equations are based on all students.'
'We could have restricted the sample for the
sophomore equations to those students who were also in
the senior sample. In so doing, however, we would have
thrown away data that could have been used in the
estimation process and, thus, would have biased our
estimates, unless we had attempted to correct for sample
selection biases, as we did in the senior equations.
Corrections for sample selection biases are based on a
number of assumptions and thus make it more likely that
the results will be called into question. To obtain the
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
24 MYERS ET AL.
As shown by Heckman (1976, 1979), Berk and
Ray (1982), and others, serious biases in the
estimates of the parameters may arise when only a
select subsample of respondents is used. To correct
for selection bias in the senior equations, we
estimated the substantive equations (see equations
[1] and [2]) and an additional equation that relates
the probability of staying in school to a set of
exogenous and predetermined variables measuring
the youth's characteristics and family background.
From this "selection" equation, we obtained
predicted values, which were then included as
explanatory variables in the senior equations. We
then used ordinary least squares (OLS) to estimate
the parameters in the equations. A detailed
illustration of this procedure is presented in
Appendix A.
The only selection process examined here is that
which determines whether a student drops out of
high school. We did not include as part of the
analysis of sample selection bias those students
who transferred from one school to another or
those who graduated early. Out of the total sample
of students, about 2,600 dropped out of school,
1,700 transferred to another school, and roughly
600 graduated early.
RESULTS
The presentation of the results is divided into
four sections. First, we present a series of
descriptive statistics pertaining to misbehavior,
achievement test scores, and grades. Second, we
focus on the direct effects of family background on
both the level of and the changes in misbehavior
and academic performance. Third, we examine the
cross-lagged relationships between misbehavior
and academic performance. Finally, we describe
the stability of students' misbehavior and academic
performance between the initial survey (1980) and
the follow-up survey (1982). (The estimates of the
probit equations used to correct for self-selection
biases are shown in Appendix B.)
Descriptive Analysis
In Table 2, we show the means and standard
deviations for student misbehavior, achievement
test scores, and grades. In general, we find modest
increases in student misbehavior between the
sophomore and senior years. Closer examination
shows that self-reported sophomore misbehavior
tends to be higher for white males than for the
other groups of students and higher for black males
than for both white and black females. This same
pattern holds for student misbehavior measured in
the follow-up survey.
White males tend to score somewhat higher on
both the reading and math achievement tests than
the three remaining groups of students. White
females have the next highest scores, followed by
black males and females.
The means for the final measure of academic
performance-grades-do not parallel those for
the achievement scores. Self-reported grades,
unlike student misbehavior and achievement scores,
are highest for white females, followed by white
males, black females, and black males. Contrast-
ing the findings for grades with the achievement
test scores shows a difference by sex: Males scored
higher on the achievement tests than females, but
females had higher grade point averages than
males.
Family Background, Misbehavior, and
Academic Performance
In this section we first describe the influence of
the exogenous variables on sophomore misbehav-
ior and academic performance (Table 3). Next, we
focus on the effects of family background on
changes in misbehavior and academic performance
(Table 4). It must be kept in mind that the
equations relating the exogenous variables to
sophomore misbehavior and academic performance
were estimated from the total sample of students.
The equations describing senior misbehavior and
academic performance were estimated from the
sample of students who remained in school
between their sophomore and senior years.
White students from one-parent families have
slightly higher levels of misbehavior than white
students from two-parent families. Further, white
sophomores from one-parent families tend to have
low achievement scores and grades. Students
whose mothers work tend to have somewhat higher
levels of sophomore misbehavior and lower
reading and math achievement scores, grades, and
educational attainment expectations than students
whose mothers do not work. Living in a
single-parent family or having a mother who works
has few significant effects for blacks.
Mother's educational attainment is positively
associated with students' academic performance
and educational attainment expectations. In nearly
all subsamples, students from large families have
high levels of misbehavior, low academic perfor-
mance, and low educational attainment expecta-
tions. As expected, family income has a positive
effect on academic performance and expectations
to attend college. However, for white males and
females, high family income is also related to high
levels of misbehavior. For black students, no
relationship between family income and misbehav-
ior is observed.
When students enrolled in academic, general,
and vocational programs are compared, we find
that, net of the other exogenous variables in the
most defensible findings, we used the complete sample to
estimate the sophomore equations.
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 25
00 (1
oo
'I (N (1 (1 W) (71 W) 0 0XoX
.~~~~~~~~~c en w
-
oo ^c ? mmF o NNmo r- -Rt
1-
Ei m
xc
qe -c
C1
,
-0 cW m a,, oN "t "toX 0
en CN en en C) C) C) en C, 00 C1 W N F o
-C,
00c>l 00 _4W) M en - ,) en oo _4
Q , Q o o ? t ^ o ? oo F C,4 ,, a,, ?c t o
=~~~~~~ C) -R
C) 00 -4 -4 O- O44oN
t: ~ ~ ~ ~~~ 00 en *l *, *) *
m~~~~~~~~~~~~~~~~~~ r-3 )0
ElE
o
? D o o
~m m ri oc en m m (- w) (-
-
oo
4- q _) C) C) 0 C) C, C) C oo?4 dt
C1
C"
4 N oc
00
XN
(N
o ^ N F
El.-FmmO?Nt
E OmOtFo
?
111
._
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
26 MYERS ET AL.
=0 ~~~~~~~~~* * ***
Icd -r- co 00 ON r- b a' Ch O - m O
C
ooIt. (1oo. oo
W
o
W
o oo.
-
o e.
a,,
o.
00 q0 0 c0 O 4n 0 ~0(14 0
* * * * *
r- 2c en ell It r- O^ 00 ^ 0 W) m r-
U) * * * * *~ ~~ ~~ ~ C* * * *)
t- C4
- -
~OC4
~,O- ~,O0(1
c
m = oo. oo. c. O. oo. n0 c 0
S * * * ~~~* * * * *
Q
ca
~~* *~- * * -**
. 0 4 c It W) 0N O tn t- 00 t
-
N \0 00 00 t CNo
o e - 1 WI N l ) O - O X- 01 X- 0 d- (-- 0
WS 0. 00. 000 0 00 00 (0 en C( -
ca * * * * ** * * *
\ O t- O O O 0 en O1 O
t C,
O
" tn en.O
s
** *~- *,- *,-
00 0
C) Q 0~~~~~~\0 0 0 tn C1 'It mc <5E r ) 00 * cnX0 C Ss c
cl s d - o O N 0 t- O
-nC, '0t( 000 -0 00 ~ ,O -
-
en (1 --
en C1 -. O
-
O. O. 4 O
-
O O) 'I O C ) C4 W) cn _
000 00 0 000C 0C )c '->0 ) ~0 0
U) ~ ~ ~ ~~ ** * * *
-'
C
10(1 ON 00 - '- t- 00 4 00
n t
I-
_n _0 O
00-
_IN all 00I 0-
C-* * * * *
,~~~~~~W ep W) C,4 en CZ ON C, t- W) W) C bX
tn 0 M n W)00 DN 0 \,
V0)
C
-
co - c "O en o _ n 0 00 X c cn _ ) C1 Xn _0 s m
FW *- *- 1-- * )C- 1- *-1 - 1-- 1--W
(U 0
Uq * * * * * *~~~~~C, *
Z~ * ***
,y co 0 s x o o r
ct
O^ mF ooo F n n XF > n
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 27
2o aA t- cn (1eVc OA cQ 00o ( en t- t- Nt- ,.- 0^
r- .o nC -- V t e v I - O t r- (1 'r- t t- N t- '
I I ~- i --
'.4 ~~~*
* * * * * *Z
C' C4
00tNtNio
Cq ot o o 0 0 o
*-d
C N 't N
i 3t
N
r^ Om
t
1 - N ?Q, - NO 00 -t t
-
a
-
1 t-N O_ ^ _ r-. oo
'0 m as - ' -- t -e I- -r Ir - -- c' I - 00 en co 00- N U
40) ONc)C sa
4t t* * * * * * * * * * *
Q .~'.00(1'.O C) en 0 . O * O*
't - "t *'. * 1 *.
en * * *
4- oc cto on N to tn tn Cf 0 N ~ o as cf) m so o-o o cre
rz
w
CJ~~~~~o qtC 0 ) xC N tD r1=
-
O en qf (N Ch ?
-
O Q 00 W ) m
32 en t- t- o oo O ONen C) -o of ONW o -s4
f
Oo (71 eW C) O N
4) ~ ~ ~ r-c n~or-0 CO 0 '0O N 0C
-
c~ 4 (' N0 0 in cN' cn~ - oW
CO *** * *
0 *
*
* en en I* *) en * W)88 8 0 N e 0e
S~~~~~~~~
8 s o c
-*o
oa 0 C> 0
_
o _n r- C4
4
*_ t' co N N 00 00 t-e 00 ( c' r _ '- N ( 1C 00o n t oo N
5
(1'"it r4 .-I 004 'f-4 --' 0 -o. (- _o 4 00_ oo 0
m
00
00o 00 0
0 00 0
0
0
_ oo 00
C 0 C
CS
- -
C)c to \0 1-0 C4 Er o obo-N 8
8
o 00 vq o ^ o o
E~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ v
CO~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C
. 0
00 . .- oo FcEu 0F ^? ib ^ ? - mX ?t ?b i)X *
U, 0 .O .O .O .O .O .O .O .O .O .O ,O O . - N O
-~ ~ ' 0 - _ I_ _ _ I_ I_ _ I_.
)Ct * * * * CO
,, tf, 00- N0 o~I o 1 - x_N _~ o N 8 _ C (1( _ 1 o(0 o o.O 00 0o___t4 =
. g o0o0.0o.o.O o. o0 .oo.oo .o o .o.-0-No
> u' * * Q * * * *
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
28
MYERS ET AL.
00 00 ~~~~~~ ~~~0 oq o00 0 0 0 0 ~ 0 0NC ~
g?0 (100 NN - - N
oo 00 C o
1
00 cnr f
tn 0*0*0*0*00ON00 ~ -c )C "f ~o c<r
Cm ci- W ci
-
q N NCq W) ci ~ C '-c Cq C ON~ ~c 0000 C cW0 W)W )
r- 00 ~ ~ ~ ~ ~ ~
0N~e
LI ~ ~ ~ ~ ~ ~
00ChW NWC 00n C C 000 Qf 0qW 0 W NCh 0- '-O"f cn q hciN00 WCO Ci14 0W-
NC11 n 0 1
C cN Cq Ch eft 0 ci 00 0 0 Nc Nr.
-
NOC
-
-'~rc cif. '1h.C 0c N- C
00nCqC - 00 -0 -0 c c cq 00 ) c) ci C c0 0 C) 0 c c 00 Ch W) Oft C
I 0 c 4 )-N q 000C C o ci
0
',
W "t cn W)C IC I) Iq C - c 0- 0 c , 0 C m
Cq Cq cn Cq n C) C cn cn 0 C4 C C',40 r- C) h W) C "ft 0
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 29
model, students in general and vocational pro-
grams misbehave more than students in academic
programs, have lower achievement scores and
grades, and are less likely to expect to attend
college.
We now turn to an analysis of changes in the
educational and behavioral outcome measures.2
Living in a single-parent family is generally not
related to changes in misbehavior or achievement
test scores. That is, the negative effects of living in
a single-parent family do not become increasingly
negative between the sophomore and senior years
of high school. However, living in a single-parent
family has negative effects on changes in grades
for white males and black females. This finding
indicates that these students in single-parent
families tend to experience less change in their
grades than students in two-parent families.
Further, white females in single-parent families
misbehave more in their senior year than in their
sophomore year and learn mathematics at a lower
rate than white females in two-parent families.
Mother's employment tends to be related to
small increases in misbehavior and lower academic
performance for white students over the two years.
Increases in mother's educational attainment are
generally related to gains in achievement for white
students but not to changes in grades or misbehav-
ior. Living in a large family tends to be associated
with smaller increases in reading achievement and
grades. More specifically, in three out of four
subsamples of students we find that, net of the
other variables in the model, students with many
siblings experience a relative decline in reading
achievement over the two-year period when
compared to those with few siblings. White males
and black females from large families tend to show
declines in grades between their sophomore and
senior years of high school.
The influence of family income on changes in
misbehavior and academic performance are mixed.
We observe positive effects for misbehavior. This
indicates that students with high family income
tend to increase their misbehavior between the
sophomore and senior years of high school at a
greater rate than those with low family income; this
is particularly true of white males and females.
Family income is positively related to changes only
in math achievement for white females and males.
Income has no effect on grades.
Finally, we note the effects of school program
and educational attainment expectations on changes
in misbehavior and academic performance. Enroll-
ment in a general or vocational curriculum has a
weak effect on changes in misbehavior. In only
two out of eight instances do we find significant
effects. In contrast, curriculum has fairly consis-
tent effects on academic performance. That is,
students in general and vocational programs
generally experience lower rates of change in their
achievement test scores and grades than students in
academic programs. As expected, students with
high educational attainment expectations experi-
ence increases in both reading and math achieve-
ment. However, educational attainment expecta-
tions generally have no impact on changes in
misbehavior or grades.
Relationship Between Misbehavior and
Academic Performance
In this section we review the findings pertaining
to the relationship between the level of academic
performance (or misbehavior) and changes in
misbehavior (or academic performance). When we
compare the results across the four samples, we
find several striking patterns (see Table 4). First,
the level of sophomore misbehavior is negatively
related to changes in academic performance in
eleven out of twelve equations. Students who have
no discipline problems score between .05 and .35
standard deviation higher in reading achievement
than students at the other extreme of the discipline
scale, depending on which race/sex group is
considered. Their scores on the math achievement
tests are between .07 and .64 standard deviation
higher, even when prior achievement is controlled.
Misbehavior in the sophomore year has a some-
what larger effect on senior achievement for black
students than for white students. The senior grades
of students who report no discipline problems are
approximately .20 to .80 standard deviation
higher than the grades of those who report serious
behavior problems.
The effects of reading and math achievement in
the sophomore year on changes in misbehavior are
mixed. For white females, high sophomore math
2
Drawing on the estimated form of equation (2), we
show why the parameter estimates reflect the effects of
exogenous and predetermined variables on changes in the
current endogenous variable. When we subtract Ylki from
each side of equation (2), we get
2ki YIki
-
(k
Y
Iki
Yl ki +
X2jB2k
or
Y2ki -Y ki
=
(a(k 1) Yl ki
+
X2iB2k
+
E2ki-
When we take the partial derivative of the new dependent
variable (Y2ki
-
Ylki) with respect to an element of
X2j,
say,
X2j1
(i.e., 8[Y2ki
-
Ylki/cIX2Ti),
we find that based on
the marginal distribution, the effect of
X2j1
equals
B2kj
(see, for example, Maddala 1984). That is, the effect of
the jth independent variable on the rate of change in the
dependent variable equals its associated parameter
estimate
(B2kj).
A similar argument follows when
difference equations are applied to situations in which a
dummy independent variable is used instead of a
continuous independent variable (e.g., single-parent
status).
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
30 MYERS ET AL.
achievement test scores are related to low rates of
change in misbehavior, but high reading achieve-
ment test scores are associated with increases in
misbehavior. For blacks and white males, achieve-
ment test scores have no effect on changes in
misbehavior. (DiPrete [1981] found the same
relationship using the cross-sectional data.) Grades
have consistent, negative effects on changes in
misbehavior for all students except black males.
For black males, the effect is negative but
insignificant. The estimates show, for example,
that students who had a D average as sophomores
are about .26 standard deviation higher on the
misbehavior scale as seniors than those who had a
B average.
Stability of Students' Misbehavior and Academic
Performance
To assess the stability of misbehavior and
academic performance between the initial and
follow-up surveys, we examined the regression
coefficient for the lagged dependent variable in
each of the senior equations (Table 4). More
specifically, we tested the null hypothesis that the
regression coefficient linking, for example, prior
achievement to current achievement equals one. If
this null hypothesis is rejected, then we can
conclude that there are significant changes between
the prior score and changes in achievement. The
results show a significant change in misbehavior
and academic performance between the sophomore
and senior years of high school for all subsamples.
A more detailed examination of the estimates of
stability indicate that white males had the most
stable misbehavior scores and that black males had
the least stable scores. White and black females
had about equal stability. For reading and math
achievement, we find little variation among
samples in the stability coefficients within tests.
On the other hand, there are considerable differ-
ences in the estimates in the grade equations. Here
we find that white males have the most stable
self-reported grades and that both black males and
females have the least stable grades. White females
are intermediate between these two extremes.
Selection Effects
As a side note, we briefly discuss the effects of
dropping out of school on each of the senior
endogenous variables. Examination of the selec-
tion effects (i.e., the parameter estimates for the
lambda variables in the senior equations) shows
that in general, there are no strong selection effects
in the senior misbehavior equations after the other
independent variables in the equations are con-
trolled. However, in all the achievement and grade
equations we find significant positive effects. A
positive parameter estimate for the sample selec-
tion correction variable shows that as the chances
of staying in school increase, there are, on
average, greater gains in academic performance.
SUMMARY AND CONCLUSIONS
The purpose of this study was to examine the
relationship between misbehavior and academic
performance and the effects of family background
on these two variables. For a sample of students
who remained in high school for two years after
their sophomore year, our analysis indicates that
those who report low grade point averages
experience greater increases in misbehavior be-
tween the base-year survey and the follow-up
survey than those who report high grade point
averages. The results for the achievement tests are
mixed, but white females with low math achieve-
ment test scores tend to experience greater
increases in misbehavior than those with high
achievement test scores.
The strong impact of grades on misbehavior and
the generally insignificant effect of achievement
test scores suggest that it is not failure to learn that
results in misbehavior; rather, failure must be
perceived. Perception of low performance relative
to classmates leads to misbehavior.
When we look at the flip-side of this relation-
ship, we see that misbehavior in the sophomore
year has negative effects on both learning and
grades. Earlier in the paper, we specified two
research hypotheses: (1) misbehavior causes poor
academic performance, and (2) poor academic
performance causes misbehavior. The results
presented here support both of these contentions.
Students from single-parent families and with
mothers who work have greater discipline prob-
lems and lower academic performance than similar
students from two-parent families with mothers
who do not work. Number of parents and mother's
employment affect not only the level of misbehav-
ior and academic performance but also the changes
in these outcomes. That is, students from single-
parent families with mothers who work tend to
experience increases in misbehavior and decreases
in performance. This is particularly true of white
males and females.
Our findings suggest that academic performance
and family situation play an important role in
determining student misbehavior. The results
suggest that failure in school as measured by
grades and math achievement influence later
misbehavior. We also found that students who tend
to misbehave experience declines in school perfor-
mance between their sophomore and senior years
of high school. Finally, misbehavior and academic
performance are influenced by family situation,
thereby lending support to the notion that family
socialization and supervision by parents must be
considered when evaluating policies pertaining to
in-school misbehavior.
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 31
APPENDIX A
AN ILLUSTRATION OF THE PROCEDURE TO
CORRECT FOR SAMPLE SELECTION BIAS
Rather than describe the general procedure for
correcting for sample selection bias, we focus on the
hypothesized specification in the population for student
misbehavior measured at the time of the follow-up
survey:
FYDISCi
=
otYBDISCi
+ X2iB2
+
E2i, (Al)
where X2i, B2, and E2i are the same as in equation (2).
The population regression function for equation (Al) can
be written as
E(FYDISCijX2i)
= (xYBDISCi
+
X2iB2
(i =
1, . . . N), (A2)
where we assume that E(E2i)
= 0. Now, suppose that we
have data on student misbehavior only for those students
who remained in school. Taking this information into
account, we can write equation (Al) as
E(FYDISCi|YBDISCi,
X2i, Di
= 1) = otYBDISCj
+
X2iB2 (A3)
+
E(E2ij Di
= 1),
where Di
= 1 if the student stayed in school and Di
= 0
if the student dropped out. When we compare equations
(A2) and (A3), we see that the expected level of student
misbehavior for the sample of in-school students at the
time of the second survey does not equal the expected
level of misbehavior in the population. One may be
tempted to estimate the misbehavior equation with the
selected sample and make inferences about the relation-
ships for the subsample of selected students using the
set-up in equation (A2). Such conditional inferences may
be faulty. That is, in the selected sample, the variables on
the right-hand side of equation (Al) are likely to be
correlated with
E2i,
thus producing biased parameter
estimates of ot and B2 if OLS is applied.
To derive estimates of ot and B2, we need to take into
account the conditional expectation of the disturbance
term in equation (A3). To accomplish this, we specify a
selection process,
Di=1 if
Zj>
0
Di= 0 otherwise, (A4)
and a selection equation,
Zi
=
X1iB1
+ Uli, (A5)
where X1i is a vector of exogenous variables, B1 is a
comformable vector of parameters to be estimated, Uli is
a random disturbance, and Uli and E2i are assumed to be
bivariate normally distributed. Given equations (A4) and
(A5), we can write equation (A3) as
E(FYDISCi|YBDISCi,
X2i, Z > 0) =
otYBDISCi
+
X2iB2
+
E(E2ijU1i
>
-X1iB1). (A6)
As Heckman (1976) shows,
E(E2ij Uli> -X1iB1)
=
(ao2/2j2)12 A
Xi
=
J(0i)/[ 1-F(0i)], (A7)
and
0i
=
(-XliBi)/(or )1/2,
where
Xi
is the inverse Mills ratio. Consistent estimates
of Bl/(ou1)"/2
and therefore Oi can be obtained by
estimating a probit equation in which Di
is the dependent
variable. Once Oi is calculated, Xi can be computed and
included in equation (A3). OLS can then be used to
estimate the parameters in equation (A3). The standard
errors from the senior equations are corrected to account
for the use of estimated values of Xi
as regressors
(Heckman 1979; Greene 1981). The senior equations
were estimated using the computer program LIMDEP
(Greene n.d.).
Two vectors of exogenous covariates (X1i
and
X2i)
are
referred to in the selection and misbehavior equations.
Ideally, these do not contain identical sets of independent
variables. If they do, we must rely on the nonlinearity of
the probit equation to secure identification. For the
analysis reported here, it is difficult to imagine an
independent variable that does not influence student
misbehavior, for example, but does influence the chances
of a student dropping out of high school. In light of this
problem, we have opted to rely on the nonlinearity of the
probit equation to conduct the analyses.
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
32 MYERS ET AL.
APPENDIX B: Probit Estimates for Selection Equation
White White Black Black
Variables Females Males Females Males
SINGLE -.295** -.380** -.094 -.017
(.068) (.067) (.101) (.104)
MWPT -.145* -.107 .087 .082
(.065) (.063) (.143) (.135)
MWFT -.261** -.188* .143 .082
(.083) (.080) (.146) (.137)
MEDUC .099** .036* .031 .006
(.016) (.016) (.030) (.029)
NUMSIB -.052** -.040** -.060** -.003
(.011) (.010) (.017) (.016)
LNINC .102* .068 .025 -.023
(.043) (.044) (.069) (.070)
EDUCEXP .196** .196** .248** .312**
(.032) (.032) (.065) (.064)
YBDISC -.283** -.261** -.400** -.315**
(.030) (.028) (.071) (.057)
YBREAD -.002 .004 .027 .014
(.007) (.007) (.017) (.016)
YBMATH .021 * * .020** .015 .021 *
(.004) (.004) (.009) (.009)
YBGRADE .302** .456** .146 .272**
(.038) (.038) (.076) (.072)
GEN -.311** -.112 -.127 -.252
(.069) (.073) (.131) (.131)
VOC -.235** - .200* - .152 - .202
(.082) (.082) (.137) (.140)
Constant - 1.438** -.768 .032 .078
(.436) (.453) (.736) (.727)
Log likelihood -1,625.1 -1,603.1 -429.16 -448.06
NOTE: Standard errors are in parentheses.
*
p<.05.
**
p<.01.
REFERENCES
Alexander, K., and E. McDill. 1976. "Selection and
Allocation within Schools: Some Causes and Conse-
quences of Curriculum Placement." American Socio-
logical Review 41:963-80.
Baker, K. 1985. "Research Evidence of a School
Discipline Program." Phi Delta Kappan 66:482-87.
Benson, C., E.A. Medrich, and S. Buckley. 1980. "A
New View of School Efficiency Household Time
Contributions to School Achievement." Pp. 169-203
in School Finance, Policies, and Practices: The 1980s,
A Decade of Conflict, edited by J.W. Guthrie.
Cambridge, MA: Balinger.
Berk, R.A., and S.C. Ray. 1982. "Selection Biases in
Sociological Data." Social Science Research 11:352-98.
Bogue, Donald J. 1985. The Population of the U.S.:
Historical Trends and Future Projections. New York:
Free Press.
Call, D.J. 1965. "Delinquency, Frustration, and Non-
Commitment." Ph.D. diss., University of Oregon.
Cloward, R.A., and L.E. Ohlin. 1960. Delinquency and
Opportunity: A Theory of Delinquent Gangs. New
York: Free Press.
Cohen, Albert K. 1955. Delinquent Boys. Glencoe, IL:
Free Press.
Coleman, James, Ernest Campbell, Carol Hobson, James
McPartland, Alexander Mood, Frederic D. Weinfeld,
and Robert York. 1966. Equality of Educational
Opportunity. Washington, DC: U.S. Government
Printing Office.
Coleman, James, Thomas Hoffer, and Sally Kilgore. 1982.
High School Achievement: Public, Catholic and Other
Private Schools Compared. New York: Basic Books.
Commission on Excellence in Education. 1983. A Nation
At Risk. Washington, DC: U.S. Government Printing
Office.
DiPrete, T. 1981. Discipline and Order in American
High Schools. Report to the National Center for
Education Statistics. Washington, DC: National Center
for Education Statistics.
Elliott, N.S. 1966. "Delinquency, School Attendance,
and Dropout." Social Problems 13:307-14.
Etaugh, C. 1974. "Effects of Maternal Employment on
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions
STUDENT DISCIPLINE AND HIGH SCHOOL PERFORMANCE 33
Children: A Review of Recent Research." Merrill-
Palmer Quarterly 20:71-98.
Gallup, Alec. 1984. "The Gallup Poll of Teachers'
Attitudes Toward the Public Schools." Phi Delta
Kappan 66:97-107.
Gold, M. 1963. Status Forces in Delinquent Boys. Ann
Arbor: University of Michigan Press.
Greene, W.H. 1981. "Sample Selection Bias as a
Specification Error: Comment." Econometrica
49:795-98.
. n.d. User's Guide to LIMDEP. New York: New
York University.
Heckman, J. 1976. "The Common Structure of Statisti-
cal Models of Truncation, Sample Selections and
Limited Dependent Variables and Simple Estimators
for Such Models." Annals of Economic and Social
Measurement 5:475-92.
. 1979. "Sample Selection Bias as a Specification
Error." Econometrica 47:153-61.
Herzog, E., and C. Sudia. 1973. "Children in Fatherless
Families." Pp. 141-232 in Review of Child Develop-
ment Research, edited by B.M. Caldwell and H.N.
Ricciuti. Chicago: University of Chicago Press.
Hetherington, E.M., K.A. Camara, and D.L. Feather-
man. 1981. Cognitive Performance, School Behavior,
and Achievement of Children from One-Parent House-
holds. Report prepared for the Families as Educators
Team, National Institute of Education. Washington,
DC: National Institute of Education.
Heyns, B. 1982. "The Influence of Parents' Work on
Children's School Achievements." Chap. 7 in Fami-
lies That Work: Children in a Changing World, edited
by S.B. Kamennan and C. Hayes. Washington, DC:
National Academy Press.
Hoffman, L.W. 1979. "Matemal Employment: 1979."
American Psychologist 34:859-65.
. 1980 The Effects of Maternal Employment on
the Academic Attitudes and Performance of School-
Aged Children. Report prepared for the Families as
Educators Team, National Institute of Education.
Washington, DC: National Institute of Education.
Jencks, C., M. Smith. H. Acland, M.J. Bane. D. Cohen,
H. Gintis, B. Heyns, and S. Michelson. 1972.
Inequality: A Reassessment of the Effect of Family and
Schooling in America. New York: Harper and Row.
Kelly, J.B., and J.S. Wallerstein. 1979. "Children of
Divorce." The National Elementary Principal 59:57-58.
Maddala, G.S. 1984. Limited-Dependent and Qualitative
Variables in Econometrics. Cambridge: Cambridge
University Press.
Merton, R.K. 1968. Social Theory and Social Structure.
Eni. ed. New York: Free Press.
Milne, A., D. Myers, A. Rosenthal, and A. Ginsburg.
1986. "Single Parents, Working Mothers, and the
Educational Achievement of School Children." Soci-
ology of Education 59:125-39.
National Center for Education Statistics. 1983a. High
School and Beyond 1980 Senior Cohort First Fol-
low-up (1982) Contractor Report. Data File User's
Manual. Washington, DC: U.S. Department of
Education.
. 1983b. High School and Beyond 1980 Sopho-
more Cohort First Follow-Up (1982) Contractor
Report. Data File User's Manual. Washington, DC:
U.S. Department of Education.
Pink, W.T. 1979. An Exploration of Relationships
Between Academic Failure, Student Social, Conflict,
and School Related Attitudes. Report prepared for the
National Institute of Education. Washington, DC:
National Institute of Education.
Polk, Kenneth. 1967. "Class, Strain, and Rebellion
Among Adolescents." Social Problems 14:320-25.
Polk, Kenneth, and David S. Halferty. 1966. "Adoles-
cence, Commitment, and Delinquency." Journal of
Research in Crime and Delinquency 3:82-96.
Purkey, S.C., and M.S. Smith. 1983. "Effective
Schools: A Review." Elementary School Journal
83:427-50.
Rosenthal, A.S., D. Myers, F. Ellman, and A. Milne.
"The Failure of Student-Parent Crossvalidation in the
High School and Beyond Survey." Pp. 385-89 in
Proceedings of the Social Statistics Section, American
Statistical Association. Washington, DC: American
Statistical Association.
Schafer, W.E., and C. Olexa. 1971. Tracking and
Opportunity: The Locking-Out Process and Beyond.
Scranton, PA: Chandler.
Shinn, M. 1978. "Father Absence and Children's
Cognitive Development." Psychological Bulletin
85:295-324.
Stinchcombe, A.L. 1964. Rebellion in a High School.
Chicago: Quadrangle Books.
U.S. Bureau of the Census. 1982. "Household and
Family Characteristics: March 1981." Current Popu-
lation Reports, ser. P-20, no. 371. Washington, DC:
U.S. Government Printing Office.
U.S. Bureau of Labor Statistics. 1983. Handbook of
Labor Statistics. Washington, DC: U.S. Government
Printing Office.
Wise, L., and D. McLaughlin. 1980. Guidebook for
Imputation of Missing Data. Technical Report No. 17.
Palo Alto: American Institutes for Research.
This content downloaded from 202.57.58.197 on Wed, 19 Mar 2014 23:51:33 PM
All use subject to JSTOR Terms and Conditions

Vous aimerez peut-être aussi