American Educational Research Journal

Summer 1992, Vol. 29, No. 2, pp. 405-424

The Influence of Family Composition

on Children's Conformity to the Student Role
Maxine Seaborn Thompson
North Carolina State University
Doris R. Entwisle, Karl L. Alexander, and M. Jane Sundius
Johns Hopkins
University
This article examines how parent configuration (two-parent, motherextended, or solo-mother) and number of siblings affect first graders' conformity to the student role as measured by their absences, latenesses, and
conduct marks. The article builds on earlier studies by examining the
different living arrangements of single parents and by directly comparing compliance indicators for African-American and White children when
socioeconomic
status, parents'
expectations,
mother's age, and
kindergarten attendance are taken into account for both. AfricanAmerican children with more sibs are more often absent or late. Both
White and African-American children with sibs get better marks in conduct at the beginning of first grade. Also, both White and AfricanAmerican children from two-parent families improve their conduct over
first grade more than do their counterparts from solo-mother homes.
However, only African Americans from mother-extended homes, not
White children from mother-extended homes, improve in conduct over
the year. Since two-parent homes generally contain more sibs, sib effects
in prior research may have inflated positive effects on schooling attributed
to two-parent family
arrangements.

MAXINE SEABORN THOMPSON is an associate professor in the Department of

Sociology at North Carolina State University, Box 8107, Raleigh, NC 27695-8107.
She specializes in social psychology and medical sociology
DORIS R. ENTWISLE is a professor in the Department of Sociology at Johns
Hopkins University, Mergenthaler Hall, Charles and 34th Sts., Baltimore, MD 21218.
She specializes in sociology of education, social psychology, and methodology.
KARL L. ALEXANDER, also a professor in the Department of Sociology at Johns
Hopkins, specializes in sociology of education and school effects and effectiveness.

M. JANE SUNDIUS is a doctoral student in the Department of Sociology at Johns

Hopkins.

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Thompson, Entwisle, Alexander, and

Sundius

ven before children enter first grade, the family exercises a major influence on their adaptation to school. For example, some parents
describe school as a place where children should follow rules, obey the
teacher, sit still, and the like, while others portray the school as a more
responsive environment where students can exercise initiative and affect
their own progress (Hess & Shipman, 1965). There can be little doubt that
much of the preparation for school takes place in the home, yet little is
known about how families of primary grade youngsters help or hinder their
children in meeting the behavioral demands of the student role (Entwisle
& Alexander, 1988). This article focuses on how the composition of the
family may affect first-grade children's conformity to the student role.
Most research on how parents affect the schooling of very young
students focuses on cognitive performance, either marks or standardized
test scores (Alexander & Entwisle, 1988). In the early primary grades, though,
compliance with school routines may be a critically important barometer
of successful or not-so-successful school adjustment, and this domain is one
over which parents exercise considerable influence. Young children are
dependent on their parents for getting them to school and getting them
there on time. Also, primary school children respond strongly to their
parents' expectations for their school conduct (Entwisle & Hayduk, 1978),
and their conduct in turn affects their performance in reading and mathematics (Entwisle & Hayduk, 1982). It can matter a great deal for children's
long-term academic prospects whether they attend school regularly and are
well-behaved because achievement patterns established early in children's
school careers tend to persist. Unfortunately, the nonacademic sources of
early academic success have prompted relatively little research.
Differences in family composition are known to affect attendance patterns and deportment of older students. In a comprehensive review of
research on one-parent households, Hetherington, Camara, and Featherman (1983) found that children from single-mother households had higher
rates of absenteeism, tardiness, and truancy and had more disruptive
classroom behaviors, problem referrals, and suspensions than did children
from two-parent homes (see also Garfinkel & McLanahan, 1986). Along
similar lines, teachers' assessments of children's social skills and personalities
are more negative for those from divorced homes compared to those from
intact homes (Santrock & Tracy, 1978). Teachers also rate children from
one-parent homes more negatively on homework completion (Epstein,
1984) and as less effective in study skills (Hetherington, Camara & Featherman, 1983) than children from two-parent homes. Hence, the general picture suggests that single parents are not as effective as those in two-parent
homes in helping their youngsters comply with the student role.
Research on how siblings affect students' school performance shows
that school achievement of students from large families is slightly but consistently lower than that of children from small families (Blake, 1989; Mer406

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

cy & Steelman, 1982). The greatest effect of sib-size is on high school graduation (Blake, 1989). In terms of proportions graduating from high school,
there is a 2 3-percentage-point difference between only children and those
coming from families of seven or more. To our knowledge, however, the
way family sib patterns intersect with parent configuration in affecting
school performance is virtually unexplored. Some of the negative effects
attributed to single parenting could reflect instead effects of sibship patterns. For one thing, only children are found more often in families of
divorced or separated parents than in two-parent families (Blake, 1981;
1989), and there are reports that only children have more difficulty getting
along with others (Falbo, 1982). Also, having many children in a family might
make it more (or less) difficult for parents to get first graders to school on
time or for children to settle into school routines. So sibs could have effects on first-graders' compliance with the student role.
The information presently available on how family type affects children's schooling, based as it is on data from the 60s and 70s, neglects what
could be several other key issues. One is the living arrangements of singleparent families. Single mothers, especially if they are African American, often
live in extended families containing other adults. Nevertheless, with two
exceptions, neither of which compares African Americans and Whites
(Kellam, Ensminger, & Turner, 1977; Furstenberg, 1976), single parents are
treated in the literature as one family type, irrespective of whether they
live alone or in extended family arrangements. This categorization could
be misleading, though, as single parents who live in extended-family arrangements may have more or different resources to draw upon in preparing children for school than do single parents who live alone (McLanahan,
1985).
In 1989, about 38% of African-American children were living with two
parents, compared to 80% of White children (U.S. Department of Commerce, 1989). Among single mothers, about the same percentage of African
Americans (41 %) as Whites (37%) lived in households that contained adults
besides the mother (U.S. Bureau of the Census, 1988), but since 5 1 % of
the African-American families are single-mother compared to 16% of White
families, the likelihood that African-American families involve extended living arrangements without a father is much greater.
Still another factor neglected in earlier investigations of how family patterns affect school performance is mother's age at the child's birth. Single
mothers tend to be younger than married mothers, and children of adolescent mothers do not do as well in school as children of older mothers (see
Hofferth, 1987, for a review).
To examine effects of family composition on children's school performance requires that we take account not only of ethnicity and mother's
age but also of the family's socioeconomic resources (Garfinkel &
McLanahan, 1986). We will also take into account the parent's performance
407

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Thompson, Entwisle, Alexander, and Sundius

expectations for the child, because of recent research showing family type
differences in parents' expectations (Thompson, Alexander, & Entwisle,
1988). Last, we need to consider the child's gender and kindergarten experience because girls are rated higher in school deportment than boys (Maccoby, 1990) and the amount of kindergarten experience children have had
also affects their performance in first grade (Entwisle, Alexander, Cadigan,
& Pallas, 1987).

The Model
The considerations reviewed above are joined in the multivariate model
shown in Figure 1, which will guide our analysis. The model is designed
to evaluate how family arrangements affect first graders' compliance with
the student role. The compliance measures are absences, tardiness, and conduct marks in the fall and spring of the year. The test variables distinguished
here include three parent configurations: mother-father, mother-extended,
and mother-alone (those who do not share households with other adults).
A fourth test variable is the number of siblings of the study child.
Control variables, known from prior research to predict youngsters'
compliance with the student role, include the following: (a) the child's sex;
(b) the family's socioeconomic status, measured here by mother's education and family economic standing; (c) the parent's performance expectations, measured here by parent's estimate of the child's school ability and

Compliance with student role

(Absences, Latenesses, Fall
conduct mark, Spring conduct mark)

Family type
(mother-alone,
mother - extended,
mother-father)
Number of sibs

OUTCOME VARIABLES
TEST VARIABLES

Porent's education
Family economic standing
Parent's expectation for marks
in reading, math, conduct
Porent's ability estimate .
School dummy variables (19)
Mother's age at child's birth
Kindergarten attendance

CONTROL VARIABLES

Checking runs also included fall CAT scores in

verbal comprehension and m a t h concepts/reasoning plus
fall marks in reading and math.

Figure 1. Family configuration and first graders' compliance with the

student role
408

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

parent's expectations for the child's future marks in reading, math, and conduct; and (d) the child's amount of kindergarten experience. Mother's age
at the child's birth is included because it is thought to be inversely related
to the child's level of school achievement and it varies by family type.
Two other factors, school differences and the child's academic achievement level, need to be discussed. The children in this sample attended 20
different schools, and family patterns varied across schools. The model,
therefore, is estimated using a set of "school dummies" to adjust for possible school differences in the outcomes. Schools, for example, could record
latenesses differently or have different standards for deportment. We would
not want such school differences to be confused with family effects. In addition, youngsters who are doing well academically may be less disruptive
or more favorably judged in conduct (halo effects), or may even be absent
less or tardy less than other children. Although achievement test scores and
marks in reading and arithmetic are not included in Figure 1, these variables
were inserted as an additional check for halo effects at the end of every
analysis, as discussed below.
Setting and Data
The Sample
The data for the analysis come from the Beginning School Study (BSS), initiated in the fall of 1982 in 20 Baltimore City elementary schools. A stratified
random sampling of schools in the City system was drawn to yield a sample about equally divided between White and African-American children
and representative of all socioeconomic levels in the school system. Kindergarten rosters for 1981-82 in these 20 schools served as initial sampling
lists. These lists were supplemented by rosters of new registrants in the fall.
Both rosters were then used to draw random samples of children from each
first-grade classroom in September of 1982. Less than 3 % of the children
thus selected were excluded because of parent refusals.
After giving permission for their first-grade child to participate in the
research, a high percentage (97%) of parents (usually the mother) also agreed
to be interviewed. Analyses presented here are based on a core sample of
484 first graders, those who remained in the study for the entire year and
for whom there are complete data on all the relevant variables. Table A-l
in the Appendix shows that for both African Americans and Whites, the
children with complete data have slightly better marks and slightly higher
parental expectations than those in the full sample, but generally differences
between the full sample and the sample analyzed in this article are very small.
Procedure
Parents were questioned in the summer or early in the fall of the year in
which their children began first grade. These interviews provided data on
household composition, kindergarten attendance, family socioeconomic
409

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Thompson, Entwisle, Alexander, and

Sundius

status, and parents' performance expectations. Information on mother's age

was secured in the following year. School marks, number of absences, tardy
days, and CAT (California Achievement Test) scores were obtained from
school records after the end of the first-grade year.
Control Variables
Ethnicity. Ethnicity was coded 0 for White and 1 for African American.
Sex. Sex was coded 0 for boys and 1 for girls.
Kindergarten attendance. Parents were asked, "Did your children attend kindergarten?" Answers were coded as (4) one full year with full-day
sessions, (3) one full year with half-day sessions, (2) one half year or less
with half-day sessions, or (1) not at all.
Parents' educational attainment. This is measured by number of
school years completed, as reported directly by parents (89% mothers).
Family economic standing. This is measured by whether or not the
child is eligible for subsidized meals at school (breakfast and/or lunch). According to guidelines used in the summer of 1984, a child in a family of
four with a yearly income of $12,870 would be eligible for full meal subsidy (breakfast and lunch), while one with an income of $18,315 would
be eligible for partial subsidy. In the coding used here, 1 represents no subsidy and 0 represents any subsidy.
Parents' general ability estimate. This variable is coded from a single
question that asked the parent: "How do you think your child compares
with other children in his/her school in terms of ability to do school work?"
The five response options ranged from "among the best" (5) to "among
the poorest" (1).
Parents' performance expectations. Parents responded to three questions asking their "best guess" for their child's first report card marks in
reading, math, and conduct in first grade. The response options corresponded
to the marking system employed on report cards (from 4 for "excellent"
to 1 for "unsatisfactory" in reading or math, and from 2 for "satisfactory"
to 1 for "needs improvement" in conduct).
Mother's age. This is the mother's age in years at the time the study
child was born.
California Achievement Test scores. In fall and spring of first grade,
systemwide testing provided California Achievement Test (CAT) scores (Level
11, Form C) in reading comprehension and mathematics concepts/reasoning.
Marks. Children's marks in reading and math were obtained from
school records. These are the following: E (excellent), G (good), S (satisfactory), or U (unsatisfactory), coded from 4 for "excellent" to 1 for
"unsatisfactory."
Test Variables
Parent configuration. This information was obtained directly from the
parent. Family living arrangements are the following: (1) mother-alone
410

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

households; (2) mother-extended households, in which mothers lived in

households with adults other than the child's father or stepfather; about
40% lived with grandparents, the remainder with other relatives; and
(3) mother-father households that could include step-parents (estimated to
be 6% in Year 2) or other relatives. Two dummy variables were constructed
for mother-extended and mother-father households, with the mother-alone
household as the baseline. A small number of mother-absent households
was identified (about 2%). This category is too small to sustain analysis and
was eliminated.
Number of siblings. This is the number of the child's brothers and
sisters as reported by the parent, when asked, "How many older brothers,
older sisters, younger brothers, and younger sisters does the child have?"
It could include half- or step-sibs as well as full sibs.
Dependent

Variables

Absences. The total number of absences for the school year was taken
from the end-of-year report card.
Latenesses. The total number of days the child was late for school was
taken from the end-of-year report card. Because the number of latenesses
was markedly skewed, latenesses were recorded as 0 if below the median
or 1 if above the median for each ethnic group.
Conduct marks. First-quarter and fourth-quarter marks in conduct were
obtained from school records. These are coded as 2 for "satisfactory" or
1 for "needs improvement."
Analysis
Table 1 shows that about three-quarters of the White children in this sample and somewhat under half of the African-American children are in motherfather homes. About a third of African Americans and one-tenth of Whites
live in mother-extended arrangements, while about one-quarter of African
Americans and one-sixth of Whites live with mothers only. Mothers who
live in extended arrangements are the youngest (African American, 20.1
years; White, 22.2 years). African Americans who live in mother-father families are the oldest of any subgroup (26.0 years).
The number of siblings is equivalent overall for the two ethnic groups,
but varies across parent configurations. Children in mother-father homes
have the most siblings, while children in mother-extended homes have the
fewest (one or less on average for both races).
The model in Figure 1 was used to examine family relationships. This
model estimated effects of family variables (parent configurations and
number of sibs) net of the child's sex, family socioeconomic status, parental performance expectations, kindergarten attendance, and mother's age.
To control for school differences, all analyses also included sets of dummy
variables (19 for the pooled sample; 13 for the African-American or White
411

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Ik

to

Table 1
Means and Standard Deviations for Variables by Family Type and Race8
Mother only b
Variable

Sex
Parent ability estimate
Parent expectationsreading
Parent expectationsmath
Parent expectationsconduct
Parent educationyears
Economic standing
Amount of kindergarten
Mother's age at child's birth
Number of siblings
Special problems
Number of absences
Number of times tardy
Number of parent conferences
Fall conduct mark
Spring conduct mark
Fall testreading comprehension
Fall testmath concepts
Fall markreading
Fall markmath

.49
3.8
2.7
2.6
1.9
11.6
.12
3.1
23.2
1.4
.15
15.5
10.5
1.2
1.64
1.63
284
286
1.6
2.0

SD
African
.50
.85
.65
.72
.33
2.1
.33
.41
5.2
1.3
.36
12.8
13.6
1.3
.48
.49
38.6
30.3
.72
.91

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Mother extended c
M
Americans
.55
3.9
2.7
2.8
1.9
11.8
.05
3.1
20.1
.96
.21
13.1
9.1
.74
1.63
1.74
276
289
1.8
2.0

Mother & Father d

Total e

SD

SD

SD

.50
.84
.65
.64
.28
1.9
.22
.47
4.5
.92
.41
10.8
12.2
1.1
.49
.44
35.7
26.8
.65
.81

.53
3.8
2.9
2.8
1.9
12.4
.38
3.4
26.0
1.8
.18
12.1
6.1
.45
1.76
1.82
290
295
1.9
2.3

.50
.86
.78
.66
.28
2.6
.49
.56
6.0
1.7
.39
13.3
12.8
.89
.43
.38
42.1
28.8
.76
.86

.53
3.8
2.8
2.8
1.9
12.0
.22
3.2
23.5
1.5
.18
13.2
8.0
.72
1.69
1.75
284
291
1.8
2.2

.50
.85
.72
.67
.29
2.3
.41
.52
5.9
1.5
.39
12.5
12.9
1.1
.46
.43
39.7
28.7
.72
.87

i
Si

S:

Whites

Sex
Parent ability estimate
Parent expectationsreading
Parent expectationsmath
Parent expectationsconduct
Parent educationyears
Economic standing
Amount of kindergarten
Mother's age at child's birth
Number of siblings
Special Problems
Number of absences
Number of times tardy
Number of parent conferences
Fall conduct mark
Spring conduct mark
Fall testreading comprehension
Fall testmath concepts
Fall markreading
Fall markmath

.46
3.2
2.3
2.5
1.8
10.4
.24
3.1*
25.1*
1.6
.27
13.9
3.8
.24
1.70
1.70
274
289
1.8
2.3

.51
.93
.77
.61
.36
2.7
.44
.63
6.3
1.4
.45
10.8
5.9
.79
.47
.47
41.1
40.4
.58
.98

.50
3.4
2.5
2.8
1.9
10.8
.27
2.9*
22.2*
1.0
.23
14.1
5.1
.62
1.85
1.73
274
289
2.1
2.5

.51
.64
.65
.65
.33
2.2
.45
.46
3.7
1.3
.43
10.2
8.2
.80
.37
.45
31.3
27.4
.63
.99

.52
3.6
2.8
2.7
1.9
12.2
.58
3.3*
24.6*
1.5
.22
12.0
2.0
.33
1.88
1.93
283
299
2.0
2.4

.50
.77
.77
.79
.34
2.9
.50
.50
5.1
1.2
.42
9.5
4.3
.70
.32
.26
42.3
3.8
.67
.78

.50
3.5
2.7
2.7
1.9
11.8
.50
3.2*
24.4*
1.5
.23
12.5
2.6
.35
1.85
1.87
281
297
2.0
2.4

.50
.79
.78
.75
.34
2.9
.50
.52
5.2
1.3
.42
9.7
5.2
.73
.36
.33
41.0
34.3
.66
.83

Means reported here are for listwise present cases. bFor mother only, African Americans = 59 (23%) and Whites = 33 (15%). cFor mother extended,
African American = 80 (31%) and Whites = 26 (11%). d For mother & father, African Americans = 117 (46%) and Whites = 169 (74%). e For total,
African Americans = 256 and Whites = 228.
*Means for mother's age and amount of kindergarten are based on a subsample (N = 176) of the total White sample (N = 228).
*v

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Sa

^e
^
J
^
.
3s

pv

<|L
>t

Thompson, Entwisle, Alexander, and

Sundius

subsamples). All analyses were first run without mother's age and kindergarten attendance, because there was a high level of missing data on these
two variables in the White single-parent subsample and this resulted in excessive sample attrition. Including or excluding these two variables did not
lead to changes in patterns of significance as reported below. The sample
sizes reported in headings of Table 1, therefore, are for Whites when age
and kindergarten data may be missing.
Absences
Average rates of absenteeism are close for Whites and African Americans
(12.5 days and 13.2 days, respectively). In both groups, children from fatherpresent homes are absent least (about 12 days), but variation in absences
by parent configuration is minimal.
In an OLS (Ordinary Least Squares) regression with number of absences
as the dependent variable, only modest amounts of variance are explained
although twice as much variance is explained for African Americans as for
Whites (Table 2). None of the parent configuration effects is significant for
African Americans or Whites, but African-American children with more sibs
are absent significantly more often (metric coefficient of 1.28, p = .03 for
African Americans). The number of sibs has a positive but not significant
effect for Whites. 2
Lateness
Latenesses are problematic because some schools start earlier in the morning than others, and record-keeping is not as careful for lateness as it is for
absences. Keeping these limitations in mind, we note that Table 1 shows
African-American students are tardy on average over 8 times in first grade
and white students are tardy about 2.6 times. This difference between races
is smaller for the students who attend integrated schools (5.8 for African
Americans; 2.7 for Whites). These distributions, however, are highly skewed.
More than 50% of first graders are late less than twice in the entire year,
while a handful are late 40 times or more.
Because of the heterogeneity in variance, separate logistic regressions
were run to explain lateness measured as a dichotomous outcome (below
or above the median) for each ethnic group (Table 3). These regressions
show that African-American children from father-present homes are only
25% as likely to be often late as those from solo-mother homes, an effect
significant beyond the .001 level.3 White children in two-parent families
are about 10% more likely than those living with solo-mothers to be late,
a nonsignificant effect.4
Number of sibs makes a difference for lateness. African-American
children with fewer sibs are significantly (p < .05) more likely to be on time
than their counterparts with more sibs. Whites with fewer sibs are also
somewhat more likely (not significantly) to be on time. The effect of number
of sibs on lateness does not differ by race. 5
414

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

Table 2
OLS Regression for Absences of African-American and
White Children Over First Grade
(Standardized Coefficients in Parentheses)

Variables
Test variables
Mother-extended 3
Mother-father 3
Number of sibs
Context variables 5
Sex
Parent education
Poverty status
Mother's age
Kindergarten
attendance
R2
R2 (adj.)

Whites
(TV = 206)

African Americans
(TV = 282)

.97
(.03)
.75
(-03)
.63
(.08)

-1.8
(-.06)
-1.2
(-.05)
1.3 **
(.15)

2.8 **
(-15)
-.04
(-.01)
-1.2
(-.06)
-.24 *
(-13)
-1.9
(-11)
.176
.061

.96
(.04)
-.65 *
(-.12)
-1.8
(-06)
-.15
(-.07)
- 6 . 8 **
(-29)
.214
.137

Mother alone is baseline. b Other control variables also included are parent ability estimate,
parent performance expectations for reading, mathematics, and conduct, plus 13 school dummies (6 segregated and 8 integrated schools for each race).
*Significant at or beyond the 10% level. **Significant at or beyond the 5% level.

Conduct
Children's conduct marks in first grade tap still another facet of compliance
with the student role. The strong relationship between gender and deportment, which is commonly found (see Entwisle & Hayduk, 1982), is visible
as soon as children start school (see Table 4). Conduct marks here span
a narrow range but, even so, in both fall and spring of first grade, girls'
marks exceed boys' by about half a standard deviation on the girls' measure.
Also, conduct marks in both fall and spring are higher for Whites than for
African Americans (1.85 vs. 1.69 for fall, and 1.87 vs. 1.75 for spring). This
difference by race is significant in fall (p < .05 in a logistic regression for
the pooled sample) and approaches significance (p = .11) in the spring.
For African Americans and Whites, the fall conduct marks for children
from father-present homes are generally higher than those for children from
415

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

I*

ON

Table 3
Logistic Regressions for Latenesses and Conduct Marks8
Lateness
White
coefficient
(TV = 206)

Variables
Test variables
Mother-extended 5
Mother-father 5
Number of sibs
Context variables 0
Sex
Parent education
Poverty status
Mother's age
Kindergarten
attendance

.28
.04
.08
.09
-.06
-.33
-.04**
-.01

Fall conduct mark

Afr.-Am.
coefficient
(TV = 282)

White
coefficient
(TV = 206)

Afr.-Am.
coefficient
(TV = 283)

Spring conduct mark

White
coefficient
(TV = 206)

Afr.-Am.
coefficient
(TV = 277)

-.03
-.69**
.15**

-.68
.33
.39**

.02
.20
.12*

-.04
1.16**
-.04

.64**
.67**
-.04

.15
-.06*
.10
-.02
-.03

.31
-.17**
.12
-.03
-.24

.68**
.00
-.04
-.01
.06

.98**
-.02
.32
-.02
.03

.24
-.03
-.11
.01
.07

These analyses were performed using the PROBIT procedure from SPSSX Version 3.1. b Mother alone is baseline. c Other control variables also included are parent ability estimate, parent performance expectations for reading, mathematics, and conduct, plus 13 school dummies (6 segregated
and 8 integrated schools for each race). For the regressions predicting spring conduct mark, fall conduct mark was also included as a control variable.
* Significant at or beyond the 10% level. ** Significant at or beyond the 5% level.

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

JS^
!

s
IS-

s.>i

Si
3
S^

3
5^

s*
Co

Family Composition and Student

Conformity

Table 4
Fall and Spring Conduct Marks by Sex and Race
African Americans
Sex

SD

Boys
Girls

1.59
1.82

Boys
Girls

1.68
1.82

Whites
N

Fall conduct mark

.493
152
1.77
.388
158
1.91
Spring conduct mark
.468
152
1.79
.383
158
1.94

SD

.418
.290

125
131

.408
.240

125
131

either solo-mother or mother-extended families, but not significantly so (see

Table 3). However, logistic regressions show that children with more sibs
in both ethnic groups tend to get higher fall conduct grades than do children
with fewer sibs (p < .05 for whites;/? < .10 for African Americans). For example, African Americans with no sibs get 1.62 in conduct in the fall, compared to 1.72 for children with sibs; Whites with no sibs get 1.78 compared to 1.89 for those with sibs. For both ethnic groups, the low conduct
marks of the only children explain much of the difference associated with
number of sibs.
Improvement in conduct marks over first grade, which would be an
indication of adjustment to school, is assessed by logistic regressions using
the model in Figure 1 plus the fall conduct mark to predict the spring conduct mark. Table 3 shows that conduct marks of children who come from
father-present homes improve significantly over the year: African-American
children's marks increase so that 82% receive a grade of satisfactory in spring
compared to 76% in fall; White children's marks also improve in that 9 3 %
receive a satisfactory grade compared to 88%. The regression shows that
these changes are significant when evaluated against changes in marks of
children from solo-mother homes (p < .05 for both ethnic groups). AfricanAmerican children in mother-extended families also improve significantly
over the year (74% vs. 6 3 % satisfactory marks, p < .05) compared to their
counterparts in solo-mother homes, while marks of African-American
children from solo-mother homes decrease a little (64% vs. 6 3 % satisfactory). By contrast, White children in mother-extended homes, who are
relatively few in number, show a decline (nonsignificant), and marks of
White children from solo-mother homes remain the same from fall to spring
(70% satisfactory both times). Number of sibs does not have a significant
effect on changes in children's marks over the year, although children with
no sibs did improve somewhat (1.73 for African Americans; 1.86 for Whites).
417

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Thompson, Entwisle, Alexander, and

Sundius

Discussion
This work builds on prior research on single parenting in several ways. Most
importantly, it focuses on effects for young children, at the point of school
entry, who were growing up in the 80s. Most of what is presently known
about effects of single parenting is based on findings for older children who
grew up in the 60s and 70s (see Garfinkel & McLanahan, 1986). Our research
also builds on prior work by taking parents' expectations and ability
estimates into account. Earlier studies contrast effects for African Americans
and Whites, controlling socioeconomic indicators but not parents' attitudes
about their children. Finally, by classifying single mothers according to their
different living arrangements and by adding sib size to the analysis, we offer
a more analytic approach to the specific family factors that may burden
children of single mothers in their early schooling. In particular, systematic
differences in family size and in mother's age between one- and two-parent
families have previously been overlooked.
In designing this analysis, we intended to study single parents who live
in extended arrangements separately from those who live alone because
of suggestions that another adult in the home may represent "another pair
of hands." That is, two adults may be able to help children more than just
one adult, irrespective of whether or not the second adult is the child's
parent (McLanahan, 1985). In this article, the main comparison showing
an advantage for children living in mother-extended homes compared to
those in mother-only homes is seen for African-American children's improvement in conduct over first grade. Our sample has relatively few White
mothers living in extended arrangements and so is not ideal for testing differences between White and African-American mother-extended families,
but these extended arrangements may have quite different implications for
the two groups. Sharing living quarters with kin is characteristic of AfricanAmerican families (Staples & Mirande, 1980), and it may help in managing
scarce economic resources. The extended family form in the AfricanAmerican community thus represents a kind of social support system that
is available to African-American single mothers, but that is not as well institutionalized among Whites. Other research also documents the benefits
for African-American mothers in extended family arrangements. These
mothers are better off than those who live alone in terms of their own
psychological well-being (Stack, 1974), and their children are better adjusted
to school (Furstenberg, 1976; Kellam et al., 1977). There is not much
research about school outcomes for children of White mothers who live
in extended arrangements as compared to their counterparts who live with
mothers alone.
In many ways, the African-American and White single parents in our
sample look similar. In both groups, mothers who live alone have more
children than those who live in extended arrangements. In both groups,
those who live alone are about 3 years older than those who live in ex418

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

tended families. Additionally, most mothers, in both groups, who live in

extended arrangements reported that they shared a residence with "other
relatives." For 40% of African Americans and 30% of Whites, these "other
relatives" include a grandmother.
The main difference between racial groups is that African-American
mothers in extended families are about 2 years younger that their White
counterparts; 58% of African-American mothers in extended families were
adolescent when the study child was born compared to 36% of the White
mothers. In light of the higher percentage of African-American adolescent
mothers living in extended arrangements and their younger ages at birth
of the study child (the youngest was 13), it seems likely that many of these
mothers continued to live where they were after they became pregnant
they never moved out of their family of origin. The White parents' ages
are enough higher (the youngest was 18 at the time the study child was
born) so that probably fewer of those in extended arrangements are there
because they remained in their family of origin. Teen parenting was not
uncommon among Whites in the late 70s when our study children were
born but still was considerably less common for Whites than for African
Americans. Among adolescents, age 15 through 19, who became mothers
in 1975-1979, the same 5-year time period in which our study children
where born, about 32% of Whites conceived and gave birth out of wedlock
compared to 80% of African Americans (O'Connell & Rogers, 1980). In light
of these differences, using age as a further control in the analysis strengthens
conclusions about ethnic differences.
Another goal of this research was to consider number of sibs along
with parent configuration. In this sample, which is randomly drawn from
a large urban school population, single mothers who live alone had more
children by the time the study child entered school than did single mothers
who live in extended arrangements, irrespective of race. Prior research on
family size shows consistently that children with more sibs do not achieve
as well in school as those with fewer sibs, but most of this research is
oriented toward larger families (Mercy & Steelman, 1982). We see both disadvantages (more absence and lateness for African-American children) and advantages (better conduct for White children) for those children with more
sibs. To our knowledge, positive effects of sibs on school performance have
not previously been reported, but should be interpreted in light of the fact
that the children in this sample had only one or two sibs on average. In
prior research, two-parent effects could have reflected sib effects. That is,
two-parent families have more children than single-parent families; so some
part of the negative effect of single parenting previously reported might
be explained by the fact that single-parent families have just one child more
frequently than other families (Blake, 1989).
Why would only children have difficulty settling into school? Blake
(1989) notes that single children not only have a high probability of com419

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Thompson, Entwisle, Alexander, and

Sundius

ing from broken families but are more likely to have had health problems
as infants. There are also "socialization" problems. Compared to other
children, only children belong to fewer organizations, report having fewer
friends, visit friends and relatives less often, and have a less active social
life (Falbo, 1982). In addition, only children probably receive more help
and nurturance from their parents than children with sibs (Sutton-Smith
& Rosenberg, 1970). Compared to children with sibs, then, it may be hard
for "onlies" to adjust to a classroom group, to share the teacher's attention
with other children, and to share communal resources like playthings and
equipment.
Better conduct in first grade could have considerable impact on
children's long-term prospects for doing well in school because children's
first marks are strong predictors of later marks (Alexander & Entwisle, 1988),
and in first grade deportment predicts academic marks. In fact, effects of
conduct in first grade on academic performance are impressive. In other
analyses of the BSS data, which include the control variables in Figure 1,
children's first conduct marks predicted their first-quarter reading marks and
improvement in their reading marks over first grade (Entwisle, Alexander,
Cadigan, & Pallas, 1987). For African-American boys, first-quarter conduct
marks also predicted first-quarter math marks. Similar findings emerged in
earlier work with another large sample of Baltimore children (Entwisle &
Hayduk, 1982) in which conduct marks predicted arithmetic marks at the
end of both the first and second semesters of first grade, net of a series
of controls that included the child's tested cognitive ability.
Our two most provocative findings related to family configuration are
(a) children in both ethnic groups who have more sibs are judged by teachers
to be better behaved when they start first grade, and (b) African-American
children from either mother-extended or father-present homes improve
more in conduct over first grade than their classmates from mother-only
homes, and for whites those from father-present homes improve. The larger
implications of our study could be substantial both for designing research
in the future and for interpreting research done in the past. The positive
effects of two-parent families may have been overestimated in previous
research in which sib effects are neglected.
Other Findings
We summarize briefly three further aspects of adjustment to the student
role that did not uncover differences related to family composition.
The first of these investigations centers on the number of special problems recorded in children's school records. These problems could be of
many types: referrals to psychologists, speech therapists, social workers,
reading resource specialists, vision or hearing specialists, physical therapists,
attendance workers, or to one of three school committees concerned with
classroom behavior problems and retention. The incidence of any of these
420

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

problems is low, so the measure used here is referral status (scored 1 if any
of the problems occurred, and 0 if otherwise). Referral status (Table 1) used as a dependent variable in logistic regressions revealed no significant
effects of any family composition variable.
The second topic concerns parent conferences. Generally, teachers request conferences because children are having some kind of difficulty and
teachers request conferences much more often for first-grade children than
for older children. Unfortunately, although all schools in this sample were
part of a single school system and ostensibly used the same guidelines to
initiate conferences, the number of conferences per child varied so widely
from school to school as to preclude any overall analysis. Still, conferences
for African-American children were most frequent for those from motheronly homes, whereas for Whites the highest frequency was for those from
mother-extended homes (see Table 1).
Finally, parental expectations were investigated. Prior research shows
that single parents who live along have lower expectations for their
children's performance in first-grade reading and math than other parents,
and these lower parental expectations account in part for the poorer performance in reading and math of first graders from single parent homes
(Thompson et al., 1988). Here, however, parents' expectations for conduct
did not serve a mediating role in explaining differences across family types
in children's conduct. There are no significant differences in parents' conduct expectations by family configuration when these expectations are explained using the model in Figure 1. Furthermore, for increases in conduct
marks over the year, logistic regressions showed that parents' expectations
for conduct have effects that look much the same for each group (positive,
not significant).

421

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

APPENDIX
TABLE A - 1 , Attrition Analysis: Total
Sample vs. Sample With Complete Data
Sample w i t h
c o m p l e t e data b

Total sample 3
Variable

SD

SD

African Americans

Sex
Parent ability estimate
Parent expectationsreading
Parent expectationsmath
Parent expectationsconduct
Parent educationyears
Economic standing
Amount of kindergarten
Mother's age at child's birth
Number of siblings
Referral status
Number of absences
Number of times tardy
Number of parent conferences
Fall conduct mark
Spring conduct mark
Fall testreading comprehension
Fall testmath concepts
Fall markreading
Fall markmath

.50
3.8
2.7
2.7
1.9
12.1
.24
3.2
23.6
1.4
.19
13.0
8.1
.66
1.68
1.74
282
290
1.8
2.1

.50
.85
.71
.66
.32
2.3
.43
.53
6.0
1.4
.39
12.1
12.9
1.0
.47
.44
38.8
28.5
.72
.83

Sex
Parent ability estimate
Parent expectationsreading
Parent expectationsmath
Parent expectationsconduct
Parent educationyears
Economic standing
Amount of kindergarten
Mother's age at child's birth
Number of siblings
Referral status
Number of absences
Number of times tardy
Number of parent conferences
Fall conduct mark
Spring conduct mark
Fall testreading comprehension
Fall testmath concepts
Fall markreading
Fall markmath

.52
3.5
2.6
2.7
1.8
11.7
.46
3.1
24.5
1.5
.25
12.9
2.3
.37
1.83
1.86
283
297
2.0
2.4

.50
.79
.80
.77
.37
2.9
.50
.55
5.2
1.2
.43
10.2
4.9
.75
.37
.34
41.9
35.2
.68
.83

.53
3.8
2.8
2.8
1.9
12.0
.22
3.2
23.5
1.5
.18
13.2
8.0
.72
1.69
1.75
284
291
1.8
2.2

.50
.85
.72
.67
.29
2.3
.41
.52
5.9
1.5
.39
12.5
12.9
1.1
.46
.43
39.7
28.7
.72
.87

.50
3.5
2.7
2.7
1.9
11.8
.50
3.2*
24.4*
1.5
.23
12.5
2.6
.35
1.85
1.87
281
297
2.0
2.4

.50
.79
.78
.75
.34
2.9
.50
.52
5.2
1.3
.42
9.7
5.2
.73
.36
.33
41.0
34.3
.66
.83

Whites

Total sample for African Americans is 393; total sample for Whites is 333. b Sample with
complete data for African Americans is 256; sample with complete data for Whites is 228.
* Means for mother's age and amount of kindergarten are based on a subsample (N = 176)
of the total White sample (N = 228).

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Family Composition and Student

Conformity

Notes
This research was supported by National Institute of Child Health and Human Development grants 1 ROl HD16302, 1 ROl HD21044, and 1 ROl HD23943 and by National Science
Foundation Grant SES-8510535. Maxine Thompson was supported by National Research Service Awards 5 F32 MH09413-02 and 5 T32 MH15735-05. We thank the children, parents,
teachers, and other personnel in the Baltimore City public schools who gave us such splendid cooperation in all phases of this research. We also thank Linda Olson and two anonymous
referees for helpful suggestions on earlier versions of the manuscript.
'The pattern of findings reported in this article did not change appreciably when the
school dummies were omitted or when children's achievement scores and marks were added.
Also, mean square errors within classrooms were compared with those between classrooms
to investigate whether clustering of students within classrooms led to distortion of levels
of test significance. The findings are sufficiently robust to allay concerns on this score as well.
2
The difference by race in the size of this effect is not significant as judged by a Mest
of the difference in the two regression coefficients.
3
Logistic regression coefficients were transformed into a measure of the change in likelihood of being often late (change in the odds ratio). For dichotomous variables, the change
in the odds when moving from one category to the other (e.g., solo-mother homes versus
father-present homes) was calculated. For continuous variables, the change in the odds was
calculated from the mean to one standard deviation above the mean.
4
Race differences in lateness for children in two-parent families are significant at 5%
level as judged by a Mest of the difference in the two regression coefficients.
5
A Mest of the difference by race in the coefficients for number of siblings was not
significant.

References
Alexander, K.L., & Entwisle, D.R. (1988). Achievement in the first two years of school:
Patterns and processes. Monographs of the Society for Research in Child Development, 53(2, Serial No. 218).
Blake, J. (1981). Family size and the quality of children. Demography, 18, 421-442.
Blake, J. (1989). Family size and achievement. Berkeley: University of California
Press.
Entwisle, D.R., & Alexander, K.L. (1988). Factors affecting achievement test scores
and marks of black and white first graders. Elementary School fournal, 88,
449-471.
Entwisle, D.R., Alexander, K., Cadigan, D. & Pallas, A.M. (1987). Kindergarten experience: Cognitive effects or socialization? American Educational Research
fournal, 24, 557-564.
Entwisle, D.R., & Hayduk, L.A. (1978). Too great expectations: The academic outlook
of young children. Baltimore: Johns Hopkins University Press.
Entwisle, D.R., & Hayduk, L.A. (1982). Early schooling: Cognitive and affective outcomes. Baltimore: Johns Hopkins University Press.
Epstein, J.L. (1984). Single parents and the schools: The effect of marital status on
parent and teacher evaluations. (Center for Social Organization of Schools,
Report No. 353). Baltimore: Johns Hopkins University.
Falbo, T. (1982). Only children in America. In M.E. Lamb and B. Sutton-Smith (Eds.),
Sibling relationships: Their nature and significance across the life span
(pp. 285-304). Hillsdale, NJ: Erlbaum.
Furstenberg, F.F. (1976). Unplanned parenthood: The social consequences of teenage
childbearing. New York: Free Press.
Garfinkel, I., & McLanahan, S.S. (1986). Single mothers and their children: A new
American dilemma. Washington, DC: Urban Institute Press.
Hess, R., & Shipman, V. (1965). Early experience and socialization of cognitive models
in children. Child Development, 36, 869-886.

423

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015

Thompson,

Entwisle,

Alexander,

and

Sundius

Hetherington, E.M., Camara, K., & Featherman, D. (1983). Cognitive performances,

school behavior, and achievement of children from one-parent households.
Washington, DC: National Institute of Education.
Hofferth, S.L. (1987). Social and economic consequences of teenage childbearing.
In L. Hofferth and C D . Hayes (Eds.), Risking the future (pp. 123-144). Washington, DC: National Academy Press.
Kellam, S., Ensminger, M., & Turner, R.J. (1977). Family structure and the mental
health of children. Archives of General Psychiatry, 34, 1012-1022.
Maccoby, E.E. (1990). Gender and relationships: A developmental account. American
Psychologist, 45, 513-520.
McLanahan, S. (1985). Family structure and the reproduction of poverty. American
Joumal of Sociology, 90, 873-901.
Mercy, J.A., & Steelman, L.C. (1982). Familial influences on the intellectual attainment of children. American Sociological Review, 47, 532-542.
O'Connell, M., & Rogers, C.C. (1980). The legitimacy status of first births to U.S.
women aged 15-24, 1939-1978. Family Planning Perspectives, 12, 16-25.
Santrock, J.W., & Tracy, R.L. (1978). The effect of children's family structure status
on the development of stereotypes by teachers, Journal of Educational Psychology, 70, 754-757.
Stack, C.B. (1974). All our kin. New York: Harper & Row.
Staples, R., & Mirande, A. (1980). Racial and cultural variations among American
families: A decennial review of the literature on minority families, Journal of
Marriage and the Family, 42, 157-173.
Sutton-Smith, B., & Rosenberg, G.B. (1970). The sibling. New York: Holt, Rinehart
& Winston.
Thompson, M., Alexander, K., & Entwisle, D. (1988). Household composition, parental expectations and school achievement. Social Forces, 67, 424-451.
U.S. Bureau of the Census. (March, 1988). Current Population Reports, Series P-20,
No. 433. Washington, DC: U.S. Government Printing Office.
U.S. Department of Commerce, U.S. Bureau of the Census. (1989). Money, income,
and poverty status in the U.S. Current Population Reports, Series P-60, No.
168. Washington, DC: U.S. Government Printing Office.

424

Downloaded from http://aerj.aera.net at UNIV ARIZONA LIBRARY on May 30, 2015