To appear in Murray A. Straus (in press, 2003) CP70
The Primordial Violence:
Corporal Punishment By Parents,
Cognitive Development, And Crime. Walnut Creek CA: AltaMira Press
CORPORAL PUNISHMENT
AND ACADEMIC ACHIEVEMENT
SCORES OF YOUNG
CHILDREN: A LONGITUDINAL STUDY
Abstract
Objective: To test
the hypothesis that use of corporal punishment (CP) by parents is associated
with a subsequent decrease in academic performance. Methods:
The frequency of hitting or spanking in the past week was measured for
622 children ages 5 and 6 in 1992.
Academic achievement was measured by the Peabody Individual Achievement
Test (PIAT) in 1992 and 1994. Multiple
regression analysis controlled for 1992 PIAT score, child’s level of antisocial
behavior, mother’s education, race and gender of child, mother's age at
birth of child, father's presence in household, number of children in the
household, amount of emotional support and cognitive stimulation the
child received. Results:
Each increase of one unit in the four
unit CP scale at Time 1 was associated with an average decrease of 2.7 points
in PIAT score at Time 2, net of all other variables. Conclusions. The finding that CP adversely affects
academic achievement is important for children and for the nation because
academic achievement is a major determinant of economic and health status and
because CP is amenable to change through public health and parent education.
KEY WORDS: corporal punishment, spanking, cognitive development, parent behavior, academic achievement, antisocial behavior
Academic performance in the sense of school grades, test scores, or years of education is a major determinant of many other behaviors and conditions that are valued in many societies. The amount of education completed, for example, is the single most important determinant of occupation and income (National Alliance of Business 1998). Level of education also affects health status and health-related behaviors such as smoking, dietary habits, seat belt use, and use of preventive health care services (prenatal care, cancer screenings, immunizations for children). In short, academic achievement affects the probability that children and adolescents will be successful and healthy in adulthood, and that they will raise successful and healthy children.
This paper is focused on the part of academic performance that is
measured by standardized tests of "academic achievement." Performance on these tests is a significant
predictor of later educational
performance and attainment (Grundmann 1997).
Research on the etiology of academic performance, including performance
on standardized tests, shows that it is correlated with major demographics such as socioeconomic status,
race/ethnicity, and gender (Guskey 1997). A second tier of
research on the etiology of academic performance has examined characteristics
of parent-child interactions such as cognitive stimulation and emotional
support (Grundmann 1997). These studies
have generally shown that parent-child interactions are significant
determinants of academic achievement and cognitive development regardless of
socioeconomic status, race, or gender of children. The
purpose of the study reported in this paper is to expand understanding of
family environment factors that could affect cognitive functioning and academic
performance by considering a pervasive but under-researched aspect of
children’s family experiences: CP by
parents. We will first review research
on the prevalence of CP and then review research that provides a basis for
hypothesizing a link between mothers' use of CP (CP) and the academic achievement of their children in early
grade school years.
CP is defined here as the use of physical force with the intention
of causing physical pain, but not injury, for purposes of correction and
control (Straus 2001a).
Examples of CP include slapping, spanking, pinching, or ear
twisting. Parent-to-child CP as defined
above is currently legal in all 50 of the United States. Other forms of CP (e.g., teacher-student,
stranger-child) have increasingly been made illegal in recent years. For example, about half the states now
prohibit CP by school personnel, and many states prohibit CP by foster parents
(the states are listed in the web site of the Center For Effective Discipline, http://www.stophitting.com)
Prevalence
And Chronicity Of CP
The prevalence and chronicity of CP vary widely by children's age
and gender. National surveys reveal
that the current prevalence of spanking toddlers (age 2-4) is over 90% (Straus and Stewart 1999). For
some in this 90%, CP is a rare event, for others it occurs several times a
day. The prevalence of CP remains high
at ages 5 and 6, and then declines steadily from ages 7 through 17. Despite the decline in CP after age 6,
almost half (43%) of the parents of 13 year-old children and more than a
quarter of parents of 16 year-old children reported having hit them in the
previous 12 months(Straus and Stewart 1999).
Information on chronicity differs from
study to study, probably because of differences in the method of obtaining the
data. In interview studies, for
example, the longer the time period about which the parent is questioned, the
lower the frequency of CP. The National
Family Violence Surveys, for example, asked how many times CP was used in the
previous year, which resulted in a mean of 8.9 times by parents of children
aged 3 to 17. However, the National
Longitudinal Survey of Youth asked the mothers how many times they had spanked in the past
week. Among the 71 percent of mothers
of 1-4 year olds who had spanked, the mean number of times that week was
3.6. This suggests an annual chronicity
in the hundreds of times. In addition,
6.8 percent of the mothers hit the focal child during the course of the
interview (Giles-Sims, Straus, and Sugarman 1995). This
suggests an even higher chronicity.
Even a college-educated sample of mothers reported using CP an average
of 2.5 times in the past week(Holden, Coleman, and Schmidt 1995).
Demographic
Differences In CP
Research consistently
shows that boys are somewhat more likely than girls to be spanked (Giles-Sims, Straus, and Sugarman 1995), especially once they reach school age (Anderson and Anderson 1976; Straus 2001a). The
most common explanation for the higher CP prevalence among young children
(particularly boys) is that they "act out" more often than older
children. The higher spanking rate among boys might also reflect cultural
expectations of the need for "toughness" in preparation for
traditional roles of male adults.
The prevalence of
CP also varies by race and ethnicity.
African-American parents are more likely to use CP than Euro-American
and Hispanic-American parents (Day, Peterson, and McCracken 1998; Giles-Sims,
Straus, and Sugarman 1995). Some
researchers have suggested that the higher CP prevalence among African-Americans is due to cultural
differences in beliefs regarding the appropriateness of CP as a form of
discipline (Rohner, Bourque, and Elordi 1996). Others
note that racial differences in CP behavior are not as significant when
socioeconomic status is taken into account(Cazenave and Straus 1990).
Parents of low SES
are also more likely to spank than are their higher SES counterparts (Day, Peterson, and McCracken 1998; Giles-Sims,
Straus, and Sugarman 1995). One recent
study shows that the mother’s level of education is strongly related to CP use(Arias and Pape 1999). It may be
that parents with lower educational attainment are less likely to practice
“cognitive” means of discipline than are the less educated.
Trends
Finally, despite the fact that CP of toddlers still exceeds 90%,
spanking rates of older children have declined significantly in recent
years. In addition, the belief that CP
is an appropriate means of discipline is decreasing. The proportion of the US population who believed that CP is
sometimes necessary declined from 94% in 1968 to 68% in 1994 (Straus and Mathur 1996), and the decrease is continuing (Straus 2001b). As
with many attitudinal changes of this kind, changes in actual spanking behavior
may occur somewhat later in time. This
would portend further decline in the prevalence and chronicity of CP by
parents.
CORPORAL PUNISHMENT AND ADVERSE OUTCOMES
Research on the effects of CP has historically focused the
association of CP with increased risk of aggression and delinquency, as
exemplified in the research of Sears, Maccoby, and Levin (1957), Feshbach (1973) and McCord (1991).
Recently, however, there has been a series of studies showing links to
other maladaptive outcomes including depression (DuRant, Getts, Cadenhead, Emans, and Woods
1995; Holmes and Robins 1988; MacMillan, Boyle, Wong, Duku, Fleming, and Walsh
1999; Straus 1994; Straus and Kantor 1991), problem drinking (Straus and Kantor 1994), and engaging in masochistic sex as adults (Huth-Bocks,
Levendosky, and Semel 2001; Kinard 1999; Widom 1989a; Widom 1989b).
Physical abuse of children is known to adversely
affect cognitive performance (Huth-Bocks,
Levendosky, and Semel 2001; Kinard 1999; Widom 1989a; Widom 1989b) It is possible that the less severe hitting of children in the
form of CP has a similar adverse effect, although probably less strong. Three recent papers pertain more directly to
the current study. The first found
evidence that CP in early childhood was significantly related to a subsequent
slower rate of cognitive development (Arias and Pape 1999). The
second showed that. The second found
that CP in adolescence was inversely related to graduation from college (Aldarondo and Sugarman 1996). A
third study showed that CP in childhood
and adolescence is associated with lower economic and occupational achievement
in adulthood(Straus and Gimpel 2001). Taken
together, these three studies support the ideas that the use of CP impacts
adversely upon areas of success in school and work and that the effects of CP
can be lifelong. A problem with the
three studies showing that CP is related to lower cognitive, academic, and
economic achievement is that these studies did not control for the level of
misbehavior at Time 1. This is a
crucial omission because parents spank in response to misbehavior, including
failure to do homework and failing grades.
In addition, other behavior problems for which parents spank, such as
lying, fighting, and disobedience are also associated with lower academic
achievement. Thus, what seems to be an
adverse effect of CP on cognitive, academic, and economic achievement could be
spurious. It could result from the
behavior problems that lead parents to spank and that also hamper academic
achievement. The research to be
reported deals with this problem by controlling for the level of misbehavior at
Time 1. As a result, if the following
hypothesis is supported, it cannot be explained by a tendency for misbehaving
children to do more poorly in school:
The more CP a child experiences at ages 5 and 6, the more likely
the child is to experience a subsequent reduction in academic achievement test
score when measured two years later.
In view of the prevalence of CP, if
the hypothesis that CP leads to a decline in academic achievement is supported,
it could have profound implications because: (1) Level
of educational attainment is one of the major determinants of health status and
health-related behaviors such as smoking, dietary habits, seat belt use, and
use of preventive health care services (prenatal care, cancer screenings,
immunizations for children); and (2) CP is a risk factor that is likely
to be amenable to change by public health intervention and education.
The
sample consists of mothers of 622 children who were age 5 and 6 in 1992 (Time
1). These women were part of the
National Longitudinal Survey of Youth (NLSY), which began interviews with a
cohort of females age 14 to 21 in 1979.
The original survey was funded by the Bureau of Labor Statistics and
focused primarily on issues related to the labor force and the work place. The same women have been interviewed every
two years since 1979. As the cohort
matured, the NLSY expanded to include questions about the children of these
women.
Five
and six year-olds were chosen for this study for three main reasons. First, the highest rate of spanking is
between ages 2 and 6 (Straus and Stewart 1999). Second, some
advocates of CP suggest that it should not be used with children over six
(Consensus statement) (Friedman and Schonberg 1996). Third, this study
is designed to assess changes in academic achievement as measured by the
Peabody Individual Achievement Test (PIAT).
The youngest children to whom the PIAT is administered are age 5. Therefore, the sample represents a cohort
with a relatively high incidence of CP, which occurs at an age considered by many as appropriate for this form of
discipline, and for whom PIAT results were available at both Time 1 and Time 2
of the study.
The
622 children represent 63% of the 985 NLSY children who were 5-6 years old in
1992. The remaining 37% (N=363) were
excluded due to missing data on one or more of the variables. Most
missing data elements were for CP in 1992 (N=86), PIAT 92 (N=179), PIAT 94
(N=154), the Cognitive Stimulation Index (N=172), and the Emotional Support
Index (N=186). The only other control
variable with more than 10 missing cases was whether or not the father was in
the household (N=56). To assess
possible selection bias due to missing data, demographic characteristics of the
sample analyzed were compared with those of the children who were excluded
because of missing data. Compared to
those excluded, children in the sample had mothers with somewhat higher levels
of education (12.6 vs. 11.9 average years), higher PIAT scores in both 1992
(54.5 vs.47.7) and 1994 (54.5 vs. 48.2), and scored significantly higher on the
emotional support index (49.1 vs. 44.8).
The proportion of African-American children in the sample was the same as that of the excluded
(29%), but there was a significantly lower proportion of Hispanic-Americans
(21.8% vs. 29.5%) and a correspondingly higher proportion of Euro-Americans
(49.1% vs. 41.2%) in the sample compared to those excluded. Finally, the sample children were
significantly more likely to have a father in the household in 1992 than were
those excluded from the sample (66.0% vs. 54.3%). There was no significant difference between the sample and
non-sample children in mother's age at birth (25 years), number of children in
the household (2.6), level of cognitive stimulation (46.2 vs. 42.1), gender,
and percent not spanked during the week for which CP data was gathered (53.4
vs. 52.0). These differences indicate
that the cases retained for analysis are somewhat higher in socioeconomic
status and higher academic achievement than the NLSY sample as whole.
Corporal
Punishment. CP was measured at Time 1 using the
following question: "Sometimes kids mind pretty well and sometimes they
don't. About how many times, if any,
have you had to spank your child in the past week?" Fifty-two percent of mothers reported no
spanking in the previous week, while 15.5% reported using CP three or more
times.
Academic
Achievement. Academic achievement at both Time 1 (1992)
and Time 2 (1994) was measured by the Peabody Individual Achievement Test
(PIAT) for Math and for Reading Recognition.
The PIAT provides a wide-range measure of academic achievement for
children five and over. It is among the
most widely used achievement tests and has high test-retest reliability and
concurrent validity (Baker, Keck, Mott, and Quinlan 1993). Percentile scores
for the math and reading components were combined to produce a summary PIAT
score for each child in each time period.
The mean PIAT score was 54.7 at both Time 1 and Time 2. The standard deviation for each year was
about 23.
Antisocial
Behavior. Children’s level of antisocial behavior
(ASB) is considered a major “cause” of the use of CP by parents. It is measured in this study by a subscale
of the NLSY Behavioral Problems Index.
Items include: Cheats or tells lies; Bullies or is cruel/mean to others;
Does not feel sorry for misbehaving; Breaks things deliberately; Is disobedient
at school; and Has trouble getting along with teachers.
Cognitive
Stimulation and Emotional Support. The measures of cognitive stimulation and
emotional support by the mother are subscales from the Home Observation for
Measurement of the Environment - Short Form (HOME-SF) inventory, which includes
subscales appropriate for each age (Caldwell and Bradley 1984). An extensive
review of these scales indicates that they are internally consistent, stable
over time, and predictive of a variety of child outcomes, including academic
achievement.
Each
scale contains similar items for 5 and 6 year olds. Examples of cognitive stimulation items include how often the child is read to, how many
books the child has, and how often a family member takes the child on an outing
(to a park, to shop, to picnic, etc.).
The emotional support scale includes items such as how much choice the child has over food at
breakfast or lunch, how many hours the TV is on in the home each day, and how
often the child eats a meal with both the mother and father (or other male
partner).
Mother’s
Level of Education. Mother’s level of educational attainment was
employed as an indicator of socioeconomic status (SES) in this study for three
reasons: (1) it directly affects children’s academic achievement; (2) it has a
strong inverse association with CP use; and (3) it is a stable indicator of
SES.
From a public-health standpoint,
income and education tend to have different and independent effects on
health-related phenomena. Education is
associated more with internalized health-related attitudes and behaviors such
as smoking and seatbelt use, while income is associated more with external
phenomena such as financial barriers to utilization of health care
resources. In this sense, CP represents
a health-related behavior that is associated more with education than with
income, and therefore is amenable to change through education.
Other
Independent Variables. Five other maternal, child, and family
characteristics were selected as control variables from the NLSY data set due
to their association with both CP and academic achievement: mother's age at
birth of child, race and gender of child, number of children in household, and
presence of father in household.
Nineteen percent of mothers had less than 12 years
education in 1992, while 14% were college graduates. About 49% of the sample children were Euro-American, 29% African-American, and 22% Hispanic-American. Just over 52% were male and about 48% were
female. The average maternal age at
child's birth was 25.1 years, with 99% of the range being between 21 and 29
years of age. Two-thirds of mothers
reported that the father was present in the household at the time of interview,
while the average number of children in the household in 1992 was 2.6, with
over 83% reporting between 2 and 4 children.
Due to the oversample of low socioeconomic respondents by
the NLSY, and because of missing data, the sample of the current study is not
representative of either that of the National Longitudinal Survey or of the US
as a whole. However, because the sample
used for this analysis is somewhat higher in socioeconomic status and academic
achievement test score than the NLSY sample as a whole, the sample used for
this study may be closer to being representative of the US population than
would be the case if there were no cases lost because of missing data.
The
data analysis began with inspecting the frequency distributions of all
variables for skewness and outliers.
Then cross tabulation and correlations were computed to provide an
initial examination of the bivariate relationships of CP with the other
variables.
The
hypothesis was tested using hierarchical regression. Three models were estimated.
The first examined the relation of Time 1 academic achievement (PIAT92)
and the eight control variables
(antisocial behavior, mother’s
level of education, child Hispanic-American, child African-American, age of
mother at child's birth, father's presence in the home, and number of children
in the home) to academic achievement at Time 2. These controls are characteristics that may explain or underlie
the association between CP and child academic achievement.
Model
2 focused on whether CP and two other aspects of parent-child interaction
(cognitive stimulation; emotional support) could provide any additional
explanation of academic achievement.
For the effect of CP to be significant, Model 2 must add to the variance
explained in academic achievement at Time 2 net of all variables in Model 1.
Model 3 added five
interaction variables. They were the
interactions of CP with: 1) the PIAT 92 score, 2) child's race
African-American, 3) mother's level of education, 4) cognitive stimulation, and
5) emotional support.
RESULTS
Prevalence of
Corporal Punishment
(Insert Table 1 about here)
Table 1 reports the one-week prevalence of spanking
of 5 and 6 year olds for the week prior to the 1992 interview. Overall, about half of mothers reported
"no spanking", while about 16% reported having spanked 3 or more
times.
Cross-tabular
analysis of CP by each control variable showed that mothers in the lowest
education group were far more likely to use CP than were their more highly
educated counterparts. They also
exhibited a much higher frequency of spanking. Euro-Americans had the lowest
rate of spanking among the three race/ethnic groups, followed by
Hispanic-Americans, then African Americans.
Boys were much more likely than
girls to have been spanked, although the proportion of boys and girls spanked 3
or more times was not significantly different.
Mothers with four or more children were much more likely to spank than
were those with fewer than four children.
They were also much more likely to report spanking 3 or more times than
were mothers with smaller families.
Father’s presence in household and older maternal age were somewhat
related to lower rates of CP use. The
last row of Table 1 is at the heart of much controversy about CP and bears
directly on the need for longitudinal research. It shows that children with a high level of antisocial
behavior(ASB) were spanked much more than children who were low to middle ASB. The fact that use of CP is associated with misbehavior
highlights the necessity of controlling for the level of ASB when examining the
relation of CP to academic achievement.
This is because, is ASB is not controlled, finding that CP is associated
with lower achievement scores could simply reflect the fact that children who
are high in ASB are more likely to be spanked and less likely to do well in
school.
(Insert Figure 1 about here)
Figure 1 shows that the more CP
experienced by the children in this sample in 1992, the lower their mean score
on the PIAT, both that year and in 1994.
The relationship between CP and PIAT score was slightly stronger in 1994
than in 1992.
The relationships in Figure 1 might
be spurious due to confounding of both CP and PIAT scores with other family
characteristics such as socioeconomic status and the age of the mother. The multiple regression analysis will deal
with this issue by including ten control variables in the model estimated.
Another limitation of Figure 1 is that the relationship between CP and PIAT score could reflect a tendency for parents to spank children who do badly in school. Just as a good report card can be the occasion for praise and reward, a bad report card can be the occasion for admonition and spanking. This use of CP is illustrated by a coach at one high school who said, “Just as a father would do, coaches review each player's report cards every six weeks. Those with bad grades line up for the punishment: licks with a wooden paddle wrapped in athletic tape. Five for an F. Three for a D. Three for bad marks in conduct” (Aimee Edmondson, “Short of the Goal" The Commercial Appeal in Memphis, January 14, 2001).
The
possibility that the causal direction is the opposite of the hypothesized
direction is also plausible because children who do poorly in school tend to
have other behavior problems which might elicit CP. The regression analysis will deal with this issue by including
Time 1 PIAT score as one of the independent variables. This permits examining whether CP is
associated with a subsequent change in PIAT scores. Parents who use CP to correct misbehavior,
including poor school performance, expect their use of CP to result in better
behavior subsequently. Our hypothesis,
however, predicts that use of CP will be associated with a subsequent change
for the worse.
Correlation Of CP and
Academic Achievement With Other Variables
(Insert Table 2 about here)
Table 2 provides the zero-order correlations of all the variables
in the study. Going down the first two
columns of Table 2 enables one to examine the correlates of PIAT scores at
Times 1 and 2. The correlation of the two PIAT scores, although fairly high (.61), also
means that the Time 1 score explains only 36% of the variance of the Time 2
score. Thus, many other factors
contributed to changes over the two-year time span.
CP was associated with PIAT scores in both 1992 and 1994. Consistent with the rationale underlying
this study, both correlations are negative, and the correlation at Time 2 is
higher (-.28 compared to -.19). Almost all of the control variables were
significantly correlated with both PIAT92 and PIAT94.
The third column of
Table 2 gives the correlates of CP with the rest of the variables in the
study. There are significant correlations
of CP with all but two of the controls: Child Hispanic, and father in
household. The strongest correlations
with CP were for Antisocial Behavior, Cognitive Stimulation and Emotional Support.
(Insert Table 3 about here)
Table 3 presents results of the hierarchical regression analysis to test the hypothesis that CP is associated with a subsequent decline in academic achievement score. Model 1 shows that academic achievement at Time 1 and mother's level of education were significantly related to increases in academic achievement score from Time 1 to Time 2, whereas African-American children experienced a decline in PIAT scores. These relationships replicate results from many studies of academic achievement.
Model
2 was estimated to determine if CP and two other measures of mother-child
interaction (cognitive stimulation, emotional support) add significantly to the
variance in PIAT scores explained by the demographic variables in Model 1. The results show that cognitive stimulation
is associated with an increase in PIAT scores from Time 1 to Time 2, while CP
is associated with a decrease. The row
for Corporal Punishment shows that for every increase of one unit in CP, there
was a decrease of 2.7 points in PIAT percentile score from 1992 to 1994
In Model 2, mother’s education, and
African-American race remain significantly related to change in PIAT
score. Adding these variables resulted
in a significant increase in the variance in academic achievement score
explained by Model 2 compared to Model 1.
Model 3 added the interaction of CP with five other variables: 1) the PIAT 92 score, 2) child's race African-American, 3) mother's level of education, 4) cognitive stimulation, and 5) emotional support. None of these interactions were significant. Moreover, adding the interactions to the model resulted in a reduction of the amount of variance explained in the dependent variable, leaving model 2 as the best fitting model.