Author Affiliations: School of Social Work, University of Michigan, Ann Arbor.
To examine the relationship of corporal punishment with children’s behavior problems while accounting for neighborhood context and while using stronger statistical methods than previous literature in this area, and to examine whether different levels of corporal punishment have different effects in different neighborhood contexts.
Longitudinal cohort study.
1943 mother-child pairs from the National Longitudinal Survey of Youth.
Main Outcome Measure
Internalizing and externalizing behavior problem scales of the Behavior Problems Index.
Results and Conclusions
Parental use of corporal punishment was associated with a 0.71 increase (P<.05) in children’s externalizing behavior problems even when several parenting behaviors, neighborhood quality, and all time-invariant variables were accounted for. The association of corporal punishment and children’s externalizing behavior problems was not dependent on neighborhood context. The research found no discernible relationship between corporal punishment and internalizing behavior problems.
In the study of family life, a longstanding issue has been the question of whether or not corporal punishment is associated with changes in the incidence of behavior problems in children. Despite the prevalence of corporal punishment as a disciplinary tactic, several researchers have suggested that corporal punishment is associated with increases in children’s behavior problems.1- 6 In a comprehensive meta-analysis, Gershoff7 found that corporal punishment was consistently associated with increases in children’s behavior problems.
In recent years, some research has focused on whether the relationship of corporal punishment and children’s behavior problems varies across neighborhoods. Some authors8- 10 have suggested that in neighborhoods where social control is low and social norms are frequently violated that strict parenting may actually be a factor that protects children from harm. Two recent research papers have extended this idea to empirically examine the effect of neighborhood context on the relationship between corporal punishment and children’s behavior problems. In her longitudinal examination of the relationship of corporal punishment and children’s behavior problems, Eamon11 used the National Longitudinal Survey of Youth (NLSY)12 to examine the effects of neighborhood characteristics on antisocial behavior in a sample of 963 10- to 12-year-olds. Eamon found evidence to suggest that “when environmental risk is high, parenting practices that are firmer and higher in control result in lower levels of young adolescent antisocial behavior.” Using a smaller community-based sample of 841 African American families with children between 10 and 12 years of age, Simons et al13 used a cross-sectional model to examine the relationship of corporal punishment and children’s behavior problems in neighborhoods that were high in crime as compared with neighborhoods with lower levels of crime. In contrast to Eamon, these authors found “no evidence that the association between corporal punishment and conduct problems varied by community prevalence of social deviance.” However, Simons et al13 did find that the relationship between corporal punishment and children’s behavior problems was not apparent in neighborhoods where the prevalence of corporal punishment was high. Interestingly, neither of these research projects focused on the relationship of corporal punishment, neighborhood context, and behavior changes in younger children, even though corporal punishment is most frequently used with younger children.12
In an important review of the empirical research literature on corporal punishment, Benjet and Kazdin14 have noted that unobserved variables may confound the relationships between corporal punishment and increases in children’s behavior problems that pervade the literature on corporal punishment. For example, at the family level, it is possible that an unobserved inability of certain parents to control their anger might account for observed relationships between parental use of corporal punishment and increases in children’s antisocial behavior. Parents with a decreased ability to control their anger might be more likely to use corporal punishment and might also transfer this decreased ability to handle anger to their children, resulting in increased behavior problems among those children. Similarly, at the neighborhood level, observed relationships of parenting, neighborhood, and children’s behavior problems may be confounded by unobserved variables, such as the availability of economic resources, that are correlated with parents’ decisions to live in a particular neighborhood. Thus, accurate assessment of the relationship between corporal punishment and behavior problems in children requires the use of statistical models that provide strong controls for unobserved variables.
The research questions this study addresses are as follows:
What is the relationship between corporal punishment and children’s behavior problems when statistical controls for unobserved variables are applied?
What is the relationship between corporal punishment and children’s behavior problems when perceptions of neighborhood quality are accounted for?
Does the relationship between corporal punishment and children’s behavior problems vary by perceived level of neighborhood quality?
Data used in this analysis are derived from the NLSY.12 Consent was obtained from the university institutional review board to make use of this data. Since 1979, the NLSY has conducted in-home interviews with a representative sample of young men and women and has collected information on their income and labor market experiences. Since 1986, the NLSY has conducted biennial in-home assessments of the children born to the women in the original NLSY sample. In 1992, the NLSY began collecting information on respondents’ evaluations of the neighborhoods in which they lived. Lastly, in 1994, the NLSY began computing summary scores for children’s internalizing and externalizing behaviors.
Therefore, for this longitudinal analysis, data were extracted from the 1994, 1996, 1998, and 2000 interviews of the NLSY. Data from the mother record and the child record were combined to form a single record for each mother-child pair for each year of the data. To conform to statistical assumptions about independent and identical distributions of error terms, 1 mother-child pair was randomly selected from each family to be part of the analysis sample.
The sample was limited because many of the measures included in the analysis were only collected for particular age ranges. For example, information on children’s behavior problems was collected only for children between 4 and 14 years of age,15 so children were included in the sample if they were observed between these ages. Therefore, this sample includes children of a broader range of ages than does prior empirical work in this area. These procedures and limitations resulted in a final sample size of 2567 mother-child pairs. For some variables, up to 25% of the data were missing. Multiple imputation of data,16- 18 described in more detail in the section “Analytic Method,” was used to compensate for missing data.
Variables used in this analysis were chosen on the basis of their relevance to the substantive questions examined in this research project, their use in the previous empirical literature on the effects of corporal punishment, and their availability in the data. Children’s basic demographic characteristics such as their sex, race, and age were easily available from the NLSY and were included as independent variables in the model.
Children’s behavior problems were assessed using the Behavior Problems Index (BPI), a 28-item measure of children’s behavior designed to be administered relatively efficiently within the context of a survey.15 The BPI has been used in hundreds of studies of children’s behavior problems.19 In administering the BPI, mothers were asked about the degree to which their child engaged in particular problem behaviors and responded with an indication of “often,” “sometimes,” or “never.” Responses were then summed into scales. Research indicates that the BPI factors fell into 2 broad subscales of internalizing behavior problems, representing constructs such as anxiety and depression, and externalizing behavior problems, representing such constructs as antisocial behavior, peer problems, and hyperactivity.15 To facilitate comparison of measurements of behavior problems across years, the Center for Human Resource Research (Columbus, Ohio) developed standardized scores that have a mean of 100 and a standard deviation of 15.15
The use of corporal punishment was measured in the NLSY in several ways. This analysis used a continuous measure of the use of corporal punishment, taken from a question in the NLSY that asked each mother about the number of times she had spanked her child in the past week. All of the questions about corporal punishment in the NLSY left it to the respondent to define what spanking means.
The NLSY provided information on the cognitive stimulation that parents provide to their children as part of the Home Observational Measure of the Environment. The measure of cognitive stimulation consisted of 7 items and included such questions as whether or not parents read to their children or involve them in intellectually stimulating activities.15 Questions from the cognitive-stimulation measure were then summed into a scale. To enable the comparison of information on cognitive stimulation across study years, the cognitive-stimulation scores for each year were standardized to have a mean of 100 and a standard deviation of 15.
Similarly, the Home Observational Measure of the Environment collected information on the emotional support that parents afford their children. Emotional support was measured through 5 items such as whether parents talk to their children in a loving manner. Questions measuring emotional support were summed into a scale. Standardized scores, again with a mean of 100 and a standard deviation of 15, were also computed for the measure of emotional support. Because the measure of emotional support included a measure of corporal punishment, simply including the emotional support scale in astatistical model where corporal punishment was also included as a separate independent variable would have created the possibility of confounding the relationship of corporal punishment and other elements of emotional support. Therefore, this analysis followed a procedure suggested by Straus et al.1 The item measuring corporal punishment was removed and the emotional support score was recomputed. The newly computed emotional support score was then standardized to have a mean of 100 and a standard deviation of 15.
The NLSY included a measure of familial poverty. This measure, computed by the NLSY staff, compared the respondent’s income with the correct poverty threshold for the respondent’s family size and the year in which the interview was conducted.20 Those respondents whose income fell below this threshold were coded as being in poverty.
In each round of data collection, the NLSY contained 9 questions on individuals’ perceptions of the strengths and problems of their neighborhoods. Starting in 1992, adult respondents in the NLSY have been asked about their perception of whether their neighborhood is a good neighborhood in which to raise children and their evaluations of other factors such as the level of neighborhood crime and whether neighborhood residents tend to follow rules. Following the lead of Eamon,11 questions from each year of the data were standardized with a mean of 0 and a standard deviation of 1 and then aggregated into a single scale for each year. The neighborhood scales from different years of data had an average Cronbach α of 0.87.
As noted earlier, the analysis sample contained a great deal of missing data. The presence of this missing data required the use of multiple imputation. Multiple imputation16- 18 allows the analyst to generate parameter estimates that are not biased by the presence of missing data. The uncertainty generated by the presence of missing data is thus incorporated in the calculation of standard errors, confidence intervals, and P values. In a multiple imputation analysis, regression type methods are used to impute plausible values for missing data.17 Data are imputed multiple times, resulting in several data sets with imputed values. The analysis is then conducted on each data set and the parameter estimates and standard errors averaged following formulas described by Schafer.18 No prior published study of the relationship between corporal punishment and antisocial behavior has employed multiple imputation techniques.
In the statistical and methodological literature, one frequently discussed concern with many statistical methods is that they are unable to account for unmeasured variables that may be correlated with the observed independent variables that are included in a statistical model.21- 23 Used with longitudinal repeated measures data, fixed effects can account for unmeasured variables that are time invariant by expanding the framework of the standard ordinary least squares regression model. Both Greene21 and Wooldridge22 provide thorough explications of the analytic details of a fixed-effects model for longitudinal data that uses multiple observations over time for each individual. Because the fixed-effect model controls for all time-invariant characteristics, separate estimates for observed time-invariant characteristics of study members, such as sex and race or a child’s initial level of antisocial behavior, are not directly available from the model. As I have demonstrated,24 fixed-effects analysis can be used to examine the relationship of corporal punishment and antisocial behavior while accounting for prior levels of antisocial behavior. However, I did not include measures of neighborhood in my fixed-effects models, nor did I employ a multivariate imputation procedure to account for the presence of missing data.24 Fixed-effect models are most useful when one has longitudinal data, when the variables of interest vary over time, and when there is reason to be concerned about the effects of unmeasured variables that might be correlated with observed regressors.
A major concern of this analysis was to examine whether or not the relationship of corporal punishment and children’s behavior problems differed across different neighborhood contexts. Therefore, a variable representing the interaction of neighborhood characteristics and the use of corporal punishment was created. This interaction term was then entered as an independent variable into the fixed-effects models that were estimated.
The sample was very diverse in terms of its demographic characteristics. As allowed by the availability of the necessary measures, children ranged in age from 4 to 14 years old. Children were almost equally divided in terms of sex. Forty-nine percent of the children were boys and 51% were girls. The sample exhibited racial and ethnic diversity: 61% of the children were white, 22% were black, and 18% were Hispanic or Latino. There was some diversity in terms of income status; approximately 13% of the sample were living below the poverty line. Neighborhood quality also showed variability with a mean ± SD score of 0.14 ± 0.60.
There was considerable variation in some variables reflecting the activities of parents. Parents used corporal punishment an average of 0.34 times in the past week. Cognitive stimulation and emotional support provided by parents showed considerable variation as did the internalizing and externalizing behaviors exhibited by children. Descriptive statistics for the sample are reported in Table 1.
The fixed-effects regression models indicated that several variables had statistically significant effects on children’s internalizing and externalizing behaviors. Results of the fixed-effects regressions are reported in Table 2 and Table 3. The fixed-effect regression models indicated that greater use of corporal punishment by parents was associated with increases in children’s externalizing, but not internalizing, behavior problems. The parameter estimate associated with the effect of 1 spanking on increases in externalizing behavior problems was 0.71. Because the standard deviation of the measure of children’s externalizing behavior problems was 14.34, the fixed-effects models suggested that each spanking was associated with a 5% of a standard-deviation increase in externalizing behavior problems. However, because fixed-effects models can examine only variation within individuals and not variation between individuals,23,24 it is possible that these models underestimate the true relationship of corporal punishment and antisocial behavior.
Other independent variables in the model also displayed an association with children’s behavior. Neighborhood quality, as assessed by survey participants, was related to children’s internalizing behavior problems but not to children’s externalizing behavior problems. Internalizing behavior problems decreased as neighborhood quality increased. Cognitive stimulation and emotional support provided by parents to their children were associated with decreases in externalizing but not internalizing behavior problems. Children’s age was associated with increases in children’s externalizing and decreases in internalizing behavior problems. Familial poverty was associated with increases in children’s internalizing behavior problems but not in children’s externalizing behavior problems.
A second major research question was whether the relationship of corporal punishment and children’s behavior problems would vary based on the quality of the neighborhood in which survey respondents resided. As noted earlier, this question was tested by developing an interaction between neighborhood quality and corporal punishment and entering this interaction term as an independent variable into the model. The interaction of corporal punishment and neighborhood quality did not have a statistically significant effect on children’s externalizing or internalizing behavior problems. The nonsignificance of this interaction term indicated that the relationship of corporal punishment and children’s externalizing behavior problems did not vary by neighborhood.
Some researchers8- 11 have proposed that the relationship between parenting and children’s behavior problems may be weaker in neighborhoods where levels of social problems are higher. The results of these findings suggest that parental use of corporal punishment was associated with increases in children’s externalizing behavior problems. Further analysis indicated that the strength of the relationship between corporal punishment and children’s externalizing behavior problems did not depend on the social characteristics of the neighborhoods in which families resided.
Eamon11 found that the relationship of corporal punishment and children’s behavior problems varied by perceived neighborhood quality. The results of Eamon’s11 research suggested that corporal punishment may actually be beneficial in neighborhoods where neighborhood problems are high. Eamon suggested that the results of her study indicated that “when environmental risk is high, authoritarian parenting strategies result in lower levels of antisocial behavior.” Although Simons et al13 did not find any effect of neighborhood deviance on the relationship of corporal punishment and children’s behavior problems, these authors did find that the relationship of corporal punishment and children’s behavior problems disappeared in neighborhoods where the prevalence of parental corporal punishment was high. Thus, both of these studies suggested that corporal punishment may not be harmful to children in certain types of neighborhoods. The findings of this research stand in contrast to this earlier research. In the models developed for this paper, corporal punishment was associated with a higher incidence of externalizing behavior problems for children in all types of neighborhoods. The findings of this research are thus in accord with the position of the American Academy of Pediatrics (Elk Grove Village, Ill) that parents should employ methods of child discipline other than corporal punishment.25 The fact that these findings differ from Eamon’s11 are likely due to the fact that the statistical analysis used in this study employed stronger statistical controls as well as the fact that this study included a wider age range of children than those included in Eamon’s11 study.
The results of the fixed-effects regressions suggest that selection bias may account for some of the interaction between neighborhood quality and corporal punishment evidenced in previous research and provide stronger evidence of the effects of corporal punishment on increases in children’s externalizing behavior problems than previous research. There is, however, an important limitation to this research. Because this research used secondary data from the NLSY, the results of this paper rely on the fact that the NLSY allows survey respondents to decide on what they define as “spanking.”
This research stands in a growing tradition of empirical research that suggests that corporal punishment may not be an effective way to reduce behavior problems in children. Indeed, the results of this analysis, like other analyses, suggest that corporal punishment is likely to be associated with increases in children’s behavior problems even when stronger statistical controls are employed. More research is also needed on the social and family antecedents of parental use of corporal punishment and the degree to which parenting practices interact with larger social forces and outcomes for children’s well-being.
Correspondence: Andrew Grogan-Kaylor, PhD, School of Social Work, 1080 S University Ave, University of Michigan, Ann Arbor, MI 48109 (firstname.lastname@example.org).
Additional Information: I had full access to all the data in the study, and I take responsibility for the integrity of the data and the accuracy of the data analysis.
Accepted for Publication: May 25, 2005.
Acknowledgment: I thank Siri Jayaratne, PhD; Daniel Saunders, PhD; Sherrie Kossoudji, PhD; and Carl Lipo, PhD, for their advice and comments, which improved the quality of the article.
Thank you for submitting a comment on this article. It will be reviewed by JAMA Pediatrics editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 5
Customize your page view by dragging & repositioning the boxes below.
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.