Criminal Justice and Behavior-2011-Morris-818-39.pdf

http://cjb.sagepub.com/Behavior

Criminal Justice and

http://cjb.sagepub.com/content/38/8/818The online version of this article can be found at:

DOI: 10.1177/0093854811406070

2011 38: 818 originally published online 17 May 2011Criminal Justice and BehaviorSara Z. Morris and Chris L. Gibson

Corporal Punishment's Influence on Children's Aggressive and Delinquent Behavior

Published by:

http://www.sagepublications.com

On behalf of:

International Association for Correctional and Forensic Psychology

can be found at:Criminal Justice and BehaviorAdditional services and information for

http://cjb.sagepub.com/cgi/alertsEmail Alerts:

http://cjb.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://cjb.sagepub.com/content/38/8/818.refs.htmlCitations:

What is This?

- May 17, 2011 OnlineFirst Version of Record

- Jun 23, 2011Version of Record >>

by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from by guest on November 5, 2013cjb.sagepub.comDownloaded from

http://cjb.sagepub.com/

http://cjb.sagepub.com/content/38/8/818

http://www.sagepublications.com

http://www.ia4cfp.org

http://cjb.sagepub.com/cgi/alerts

http://cjb.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://cjb.sagepub.com/content/38/8/818.refs.html

http://cjb.sagepub.com/content/38/8/818.full.pdf

http://cjb.sagepub.com/content/early/2011/05/17/0093854811406070.full.pdf

http://online.sagepub.com/site/sphelp/vorhelp.xhtml
























818

CRIMINAL JUSTICE AND BEHAVIOR, Vol. XX, No. X, Month 2007 818-XXXDOI: © 2007 American Association for Correctional and Forensic Psychology

CRIMINAL JUSTICE AND BEHAVIOR, Vol. 38 No. 8, August 2011 818-839DOI: 10.1177/0093854811406070© 2011 International Association for Correctional and Forensic Psychology

AUTHORS’ NOTE: Please direct all correspondence to Sara Z. Morris, University of Georgia, 420 Baldwin Hall, Athens, GA 30602; email: [email protected].

CORPORAL PUNISHMENT’S INFLUENCE ON CHILDREN’S AGGRESSIVE AND DELINQUENT BEHAVIORSARA Z. MORRISUniversity of Georgia

CHRIS L. GIBSONUniversity of Florida

Studies show that children subjected to corporal punishment may engage in more aggression and delinquent behaviors than those who are not. Past research, however, is limited methodologically. This is largely the result of a lack of matched corpo-rally punished and nonpunished children. To address this limitation, a propensity score matching analysis was used to esti-mate the effects of corporal punishment on children’s behaviors. Using data from the longitudinal study of the Project on Human Development in Chicago Neighborhoods, findings indicate that (a) a large amount of selection bias exists, indicating that child and family characteristics of those subjected to corporal punishment are substantially different from characteristics of those not punished, and (b) when children exposed to corporal punishment (vs. those who are not) are matched on their propensities of being punished, the relationship between punishment and subsequent aggression and delinquency become statistically nonsignificant and substantively small. Findings are discussed in light of past research on corporal punishment, and limitations of the current study and ways of overcoming them in the future are discussed.

Keywords: corporal punishment; propensity score matching; delinquency; aggression; PHDCN

In the United States, the use of corporal punishment, defined as “physical force with the intention of causing a child to experience pain, but not injury, for the purpose of correc-

tion or control of the child’s behavior” (Straus, 1994, p. 4), is hotly debated. Views range from a belief that corporal punishment is normal and acceptable to the belief that any type of physical punishment is abuse. Regardless of these views, 94% of parents use some type of corporal punishment to discipline their toddlers aged 4 to 5, and just more than 40% of parents still use corporal punishment to discipline their children as old as age 13 (Straus & Stewart, 1999). Even as early as 6 months old, a substantial number of caregivers report using corporal punishment on children (Slade & Wissow, 2004).

Since a substantial number of parents use corporal punishment and feel that it is an appropriate way to discipline their children, it is important to understand both the inten-tional and unintentional effects it may have on children. Studies show that several adverse outcomes are a result of corporal punishment—for example, antisocial behavior such as aggression and delinquency (Brennar & Fox, 1998; Grogan-Kaylor, 2005a, 2005b; Simons, Wu, Lin, Gordon, & Conger, 2000), depressive symptoms (Straus & Kantor, 1994; H. A. Turner & Muller, 2004), suicidal ideation (Straus & Kantor, 1994), and psychological dis-tress (H. A. Turner & Finkelhor, 1996).

Given that much research on corporal punishment often uses nonexperimental data (e.g., Grogan-Kaylor, 2005a, 2005b; Simons et al., 2000; Straus & Kantor, 1994), it is possible

Morris, Gibson / CORPORAL PUNISHMENT 819

that those who are punished are substantively different from those who are not on many family, demographic, and individual characteristics. Of course, random assignment to being corporally punished would help ensure that no systematic differences exist between the two groups of children, but an experiment of this type is unethical, impractical, and unachiev-able. Therefore, the dilemma of estimating effects of corporal punishment on children’s behavioral outcomes becomes evident; statistically equivalent groups must be formed.

To address this problem the current study achieves matched samples of corporally pun-ished and nonpunished children using a propensity score matching (PSM) analysis. This method allows children with similar propensities of being punished, with some having been punished and others not, to be matched on many covariates and then compared on future aggression and delinquency outcomes. This approach also improves on more standard regression analyses by correcting for selection bias based on observed characteristics before assessing the association between corporal punishment and later behavior problems. Such an approach is often used on nonexperimental data for estimating the counterfactual question: What would have been the aggression and delinquency outcomes of corporally punished children had they not been punished?

Data for the current study are drawn from the Project on Human Development in Chicago Neighborhoods (PHDCN) longitudinal study to answer the following question: Does corporal punishment have an effect on children and adolescent’s future aggressive and delinquent behaviors? We proceed by reviewing past research on the association between corporal punishment and children’s externalizing behaviors as well as research on factors that increase the likelihood of corporal punishment being used on children. Before presenting our results, we discuss the utility of PSM for addressing our research question.

CORPORAL PUNISHMENT’S ASSOCIATION WITH EXTERNALIZING BEHAVIORS

Scholars suggest that using violence to discipline children may teach them that aggres-sive and delinquent behaviors are the correct way to reach their desired goals and can lead to an increased occurrence of these behaviors (Straus, 1996). For instance, McCord (2005) suggests that children begin to learn that physical violence and causing physical pain are acceptable ways of acting under some circumstances. Children who are spanked, or more generally have harsh physical punishment short of injury used on them as a form of disci-pline, may internalize the belief that aggression and antisocial behaviors are appropriate ways to deal with conflict, thus in turn leading to increased levels of aggression and delin-quency.

Two popular viewpoints on the effectiveness and appropriateness of corporal punish-ment are often drawn from the research literature. First, many studies find associations between the use of physical punishment and antisocial behavior (Berlin et al., 2009; McKee et al., 2007). Second, other studies that either focus on the use of alternative disciplinary tactics in conjunction with spanking or distinguish between mild and harsh punishment find that these factors determine whether or not spanking is associated with negative behavioral outcomes (Lansford et al., 2009; Larzelere, 2000).

Recently, three meta-analyses have summarized the findings on the relationship between corporal punishment and behavioral problems. Larzelere and Kuhn (2005) reviewed the

820 CRIMINAL JUSTICE AND BEHAVIOR

effect sizes of 26 past studies on corporal punishment. Findings showed, in contrast to many others (e.g., Grogan-Kaylor, 2005a; Straus & Kantor, 1994; Straus & Mouradian, 1998; H. A. Turner & Muller, 2004), that customary corporal punishment does not have a significant effect on behavior problems. They argue that in many studies more severe abuse (punishment that could be classified as child abuse) is combined together with customary punishment (i.e., punishment intended to change behavior and cause pain without injury), indicating a stronger effect than what is actually present. According to these authors, cor-poral punishment had more adverse consequences when it was used more severely or as the primary method of punishment.

Gershoff (2002) also conducted a meta-analysis that included 88 studies investigating the association between corporal punishment and childhood “behavioral experiences,” a term used instead of “outcomes” to indicate that the majority of studies lack at least one criterion needed to show causality (i.e., lack of time ordering). Results show that corporal punishment is related to a large number of negative behaviors. These include increased antisocial behavior, increased risk of being a victim of child abuse, increased risk of being a perpetrator of abuse as an adult, increased aggression, and decreased quality of parent–child relationships. These results were present even though she specifically did not include studies that aggregate together more serious abuse with more common corporal punish-ment, the caution given by other researchers (Larzelere & Kuhn, 2005). The only desirable outcome shown to be associated with corporal punishment was immediate compliance, but even this was inconsistent across studies.

Last, in an analysis of effect sizes taken from 70 studies, Paolucci and Violato (2004) showed that those children who experienced corporal punishment are at a very small increased risk for emotional or behavioral problems and no increased risk for cognitive problems. However, because of insufficient data in many of the studies they used, they were unable to assess effect sizes based on important control variables such as age, gender, and quality of parent–child relationship.

The relationship studied most often in individual-level studies is that between corporal punishment and antisocial behavior. Using measures of antisocial behavior drawn from the Behavior Problems Index and the Behavior Problems Checklist, there is evidence indicat-ing that children who are physically punished by their primary caregivers exhibit increases in antisocial behavior. This relationship is observed even when adjustments are made for previous behavior problems using longitudinal data (McLoyd & Smith, 2002; Simons et al., 2000; Slade & Wissow, 2004). In addition, a variety of measures of corporal punishment have been used, indicating that these findings are not unique to one specific measure of punishment (Grogan-Kaylor, 2004, 2005b; Simons et al., 2000; Straus & Mouradian, 1998; Straus, Sugarman, & Giles-Sims, 1997).

Although the studies noted above, as well as others (Grogan-Kaylor, 2005b; Simons et al., 2000), have described the negative outcomes associated with corporal punishment, many of them share a similar methodological limitation: Most are unable to discern the direction of causality. Are children who exhibit more behavioral problems and aggressive temperaments among other characteristics more likely to experience corporal punishment? If so, can this explain why studies observe that corporal punishment is correlated with negative behavioral outcomes of children? For instance, it may be that children who have corporal punishment used on them are more at risk for engaging in externalizing behavior problems, rendering the relationship between corporal punishment and behavior problems


spurious. Most studies are unable to clearly explain this relationship because of several reasons. These include having only nonexperimental data, using recall data, and not being able to establish time ordering in analyses (Grogan-Kaylor, 2005a, 2005b; Molnar, Buka, Brennan, Holton, & Earls, 2003; Rodriguez, 2003; Simons et al., 2000; Straus & Kantor, 1994; H. A. Turner & Muller, 2004; Wissow, 2001). The current study attempts to over-come some of these past limitations.

CORPORAL PUNISHMENT AND PROPENSITY SCORE MATCHING (PSM)

Although many studies, including some meta-analyses, conclude that corporal punish-ment is related to problem behaviors among children, it would be premature to reach the conclusion that corporal punishment leads to or causes subsequent negative behavioral outcomes for several reasons. First, results from several meta-analyses do not consistently confirm or deny that there is an association between corporal punishment and antisocial behaviors. Second, although several studies have used longitudinal data with adjustments for prior antisocial and externalizing behaviors, selection bias is still a limitation. That is, none have matched children who were corporally punished to those not punished on many key background characteristics so that two similar groups could be compared on future behavioral outcomes. Most studies have used some form of regression analysis to statisti-cally control for other influences that could perhaps render the relationship between corpo-ral punishment and behavioral problems spurious (Grogan-Kaylor, 2005a, 2005b; Simons et al., 2000). Although statistical control has been the traditional method used to account for selection bias, it may be inadequate for reasons that are explained later.

Studies that assessed the association between corporal punishment and children’s behav-iors have used nonexperimental data, which make it difficult to know whether children who are exposed to punishment are similar to those who are not punished, even when statistical controls are accounted for in regression models. In such studies respondents are not ran-domly assigned to be punished, which results in systematic differences between respon-dents where some are more likely to be punished than others; thus, children who are punished can be quite different from those who are not across several domains (e.g., family conditions, parental characteristics, behavior, personality, demographics, etc.). If differ-ences are not identified and addressed statistically, systematic bias can produce inaccurate conclusions.

Although regression has been a commonly used method for dealing with selection bias in corporal punishment studies, other strategies exist that may have some advantages over regression methods, one being PSM. Although PSM is not a perfect methodology by any means, we believe it is an appropriate method to apply to the study of corporal punishment and can be used to address some limitations of regression analysis. First, regression models rely on the assumption of linearity between the covariates and outcomes. If this relationship is not present, or if the distributions of the treatment and control groups are very different, then the results estimated by a regression model can be unreliable.

Second, regression models are inadequate for taking into account the distribution of covariates across the treatment and control groups. For example, punished (treatment) ver-sus not punished (control) groups may differ in their mean levels on covariates, and the distribution of those covariates could overlap very little between the two groups. According


to Foster (2003), “In that case, regression essentially projects the behavior of individuals in one group outside the observed range to form a comparison for the other at common values of the covariate” (p. 1185). Unlike regression, with PSM each respondent is matched based on his or her conditional probability of experiencing the treatment condition, and if there is not an adequate match the respondent is dropped (Newgard, Hedges, Arthur, & Mullins, 2004). In a PSM framework, only individuals with adequate matches are compared with one another.

Third, regression techniques are often limited by the number of covariates that can be included in a model as well as potential multicollinearity issues among covariates. Compared to regression methods, PSM allows for a larger number of covariates to be con-sidered while still being parsimonious because the effects are combined into one score (i.e., conditional probability). This large number of covariates can be included without danger of multicollinearity problems found in regression models (Hahs-Vaughn & Onwuegbuzie, 2006). Multicollinearity can be a problem in regression analysis if the intent is to estimate how each independent variable affects the dependent variable separately. Covariates are used in PSM only to create the propensity score, eliminating multicollinearity as a potential problem (Newgard et al., 2004).

Fixed effects models represent another method that can address some shortcomings of regression. However, this method has its own limitations that can be partially overcome with PSM as well. For example, although fixed effects models can account for unobserved characteristics, they can do so only for those that are “fixed” and do not change over time. Fixed-effect models are most appropriate for data having large amounts of variation and measurements taken at many different time points (Hill, Waldfogel, Brooks-Gunn, & Han, 2005). Another disadvantage of fixed-effect models is that they cannot estimate separate coefficients for time-invariant characteristics such as gender or ethnicity (Grogan-Kaylor, 2004). PSM shares some of the limitations of these models, in that it can account for only observed differences, and once the propensity score is estimated the effects of each covari-ate cannot be disentangled from one another. It improves on them, however, because the covariates can either be “fixed” or change over time. As such, the current study uses PSM to estimate the influence of corporal punishment on aggression and delinquency outcomes.

Although not a magic solution, PSM is one way to account for selection bias when ran-dom assignment is impossible because of ethical constraints. Since children cannot be randomly assigned to receive corporal punishment, one way to account for selection bias is to match corporally punished children to non–corporally punished children across many observed variables prior to estimating an effect of corporal punishment on subsequent behaviors. This can eliminate the influence of observed confounders so that individuals are homogeneous, with the exception that one group is exposed to a treatment, corporal punish-ment, and others not (Rosenbaum & Rubin, 1983). To our knowledge, studies that have been conducted on the relationship between corporal punishment and behavioral outcomes of children have not used matching procedures, which puts the current study in a position to shed new light on this research topic.

Prior to estimating the influence of corporal punishment, several steps are followed in PSM. First, the influences of observed covariates on the likelihood of receiving corporal punishment are combined together into a conditional probability ranging between 0 and 1, typically estimated using a logistic regression. In the current study, the conditional proba-bilities represent the propensity for being corporally punished, with 1 being 100% chance


and 0 reflecting no chance at all. Once propensity scores are calculated, matching algo-rithms are used to pair corporally punished children to children who have similar probabil-ities of being punished but who have not actually received corporal punishment.

Accuracy of the estimation of effects in propensity models is contingent on the quality of matching procedures used. We use both nearest-neighbor and kernel matching to assess robustness of our results. These two methods differ in the way matches are identified—nearest-neighbor matching relies on the absolute closest match, whereas kernel matching involves a weighting process that takes into account the distance between propensity scores (Sianesi, 2001). Matching processes such as these are performed before estimating the effect of corporal punishment on children’s subsequent aggressive and delinquent behaviors.

Despite the numerous advantages of using PSM, we acknowledge the limitations of using this technique. First, PSM is only as good as the covariates used to create the propen-sity scores. Unlike random assignment in a classical experiment, propensity score analysis ensures that there will be no systematic differences between treated and untreated groups for observed covariates (Rubin, 1997). Therefore, it is important to include any covariate that may be related to corporal punishment. As stated by Rubin and Thomas (1996), “Unless a variable can be excluded because there is consensus that it is unrelated to out-come or is not a proper covariate, it is advisable to include it in the propensity score model even if it is not statistically significant” (p. 253). Second, propensity score analysis per-forms less optimally with smaller samples (Rubin, 1997). Third, including irrelevant covariates may reduce model efficiency; however, this possibility diminishes as the sample size becomes larger (Rubin, 1997). In light of these limitations, PSM is still considered the optimal alternative solution when estimating causal effects using observational data (Dehejia & Wahba, 2002; Rosenbaum & Rubin, 1983; Rubin, 1997).

METHOD

DATA

Data for the current study are drawn from Waves 1 and 2 of the longitudinal portion of the PHDCN; although Wave 3 data would have also been useful for this project, they were not available to the authors. The PHDCN is an interdisciplinary study that was initiated to investigate the contextualized pathways by which children develop both positive and nega-tive behaviors. For the longitudinal portion of the study, a stratified probability sample was selected from 343 neighborhood clusters to achieve a representative sample of 80 neighbor-hood clusters that reflected the racial, ethnic, and socioeconomic conditions of the original clusters. The longitudinal portion of the PHDCN is fulfilled by the existence of three dif-ferent waves of data collection on multiple children and adolescent cohorts residing in the 80 sampled neighborhood clusters, each collected approximately 2.5 years apart from one another (Sampson, 1997; Sampson, Morenoff, & Raudenbush, 2005). Data collection for the longitudinal portion began in 1994–1995, the second wave of collection occurred between 1997 and 1999, and the third wave of data was collected between 2000 and 2001. Data were collected from seven different cohorts of children, approximately ages birth (0), 3, 6, 9, 12, 15, and 18, and their primary caregivers. These data were collected by in-home interviews of respondents and their caregivers. The current study uses data collected during


Waves 1 and 2 for the 6-, 9-, 12-, and 15-year-old cohorts. The original sample included 1,987 respondents; the analysis sample for the current study uses the 1,346 respondents who have valid information on the variables used in the current study.

As shown in Table 1, the analysis sample is 50% male and 50% female and has an aver-age age of 10 years old. The majority of respondents (86%) are cared for primarily by their mother, and 66% of the primary caregivers are married. The sample is 18% White, 48% Hispanic, 31% Black, and 3% “Other” ethnicity. The average family size is five members, and the average number of siblings is two. Of the sample, 26% received public assistance in the past year. Approximately 68% of the sample has experienced corporal punishment at least once in the past year. Although this prevalence rate is lower than estimates in other studies, this is likely the result of the age of our sample. Toddlers (those aged 2 to 5) are most likely to be spanked; the youngest child in our sample is 5 years old. Within the 6-year-old cohort, 79% had been spanked, whereas only 52% of the 15-year-old cohort had been spanked. Overall, of these children who experienced corporal punishment, 55% had been “slapped or spanked,” 41% had been “pushed or grabbed,” and 12% “had something thrown at them.” Approximately 9% of the sample experienced more than one of these acts.

Comparisons were made between the original and analysis samples across several vari-ables to assess if any differences were apparent. The distributions of demographic character-istics were very similar in both. For instance, 49% of the original sample is male, and the average age is 10 years old. In addition, 55% of the primary caregivers are married, and 84% are the mother of the child. Furthermore, 14% of the sample is White, 47% Hispanic, 37% Black, and 3% “Other” ethnicity. Regarding public assistance, 29% of these children and their families have received it in the past year. The percentage of the original sample that experienced each act described by the corporal punishment scale is almost identical to those found in the analysis sample. Specifically, 11% had something thrown at them at least once, and 12% had experienced this at least once. In addition, 39% of the full sample had been pushed or grabbed at least once, and 41% of the analysis sample experienced this at least once. Finally, 60% had been slapped, whereas 55% of the analysis sample experienced this at least once. Overall, comparisons of the two samples are very similar on important vari-ables, suggesting that the sample used in the current study is reflective of the larger sample.

VARIABLES

Dependent variables: Aggression and delinquency. Subscales from the externalizing behaviors scale of the Reduced Child-Behavioral Checklist (CBCL) from Wave 2 were used as outcomes, one measuring delinquent behavior and the other aggressive behavior (Achenbach, 1991). Generally, these subscales assess rule-breaking behaviors and other behaviors that are associated with disorders such as conduct disorder and oppositional defi-ant disorder (Sourander & Helstela, 2005). More specifically, the aggressive behavior subscale refers to actions termed hostile or destructive, and the delinquent behavior sub-scale is similar but includes more serious offenses.

Questions making up each subscale have response categories of 0, 1, or 2, indicating the specified behavior is not true, sometimes true, or often true of the child, respectively. These scales consist of summated responses from primary caregivers, where higher scores on each measure indicate more delinquency and aggression. The aggressive behavior scale contains 13 items and the delinquent behavior scale has 11 items. Cronbach’s reliability


coefficients for the aggressive behavior and delinquency scales are .87 and .68, respec-tively. These measures are taken from Wave 2, which was gathered approximately 2 years after the Wave 1 measure of corporal punishment.

Treatment variable: Corporal punishment. As in past research (Kantor & Jasinski, 1997; Straus et al., 1997; Straus, Hamby, Finkelhor, Moore, & Runyan, 1998), the treatment con-dition was measured using three items from the Conflict Tactics Scale for Parent-Child

TABLE 1: Descriptive Statistics

M SD Min Max

Aggression 5.65 4.97 0.00 26.00Delinquent 2.75 2.84 0.00 17.00Corporal punishment 0.68 0.47 0.00 1.00CovariatesDemographics Respondent age 10.04 3.19 5.04 16.37 Respondent gender (1 = male) 0.50 0.50 0.00 1.00 White 0.18 0.38 0.00 1.00 Black 0.31 0.46 0.00 1.00 Hispanic 0.48 0.50 0.00 1.00 Other 0.03 0.18 0.00 1.00 SES 0.08 1.44 –2.92 3.52 Family size 5.33 1.87 2.00 14.00 Number of siblings younger than 19 2.18 1.56 0.00 10.00 Public assistance 0.26 0.44 0.00 1.00Primary caregiver (PC) characteristics PC age 36.36 7.12 20.70 67.95 PC gender (1 = male) 0.09 0.29 0.00 1.00 PC less than high school 0.21 0.41 0.00 1.00 PC some high school 0.21 0.41 0.00 1.00 PC graduated high school 0.13 0.33 0.00 1.00 PC education past high school 0.46 0.50 0.00 1.00 PC mom 0.86 0.35 0.00 1.00 PC dad 0.09 0.28 0.00 1.00 PC other female 0.05 0.21 0.00 1.00 PC other male 0.01 0.07 0.00 1.00 PC married (1 = yes) 0.66 0.48 0.00 1.00 PC employed (1 = yes) 0.61 0.49 0.00 1.00Supervision or monitoring 8.38 0.89 3.00 9.00 Warmth 6.95 1.92 0.00 9.00 PC verbal skills 3.89 0.44 0.00 4.00 Provision of social relations 20.55 4.36 15.00 38.00 Conflict Tactics Scale 2.24 5.30 0.00 46.00 Conflict Tactics Scale—spouse 1.94 4.51 0.00 48.00Child characteristics Wave 1 internalizing 7.36 6.64 0.00 45.00 Wave 1 externalizing 11.16 8.78 0.00 54.00 Impulsivity 54.50 11.48 25.00 93.00 School help for emotional problems 0.03 0.18 0.00 1.00 Truant past year 0.04 0.20 0.00 1.00Family characteristics Family member with criminal record 0.30 0.46 0.00 1.00 Family member with nerve problems 0.19 0.39 0.00 1.00 Family member with legal problems 0.17 0.38 0.00 1.00 Family member attempt suicide 0.13 0.34 0.00 1.00


(CTSPC; Kantor & Jasinski, 1997). The full scale measures how conflict is dealt with between caregivers and their children, ranging from “talking things through calmly,” to spanking, to more severe actions. An index was created from three of the questions on the CTSPC concerning corporal punishment. The questions ask primary caregivers how many times in the past year they “threw something at him/her,” “pushed, grabbed, or shoved him/her,” and “slapped or spanked him/her.” Straus et al. (1998) found that this measure of corporal punishment had adequate construct validity; in accordance with a large amount of past research it was correlated in the expected direction based on age and gender of parent, age of child, and ethnicity. For each item, responses were collapsed into a dichotomy, with 1 indicating at least one time in the past year and 0 indicating none in the last year, and were then summed. Those that have a score of 1, 2, or 3 are collapsed into the treatment group, as they were exposed to corporal punishment. Those having a score of 0 are in the control group, indicating no exposure to corporal punishment. Although two of the items on this scale may be interpreted by some as harsher than normative corporal punishment, our goal was to study this construct with this specific measure because of its wide use by those conducting research on corporal punishment.

Approximately 68% of primary caregivers reported using at least one form of corporal punishment on their child within the last year. Furthermore, of those children punished (treated group), punishment was not an infrequent occurrence. In fact, for those in the treated group, approximately 44% had been slapped by a primary caregiver between 3 and more than 20 times in the past year, and approximately 31% had been pushed or grabbed by a primary caregiver between 3 and more than 20 times in the past year. Throwing something at a child was less frequent, with approximately 6% of the punished children having something thrown at them by a primary caregiver between 3 and more than 20 times in the past year.

Covariates: Variables used to create propensity of corporal punishment. Several charac-teristics increase the likelihood that children will receive corporal punishment from care-givers, including demographic, parenting, behavioral, and family factors. Studies show that younger children and males are more likely to experience corporal punishment (Day, Peterson, & McCracken, 1998; Dietz, 2000; Giles-Sims, Straus, & Sugarman, 1995; Grogan-Kaylor & Otis, 2007). African American children are more likely to receive corpo-ral punishment (Dietz, 2000; Grogan-Kaylor & Otis, 2007), and parents who have less education and lower incomes are also more likely to physically punish their children (Day et al., 1998; Dietz, 2000). Unmarried mothers and those who identified themselves as Protestants are also more likely to spank (Giles-Sims et al., 1995; Grogan-Kaylor & Otis, 2007). Last, several characteristics of the family and the parent–child relations (Grogan-Kaylor, 2004, 2005b; Simons et al., 2000; Straus et al., 1997; Straus & Mouradian, 1998; i.e., how much they argued, household size, the amount of social support available) also increase the likelihood of receiving corporal punishment (Day et al., 1998). These corre-lates, along with prior behavioral measures and personality characteristics of children that are related to punishment and future misbehavior, are included as covariates in the PSM analysis. All measures of covariates were taken from Wave 1 assessment.

Demographics. Demographic variables of respondents were included because research shows that the use of corporal punishment varies across gender, race, and socioeconomic status (SES; Day et al., 1998; Dietz, 2000; Giles-Sims et al., 1995; Grogan-Kaylor & Otis,


2007). Sex of the respondent is coded 0 (female) or 1 (male). Race was coded as four dummy variables indicating White, Black, Hispanic, or Other, with White being the refer-ence category. SES of the child’s family is coded on a continuous scale, with higher scores indicating higher family SES. This variable was created using the principal component of three variables: household income, maximum education level of primary caregiver and partner, and socioeconomic index for primary caregiver’s and partner’s jobs (see Gibson, Sullivan, Jones, & Piquero, 2010).

Primary caregiver characteristics. Primary caregiver characteristics include age, sex, education, employment status, and relationship status. Gender of the primary caregiver is coded as 0 (female) or 1 (male). Education of primary caregiver was measured using a set of dummy variables, indicating whether or not the primary caregiver had less than a high school education, had finished some high school, had finished high school, or had educa-tion past high school; the reference category is “less than high school education.” Employment status of the primary caregiver was recoded from three categories (employed, unemployed for fewer than 5 years, and unemployed for more than 5 years) to a dichoto-mous variable indicating 0 (unemployed) or 1 (employed). Relationship status of the pri-mary caregiver to the respondent was recoded into a series of four dummy variables. These categories were “mother is primary caregiver,” “other female,” “father,” and “other male.” The reference category of “mother is primary caregiver” was excluded when estimating propensity scores for comparison purposes.

The physical assault scale from the Conflict Tactics Scale for Partner/Spouse is also included as a predictor of corporal punishment. There could also be behavioral conse-quences for a child who witnesses violence in his or her home, so it is an important control. These questions address violence between partners ranging from hitting to more serious violence such as firing a gun or using a knife. Questions include how many times the respondent had “thrown something at him/her,” “pushed or shoved him/her,” and “beat him/her up.” Each question has response categories of 0 (never), 1 (once), 2 (twice), 3 (3–5 times), 4 (6–10 times), 5 (11–20 times), and 6 (> 20 times; Straus, 1979). Each question first asks the primary caregiver how many times he or she has exhibited a particular behavior in the past year, then asks how many times his or her spouse has. Two scales were created from these questions, one assessing the primary caregiver’s behavior and the other assess-ing the spouse’s. This scale has a reliability coefficient of .86 for the primary caregiver and .85 for items about the caregiver’s partner.

Parenting and family variables are also used as predictors, including measures of paren-tal warmth, parental verbal skills, and supervision and monitoring from the Home Observation for Measurement of the Environment (HOME) instrument (Leventhal, Selner-O’Hagan, Brooks-Gunn, Bingenheimer, & Earls, 2004), whether or not the family is pro-vided with public assistance, the size of household, the number of siblings, and the amount of social support available.

The HOME instrument is used to assess aspects of developmental conditions and actions by caregivers that affect a child’s well-being (Leventhal et al., 2004). Three subscales are used from this instrument. The parental warmth index consists of observational items coded dichotomously (0 = interviewer did not witness it, 1 = he or she did) such as “parent encour-ages child to contribute” and “parent praises child twice during visit.” Higher scores on this index indicate more warmth. The parental verbal skills index is made up of questions that


assess the verbal and communication skills of the primary caregiver. Higher scores indicate more verbal and communication skills. Parental warmth and verbal skills items were mea-sured and coded by interviewers during in-home interviews. The supervision and monitor-ing measure contains 13 questions asked to the primary caregiver that relate to how well he or she supervises the respondent, such as “subject has a set time (curfew) to be home on school nights” and “establishes rules for behavior with peers and asks questions to deter-mine whether or not they are being followed.” Although this scale’s internal consistency is low, the face validity of this scale indicates that all questions pertain to supervision and monitoring of the child. Based on this, the scale is still included in the model. For the supervision/monitoring scale, higher scores indicate more supervision.

Traits and behaviors. Trait and behavioral characteristics of respondents include impul-sivity and externalizing behaviors. The Emotionality, Activity, Sociability, Impulsivity temperament instrument (EASI), administered to primary caregivers regarding their chil-dren, was designed to measure various personality traits in children (Buss & Plomin, 1984). The impulsivity scale from the EASI is used in the current study, which is made up of four different subscales measuring inhibitory control, sensation seeking behavior, decision time, and persistence. Buss and Plomin (1984) refer to impulsivity as a child’s tendency to resist urges rather than giving in to them quickly. Inhibitory control refers to the ability to delay an immediate response that arises or to resist temptations. Sensation-seeking behavior is characterized by unpredictability and the desire to constantly try new things and have new experiences. Decision time manifests itself in an impulsive person by quick actions without contemplating consequences or other possible choices. Persistence refers to the desire to seek novelty as soon as an activity becomes boring or dull. Examples of questions from this scale include “I have trouble controlling my impulses” (inhibitory control), “I often say the first thing that comes into my head” (decision time), “I’ll try anything once” (sensation seeking), and “I tend to give up easily” (persistence). The scale consists of 20 items and response categories include 1 (uncharacteristic of my child), 2 (somewhat uncharacteris-tic), 3 (neither), 4 (somewhat characteristic), and 5 (characteristic; Buss & Plomin, 1984). Responses to items were summed for each individual, with higher scores indicating more impulsivity. The impulsivity scale has an alpha coefficient of .76.

Children’s prior levels of behavior were measured using the internalizing and external-izing behavior scales from the CBCL at Wave 1 to adequately control for past negative behaviors and psychological conditions, which includes both the delinquency and aggres-sion subscales from the externalizing scale and depression/anxiety, somatic complaints, and withdrawal from the internalizing scale (Achenbach, 1991). It is expected that kids who misbehave and act out are more likely to be subjected to corporal punishment. Responses to items for the externalizing scale were summed to create a measure that has a reliability coefficient of .89, and similarly the internalizing measure has a reliability coefficient of .89. Higher scores on both measures indicate more externalizing and internalizing behaviors.

Other behavioral indicators that may put children at risk for corporal punishment were included such as the number of times the child skipped school or class, whether or not the child has had special education for behavioral or emotional problems, and how many days the child has missed school or work because of mental or physical health. In addition, a dichotomous question was included asking whether or not the child has ever had special education for behavioral or emotional problems. This was coded as 0 (no) or 1 (yes).


Family characteristics. Several other covariates were included that measure family characteristics including whether or not the respondent has a family member with a criminal record, a family member with frequent legal problems, a family member with nerve problems, or a family member who has attempted suicide. Each of these items were coded 0 (no) or 1 (yes). Public assistance is measured dichotomously by a question asking the primary caregiver if the family has received public assistance in the past year. Both family size and number of siblings younger than 19 are included as covariates. Both family size and number of siblings are measured continuously, with higher numbers indicating more siblings or more family members in the household.

The Provision of Social Relations Scale is used to assess the amount of social support available to the primary caregiver from family and friends and was designed to measure how much support one receives from others (R. J. Turner, Frankel, & Levin, 1983). There are two different subscales, one on family support and the other on friend support. These are designed to be combined together to form a measure of overall social support. Respondents answer 1 (very true), 2 (somewhat true), or 3 (not true) to questions such as “I have at least one friend that I could tell anything to” and “No matter what happens, I know that my family will always be there for me should I need them.” Responses to all questions on both subscales are summated to create a scale score for each individual, with higher scores indicating a lack of (or less) support (R. J. Turner et al., 1983). The reliability coefficient for this scale is .73.

ANALYSIS

To estimate the influence of exposure to corporal punishment on aggressive and delin-quent behaviors, we proceeded as follows. Stata 9.0 was used to conduct all analyses, specifically the PSMATCH2 program within Stata (Leuven & Sianesi, 2003). First, a series of independent samples t tests were estimated to assess whether significant differ-ences on covariates existed for the corporally punished and nonpunished groups prior to matching. As such, this would indicate whether the group that had received corporal punishment was significantly different from the group not punished on each covariate discussed earlier. Second, although not reported, a logistic regression was estimated to obtain propensity scores or conditional probabilities for being corporally punished. Third, a nearest-neighbor matching algorithm was used to pair treated and untreated groups on their propensity scores. Fourth, we assessed whether the matching procedure was successful by re-estimating independent samples t tests for the matched sample. If successful, significant difference should no longer exist between the two groups on covariates and the mean differences for each covariate should be substantially reduced for the two groups. Then, the influence of corporal punishment on aggression and delin-quency was estimated using the matched sample. Lastly, we confirmed results from our matching analysis by using a stratification approach where individuals were grouped into four categories with similar propensities to be punished but some were and some were not actually exposed to corporal punishment. This method allowed us to retain all respon-dents, as opposed to corporally punished respondents being dropped for not having an adequate match.


RESULTS

PRE- AND POSTMATCHING T TESTS

Results from prematching t tests are shown in Table 2. Several covariates exhibited sta-tistically significant differences between the treated (exposed to corporal punishment) and untreated (not exposed to corporal punishment) groups. Children subjected to corporal punishment were significantly younger, were more likely to be on public assistance, had a male primary caregiver, had primary caregivers with higher scores on the provision of social relations scale, and had primary caregivers with more verbal skills. In addition, cor-porally punished children were less likely to have “other female” defined as their primary caregiver than those in the control group. A statistically significant difference was observed for age of primary caregiver. The caregivers of children subjected to corporal punishment were approximately 3 years younger than caregivers of those not subjected to corporal punishment. The scores on the Conflict Tactics Scale for primary caregivers, as well as those of their spouses, were significantly higher for children receiving corporal punish-ment, indicating that children who were punished had more conflict in the home than their counterparts. Corporally punished children had much higher scores on internalizing and externalizing behavior scales at Wave 1, were more impulsive, and were less likely to have “dad” defined as their primary caregiver than those who were not subjected to corporal punishment. Finally, those who were corporally punished were significantly more likely to have received help in school for emotional problems and to have had a family member with frequent legal problems, with nerve problems, with a criminal record, and who had attempted suicide.

Results discussed thus far indicate that a sizeable amount of selection bias exists between children subjected to corporal punishment and those who were not. On the other hand, some covariates did not show statistically significant differences between the two groups, including ethnicity, SES, family size, number of siblings, education, “other male” defined as primary caregiver, primary caregiver employment and marital status, supervision/monitoring scale, parental warmth scale, and truancy.

Next, we estimated propensity scores using a logistic regression (results can be provided on request). Logistic regression creates a conditional probability of experiencing corporal punishment, as a function of all covariates discussed earlier. Previous research has shown that many of these covariates have an impact on whether or not a child experiences corpo-ral punishment. Although the number of factors investigated is large, even if a covariate is only weakly related to exposure to corporal punishment, the weakness of this relationship is overcome by the reduction in bias it provides (Newgard et al., 2004). After investigating common support and confirming overlap in propensity scores across groups, a matching procedure was performed using nearest neighbor one-to-one matching without replace-ment. This procedure resulted in a sample of 870 matched respondents. As mentioned earlier, one potential problem with nearest-neighbor matching is that being the nearest propensity score match does not guarantee that the pair is identical in their propensity scores. Acknowledging this, we restricted the range to be .01 units for the caliper for match-ing pairs on their propensity scores so that for treated and untreated respondents to be paired there must be a difference of .01 or less. We also conducted the same analysis using a .05 for the caliper. Results did not differ from what are reported here.


TABLE 2: Pre- and Postmatching t Tests Using Nearest-Neighbor Matching

Full Sample Matched Sample

Treated (n = 911)

Untreated (n = 435)

t Value

Treated (n = 435)

Untreated (n = 435)

t Value

Demographics Respondent age 9.58 11.01 –7.86* 10.63 11.01 –0.89 Respondent gender (1 = male) 0.51 0.46 1.73 0.50 0.46 0.44 Black 0.32 0.29 1.19 0.29 0.29 0.24 Hispanic 0.48 0.49 –0.53 0.48 0.49 –0.72 Other 0.03 0.04 –1.44 0.34 0.44 –0.24 SES 0.09 0.06 0.27 0.13 0.06 0.87 Family size 5.33 5.33 –0.04 5.20 5.34 –0.92 Number of siblings younger than 19 2.14 2.25 –1.12 2.20 2.25 –0.20 Public assistance 0.28 0.23 2.00* 0.24 0.23 0.37Primary caregiver (PC) characteristics PC age 35.47 38.25 –6.80* 37.78 38.25 –0.15 PC gender (1 = male) 0.08 0.13 –2.99* 0.10 0.13 –0.70 PC some high school 0.22 0.19 1.41 0.19 0.19 0.05 PC graduated high school 0.12 0.14 –0.60 0.14 0.14 0.16 PC education past high school 0.47 0.44 0.91 0.46 0.44 0.55 PC dad 0.07 0.12 –3.12* 0.10 0.12 –0.50 PC other female 0.04 0.06 –2.01* 0.06 0.06 –0.32 PC other male 0.01 0.00 0.21 0.00 0.00 –1.31 PC married (1 = yes) 0.64 0.68 –1.22 0.68 0.68 0.25 PC employed (1 = yes) 0.62 0.60 0.56 0.58 0.60 0.09 Supervision or monitoring 8.40 8.34 1.12 8.41 8.34 1.14 Warmth 6.95 6.94 0.13 7.02 6.94 0.28PC verbal skills 3.91 3.84 2.84* 3.88 3.84 0.90 Provision of social relations 20.76 20.12 2.55* 20.23 20.12 0.15 Conflict Tactics Scale 2.62 1.44 3.83* 1.46 1.44 –0.32 Conflict Tactics Scale (spouse) 2.29 1.21 4.14* 1.12 1.21 –0.62Child characteristics Wave 1 internalizing 8.08 5.82 5.92* 5.93 5.82 –0.56 Wave 1 externalizing 13.05 7.18 12.05* 7.22 7.18 –1.18 Impulsivity 56.40 50.52 9.04* 51.14 50.52 0.16 School help for emotional problems 0.04 0.01 2.96* 0.01 0.01 –0.94 Truant past year 0.05 0.03 1.57* 0.04 0.03 0.99Family characteristics Family member with criminal record 0.33 0.23 3.97* 0.22 0.23 –0.28 Family member with nerve problems 0.21 0.14 2.81* 0.16 0.14 –0.11 Family member with legal problems 0.19 0.14 2.21* 0.15 0.14 0.29 Family member attempt suicide 0.14 0.10 1.98* 0.11 0.11 –0.43

*p < .05.

Once matching was completed, another set of t tests was performed to ensure that bal-ance was achieved and significant differences between groups did not remain. As shown in the last three columns of Table 2, the matching procedure was successful in balancing the two groups. Once children were matched on their propensity of receiving corporal punish-ment, significant differences no longer existed between treated and untreated groups on the Wave 1 covariates. Substantive differences were also considerably reduced for many covariates after matching. For instance, the 6.68-point difference in externalizing behavior scores at Wave 1 for the treated and untreated groups was reduced to 0.4. Similarly, the 5.88-point difference for impulsivity at Wave 1 was reduced to 0.62. Overall, these results


confirm that our matching procedure was successful in reducing bias and balancing the two groups on various behavioral and other background characteristics. Results assessing the influence of corporal punishment on aggression and delinquency measured at Wave 2 are now discussed.

TREATED AND UNTREATED GROUP COMPARISONS

Figures 1 and 2 display results from our matching models using one-to-one nearest neighbor matching. For comparison purposes, each contains results for unmatched and matched samples with both treated (punished) and untreated (not punished) group means for aggression and delinquency scores at Wave 2. As shown in Figure 1, a statistically sig-nificant difference exists between average aggression scores for the unmatched sample. Specifically, the corporally punished and nonpunished groups’ mean aggression scores significantly differed by 2.33 points, indicating that those who received corporal punish-ment at Wave 1 exhibited more aggression at Wave 2. Using the matched sample, signifi-cant differences for aggression were no longer present and the mean difference was only 0.16, suggesting that children subjected to corporal punishment at Wave 1 had very similar mean aggression scores at Wave 2 when compared to their counterparts not subjected to corporal punishment. As shown in Figure 2, a similar pattern was observed for delinquency. For the unmatched sample, a statistically significant difference exists between mean delin-quency scores for the punished and not punished groups. Children receiving corporal pun-ishment, versus those who did not, show a mean difference of 0.97, indicating that corporally punished children have, on average, higher delinquency scores. In turn, the matched sample shows no statistically significant difference. As mentioned above, we employed both kernel and nearest-neighbor matching to assess the robustness of our results. Results for all analyses were very similar, and for the sake of parsimony we present only nearest-neighbor results; kernel matching results are available on request.

It is important to acknowledge the fact that children from the PHDCN are nested within neighborhood clusters. The problem that results from nested observations is a lack of inde-pendence or the possibility of correlated error terms, as kids from the same neighborhoods are more similar to one another than they are to kids from other neighborhoods. As such, it can lead to unreliable standard errors that may affect significance tests. Therefore, our results were confirmed by estimating models with robust standard errors using a “cluster” function in Stata 9.

Finally, an analysis is presented where propensity scores were grouped into quartiles so that four groups of respondents were created. Within each grouping respondents had simi-lar propensities of being corporally punished, but some actually were punished (treated) and others were not (untreated). Importantly, these analyses included all children in the analysis sample as opposed to those only having successful matches. Figures 3 and 4 rep-resent mean differences between corporally punished and nonpunished children within each propensity score classification. For instance, Group 1 contains children with the low-est propensity for being punished and Group 4 contains those children with the highest propensities for being punished. First, a linear trend is observed in that average scores for aggression and delinquency, for both treated and untreated groups, tend to increase as the propensity for receiving punishment increases. Second, only one statistically significant difference for mean aggression and delinquency scores existed between punished and not


Figure 1: Children’s Wave 2 Aggressive Behavior Results Before and After Matching (nearest-neighbor matching)*Difference is significant at the p < .05 level.

012345678

Val

ue

Unmatched* MatchedCondition

PunishedNot Punished

6.41

4.08 4.084.24

Figure 2: Children’s Wave 2 Delinquent Behavior Results Before and After Matching (nearest-neighbor matching)*Difference is significant at the p < .05 level.

00.5

11.5

22.5

33.5

Val

ue

Unmatched* MatchedCondition

PunishedNot Punished

3.17

2.10 1.97 2.10

punished groups within propensity groupings. Those receiving punishment exhibited more aggression than those not receiving punishment in the group that had the lowest propensi-ties for receiving punishment (Group 1). Finally, groups within the other strata did not show statistically significant differences in aggression or delinquency at Wave 2.


Figure 3: Mean Scores for Aggressive Behavior for Analysis Sample by Corporal Punishment and Propensity to Experience PunishmentNote. Group 1 has 197 untreated and 140 treated respondents; Group 2 has 133 untreated and 203 treated respondents; Group 3 has 82 untreated and 255 treated respondents; Group 4 has 23 untreated and 313 treated respondents.*Difference is significant at the p < .05 level.

0

2

4

6

8

10M

ean

Scor

e fo

rA

ggre

ssiv

eB

ehav

ior

1st *(.053-.538)

2nd(.538-.699)

3rd(.700-.837)

4th(.837-.996)

Propensity Group

3.44

2.62

4.00 4.12

6.31

9.37 9.48 Punished

Not Punished

5.98

Figure 4: Mean Scores for Delinquent Behavior for Analysis Sample by Corporal Punishment and Propensity to Experience PunishmentNote. Group 1 has 197 untreated and 140 treated respondents; Group 2 has 133 untreated and 203 treated respondents; Group 3 has 82 untreated and 255 treated respondents; Group 4 has 23 untreated and 313 treated respondents.*Difference is significant at the p < .05 level.

0

1

2

3

4

5

6

Mea

n Sc

ore

for

Del

inqu

ent

Beh

avio

r

1st(.053-.538)

2nd(.538-.699)

3rd(.700-.837)

4th(.837-.996)

Propensity Group

1.32

2.80

2.11

3.13

4.65 4.96 PunishedNot Punished

1.88 1.74


DISCUSSION AND CONCLUSIONS

It was argued in the current study that research on the influence of corporal punishment on behavioral problems cannot be confirmed or denied because of limitations of past stud-ies (Grogan-Kaylor, 2004, 2005a; Straus & Mouradian, 1998), namely, the absence of random assignment as the result of having only nonexperimental data. As such, we assessed the effect of corporal punishment on children’s aggressive and delinquent behaviors using a PSM approach on a longitudinal sample of children residing in Chicago, Illinois. The current study, albeit far from definitive, has moved closer toward addressing some limita-tions of past studies by matching children based on their propensity for being punished before comparing their subsequent aggression and delinquency.

Generally, our results are contradictory to past research that suggests the use of corporal punishment on children will lead to future misbehavior (Larzelere & Kuhn, 2005) and more in line with studies concluding that corporal punishment is marginally or not at all related to children’s misbehavior. First, a large amount of selection bias was present before matching occurred, indicating that children displaying negative behaviors and having particular indi-vidual, personality, and family characteristics were much more likely to have corporal pun-ishment used on them by their primary caregivers. Second, after successfully matching the two groups of children on various background characteristics, evidence emerged that indi-cated selection dynamics account for the “corporal punishment effect.” Specifically, once matched, those subjected to corporal punishment did not differ on their aggression or delin-quency scores at Wave 2 when compared to those who did not receive corporal punishment. These results were upheld using several matching procedures and methods of estimation.

One finding that we hope to explore more in the future appeared in the stratification analysis. Unlike the other matching techniques that excluded respondents without matches, this procedure utilized the entire analysis sample, which was divided into quartiles of indi-viduals based on the similarity of their propensities for being punished. This analysis indi-cated that the group with the lowest propensity scores (those who were least likely to be punished) exhibited differences in their problem behaviors depending on whether or not they were subjected to corporal punishment. The children who were corporally punished in this group showed a higher mean level of aggression than those who were not punished. Again, this group of children contains those who have the least amount of “risk” for being corporally punished and perhaps represents a group of individuals with more privilege and social capital. It is possible that experiencing physical punishment from a parent is a rarer and more traumatic experience for these children. A closer investigation of this group is in order for the future to understand the mechanisms behind this difference.

Although the current analyses shed light on some of the potential consequences of physical punishment, other types of discipline may play an important role in behavioral outcomes as well. For example, several studies have shown that outcomes are worse when the physical punishment is excessively harsh (Larzelere, Cox, & Smith, 2010; Larzelere & Kuhn, 2005). The current study focused only on corporal punishment, but we acknowledge the utility of comparing disciplinary methods for future research endeavors.

Results from the current study provide another piece of evidence that contributes to the broader body of research on the effects of corporal punishment. Our results demonstrate the need for more attention to the potential consequences of corporal punishment, especially in light of the current study’s limitations discussed below.


First, our use of a dichotomous corporal punishment measure presents certain limita-tions; children experiencing one or more of three types of corporal punishment were clas-sified as being in the same group. Some may argue that this measure is limited because it indicates that exposed children were exposed once or at a low frequency, but this is untrue. As reported earlier, those who were punished had a high frequency of being slapped and pushed or grabbed. Nonetheless, information on frequency and type of punishment used may affect our results. Some studies have assessed how these dimensions of corporal pun-ishment affect children. For instance, in their meta-analysis, Larzelere and Kuhn (2005) found that the effect sizes for physical punishment were worse than other discipline prac-tices when it was severe or used as the predominant disciplinary method. In a supplemental analysis, we investigated these issues using multivariate regression models to assess whether the frequency of each type of corporal punishment was associated with delin-quency and aggression. Findings indicated that after statistically controlling for covariates, neither the frequency nor type of corporal punishment had a statistically significant influ-ence on Wave 2 aggression or delinquency.

Second, the current study was limited by only assessing the direct influences of corporal punishment on behavior. It is possible that this may not be the only potentially important relationship. Perhaps the causal influences of corporal punishment on children’s behavior are conditioned by other variables, such as parenting practices and demographics (i.e., race, gender, and age) of children and caregivers. For instance, Simons et al. (2000) found that a lack of warmth from parents increased the negative effects of corporal punishment. As such, the effects might be observed only under certain parenting conditions or for particular groups of children. We did estimate split group PSM models to assess the effects of corpo-ral punishment on aggression and delinquency measures across gender, race, and age cohort. Findings indicated that no differences existed; we did not find corporal punishment to have statistically significant effects. We believe this is an interesting finding given the debate regarding the role of culture in predicting corporal punishment and its consequences. Nonetheless, interactions between corporal punishment and other variables mentioned above should be investigated more thoroughly.

Third, the current study was further limited by not including toddlers in the analysis. Research has indicated that younger children (especially toddlers) experience more corpo-ral punishment (Dietz, 2000; Giles-Sims et al., 1995; Grogan-Kaylor & Otis, 2007). Dietz (2000) found that children younger than 6 years old were more likely to be spanked, and Giles-Sims et al. (1995) showed that children between 3 and 5 years old were spanked more frequently than other age groups. Because of inconsistencies in measurement instruments across cohorts, the 3-year-old cohort was not included in this study. Several important covariates, as well as the outcome variables taken from the CBCL, were measured differ-ently with this cohort. Furthermore, this cohort had the highest percentage of respondents experiencing corporal punishment out of all cohorts. In the future we plan to conduct sepa-rate analyses for the 3-year-old cohort.

Fourth, although PSM has several advantages over traditional regression techniques, it is not a perfect solution to estimating causal effects (Dehejia & Wahba, 2002; Rosenbaum & Rubin, 1983); this methodological approach does have its own limitations. For instance, PSM can be performed only using observable and measured characteristics. By relying on guidance from past research, we are confident that most major background characteristics that could influence our results were incorporated. Nonetheless, we agree with King, Massoglia, and Macmillan (2007) that “propensity score matching models represent a


method, not the method, for assessing treatment effects” (p. 58). That is, these models should be viewed as only one way to compensate for a lack of random assignment. Other strategies should be carefully explored in the future.

Fifth, our analyses investigated differences between children measured at two waves in an interval of 2.5 years as opposed to assessing intraindividual change, limiting the conclu-sions that can be drawn about trajectories of aggression and delinquency. Exploring trajec-tories of externalizing behaviors of children over a longer developmental time period is warranted. Perhaps children exposed to corporal punishment may have quite different lon-gitudinal patterns of externalizing behaviors compared to their counterparts not exposed. Growth curve analyses could perhaps shed light on how children’s behavioral trajectories are influenced by corporal punishment. Furthermore, the effects of corporal punishment may differ based on age, although when we split our analysis by age cohorts we did not find differences. For instance, Grogan-Kaylor (2005a) also found that corporal punishment had more of an effect on antisocial behavior when it was used on teenagers.

Sixth, our treatment indicator results in a group of punished children who have been corporally punished more recently but does not necessarily include those punished at an early stage in life. As a result, the control group likely contains some individuals who were punished earlier in life but not within a year of the point in time that the data were collected. We acknowledge this as an inherent limitation of our data but feel that we share this limita-tion with most other studies that ask respondents about corporal punishment experienced within a specific time. We believe that by investigating the youngest cohort of individuals in the sample in the future we can more fully understand the implications of this concern.

Seventh, it may be perceived as a limitation of our analysis that making a simple distinction between those who were and were not punished is not the most interesting or appropriate distinction and comparing those who had a high frequency of punishment to those who did not is more important. As such, we have conducted several additional sensitivity analyses that use different cut points for the three-item corporal punishment measure. For instance, we dichotomized the measure at the highest 15th percentile (those who were punished the most frequently), and this served as our treatment group compared to the other 85% of our sample. Similar to analyses of the measure used in the current study, we found no statistically sig-nificant effects of corporal punishment on Wave 2 aggression or delinquency measures. Ultimately, we did not find an effect of corporal punishment regardless of whether the respon-dents had more frequent or less frequent exposure to corporal punishment. In the future we hope to more fully investigate this issue, perhaps using a “dose-response” framework such as the one outlined by Loughran and colleagues (2009) to further understand how frequency of punishment affects behavior. In conclusion, we feel this study has added to the extant research on corporal punishment and is an improvement over several studies that have explored the relationship between corporal punishment and children’s behaviors. The findings should be interpreted with the limitations presented above in mind but nonetheless as important. This study has laid the foundation for a modest research agenda on corporal punishment, with hopes that answers to some of these questions will be ascertained.

REFERENCES

Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4-18 and 1991 profile. Burlington: University of Vermont, Department of Psychiatry.


Berlin, L.J., Ispa, J.M., Fine, M.A., Malone, P.S., Brooks-Gunn, J., Brady-Smith, C., Ayoub, C., Bai, Y. (2009). Correlates and Consequences of Spanking and Verbal Punishment for Low-Income White, African-American, and Mexican American Toddlers. Child Development, 80: 1403–1420.

Brennar, V., & Fox, R. A. (1998). Parental discipline and behavior problems in young children. Journal of Genetic Psychology, 159, 251–256.

Buss, A., & Plomin, R. (1984). Temperament: Early developing personality traits. Hillsdale, NJ: Lawrence Erlbaum.Day, R. D., Peterson, G. W., & McCracken, C. (1998). Predicting spanking of younger and older children by mothers and

fathers. Journal of Marriage and the Family, 60, 79–94.Dehejia, R. H., & Wahba, S. (2002). Propensity score-matching methods for nonexperimental causal studies. Review of

Economics and Statistics, 84, 151–161.Dietz, T. L. (2000). Disciplining children: Characteristics associated with the use of corporal punishment. Child Abuse &

Neglect, 24, 1529–1542.Foster, E. M. (2003). Propensity score matching: An illustrative analysis of dose response. Medical Care, 41, 1183–1192.Gershoff, E. T. (2002). Corporal punishment by parents and associated child behavior experiences: A meta-analytic and

theoretical review. Psychological Bulletin, 128, 539–579.Gibson, C. L., Sullivan, C., Jones, S., & Piquero, A. (2010). Does it take a village? Assessing neighborhood effects on chil-

dren’s self-control. Journal of Research in Crime and Delinquency, 47, 31–62.Giles-Sims, J., Straus, M. A., & Sugarman, D. B. (1995). Child, maternal, and family characteristics associated with spanking.

Family Relations, 44, 170–176.Grogan-Kaylor, A. (2004). The effect of corporal punishment on antisocial behavior in children. Social Work Research, 28,

153–162.Grogan-Kaylor, A. (2005a). Corporal punishment and the growth trajectory of children’s antisocial behavior. Child

Maltreatment, 10, 283–292.Grogan-Kaylor, A. (2005b). Relationship of corporal punishment and antisocial behavior by neighborhood. Archives of

Pediatric and Adolescent Medicine, 159, 938–942.Grogan-Kaylor, A., & Otis, M. D. (2007). The predictors of parental use of corporal punishment. Family Relations, 56, 80–91.Hahs-Vaughn, D. L., & Onwuegbuzie, A .J. (2006). Estimating and using propensity score analysis with complex samples.

Journal of Experimental Education, 75, 31–65.Hill, J. L., Waldfogel, J., Brooks-Gunn, J., & Han, W. J. (2005). Maternal employment and child development: A fresh look

using newer methods. Developmental Psychology, 41, 833–850.Kantor, G. K., & Jasinski, J. L. (Eds.). (1997). Out of the darkness: Contemporary perspectives on family violence. Thousand

Oaks, CA: Sage.King, R. D., Massoglia, M., & Macmillan, R. (2007). The context of marriage and crime: Gender, propensity to marry, and

offending in early adulthood. Criminology, 45, 33–65.Lansford, J.E., Criss, M.M., Dodge, K.A, Shaw, D.S., Pettit, G.S., & Bates, J.E. (2009). Trajectories of physical discipline:

Early childhood antecedents and developmental outcomes. Child Development, 80: 1385–1402.Larzelere, R. E. (2000). Child outcomes of nonabusive and customary physical punishment by parents: An updated literature

review. Journal of Child Psychology and Psychiatry, 3, 199–221.Larzelere, R. E., Cox, R. B., & Smith, G. L. (2010). Do nonphysical punishments reduce antisocial behavior more than spank-

ing? A comparison using the strongest previous causal evidence against spanking. BMC Pediatrics, 10, 10–27.Larzelere, R. E., & Kuhn, B. R. (2005). Comparing child outcomes of physical punishment and alternative disciplinary tac-

tics: A meta-analysis. Clinical Child and family Psychology Review, 8, 1–37.Leuven, E., & Sianesi, B. (2003). PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching,

common support graphing, and covariate imbalance testing (Version 1.2.3). Retrieved from http://ideas.repec.org/c/boc/bocode/s432001.html

Leventhal, T., Selner-O’Hagan, M. B., Brooks-Gunn, J., Bingenheimer, J. B., & Earls, F. E. (2004). The homelife interview from the Project on Human Development in Chicago Neighborhoods: Assessment of parenting and home environment for 3- to 15-year-olds. Parenting: Science and Practice, 4, 211–241.

Loughran, T. A., Mulvey, E. P., Schubert, C. A., Fagan, J., Piquero, A. R., & Losoya, S. H. (2009). Estimating a dose-response relationship between length of stay and future recidivism in serious juvenile offenders. Criminology, 47, 699–740.

McCord, J. (2005). Unintended consequences of punishment. In M. Donnelly & M. A. Straus (Eds.), Corporal punishment of children in theoretical perspective (pp. 165–169). New Haven, CT: Yale University Press.

McKee, L., Roland, E., Coffelt, N., Olson, A., Forehand, R., Massari, C., . . . Zens, M. (2007). Harsh discipline and child problem behaviors: The roles of positive parenting and gender. Journal of Family Violence, 22, 187–196.

McLoyd, V. C., & Smith, J. (2002). Physical discipline and behavior problems in African American, European American, and Hispanic children: Emotional support as a moderator. Journal of Marriage and Family, 64, 40–53.

Molnar, B. E., Buka, S. L., Brennan, R. T., Holton, J. K., & Earls, F. E. (2003). A multilevel study of neighborhoods and parent-to-child physical aggression: Results from the Project on Human Development in Chicago Neighborhoods. Child Maltreatment, 8, 84–97.


Newgard, C. D., Hedges, J. R., Arthur, M., & Mullins, R. J. (2004). Advanced statistics: The propensity score—A method for estimating treatment effect in observational research. Academy of Emergency Medicine, 11, 953–961.

Paolucci, E. O., & Violato, C. (2004). A meta-analysis of the published research on the affective, cognitive, and behavioral effects of corporal punishment. Journal of Psychology, 138, 197–221.

Rodriguez, C. M. (2003). Parental discipline and abuse potential affects on child depression, anxiety, and attributions. Journal of Marriage and Family, 65, 809–817.

Rosenbaum, P. R., & Rubin, D. B. (1983). Reducing bias in observational studies using subclassification on the propensity score. Journal of the American Statistical Association, 79, 516–524.

Rubin, D. B. (1997). Estimating causal effects from large data sets using propensity scores. Annals of Internal Medicine, 127, 757–763.

Rubin, D., & Thomas, N. (1996). Matching using estimated propensity scores: Relating theory to practice. Biometrics, 52, 249–264.

Sampson, R. J. (1997). The embeddedness of child and adolescent development: A community-level perspective on urban violence. In J. McCord (Ed.), Violence and childhood in the inner city (pp. 31–77). Cambridge, UK: Cambridge University Press.

Sampson, R. J., Morenoff, J. D., & Raudenbush, S. (2005). Social anatomy of racial and ethnic disparities in violence. American Journal of Public Health, 95, 224–232.

Sianesi, B. (2001, May). Implementing propensity score matching estimators with STATA. Presented at the U.K. Stata Users Group VII meeting, London, UK.

Simons, R. L., Wu, C., Lin, K., Gordon, L., & Conger, R. D. (2000). A cross-cultural examination of the link between cor-poral punishment and adolescent antisocial behavior. Criminology, 38, 47–79.

Slade, E. P., & Wissow, L. S. (2004). Spanking in early childhood and later behavior problems: A prospective study of infants and young toddlers. Pediatrics, 113, 1321–1330.

Sourander, A., & Helstela, L. (2005). Childhood predictors of externalizing and internalizing problems in adolescence: A prospective follow-up study from age 8 to 16. European Child & Adolescent Psychiatry, 14, 415–423.

Straus, M. A. (1994). Beating the devil out of them: Physical punishment in American families. San Francisco, CA: Jossey-Bass.

Straus, M. A. (1996). Spanking and the making of a violent society. Pediatrics, 98, 837–842.Straus, M.A. (1979). Measuring Intrafamily Violence and Conflict: The Conflict Tactics (CT) Scales. Journal of Marriage and

Family, 41: 75–88.Straus, M. A., Hamby, S. L., Finkelhor, D., Moore, D. W., & Runyan, D. (1998). Identification of child maltreatment with the

Parent-Child Conflict Tactics Scales: Development and psychometric data for a national sample of American parents. Child Abuse & Neglect, 22, 249–270.

Straus, M. A., & Kantor, G. K. (1994). Corporal punishment of adolescents by parents: A risk factor in the epidemiology of depression, suicide, alcohol abuse, child Abuse, and wife beating. Adolescence, 29, 543–563.

Straus, M. A., & Mouradian, V. E. (1998). Impulsive corporal punishment by mothers and antisocial behavior and impulsive-ness of children. Behavioral Sciences and the Law, 16, 353–374.

Straus, M. A., & Stewart, J. H. (1999). Corporal punishment by American parents: National data on prevalence, chronicity, severity, and duration, in relation to child and family characteristics. Clinical Child and Family Psychology Review, 2, 55–70.

Straus, M. A., Sugarman, D., & Giles-Sims, J. (1997). Spanking by parents and subsequent antisocial behavior of children. Archives of Pediatrics and Adolescent Medicine, 151, 761–767.

Turner, H. A., & Finkelhor, D. (1996). Corporal punishment as a stressor among youth. Journal of Marriage and the Family, 58, 155–166.

Turner, H. A., & Muller, P. A. (2004). Long-term effects of child corporal punishment on depressive symptoms in young adults. Journal of Family Issues, 25, 761–782.

Turner, R. J., Frankel, B., & Levin, D. (1983). Social support: Conceptualization, measurement, and implications for mental health. In J. R. Greenley (Ed.), Research in community and mental health (Vol. 3, pp. 67–111). Greenwich, CT: JAI.

Wissow, L. (2001). Ethnicity, income, and parenting contexts of physical punishment in a national sample of families with young children. Child Maltreatment, 6, 118–129.

Sara Z. Morris is a PhD candidate in the Sociology Department at the University of Georgia. Her research interests include juvenile delinquency, the effect of parental socialization on behavior problems in children, and quantitative methodologies. Most recently, her work has focused on the effects of harsh discipline on both internalizing and externalizing behaviors in children. She will be starting as an assistant professor at the University of West Florida in August 2011.

Chris L. Gibson is an assistant professor in the Department of Sociology and Criminology & Law at the University of Florida and a W. E. B. Du Bois Fellow of the National Institute of Justice. His research focuses on the intersection among neighborhood influences, a life-course perspective, and biosocial criminology to study how characteristics of individuals and their environments lead to involvement in antisocial behavior and violent behavior.

Date post:	23-Dec-2015
Category:	Documents
Upload:	danacapastraru
View:	9 times
Download:	4 times

Criminal Justice and Behavior-2011-Morris-818-39.pdf

Documents