Trends in Childhood Violence and Abuse Exposure: Evidence From 2 National Surveys

Objective  To assess trends in children's exposure to abuse, violence, and crime victimizations.

Design  An analysis based on a comparison of 2 cross-sectional national telephone surveys using identical questions conducted in 2003 and 2008.

Setting  Telephone interview.

Participants  Experiences of children aged 2 to 17 years (2030 children in 2003 and 4046 children in 2008) were assessed through interviews with their caretakers and the children themselves.

Outcome Measure  Responses to the Juvenile Victimization Questionnaire.

Results  Several types of child victimization were reported significantly less often in 2008 than in 2003: physical assaults, sexual assaults, and peer and sibling victimizations, including physical bullying. There were also significant declines in psychological and emotional abuse by caregivers, exposure to community violence, and the crime of theft. Physical abuse and neglect by caregivers did not decline, and witnessing the abuse of a sibling increased.

Conclusion  The declines apparent in this analysis parallel evidence from other sources, including police data, child welfare data, and the National Crime Victimization Survey, suggesting reductions in various types of childhood victimization in recent years.

There have been marked declines in several indicators of childhood exposure to violence and abuse from the early 1990s to the recent past. For example, physical abuse substantiated by state child welfare authorities dropped 52% from 1990 to 2007, and substantiated sexual abuse dropped 53% during this same period.¹ The National Crime Victimization Survey reported that among 12- to 17-year-old adolescents, rates of aggravated assault, robbery, and sexual assault dropped 69%, 62%, and 52%, respectively, from 1993 to 2005.² Criminal victimizations in school also dropped 60% from 1995 to 2005.³ A survey of all sixth- ninth-, and 12th-grade students in Minnesota schools found a 28% decline between 1992 and 2007 in those reporting sexual abuse by a family member and a 30% decline in abuse by a non–family member.⁴ The Federal Bureau of Investigation reported that homicides of 14- to 17-year-old adolescents dropped 60% from 1993 to 2005 and those with victims aged 13 years and younger dropped 36%.⁴

While there is some debate about the meaning of these trends,⁵ support for the idea that they represent true declines and not enforcement, administrative, or statistical artifacts comes from a variety of sources.⁶ The declines appear in self-report studies directly from adolescents as well as in agency statistics.^1,2 A study of the patterns suggested that they did not bear the hallmarks of declines due to changing standards or a decreased willingness to report victimizations.⁶ The declines occurred in a broader context of declining societal crime and violence, which was widely accepted as real.^7-9 Moreover, other closely related child-welfare indicators—such as suicide, running away, juvenile delinquency, and teenage pregnancy¹⁰—have also improved during the same period, a convergence suggesting that this trend is valid.

The declines have not occurred at a constant rate over time. Many of the sharpest drops in rates occurred during the 1990s in contrast to the trends in the mid-2000s, which were characterized by smaller declines or even small increases. Substantiated sexual abuse dropped at an average annual rate of 6% per year between 1992 and 2001, but the average decline was only 2% per year between 2002 and 2006. The most dramatic declines in juvenile-victim homicide occurred between 1993 and 2000, after which rates plateaued or even increased slightly.¹¹ Other indicators, however, such as the National Crime Victimization Survey, showed continuing declines during the 2000s.² Thus, there are questions about whether declines continued during the 2000s and for what kinds of crime and violence.

Some answers to these questions can be obtained from 2 national surveys that were conducted 5 years apart during the 2000s, used similar methods and questions, and inquired about a broad range of violence and abuse exposures. This article will report trend information comparing the 2003 data from the Developmental Victimization Survey (DVS) and 2008 data from the National Study of Children Exposed to Violence to assess whether declines have continued during this period.

This analysis compared juvenile victimization data obtained from 2 very similar national surveys conducted 5 years apart. The earlier survey, the DVS, collected data on 2030 children aged 2 to 17 years between December 2002 and February 2003. The National Survey of Children's Exposure to Violence (NATSCEV) obtained information between January and May 2008 on 4046 children aged 2 to 17 years.

Data on experiences of violence and abuse in both surveys were obtained using very similar items from the Juvenile Victimization Questionnaire (JVQ). The JVQ is a comprehensive instrument designed to screen for a wide range of victimization events, covering such general areas of concern as physical assault, property victimization, child maltreatment, peer and sibling victimization, sexual victimization, witnessing violence, and indirect exposure to violence.¹² Both surveys asked the same questions about 34 separate victimization types and collected similar demographic and background information.

The sampling methodology and procedures for both studies are described in detail elsewhere^13,14 and will only be summarized here. Although data collection occurred 5 years apart, the studies were parallel in most respects. Both relied on a nationwide sampling of residential telephone numbers and used list-assisted random-digit dialing for sample selection and data collection. List-assisted random-digit dialing is the generally accepted method for scientific sampling of households with telephones in the United States. The sampling frame is landline residential telephone banks with 1 or more listed (available for directory publication or directory assistance) residential telephone numbers.¹⁵ A recent study¹⁶ has confirmed that 95% of working residential landline telephone numbers are still located in their listed telephone banks. In both studies, a short interview was conducted with an adult caregiver (usually a parent) to obtain family background information. Then, 1 child was randomly selected from all eligible children living in the household by selecting the child with the most recent birthday. If the selected child was aged 10 to 17 years, the primary interview was conducted with the child. If the selected child was aged 2 to 9 years, it was conducted with the caregiver who was “most familiar with the child's daily routine and experiences.” For the earlier survey (DVS), interviews were completed with 79% of the eligible persons contracted. The sample for the more recent survey (NATSCEV) combined a national cross-section portion (n = 3053) with an oversampled portion drawing from telephone exchanges with high proportions of Hispanic, African American, or low-income households (n = 1496). Interviews were completed with 71% of the eligible persons contacted in the cross-section and 63% of those contacted in the oversample.

Consent was obtained from the parent and assent from the child. All procedures in both studies had been approved by the institutional review board at the University of New Hampshire. In both surveys, weights were calculated and applied to control for the differential probability of selection of cases owing to sampling procedures and variations in within-household eligibility and to adjust the sample counts to match existing national demographic profiles in terms of child age, gender, race/ethnicity, and household income.

Exposures were assessed with the 34 JVQ items noted earlier (Table). Each item asked whether the child experienced a specifically described event. The items included exposure to abuse, violent acts, crimes, and acts that would be considered crimes if committed among adults, though in some cases not considered criminal when occurring among children (eg, hitting by peers or siblings). Virtually all of the exposures that were included have been studied individually as traumas or threats to children's welfare in other research. The same items were used for the self-report interview (children aged 10-17 years) and the caregiver interview (children aged 2-9 years), with slight wording changes to appropriately identify the subject of the question (eg, “Did anyone use force to take something away from your child/you that he or she was/you were carrying or wearing?”). The JVQ has shown evidence of good test-retest reliability and construct validity across a wide spectrum of developmental stages.¹²

To facilitate analysis and help clarify trends, the 34 primary victimizations were organized into 8 general domains, each reflecting distinctive aspects of juvenile victimization. Eight aggregate measures were constructed to represent each child's overall experience within each domain, with each measure recording whether a child had experienced any victimization of that type. The aggregates are physical assault, property victimization, maltreatment, peer-sibling victimization, sexual victimization, sexual assault, witnessing family violence, and exposure to community violence. For example, physical assault documents whether a child experienced any of the relevant physical assault screeners, and maltreatment records any maltreatment episode (aggregate measures and associated screeners are presented in the Table). One screener, “sex with an adult,” was allowed to stand alone because variability in state laws about age of consent means that in some cases these exposures constituted statutory sex offenses and in others they did not.

The victimizations of interest for the present analysis were those that occurred within the year prior to the interview. Information on victimization timing was obtained somewhat differently in the DVS and NATSCEV. For the DVS, each question asked specifically about experiences “in the past year.” The NATSCEV asked each question with the phrase “At any time in your child’s/your life . . . ,” which was followed with an additional question designed to isolate past-year episodes from victimizations that might have occurred earlier in the child's life (“Thinking of [the last time/when] this happened . . . did it happen in the last year?”). In both surveys, cues were provided to assist the respondent in setting and applying the past-year time frame. A measure of socioeconomic status was computed by combining the measure of the child's household income with the measure of the highest level of education achieved by either of the child's caregivers.

Victimization rates calculated for the DVS sample were compared with those calculated from the NATSCEV sample to determine whether significant changes had occurred in victimization levels during the 5-year interval. For both surveys, the rate for each victimization was computed as the proportion of children in the sample who reported that type of victimization experience in the year preceding their interview. Rates are reported as percentages of each sample base. For aggregate victimizations, the rate was the proportion of children in each survey who had any victimization of each aggregate type. Weighted data were used for both surveys to ensure that each sample was as representative of national patterns as possible.

Assessment of significant differences in each set of rates was based on a 2-sample difference-of-proportions test. Because the surveys were similar in design but not completely identical, we chose to take a conservative stance in comparing victimization rates. For purposes of this analysis, differences were statistically significant at P < .01 (α).

Several types of victimization were reported less often in 2008 than in 2003 (Table). The annual prevalence of any physical assault dropped from 53.1% to 49.6% of children and any peer-sibling victimization in the aggregate dropped from 58.8% to 52.7%. Both of these were affected by the drop in the single item on peer and sibling assault (an item included in both aggregates). The peer/sibling victimization aggregate decline was also strongly affected by the large drop in physical bullying, and a smaller decline (significant at P < .05 but not P < .01) in emotional bullying. Any property victimization did not decline, but theft and burglary of family households did individually. The small decline in any maltreatment was not significant, but the item on psychological and emotional abuse, which makes the largest contribution to maltreatment, did decline significantly. The annual prevalence of any sexual assault declined from 3.3% to 2.0% of children. Sex with an adult (tracking possible statutory sex offenses) also declined. Any exposure to community violence also declined from 29.7% to 24.8% of children owing to a particularly strong drop in witnessing assaults, with and without weapons.

Notable types of individual victimization that did not show significant differences in the 2 surveys included kidnapping and bias attacks (assaults targeting race, religion, or perceived sexual orientation), physical abuse and neglect by caregivers, dating violence, sexual abuse by known and unknown adults, the witnessing of domestic violence, and exposures to a shooting. Two individual items showed significant increases: robbery and witnessing the physical abuse of a sibling.

Analysis of the trends by gender and age showed no differences in the patterns. Analysis by socioeconomic status showed considerably stronger declines for children from households of lower socioeconomic status than for other children. Total victimization declined for children of lower socioeconomic status by 19% (Z = 11.48, P < .001) but did not change significantly for the other children (Z = −1.94, P = .05).

The victimization rates in the 2 surveys showed a clear pattern of mostly lower rates for 2008 than 2003. The big declines in peer-sibling assaults and bullying were noteworthy, as was the decline in the witnessing of weapon and nonweapon assaults. Psychological/emotional maltreatment by caregivers appeared to have lessened, though other forms of child maltreatment did not. Sexual assault appeared to decline, as did thefts. These declines are consistent with trends apparent in the National Crime Victimization Survey, which since 2003 has shown marked drops in 12- to 17-year-old adolescents reporting assaults, sexual assault, and property crime.

Among the victimizations that did not decline were adult-instigated victimizations, including physical abuse and neglect, and exposure to adult domestic violence. Child exposure to adult physical abuse actually showed an increase. These patterns are somewhat at odds with data on child maltreatment substantiations, for example, which have indicated some declines in physical abuse during the 2002-2006 period. However, trends in substantiated abuse reflect agency data, which are influenced by many factors besides real underlying rate changes. Moreover, the number of cases of reportable child maltreatment appearing in general population surveys of the sort compared here may be too small to allow the detection of the relatively modest trends like the 17% decline in physical abuse shown by the federal child abuse data during this period.

However, there are possible reasons why various forms of peer victimization may have declined more than those perpetrated by adults. In the wake of several high-profile school shooting episodes in the late 1990s, many schools began to rapidly adopt anti-bullying and other violence-prevention policies and programs. Formal evaluations of some of these programs have shown them to be successful in reducing violence and bullying.^17,18 Declines in peer-related victimization during the mid-2000s may be one of the effects of this social-problem mobilization.

The finding that the declines were greater among children of lower socioeconomic status households is plausible as well. Various theories about recent declines in crime rates have mentioned such factors as increasing police staffing and activity (including at schools),^7,19 growing immigration among groups with a high motivation for economic advancement, and the development and dissemination of psychiatric medications to segments of the population previously not amenable to mental health services.⁴ These factors might be expected to have more impact on low–socioeconomic status communities and children.

However, the trend data in this study need to be interpreted with an important caveat concerning at least 1 methodological difference that could be an alternative explanation for any drop. In the 2003 survey, respondents were asked the questions about victimizations in the past year first, before being asked in some selected instances about any lifetime exposure to a subset of the victimization screeners. In the 2008 survey, by contrast, respondents were asked about lifetime exposures for all screeners first and then asked to specify if any of these occurred in the last year. It is possible that respondents in 2003, being asked first about a 1-year time frame, eager to appear helpful or responsive and not knowing they would have another opportunity to disclose them later, reported some episodes within the past year that had actually occurred earlier, a process known in epidemiology as “telescoping.” In 2008, such respondents would have had an opportunity to disclose those episodes in the first set of inquiries about lifetime exposure and then might have been more accurate in their judgment about whether they occurred in the past year or not. This process could have resulted in lower past-year estimates in 2008 than in 2003.

On the other hand, the pattern of findings in 2003 and 2008 does not strongly support the telescoping explanation for the changes between the surveys. A telescoping difference might have been expected to show up across all screeners, though some of the screeners actually showed increases between 2003 and 2008, and many showed little change. In addition, it is not clear why the telescoping would have applied more to peer victimizations than adult-perpetrated ones. Another argument against a telescoping explanation is that delinquent behaviors in the past year (measured separately and in exactly the same way in 2003 and 2008) also declined (data not presented here). This is consistent with a real decline in victimization.

The data analyzed in this study can be read as a possible additional confirmation that declines have occurred in some forms of child victimization during the recent past. If that is true, it is certainly encouraging news for those working to create safer environments for children, an endeavor that has mobilized a considerable quantity of resources in the last few decades.

Correspondence: David Finkelhor, PhD, Crimes Against Children Research Center, University of New Hampshire, 126 Horton Social Science Center, 20 College Rd, Durham, NH 03824 (david.finkelhor@unh.edu).

Accepted for Publication: September 20, 2009

Author Contributions:Study concept and design: Finkelhor, Turner, and Hamby. Acquisition of data: Finkelhor, Turner, and Ormrod. Analysis and interpretation of data: Turner, Ormrod, and Hamby. Drafting of the manuscript: Ormrod. Critical revision of the manuscript for important intellectual content: Finkelhor, Turner, Ormrod, and Hamby. Statistical analysis: Finkelhor and Ormrod. Obtained funding: Finkelhor and Turner. Administrative, technical, and material support: Finkelhor and Ormrod. Questionnaire management and methodological expertise: Hamby.

Financial Disclosure: None reported.

Funding/Support: All of the funds for this program were derived from federal sources. This project was supported by grant 2006-JW-BX-0003 from the Office of Juvenile Justice and Delinquency Prevention, Office of Justice Programs, US Department of Justice. The total amount of federal funding involved is $2 709 912.

Disclaimer: The points of view and opinions in this document are those of the authors and do not necessarily represent the official position or policies of the US Department of Justice.