The analysis drew on an unusual and methodologically valuable data source. The National Health Interview Survey, the nation’s flagship household health interview, includes a follow-back component called NHIS-Teen, in which a randomly selected teenager aged 12 to 17 in each participating household answers questions directly. In the 2021-2022 cycle, 1,127 parent-teenager pairs provided usable responses, allowing the researchers to place two informants from the same family side by side on three distinct measures of bullying involvement: in-person bullying victimization, cyberbullying victimization, and bullying perpetration.

The technical challenge in such work is not simply asking two people the same question; it is quantifying how well their answers agree beyond what chance would predict. The team employed three complementary statistics. Percent agreement offers a raw measure of concordance but is inflated when the behavior in question is rare, because most pairs agree trivially by both saying no. Weighted kappa, the classic chance-corrected coefficient introduced by Landis and Koch, adjusts for chance agreement and categorizes concordance on a scale from slight to almost perfect. Prevalence-adjusted bias-adjusted kappa, or PABAK, further corrects for imbalances in how often each response category occurs and for systematic tendencies of one informant to report more or less than the other. Applying all three measures, the researchers found weighted kappa values ranging from fair to moderate for each form of bullying involvement, indicating that parents and teenagers agreed far from perfectly.

The direction of the disagreement was strikingly consistent. Across all three bullying measures, prevalence estimates built on teenager self-report exceeded those built on parent report. In other words, when a family contained a teenager who had been bullied in person, bullied online, or had bullied others, the parent was often unaware of it, and the teenager was seldom over-reporting relative to the parent. This pattern matches the authors’ hypotheses, which were grounded in prior literature showing that children and adolescents disclose victimization selectively, and that some experiences, particularly those occurring in digital spaces parents cannot observe, remain invisible to the adults in the household.

Importantly, the extent of parent-teenager concordance was not uniform across the population. The researchers used bivariate logistic regression to identify sociodemographic characteristics associated with agreement, and found that concordance of bullying involvement varied by race and Hispanic origin. This finding carries practical weight: if agreement between parent and teenager differs across groups, then surveys that rely exclusively on parent proxy report may distort prevalence comparisons between racial and ethnic groups, potentially obscuring or exaggerating real disparities in bullying exposure.

The study went beyond agreement statistics to ask a deeper question: does it matter, for research on health consequences, who reports the bullying? Bullying victimization and perpetration have been linked in previous work to anxiety, depression, and increased suicidal thoughts and behaviors, and meta-analyses have documented these associations across dozens of studies. The NHIS-Teen team fit logistic regression models relating each of the three bullying measures to health outcomes, once using parent reports and once using teenager reports. The associations pointed in the same direction regardless of who served as the informant, but the magnitude of the associations differed between the two reporting sources, contradicting the hypothesis that parent-reported associations would uniformly be stronger.

That divergence in magnitude matters for how the scientific literature is read. If estimates of the harm associated with bullying shift depending on whether a parent or the teenager supplies the exposure information, then meta-analyses that pool studies mixing informants may carry hidden heterogeneity. More fundamentally, the results suggest that parent proxy reports are not interchangeable with adolescent self-reports, even though parent reports remain common in national child health surveys, largely because young children cannot reliably self-report and because proxy reporting is cheaper and faster to collect.

Why do parents miss so much? Developmental research offers several converging explanations. Adolescents gain increasing autonomy and spend large portions of their day away from parental observation, in classrooms, hallways, buses, and increasingly in online environments where interactions leave no physical trace. Disclosure to parents is voluntary and depends on trust, on the adolescent’s appraisal of whether telling will help, and on fears of escalation or confiscation of devices. Prior studies of children, parents, and teachers, including work in school-based samples, have similarly documented only modest agreement about victimization, with children’s own reports generally capturing more episodes. The present analysis extends that picture to a nationally representative household sample, strengthening the case that the discrepancy is a broad structural feature of how bullying is observed rather than an artifact of any single school or community.

The authors emphasize that their findings do not render parent reports useless. Parent and teenager reports may capture partially overlapping but non-identical constructs: a parent’s report reflects the family’s shared awareness of the child’s social world, which is itself meaningful for intervention, while a teenager’s report reflects lived experience more directly. The concordance statistics quantify the gap between these two vantage points. Weighted kappa in the fair-to-moderate range signals that neither source can stand in for the other without information loss. The authors suggest that future work examine how parent-teenager communication shapes social and emotional health, and whether families with more open communication show higher concordance, turning agreement itself into a potential marker of family functioning.

For policymakers and survey designers, the implications are concrete. National monitoring systems that track bullying solely through parent interviews may understate prevalence and misestimate subgroup differences, weakening the evidence base for school and public health programs. Building direct adolescent self-report into surveillance, as NHIS-Teen does, offers a model: the follow-back design preserves the efficiency of the household survey while recovering the teenage perspective. As bullying increasingly migrates into digital arenas that parents rarely see, the cost of relying on parental observation alone will likely grow, and studies like this one provide the quantitative justification for listening to teenagers directly.

Subject of Research: Parent-teenager concordance in reporting bullying involvement among U.S. adolescents

Article Title: Parent- and Teenager-Reported Concordance of Bullying Involvement in a National Sample of Teenagers

Article References: Ng, A. E., Bottoms-McClain, L., Black, L. I., Zablotsky, B., Bose, J., & Blumberg, S. J. (2026). Parent- and Teenager-Reported Concordance of Bullying Involvement in a National Sample of Teenagers. Journal of Child and Family Studies. https://doi.org/10.1007/s10826-026-03331-z

Image Credits: AI Generated

DOI: 10.1007/s10826-026-03331-z

Keywords: bullying, cyberbullying, concordance, parent-report, teenager self-report, NHIS-Teen, adolescent health, victimization, perpetration, weighted kappa, Parent-, Teenager-Reported