Depression May Drive Loneliness in Youth After Disaster, Landmark Study Finds

In the months after a devastating earthquake struck Japan’s Noto Peninsula, researchers fanned out across the city of Wajima to ask children and adolescents a deceptively simple set of questions: Do you feel lonely? Do you feel like crying? Do you hit or kick other children? Do you try to help classmates in difficulty? The answers, gathered three times over eight months from hundreds of students, have now yielded one of the most detailed longitudinal portraits ever assembled of how loneliness, emotional distress, and social behavior intertwine in young people living through a collective catastrophe.

The study, conducted by a team led by Yasuo Murayama of Kanazawa University and published in Child Psychiatry & Human Development, tracked 634 students in grades 3 through 9 across all public elementary and junior high schools in Wajima City, a community that suffered extensive damage in the January 2024 Noto Peninsula Earthquake. Data were collected at three time points between July 2024 and February 2025, roughly three months apart, as part of a broader mental health support program run jointly by the Ishikawa Kokoro no Care Center and Kanazawa University’s Research Center for Child Mental Development. Participation was voluntary, anonymous, and approved by the Chukyo University Research Ethics Committee.

What makes the research technically distinctive is its modeling strategy. The investigators employed a three-wave cross-lagged panel model, a structural equation technique that estimates whether each variable predicts change in the others over time while statistically controlling for each variable’s own prior level. The researchers deliberately chose the traditional cross-lagged panel model over the newer random-intercept variant, arguing that in a period when situational stress was widely shared across the community, a random intercept would risk conflating stable trait-like differences with persistent contextual effects of the disaster. The final model fit the data exceptionally well, with a comparative fit index of 1.00 and a root mean square error of approximation of just 0.011.

The study’s central and most striking finding concerns an asymmetry in the relationship between loneliness and depressive symptoms. While the two constructs showed bidirectional longitudinal associations, the pathway running from depressive symptoms to later loneliness was substantially stronger and more consistent than the reverse. Depressive symptoms at one wave predicted loneliness at the next with standardized coefficients ranging from 0.15 to 0.23, both statistically significant at p less than 0.001 and qualifying as large effect sizes under recent guidelines for cross-lagged effects. In contrast, loneliness predicted subsequent depressive symptoms with coefficients of only 0.06 to 0.08, one of which was merely marginally significant.

This directional pattern partially diverges from earlier meta-analytic work, which reported relatively comparable reciprocal effects between loneliness and depression in the general population. The authors propose that the post-disaster context may explain the difference. Cognitive models of depression hold that depressive symptoms are associated with negative biases in how people interpret their social environments, and in a community where nearly everyone is struggling, awareness that others face similar hardships may blunt the usual impact of loneliness on mood. Meanwhile, elevated depressive symptoms may still distort individuals’ perceptions of their relationships, gradually deepening feelings of isolation. Consistent with this interpretation, research conducted during the COVID-19 pandemic similarly found stronger longitudinal effects from depression to loneliness than the reverse in collectively affected populations.

The behavioral findings were equally revealing. Aggressive behavior consistently predicted higher subsequent loneliness, with standardized coefficients of roughly 0.08 to 0.09 at both intervals, and it also forecast declines in prosocial behavior with coefficients between negative 0.11 and negative 0.12. Notably, the reverse pathway from loneliness to later aggression was not supported, a result that contrasts with laboratory and animal research suggesting social isolation heightens aggression. Instead, the data suggest that in the complex social ecology of real classrooms, it is aggressive conduct that erodes future social standing and connection, echoing a five-year longitudinal study of first-grade children that found aggression predicted reductions in prosocial behavior while the reverse was not significant.

One unexpected result concerns prosocial behavior. In the first interval, higher prosocial behavior actually predicted a small increase in subsequent loneliness, a finding that ran counter to the common assumption that helping others buffers against isolation. The researchers offer a tentative hypothesis: in a post-disaster community under collective stress, children who frequently engage in helping behavior, or who perceive themselves as doing so, may focus excessively on others’ needs at the expense of their own, potentially receiving less support in return. They note prior evidence that self-sacrificing altruism is sometimes linked with interpersonal difficulties, but they emphasize this interpretation should be regarded as a preliminary hypothesis, since individual disaster exposure was not directly assessed.

The temporal stability of the constructs added further nuance. Between the first and second waves, depressive symptoms and aggressive behavior showed notably high stability, with autoregressive coefficients of 0.680 and 0.697 respectively, while loneliness was considerably less stable at 0.345. Loneliness, the authors suggest, is a socially constructed and cognitively mediated experience that is more state-dependent and therefore more malleable over time, whereas emotional and behavioral problems may have been particularly entrenched during the recovery period. Analyses comparing students with complete data to those with missing waves found no significant baseline differences on any measure, suggesting attrition did not bias the results, and exploratory multi-group analyses indicated the longitudinal associations were largely invariant across sex and grade level.

The findings carry concrete implications for practitioners supporting young people after disasters. Because the study’s evidence aligns with the developmental cascade framework, which views development as a system of mutually reinforcing interactions across emotional, behavioral, and social-cognitive domains, interventions targeting a single domain may produce cascading benefits elsewhere. The authors argue that early identification and treatment of depressive symptoms should be a priority, given their central and stable role, while the lower temporal stability of loneliness suggests that efforts to foster a shared sense of social connectedness within affected communities may be especially valuable. Even modest, coordinated multi-domain support, they contend, may contribute meaningfully to psychosocial recovery in resource-limited settings.

The researchers also acknowledge important limitations. The observational design precludes causal conclusions, the traditional cross-lagged panel model cannot cleanly separate within-person change from stable between-person differences, and the exclusive reliance on self-report measures may have inflated associations through shared method variance. Socioeconomic status was not assessed, a deliberate ethical choice given the sensitivity of income questions for families whose livelihoods were disrupted by the earthquake, leaving open the possibility of residual confounding. Because the sample was drawn exclusively from a disaster-affected, ethnically homogeneous Japanese community and lacked a comparison group, the authors caution that the observed asymmetry between depression and loneliness may reflect post-disaster circumstances or broader developmental processes, and they frame several of their interpretations as hypothesis-generating rather than definitive. Future research incorporating perceived social support, multi-informant measures, and earlier post-disaster time points could help clarify how young people’s emotional lives and social worlds reshape one another in the long aftermath of catastrophe.

The Wajima findings gain additional weight when placed against the broader epidemiological picture of youth loneliness. A meta-analysis drawing on data from 113 countries has estimated that roughly one in ten adolescents aged 12 to 17 experiences loneliness, while earlier surveys in Japan suggested that nearly one-third of children and adolescents reported the feeling. Loneliness, in this literature, is understood not as objective isolation but as a subjective aversive state arising from a perceived gap between desired and actual relationships, which helps explain why it can persist even in densely connected communities like Wajima.

Adolescence is considered a period of particular vulnerability because peer relationships, social evaluation, and identity formation all reach heightened salience, making the grade 3 through 9 span covered by the study developmentally informative. The pandemic-era research cited by the authors offers a useful parallel: adolescents worldwide reported elevated loneliness during COVID-19, and studies of collectively affected populations likewise found stronger longitudinal effects running from depression to loneliness than the reverse, mirroring the asymmetry observed after the earthquake.

The developmental cascade framework that anchors the study emerged from empirical demonstrations that internalizing problems, externalizing problems, and interpersonal quality influence one another reciprocally over time, meaning no single domain should be examined in isolation. Prior prospective work has shown depressive symptoms predicting later aggression, prosocial behavior buffering later depressive symptoms, and externalizing problems eroding later prosociality. By simultaneously modeling all four constructs, the Wajima study addresses a gap left open by research that considered these variables only in pairs, and it suggests that the cascade processes documented in general populations remain operative even under the unusual strain of a large-scale natural disaster.

Subject of Research: Longitudinal associations among loneliness, depressive symptoms, and social behaviors in youth after a major earthquake

Article Title: Longitudinal Interrelations Among Loneliness, Depressive Symptoms, and Social Behaviors in Youth: A Cross-Lagged Panel Study in a Post-Disaster Context

Article References: Murayama, Y., Nishiura, M., Soma, D., Kameya, M., Tanaka, S., Tsujii, M., & Kikuchi, M. (2026). Longitudinal Interrelations Among Loneliness, Depressive Symptoms, and Social Behaviors in Youth: A Cross-Lagged Panel Study in a Post-Disaster Context. Child Psychiatry & Human Development. https://doi.org/10.1007/s10578-026-02096-2

Image Credits: AI Generated

DOI: 10.1007/s10578-026-02096-2

Keywords: loneliness, depressive symptoms, aggressive behavior, prosocial behavior, children and adolescents, cross-lagged panel model, developmental cascade, post-disaster mental health, psychosocial adjustment, Noto Peninsula Earthquake, Japan, longitudinal study