Background: Social capital is dynamic; however, little is known about the association of its dynamics with health resilience in a pandemic. We examined the longitudinal association of pre-COVID-19 cognitive social capital trajectories with peri-COVID-19 depression and the moderating effect of province-level COVID-19 severity on the association in China. Methods: We employed four-wave adult (≥ 16 years) data (follow-up n=6,228) from the biennial China Family Panel Studies between 2014 and 2020 with peri-COVID-19 depression in 2020 measured by the 8-item Center for Epidemiological Studies-Depression Scale. Pre-COVID-19 cognitive social capital from 2014 to 2018 included dichotomized (high/low) generalized trust, trust in neighbors, trust in local government officials, and reciprocity, with each of them categorized into five trajectories: persistently low, decreased, fluctuated, increased, and persistently high. Province-level COVID-19 severity in 2020 was a factor score constructed by the number of COVID-19 cases and deaths in each province. Mixed-effects linear regression was conducted to answer our research question. Results: Persistently low generalized trust (β: 0.46; 95% CI: 0.15,0.78), persistently low (β: 0.57; 95% CI: 0.22, 0.92), decreased (β: 0.36; 95% CI: 0.07, 0.65), and increased (β: 0.40; 95% CI: 0.12, 0.68) trust in neighbors, and persistently low (β: 0.39; 95% CI: 0.02, 0.77) and decreased (β: 0.68; 95% CI: 0.38, 0.97) reciprocity, compared with their persistently high trajectories, were associated with increased peri-COVID-19 depression. Province-level COVID-19 severity did not significantly moderate the association between pre-COVID-19 cognitive social capital trajectories and peri-COVID-19 depression. Conclusions: Long-term strategies to maintain high cognitive social capital are needed to protect mental health against a pandemic, regardless of the expected severity of the pandemic.