We evaluate the impact of contagion and common exposures on banks’ capital using a structural regression framework derived from the balance sheet identity and inspired by the structural VAR literature. Contagion arises through bilateral exposures, fire sales, rollover risk, and market-based sentiment, while common exposures reflect overlapping portfolio holdings. We estimate the model using granular regulatory balance sheet and interbank exposure data for the Canadian banking sector. Our results yield three key insights. First, contagion driven by bilateral contractual exposures remains relatively stable over time until the onset of quantitative easing. In contrast, non-contractual contagion channels are less stable and move with market conditions. Second, we observe an increase in common exposure risk along with a decrease in contagion risk, following unprecedented fiscal and monetary policy measures in the COVID-19 pandemic. Third, we demonstrate how our framework complements traditional bank stress-testing approaches that focus on individual institutions by analysing second-round effects. In a policy application, we simulate targeted bailouts and show that their effectiveness in stabilizing the system is related to the interconnectedness of the rescued institution.