Abstract: This paper examines whether standard measures of systemic risk are prone to survivorship bias. Using data on more than 2,500 financial institutions across 32 countries over 1995-2025, we show that relying only on surviving firms leads to a systematic overstatement of systemic risk of about 13 basis points, or 5%, at the country level. The bias can be decomposed into two complementary effects. A firm-level weighting effect (WE), unique to systemic risk estimation, arises because institutional exits reassign system-index weights to surviving firms, mechanically inflating their measured contributions. A country-level omission effect (OE) reflects the exclusion of exiting institutions that are, on average, less systemically important, partly because many are "too-small-to-save" and because their observed histories end before major systemic events. A Monte Carlo framework shows how institutional size, interconnectedness, and exit timing jointly determine the sign and magnitude of the bias, and indicates that the effect can become negative when very large and highly interconnected institutions exit. The results also imply that survivoronly datasets systematically overstate the importance of long-lived institutions while understating risks associated with early-exiting and newly established firms. Overall, ignoring survivorship bias can distort stress testing, crosscountry comparisons, and the identification of systemically important institutions, with direct implications for macroprudential monitoring.

Abstract: Cyber risk has become a growing concern for financial markets and regulators, yet its implications for financial stability remain insufficiently understood. Using publicly disclosed data breaches among S&P 500 firms over 2010-2024, this paper examines the systemic footprint of cyber incidents. We find that data breaches are associated with an increase of approximately 3% in firms’ contribution to systemic risk, with the effect concentrated in economically meaningful incidents, while minor breaches exhibit no detectable impact. The effect is primarily driven by hacking-related events and is short-lived, concentrated in the immediate aftermath of the breach. We find little evidence of sectoral heterogeneity, indicating that the association extends beyond financial institutions to non-financial firms. We further show that the systemic impact of breaches is state-dependent and has strengthened over time, with larger effects observed in periods of elevated market volatility and for delayed disclosures, while the marginal impact of individual breaches is attenuated when cyber incidents are more widespread. The results are robust across specifications, including a stacked difference-in-differences design that delivers comparable estimates and shows no evidence of pre-trends. Overall, the findings are consistent with the view that cyber incidents are not purely idiosyncratic shocks but may have broader systemic implications, highlighting the importance of incorporating cyber risk into systemic risk monitoring and macroprudential policy frameworks.

Abstract:  We examine how bank branch closures affect non-tradable local demand using census-tract nighttime lights (NTL) in the United States from 1994 to 2020. To isolate plausibly exogenous variation in branch closures, we exploit within-county, tract-level exposure to post-merger branch consolidation, that is, the closure of redundant branches arising from greater network overlap rather than from local economic conditions. Instrumental-variable estimates show that closures reduce NTL, indicating declines in non-tradable local demand. Effects are strongest in areas with older, lower-income, lower-density, and lower Black populations, as well as in high-mobility counties. Spillover tests indicate that the impacts are highly localized, fading quickly beyond adjacent tracts.