Systemic Cyber Risk (with Ahmed Tahoun)
Granular Credit Risk (with Sigurd Galaasen, Ragnar Juelsrud, and Hélène Rey)We provide the first causal quantification of single-name concentration risk for banks using a novel administrative matched bank-firm dataset from Norway. Our identification strategy is twofold. At the level of the loan, we compare loan outcomes of firms that borrow from the same bank, are from the same ZIP code and industry, but experience different idiosyncratic shocks. At the bank level, we first document that the loan size distribution is thick tailed. We then exploit that heavy tail and construct a Gabaix-Koijen (2019) granular instrumental variable (GIV) that purges out confounding bank-side factors. We find that GIV-instrumented firm shocks have a strong impact on bank-level returns and writedowns. The effect is strongly concave, consistent with the payoff structure of the debt contract. Using detailed data on banks' non-interest income we find that granular credit risk goes largely unhedged. In the aggregate, granular credit risk can explain up to 20\% of the variation in total banking sector returns on corporate loans. Our results are useful for calibrating the granularity adjustment in the regulation of credit concentration risk.
The Cross-Section of Risk-Taking and Asset Prices (with Nuno Coimbra and Hélène Rey)The distribution of institutional investor risk-taking carries significant explanatory power for the cross-section of asset returns. We compute an investor-level Value-at-Risk (VaR) measure - our proxy for ex-ante riskiness - from a structural model with stochastic volatility that we estimate with a particle filter. Our pricing factor - CrossRisk - is then constructed from shocks to the procyclical dispersion of the time-varying VaR distribution. CrossRisk is able to price equity, bond, CDS, options, currency, and commodity market portfolios comparably to numerous single and multi-factor benchmarks. We show that the mechanism behind our results is the extensive margin - dynamic entry and exit of investors into the risky market. A synthetic high minus-low CrossRisk beta pre-sorted equity portfolio built on the full universe of CRSP firms has an annualized returns spread of 5.8%.