Backtesting Macroprudential Stress Tests

Abstract

In this paper, we consider models of price-mediated contagion in a banking networkof common asset holdings. For these models, the literature proposed two alternativeclasses of liquidation dynamics:threshold dynamics(banks liquidate their invest-ment portfolios only after they have defaulted), andleverage targeting dynamics(banks constantly rebalance their portfolios to maintain a target leverage ratio).We introduce a one-parameter family of non-linear liquidation functions that inter-polates between these two extremes. We then test the capability of these modelsto predict actual bank defaults (and survivals) in the United States for the years2008-10. We show that the model performance depends on the type of shock be-ing imposed (idiosyncratic versus systematic). We identify the two most relevantasset classes, for which the model has predictive power when these asset classes areexposed to an initial shock. In these cases, the model performs better than alter-native benchmarks that do not account for the network of common asset holdings,irrespective of the assumed liquidation dynamics. We also show how the best per-forming liquidation dynamics depend on the combination of the initial shock leveland the market impact parameter, on the cross-sectional variation in the marketimpact parameter, and on the number of asset liquidation rounds.