Using structural models to understand macroeconomic tail risks

Understanding asymmetric risks in macroeconomic variables is challenging. Most structural models used for policy analysis are linearised and therefore cannot generate asymmetries such as those documented in the empirical growth-at-risk (GaR) literature. This report examines how structural models can incorporate non-linearities to generate tail risks. The first part reviews the various extensions to dynamic stochastic general equilibrium (DSGE) models and the computational challenges involved in accounting for risk distributions. This includes the use of occasionally binding constraints and more recent developments, such as deep learning, to solve non-linear versions of DSGEs. The second part shows how the New Keynesian DSGE model, augmented with the vulnerability channel as proposed by Adrian et al. (2020a, b), satisfactorily replicates key empirical facts from the GaR literature for the euro area. Furthermore, introducing a vulnerability channel into an open-economy set-up and a medium-sized DSGE highlights the importance of foreign financial shocks and financial frictions, respectively. Other non-linearities arising from financial frictions are also addressed, such as borrowing constraints that are conditional on an asset’s value, and the way macroprudential policies acting against those constraints can help stabilise the economy and generate positive spillovers to monetary policy. Finally, the report examines how other types of tail risk beyond financial frictions – such as the recent asymmetric supply-side shocks – can be incorporated into macroeconomic models used for policy analysis. JEL Classification: E70, D50, G10, G12, E52