You are here
How Bad Can Financial Crises Be? A GDP Tail Risk Assessment for Portugal
2022
Authors
Duarte Maia
Marina Feliciano
Ivan De Lorenzo Buratta
Publication Year
2022
JEL Code
C53 - Forecasting and Other Model Applications
E01 - Measurement and Data on National Income and Product Accounts and Wealth - Environmental Accounts
E17 - Forecasting and Simulation
E27 - Forecasting and Simulation
E32 - Business Fluctuations; Cycles
E44 - Financial Markets and the Macroeconomy
G01 - Financial Crises
Abstract
By monitoring the evolution of risks to economic activity over time, we quantify the likelihood and severity of future negative economic growth. Following the Growth-at-risk approach, we explore the non-linear relationship between the current financial situation and the distribution of future GDP growth for Portugal. We find that both financial vulnerability and risk have a negative effect on the left tail of the one-year-ahead GDP growth distribution. Financial
vulnerability has the largest impact on GDP growth at the medium to long term horizon while financial risk is only significant at the short term horizon. The GDP-at-risk measure signals economic recessions, no matter whether fueled by financial stress or imbalances, reaching negative values before 2008 and stagnating at low levels before the European Sovereign Debt Crisis. To provide policymakers with better tools to signal an increase in the likelihood of a crisis, we compute a set of complementary risk measures. Among those analyzed, the distance between the tails of the conditional distribution of GDP growth complements GDP-at-risk in anticipating economic recessions since it signals the Great Financial Crisis with a clear downward trend before 2008. The moments of the GDP growth distribution have some power in signalling recessions, as they identify changes in the characteristics of the distribution. Finally, we argue that the expected shortfall and longrise can complement the GDP-at-risk measure since they encompass information which is not limited to a single percentile of the distribution.
Document link