A growth-at-risk model for the German economy

This paper describes a Growth-at-Risk (GaR) model of the Bundesbank for Germany. This model takes the form of a quantile regression that quantifies downside risk to German GDP growth associated with financial developments. A systematic comparison of diverse model specifications is performed to select the most suitable GaR model based on economic criteria and out-of-sample predictive performance. The preferred model relates the 10% quantile of the conditional distribution of GDP growth to financial stress in Germany as captured by the Country-Level Index of Financial Stress (CLIFS), as well as US financial conditions as meas- ured by the National Financial Conditions Index (NFCI) for the USA. In addition, the preferred specification includes GDP growth of the two preceding periods to account for serial dependence and a business confidence indicator (BCI) of German companies, which underscores that economic sentiment also matters for downside risk to growth. The evaluation shows that the 10% quantile coefficients are more stable than those of the 5% quantile, making the 10% quantile a more robust measure of downside risk for German GDP. Data from the COVID period are excluded, as the pandemic was not a financial system-driven crisis. Estimation results show that financial stress, measured by both CLIFS and NFCI, contributed most strongly to downside risk to GDP growth during the 2007/2008 Global Financial Crisis. The CLIFS also significantly increased downside risk in the early 2000s and following the Russian invasion of Ukraine. In recent years, historically low financial stress has corresponded to moderate downside risk, with economic sentiment acting as the main amplifier.