ALBERT

Description: We study yield spreads between government bonds in the European Monetary Union. This segment of the global fixed income market is of particular importance for insurance companies in Europe. Our empirical research strategy is inspired by Gunay (2020) who has analyzed the relationship between credit and liquidity risk in the United States using Granger causality tests. More specifically, we employ the procedure developed by Toda and Yamamoto (1995) to test for Granger causality among yield spreads in five different member countries of the European Monetary Union (namely Austria, Belgium, France, Italy and Ireland) relative to Germany. We examine interest rate data from bonds with three different maturities (5, 10 and 30 years). Given the importance of long-term bonds as asset class for European life insurers and pension funds, the empirical results from the often ignored market for government bonds with a maturity of 30 years should be of interest. With regard to long-term sovereign debt, there is no evidence for Granger causality among the time series examined here. Consequently, the risk premia required by investors to hold government bonds of one specific member country of the EMU do not help to forecast the risk premia that have to be paid by other countries. Given the structure of their liabilities, this empirical finding should be of high relevance for portfolio and risk managers in the European life insurance industry and in pension funds. With regard to the yield spreads to be observed in the market for 10-year government bonds, there seems to be no clear picture. Focusing on fixed income securities with a maturity of 5 years, there is one very interesting empirical finding. The test results reported here seem to imply that there is unidirectional Granger causality running from the yield spreads in all other four countries to Austria. Given that Austria is a comparably small country which is assumed to be in a fiscally stable position, this result could be interpreted as evidence for credit risk premia as being helpful to forecast liquidity risk premia in the market for medium-term government bonds issued by member states of the European Monetary Union.

Description: ZusammenfassungDie Möglichkeit einer Prämienanpassung in der deutschen PKV ist vom Wert des sogenannten auslösenden Faktors abhängig, der mittels einer linearen Extrapolation der Schadenquotienten der vergangenen drei Jahre berechnet wird. Seine frühzeitige, verlässliche Vorhersage ist aus Sicht des Risikomanagements von großer Bedeutung. Wir untersuchen deshalb vielfältige Vorhersageansätze, die von klassischen Zeitreihenansätzen und Regression über neuronale Netze bis hin zu hybriden Modellen reichen. Während bei den klassischen Methoden Regression mit ARIMA-Fehlern am besten abschneidet, zeigt ein neuronales Netz, das mit Zeitreihenvorhersage kombiniert oder auf desaisonalisierten und trendbereinigten Daten trainiert wurde, das insgesamt beste Verhalten.