CDS and the forecasting of bank default

File pdf 1.4MB

The last financial crisis hit worldwide economies with an unprecedented magnitude. In order to predict bank default or banking crises, some empirical papers have been written ever using sophisticated tools that can use models from operational research. Based on a short analysis of the forecasting power of several types of financial products, we conclude that CDS characteristics are the best measure to forecast and thus ideally prevent the potential default of a bank. Thanks to the economics of CDS and the results of other empirical studies, we show that CDS spreads are undoubtedly a good, though not a perfect proxy for bank risk, even though they are more sensitive to information changes than other products.
So, by creating and using a specific trigger based on CDS and the appropriate response, in the case of the trigger being activated, we examine if we could prevent the default of a bank. But as CDS spread cannot be taken as a perfect proxy for the true probability of default of the underlying corporate entity, we had to investigate further to try to find another benchmark. Initially, a good candidate appeared to be the Markit 5-year iTraxx Senior Financial index. So, using the CDS of each bank and this index, we first applied the following procedure: an intervention should be triggered whenever the CDS price is above 100 bps for at least 20 of the last 30 trading days.
We studied 50 among the TOP 100 European banks from a 63 European bank sample for the period from 2007 to 2013 and examined a few in detail results. We found that one or even two triggers at 100 bps gave disappointing results, as most of the banks went over this limit during the second year of our period of study. Subsequently, we were better able to see how to manage some properties that we had identified by using a meta-rule that took account of the lapse of time between two thresholds based on CDS spreads in order to forecast any significant financial distress for a bank. This gave more reliable results for forecasting bank default than our less sophisticated first approach and could be useful to regulators.

Author(s):

Eric Thorez    
Paris Dauphine University - PSL
France