Problematic Calibration of the EU Banking Sector Stress Test for Ireland
This post was written by Gregory Connor
The details for the calibration of the EU-wide bank stress test are now available. Looking only at Ireland, and only at one of the key variables in the stress test, the calibration looks problematic. It may be coincidental that the Irish adverse scenario has been badly chosen; it might be that all the other member countries have reasonable calibrations. If the others are as problematic as in the Irish case, this is not a reliable EU banking sector stress test.
Under the adverse scenario, Irish property prices are assumed to suffer a cumulative three-year drop of 3.03%; equivalent to a decline of 1.02% each year for three years in a row. Over the period covered by CSO data, 2005-2013, Irish residential property prices had an annual sample volatility of 11.7%. This in turn implies (under reasonable assumptions) a three-year volatility of 20.27%. In risk analysis it is conventional analytical shorthand to measure adverse outcomes in “x-sigma” units defined as the outcome as a multiple of the standard deviation. For an adverse scenario calibration, the assumed outcome is usually roughly a two-sigma or three-sigma event. Using a four-sigma shock would not be unusual (due to fat tails in some probability distributions). The EBA has calibrated the adverse price shock as a 0.1492-sigma event. That is not credible as an adverse scenario in a stress test.
Keep in mind that the stress test is meant to reassure market participants that even in an adverse scenario the Irish banks are sound. This test reassures us that if property prices fall by as much as one percent a year over the next three years, the banks have enough capital. In the case of a two-percent fall, there are no promises.
As a caveat, this does not mean that the Irish banks need equity capital. They have already had a credible stress test (in 2011) and a big capital injection. Also, the Irish property market although very volatile has a maximum likelihood price change which is positive over the next three years. However the asset class also has considerable “downside” potential and continued high volatility. Conventionally, at least in the case of portfolio risk analysis, the unconditional mean of a stressed variable is set equal to zero for risk analysis purposes. The EBA has chosen to build in a big positive benchmark price rise for Irish property assets, and this is part of the reason that the adverse scenario is unacceptably mild. In any case, this calibration is extremely mild as an adverse scenario and not reassuring for the EU-wide test.