European Stress Tests

The announcement of the results of Europe’s bank stress tests on July 23rd was greeted with relief and rallies in bank shares and the euro. It also sparked widespread criticism of the adequacy of the tests and cynicism regarding Europe’s cosseting of the ninety-one banks (accounting for 65% of European bank assets) that were tested. Several commentators drew unfavorable comparisons with the stress tests that were applied to U.S. banks earlier in the year. Both reactions were thoroughly predictable, but it is interesting to consider why this should be so.

Relief was largely a matter of how weak confidence in Europe has been in the wake of the Greek crisis. All the banks that failed the test – Germany’s Hypo Real Estate, Greece’s ATEBank and five Spanish savings banks – were thought to be in difficulty well before the announcement of the results. One of the Spanish cajas that failed the test even anticipated its result by issuing €450 million of convertible debt to J.D. Flowers. So the test results did nothing to add to the general gloom, and investors’ reflex reaction was to mark up prices. I am reminded of an interesting market commentary that ran on the DJ Newswire many years ago: “The market rose on lack of bad news.” Critical reactions were equally reflexive. It is to be expected that there will always be those who feel that the stress scenarios applied to the banks were not stringent enough, and who, in consequence, suspect the authorities’ motives. The justice of their objections to the E.C.B.’s methodology – among them its unwillingness to extend price haircuts to banks’ held-to-maturity positions and its debatable treatment of sovereign exposures – are to some extent neither here nor there. The simple fact that tests were performed made criticism along these lines inevitable.

All of which begs the question, how stressful should stress tests be? At some level of stress, every bank is unsound, and given empirical evidence that >20σ events can occur in financial markets, there is no a priori case for calibrating tests to any given level of stress. For all that it is recurrent and even patterned, stress is inherently improbable – otherwise it would not be stress, but business-as-usual. This is why VaR and similar measures of banks’ quotidian risks are supplemented by stress tests in the first place. Cynics argue that authorities calibrate their tests to a stress level that will guarantee that their results do no damage to confidence in the banking system. There is inevitably some truth in this: it is hardly the authorities’ intention to induce stress by testing for it. They must obviously weigh the danger of diminishing confidence through excessive stringency against the risk of discrediting the stress testing exercise. But it is not remarkable, in the current state of European banking, that a 92% pass rate, the revelation of no unexpected weaknesses in the system and an estimated aggregate capital shortfall of only €3.5 billion suggest to some commentators that the tests might appropriately have been more stringent.

The challenge to the authorities is to determine the failure rate that is consistent with the maintenance of confidence in the banking system, and what and how many unexpected test failures confidence can tolerate. The alternative is to refrain from public release of stress test results. Since public release of the test results is itself a confidence-building exercise, it is unreasonable to expect them to behave otherwise. However, it should give us pause that European authorities apparently feel that confidence is so fragile that calling attention only to well-recognized weaknesses that require only a small capital infusion to correct is all that confidence in their banking system can withstand.