Authors: Willem Pieter De Groen and Daniel Gros
Series: CEPS Commentary
The latest round of stress test conducted by the European Banking Authority for 51 large banking groups showed huge differences in the degree to which a shock, or so-called adverse scenario, would impact different banks. We find that banks with a high exposure to sovereign risk and especially those with many non-performing loans tend to be less resilient. By contrast, national factors, such as growth rates, had little impact on the outcome. 
Since the outbreak of the Great Financial Crisis, it has become customary to subject banks to so-called stress tests. In Europe, these tests are managed by the European Banking Authority (EBA) for systemic banking groups, which account for the bulk of Europe’s banking system.
The results of the latest round were published in July and led to yet another bout of volatility in bank shares.
The stress tests are supposed to show how much banks would lose if the economy and financial market tanked. The results suggest that overall the European banking system has become much more resilient. Few banks would need to be recapitalised even under a so-called ‘adverse’ scenario, which includes a cumulative reduction in growth of 7%, an increase in risk premia and a large drop in house prices.
Technically speaking, the result was that the 51 banks would lose on average 3.4% of the “fully-loaded common equity tier 1” (CET1)’ under the stress scenario. Given that the average CET1 ratio, which is a key regulatory indicator, stands at 12.6%, this seems manageable.
What is surprising, however, is that the impact of the adverse scenario varies hugely across banks. At one extreme one finds the Italian Banca Monte dei Paschi di Siena (MPS), which would lose 14.5% CET1. At the other extreme, there is the only Norwegian bank in the sample – DNB – which would lose less than 0.1% of its CET1.
What explains these huge differences?
One seemingly obvious answer might be the (national) economy. The Norwegian economy is doing clearly much better than the Italian. But an initial attempt to relate the stress-test results to national growth rates did not yield any significant relationships.
This empirical result should not be surprising since the underlying argument turns out to be wrong. First of all, the current state of the economy is already visible in the starting balance sheet. The stress test is designed to measure the impact of a change. What matters is thus not the starting point of the economy, but by how much it deteriorates under the so-called stressed scenario. To put it differently: a stress test does not measure strength, but resilience. And if one looks at the details of the stressed scenario, one finds that often the weaker economies are actually (assumed to) do better than the stronger ones. The stress test assumes, for example, that Germany’s (nominal) GDP declines by 3.7%, much more than the other large euro-area countries France (-0.1%), Italy (-1.0%) and Spain (-0.4%).
There are thus two reasons why the often-heard argument that the weakness of Italian banks is due to the country’s weak economy does not explain why Italian banks did not perform well under the stress tests: these tests measure resilience to shocks, not strength per se. Moreover, the 2016 tests actually assume that the economic stress is stronger in Germany than in Italy.
Given that national variables do not seem to explain the stress test results, we turned our attention to key balance-sheet indicators of the individual banks, concentrating on two with particular policy relevance: the share of non-performing loans (NPL) in total exposures and the importance of sovereign exposure in the overall balance sheet.
Sovereign exposure is regarded as riskless by regulation and many have argued that this should remain the case because if a bank holds many government bonds it should be less affected by economic shocks. However, this is not what we found. A consistent finding of our work is that the higher the proportion of sovereign exposure, the stronger the impact of the adverse scenario on the bank. In other words, more sovereign exposure makes a bank more susceptible to risk. This finding should be taken into account in the ongoing discussions on the regulatory treatment of sovereign debt in the euro area.
Over the last few months there has been much discussion about what to do with the NPLs, which represent in particular a problem in Italy. Loans that have already been classified as non-performing, represent an inheritance from the past. One would thus expect that a bank that starts the stress test with more NPLs would not necessarily have a worse result than banks with fewer NPLs. But this not what we found.
The single-most important predictor of the impact of the adverse scenario on the capital position of a bank is its share of non-performing loans at the start of the test. Moreover, we found that the impact of non-performing loans on the stress test results grows with the size of NPLs. A bank that has twice as many NPLs shows, ceteris paribus, an almost four times larger loss under the adverse scenario. A high level of NPLs, even if appropriately provisioned for, thus reduces the ability of a bank to withstand shocks.
Overall, our empirical results suggest that one could have predicted over 60% of the variability of the results of the stress tests across banks with just two indicators: NPLs and sovereign exposure. The results of this stress test thus suggest that overall, the European banking system has become more resilient. But there remain significant pockets of weakness, especially for banks that combine high levels of NPLs with high degrees of sovereign exposure.
Formally, the 2016 stress tests were less important than those of the past. In previous rounds and other capital exercises, the EBA used a threshold to determine whether a bank had to raise capital. This time the results of the exercise feed ‘only’ into the Supervisory Review and Evaluation Process (SREP), which the direct supervisors of the banks (in the Euro-area de facto now an arm of the ECB) use to determine the add-ons to the legislative capital requirements.
The ECB should take the results reported into account when it performs the SREP later this year. Sovereign exposure should be regarded as a risk factor, and NPLs should not be looked at only with an eye to provisioning, but as an indicator of additional high risks.
Willem Pieter De Groen is Research Fellow in the Economy and Finance research unit at CEPS. Daniel Gros is Director of CEPS.
CEPS Commentaries offer concise, policy-oriented insights into topical issues in European affairs. The views expressed are attributable only to the authors in a personal capacity and not to any institution with which they are associated.
© CEPS 2016
 See also related companion analysis by Willem Pieter De Groen (2016), “The EBA EU-wide Stress Test 2016: Deciphering the black box”, CEPS Policy Brief No. 346, CEPS, Brussels, August.
 This scenario assumes that GDP growth will not necessary be negative, but will turn out to be 7.1% lower than presently forecast by the European Commission. For further details, see http://www.eba.europa.eu/-/eba-launches-2016-eu-wide-stress-test-exercise
 See Willem Pieter De Groen (2016), “A closer look at Banca Monte dei Paschi: Living on the edge”, CEPS Policy Brief No. 345, CEPS, Brussels, July.
 See Daniel Gros (2014), “Banking Union with a Sovereign Virus: The self-serving regulatory treatment of sovereign debt in the euro area, CEPS Policy Brief No. 289, CEPS, Brussels, March (www.ceps.eu/publications/banking-union-sovereign-virus-self-serving-regulatory-treatment-sovereign-debt-euro) and the 2015 report of the Advisory Scientific Committee to the ESRB (www.esrb.europa.eu/pub/pdf/other/esrbreportregulatorytreatmentsovereignexposures032015.en.pdf).
 Besides those in Italy, banks in Cyprus, Greece, Slovenia and Portugal also have large stocks of NPLs.