The EU Bank Stress Tests—What Have We Learned?by Jacob Funk Kirkegaard | July 26th, 2010 | 05:51 pm
On Friday, July 23, the Committee of European Banking Supervisors (CEBS), along with national EU bank regulators and—crucially—most of the affected individual banks, finally released new data to the public regarding the stress testing of the European banking system.
What have we learned?
Appearances matter—especially for bank stress tests. So first it is clear that whoever has been in charge of media relations and communication at CEBS, the EU Commission, and elsewhere must immediately be fired (and denied any EU pension)! Despite the still fresh memory of what happened at the publication of the US Supervisory Capital Assessment Program (SCAP) tests in May 2009, which were roundly (but erroneously) criticized in the financial media and by pundits as too lenient toward the 19 banks tested, CEBS and other EU authorities carried out an inadequate and incoherent media strategy that condemned the EU stress tests to a similar fate. This is very unfortunate, as there is a lot of new value in the new data.
The public relations mistakes do not affect the longer-term value from these stress tests, however. Now that the results are out, financial markets will shift their concerns away from European sovereign governments (Europe is not bust!) and toward individual European banks. Despite the many flaws of the stress test process and the fact that the EU banking system has overall NOT received anything near a clean bill of health, EU leaders have probably met their key objective in launching this exercise.
When looking in more detail at the stress tests, it is worth recalling some often forgotten circumstances preceding them:
The EU stress tests come after a lot of government capital support has already been provided. The fact that only 7 out of 91 EU banks failed and only required an additional €3.5 billion in new capital, when compared to 10 out of 19 US banks and $75 billion in new capital in 2009, is widely taken as a proof that the EU stress tests were far weaker than the (already perceived weak) US SCAP stress tests. But this assumption ignores the fact that since late 2008, EU governments have provided €170 billion in government capital support to the 91 banks included in the stress tests, or approximately 1.2 percentage points [pdf] of their aggregate Tier 1 capital ratio at the time of the stress tests.
A large part of the reason why so few EU banks failed and needed to raise so little capital is that their governments had already provided a lot of support pre–stress tests. Without this earlier support, many more EU banks would have failed and substantially more capital would have been needed after the stress tests were published. It is unfortunate that EU regulators and governments have not emphasized this fact, though perhaps they were trying to avoid reminding their electorates about their earlier and unpopular largesse toward their banks. Many EU banks have increased their capital cushions since the beginning of the crisis. As pointed out in the Bank of England Financial Stability Report from June 2010, average Tier 1 capital ratios in the German, French, Italian, Spanish, and UK banking systems increased from 2008 to 2009 by 1.77, 1.59, 1.19, 1.22, and 2.98 percentage points of Tier 1 capital, respectively.1 These earlier capital increases mean that EU banks went into the stress tests in July 2010 in far better shape than they would have otherwise.
Combined with the fact that many EU banks merely scraped by with just above the 6 percent Tier 1 capital threshold, these earlier capital increases make a lot of difference. For example, if Germany’s participating banks had had on average 1.77 percentage points lower Tier 1 capital at the time of the stress tests (i.e., had they not benefitted from the increases in 2008–09), four more banks would have failed the tests.2
Obviously, EU leaders allowed these stress tests to go ahead in 2010, knowing that their banks had increased the capital cushion substantially. Timing is everything—also in bank stress tests.
The EU stress tests’ macro assumptions vary among participating countries. The only substantial information provided by CEBS and national regulators about the macroeconomic assumptions used in the stress tests was the aggregate 3 percent deviation in GDP growth assumption provided on July 7. This is unforgivable as a communication strategy, even if it reflects the accelerated timetable for publication set in June 2010. The lack of prereleased stress test assumptions virtually assures that insufficient attention will be paid to the surprisingly large country differences in the application of the stress test assumptions. This is very unfortunate, considering the wide divergence in the macroeconomic circumstances among the participating countries.
Applying a “one size fits all” set of macroeconomic assumptions within the European Union would virtually assure that the lowest common denominator ruled. Fortunately, though, this is not the case, as it is illustrated in tables 1 and 2.
Table 1 [pdf] shows all the country-specific GDP and unemployment growth assumptions utilized in national EU stress tests, as well as the assumptions relied upon in the 2009 US SCAP tests. A striking feature is the large national divergences in the “GDP growth stress level,” ranging from a 4.3 percent cumulative decline in GDP growth in Slovakia to only a 1.7 percent cumulative decline in Sweden and the Czech Republic. The difference reflects varying national economic circumstances. It is appropriate, for example, to assume that German GDP could decline a relatively stiff cumulative 3.3 percent, when compared to just 2.1 percent in France. As can be seen in the data for 2009, the export-dependent German economy saw a decline of 4.9 percent, compared to just 2.6 percent in more stable France.
Table 1 also illustrates how the EU stress tests in many countries assume a double dip recession, following the deep initial EU recession in 2009. This is in contrast to the US SCAP tests, which modeled only the depth of the “first dip” in the United States and (correctly) assumed a return to US GDP growth in 2010 even in the adverse scenario.
Table 2 [pdf] shows the differences in assumptions concerning the national commercial and residential real estate markets among EU members. As one would expect, given the extremely different circumstances among EU real estate markets, the differences shown in table 2 are far larger than was the case in table 1. Spanish regulators, clearly aware of Spain’s troubled real estate sector, assumed an outlook far worse than in any other country in the European Union. Ireland, which shares a housing bubble experience similar to that of Spain, has also assumed a very bad outlook. Both countries assume two-year real estate price declines as large (or larger) than was the case in the US SCAP tests, on top of the declines already experienced in 2008 and 2009.
The German adverse assumption of a 10 percent annual decline in all real estate prices in both 2010 and 2011 also looks fairly tough, especially considering that Germany—unlike most other EU members—has experienced no residential real estate bubble. The tests make similar adverse assumptions for the UK real estate market. Indeed, the benchmark assumptions of small residential real estate price increases in the UK in both 2010 and 2011 seem decisively bullish and must assume an extremely accommodating monetary policy in the United Kingdom (such as expanding quantitative easing?), given the fiscal policy retrenchment announced by the UK government over the same period.
Benchmark assumptions of real estate price increases in, for instance, Denmark, combined with only modest price declines in the adverse scenario, similarly make the Danish stress test decisively soft.
Concerns over EU Banks Sovereign Debt Exposures Can Now Be Largely Dismissed, but not Because of the Stress Tests. The principal tail-end event hovering over the EU banking system is the extent of European banks’ exposure to the EU sovereign debt crisis, in particular Greek sovereign debt. Obviously, this issue has been politically tricky for EU policymakers. They have been reluctant to articulate the unthinkable (e.g., a default) for Greece. At the same time, they are trying to shore up confidence in the country’s economy. The political compromise has been to assume a general deterioration of EU sovereign debt markets in the form of rising yields for 5-year bonds, and then to deduct haircuts based on these rising yields. These are presented in table 3, reproduced from the CEBS summary report.
Table 3 [pdf] shows how the implied adverse haircut for Greek sovereign bonds at the end of 2011 would be 23.1 percent, 14.1 percent for Portuguese bonds and so forth, down to 4.7 percent for German bonds and 4.2 percent for Slovenian bonds. In many ways, this political compromise is a bizarre exercise that assumes that EU government bonds go up and down together. As this crisis has shown, that is demonstratively not the case. Safe-haven German bonds have strengthened dramatically, as those of peripheral countries have weakened. The stress tests should have overlooked political inconvenience and assumed that German bond yields would decline as those of Greece and other peripheral eurozone members decline.
A more serious critique is that these haircuts are only applied to EU banks’ “trading books,” i.e., the part of banks’ balance sheet that is marked to market. Government bonds that are assumed as held to maturity (in the so-called “banking books”) by banks escape the haircut. As around 90 percent of government bonds are typically assumed in the category of held to maturity, this accounting issue has huge implications on the severity and credibility of the stress tests. As a sovereign default scenario would not distinguish between where banks have assigned their bonds, not applying haircuts to sovereign bonds held in banks’ banking books has the potential of being seen as assuming away the entire issue of a Greek default or restructuring.
This politically determined outcome of the stress tests damages the likely perception of the results in the financial media and among superficial pundits. However, here it is important to distinguish between the stress test results themselves and the potential benefit provided from the new data.
While the stress tests themselves are of limited use, the release of participating individual banks’ holdings of sovereign bonds in both their trading and banking books (i.e., all of them) is highly beneficial, especially if one is trying to calculate the potential effects of a Greek default. The data enable analysts to construct their own scenarios for how future sovereign crises will affect the health of EU banks.
At the time of writing, sovereign bond holdings for 84 of 91 participating banks are publicly available. Only six German banks—Deutsche Bank, Deutsche Postbank, DZ Bank, Hypo Real Estate (HRE failed the tests), Landesbank Berlin and WGZ Bank—and one Greek bank, the (failed) ATEBank have not released these data. This failure is inexplicable. One can only hope that market stigma will force them to come clean rather than be seen as having something to hide. The lack of transparency from Deutsche Bank, which has global aspirations and a high-profile CEO, is particularly puzzling. Why would Deutsche Bank want to be categorized with the European Union’s failed banks?
It is regretful that legal issues and requirements for “voluntary data release” of this very important information by the participating banks have prevented CEBS from centrally presenting them at their website. The availability on the Internet of these data points should have been accompanied by direct links to all respective banks provided by CEBS or others. The failure to do so is another example of the flawed media and communication handling of the EU stress test.
As a service to RealTime readers, the sovereign exposure data for the 84 available EU banks is available here in table 4 [xlsx]. Note that most of the sovereign data is from March 31, 2010, and hence predates recent changes in European Central Bank (ECB) policies. Any deterioration of the ECB balance sheet as a result of its change in policies therefore will have proportionately reduced the sovereign bank holdings of Greek bonds, for example. Only some of the German data is dated May 31, 2010. These German banks may—not without irony—therefore have benefitted from selling bonds directly to the ECB before releasing their sovereign holdings data.
With the individual bank sovereign holdings available, it is clear from table 4 how national banks tend overwhelmingly to hold the bonds of their own governments. This example of home bias is unsurprising, even if its extent more than 10 years after the introduction of the euro is perhaps surprising.
What do the data tell us about total sovereign exposures of participating EU banks? As mentioned, analysts can assume any sovereign outcome and further stress test–participating banks accordingly. Despite the weak assumptions in the published stress test results, the availability of information about the individual banks’ sovereign holdings negates this critique of the entire stress test exercise. The only problem is that analysts and pundits will have to do their own stress tests. Undoubtedly and regretfully this will prove a prohibitive obstacle for most journalists and pundits, who seem likely to therefore continue with a superficial critique. A proper media and communication strategy could have addressed the criticism.
Since the beginning of the sovereign crisis, the European Union has managed to put together a credible policy response. It has come in the form of the Greek bailout, subsequent establishment of the European Financial Stability Facility (EFSF), accompanying austerity packages and structural reforms in EU member states and the results of these stress tests themselves. Most sovereign exposure concerns today therefore ought to center around Greece alone, and whether or not it will ultimately have to restructure its debts. As another thought experiment, this RealTime piece will therefore, relying on the sovereign bank exposure data, do a small, quick-and-dirty additional stress test of the effect of a Greek default on Europe’s banks.
The scenario will go beyond the EU stress test adverse scenario and assume that atavistic Greeks, will become fed up with austerity measures after following the IMF program until the end of 2011 and then repudiate all Greek sovereign debt. Note that Greece may then have a primary surplus thanks to the IMF and may seize on what is perceived to be an opportune time to walk away from its debt obligations. It’s a highly simplistic exercise that assumes that the effects of a debt repudiation are contained to Greece, with no dynamic effects or spillovers to other countries. The effects of such a scenario (which is labeled “Adverse Scenario”) are presented in table 5 [xlsx].
Table 5 shows several things. First, the revealed location of Greek sovereign debts affects 84 EU bank balance sheets with a total of about €81.5 billion, with about 90 percent of these in the banking books. This is about 60 percent of the total outstanding claims (ultimate risk basis) of European banks of $183 billion (i.e., approximately €135 billion at the end-Q1 foreign exchange rate) against Greece, as reported by the Bank for International Settlements (BIS) at the end of Q1 2010. Substantial nonsovereign exposures related to Greece therefore also exist against Greece.
Table 5 shows how this adverse scenario, on top of the existing EU stress tests, results in the predictable collapse of the entire Greek banking sector. No EU country’s domestic banking sector could likely survive the complete repudiation of its sovereign’s debt. Table 5 then further shows how the immediate contagion is limited to only Cyprus, where the two participating banks also collapse.
However, no other EU bank will collapse from a Greek default alone. A significant deterioration is visible in the Portuguese banking sector, but, apart from that, only Belgium’s Dexia Bank, France’s SocGen, Germany’s Commerzbank and Luxembourg’s Banque Raiffeisen suffer a decline in Tier 1 of 1 percentage point or more, and none of these banks (including in Portugal) drop below the 6 percent threshold.
Superficially therefore table 5 suggests that a complete but contained Greek default has relatively modest spillover effects through the EU banking system. Concerns about the catastrophic impact of a Greek default across the European economy thus seem exaggerated. This is a finding that should bode well for longer-term market reaction to these stress tests.
The results of the bank stress tests and particularly the sovereign debt holdings data should also serve to reopen the European wholesale funding markets for many participating banks. This should especially be the case for the stronger Spanish banks shown now to be well-capitalized by the generally rigorous Spanish tests and holding limited sovereign exposures to Greece, as these have widely been perceived to have been shut out of these markets in recent months. Many banks should therefore have to rely less on ECB liquidity in the future, a key objective for policymakers. Given the generally very diverse capital levels revealed with many EU banks close to the 6 percent threshold, it is unclear to what extent these stress tests will serve to reduce the London Interbank Offered Rate (LIBOR) or Euro Interbank Offered Rate (EURIBOR) interest rates charged by banks to one another.
As for the longer-term implications of the EU stress tests, the real impact will likely come, not from the flawed results themselves, but from the increased transparency resulting from the process. As such, one should not expect a huge rally in EU bank stocks (as we saw after the US SCAP tests were published). The EU stress tests have been too weak to constitute a “turning point” in this regard.
It is clear, moreover, that while many EU banks have passed the stress tests, not least due to the capital already raised, their passing grade was narrowly met. The limited number of “failures” resulting from these tests is no indication that the EU banking system is well capitalized. On the contrary, while perhaps not on the brink of collapse, a lot of EU banks are distinctly poorly capitalized and these stress tests have illustrated just who they are. With regulators refusing to demand that they raise more capital, financial markets may convince them to do so.
The leaders of the world’s major economies have pledged a substantial increase in the global bank capital requirements by the time they meet at the G-20 in Seoul later this year. Their commitment poses a daunting challenge for large parts of the EU banking system as it faces market demands to raise lot more capital. European governments concerned over constraints to their banks’ ability to lend should not simply blindly oppose any tightening in the global bank capital requirements. Rather they should take appropriate measures to avoid such an outcome.
This author considers the passing of so many of Europe’s banks of these stress tests a lost opportunity to consolidate the banking sector and expunge unwarranted regional government meddling in the banking sector, especially in Germany’s many and weak Landesbanken. Instead one must hope that government measures will ensure that Europe’s banks adhere to a new Basel III capital requirement regime that forces needed consolidation in select parts of the European banking system.
Basel III will therefore hopefully finish what the stress tests—most noticeably in Spain—have begun.
2. In the case of Spain, without the 1.22 percentage points of Tier 1 capital added in aggregate 2008 and 2009, the number of failures among stress tested banks would have been 19 out of the 27 participating banks. Similarly three out of the five participating Italian banks would have failed with 1.19 percentage point lower Tier 1. Only in France and the United Kingdom would all the banks have passed without increases in Tier 1 during 2008.