JP Morgan Pours More Cold Water On Stress Test Credibility, Sees Anything Less Than 25% Haircut On Greek Bonds As A Joke
Just as Europe was happily gloating that it managed to push markets up with the second stress test rumor of a Greek debt haircut of 16-17%, here comes JPM's Francesca Tondi dashing European bank regulators' hope, and stating that anything below 30% on Greek bonds, 20% on Portugal, 15% on Spain, and 10% on Italy, Ireland and Hungary (in the base case scenario), is a joke. Below is the sensitivity table that serves as the framework for JPM's stress test. As can be seen, even the optimistic scenario is far more draconian than what Europe is planning on using, thereby discrediting the test, whose assumptions are now likely going to be revised yet again to be seen as anything remotely credible.
Below are JPM's haircut levels on debt by country:
Here is JPM's stress test summary:
With heightened debate over a stress test exercise in Europe, which regulators could publish by the end of the month, for illustrative purposes, we model a stress scenario for our universe of 36 listed European commercial banks, which include a stress test of higher credit losses consistent with a recession scenario, as well as an additional charge for "haircuts" on the banks’ sovereign bond holdings. Our analysis does not include the unlisted banking sector.
Ultimately, the scope of the exercise, which at the European level is "orchestrated" by CEBS, is for regulators (but also market participants when published) to obtain reassurance that banks have sufficient capital buffers to withstand a stress; such a minimal “watermark” capital level has not been publicly defined but we assume that it is a core Equity Tier 1 ratio of 6%. In our view, the methodology used will be as relevant as the actual conclusions. Based on our illustrative scenarios, we draw the following conclusions:
1) -Modeling a macro recession over the next 18 months, would increase LLP and erode earnings, but would not “tip the sector over the edge” as 25 banks out of 36 would pass the test. Our universe would “only” require €30bln of additional capital (Core ratio from 8.8% to 7.3%); concentrated in “only” 11 banks (largely based in Germany, Ireland, Greece), which however, in our view, have a capital shortfall of c. €15-20bln even before a stress test.
2) -We also flag that even under a recession scenario 10 European banks not only do not need further capital, but also still report a profit: among the large banks, these include SAN, Stan Chart, BBVA, DBNor, HSBC, BNP, Nordea and ISP.
3) -We add to our analysis also some scenarios of "haircuts" on the banks’ sovereign bond holdings, which market participants have
requested for some time, although regulators are still debating whether to include in the actual stress test. Having applied a haircut (of 6% on avg, with ranges from 5% to 30%) to the banks’ sovereign bond portfolio, our modelling indicates that our universe of banks may need an additional €24bln of capital, bringing the total potential capital need to €54bln, spread over 21 banks. Note that to date, European banks have already received, and still can use as a buffer, €35bln in the form of government support, which at least goes some way towards providing a cushion, should the stress materialise.
4) -Putting a “stress capital need” into context highlights that the market has already largely discounted a stress scenario - Ultimately, even the €57bln capital requirement in our more comprehensive scenarios (of macro recession and sovereign haircut), represents c. 8% of our universe market cap, vs the c. 16% value "destruction" suffered by our universe in the last 3 months (admittedly partly due to liquidity issues). Trading on P/NAV of 1.0-1.5x even after NAV erosion and recapitalization, we find the sector valuation to be sufficiently discounted, but especially flag as attractive BNP, ISP, DBNor, Soc Gen, UCG and HSBC (1.9x P/NAV) based on a combination of attractive valuation, recent (in our view) ’unjustified’ market performance and resilience
to the stress test.