Stress Free Europe

Last week the European Banking Authority announced new stress test requirements for banks, and while some reports claimed that the standards are being lowered because only a 5% stress is now required, others pointed out that these new stresses directly affect Tier 1 capital which is supposed to absorb the stress.  While there is confusion about whether this new requirement is more stringent or less stringent than last year’s stress tests, that whole debate misses the point:  Are either of these tests sufficient?

My answer is a clear NO.  The whole point of stress testing – the simulation of market movements and resultant revaluation of a fund’s assets – is to understand how sensitive a fund is to a number of factors simultaneously and to determine just how bad things can get.  The critical piece is to move markets by a variety of amounts: say 5%, 10%, 15%, 20%, etc. and to then understand the fund’s responses to those movements.   Let’s say that in response to those market changes, Fund A’s revaluation estimates are -2.5%, -5%, -7.5% and -10%, while Fund B’s changes are closer to -2.5%, -4%, -20%, -35%.  Clearly, Fund A has a linear response to this market while Fund B has a very non-linear response (bigger movements in the market result in disproportionate movements in the fund vale).  Instead of using all those different potential shifts, what happens if a stress test is done only at one of those points – say the 10% simulation?  In this case, a very different picture is painted by the results: it looks like Fund B does better than Fund A.  To complicate matters, no one really knows the likelihood of the markets having exactly a 10% move.  Perhaps a 5% move or a 20% move is more likely.  So by pinning down one single market change (in this case 10%), and not considering any other possibilities, the stress test only captures one small portion of the fund’s possible movements.  Conclusion: unless a stress tests simulates many different potential market movements it does not supply nearly as useful – or potentially correct – information.  Doing a stress test at just one point (like 10%) is of marginal or no value.

And this is exactly what the European Banking Authority is likely to require.  Reports indicate that the EBA is likely to require a 17% loss on Greek debt, a 5% loss on broad markets, and other very specific changes in market conditions.  Over specifying the stress is just wrong – it provides no material value.

Figure 1: Stress Tests should be done at a variety of values, not just one or two.

While the EBA may have good reason to believe that Greek debt is most likely to lose 17%, it makes a lot more sense to test the banks at several intervals – perhaps 5% through 30% in multiples of 5%.  Instead of going on about how bad the European approach is, let me instead talk about the right way to do this. The figure shows an example of a good stress test, where interest rates are stressed along the left/right and credit spreads are stressed up/down.  The gray square toward the middle is the current fund value, and every combination of interest rate move and credit spread move has a simulated revaluation.  You can immediately tell where the fund gets into danger.

It’s also interesting to note that the stress tests that were in place last year were passed by almost all the banks across Europe.  Especially interesting is that the banks in Ireland all passed, just a few short months before a massive bailout saved that country from financial ruin.  These ‘new and improved’ stress tests (my words, not theirs), it seems, don’t do much, if anything, to ensure financial viability or stability.  It seems that Europe just hasn’t learned its lesson yet.  C’est la vie!

One Response to Stress Free Europe

  1. Tore Olafsen says:

    Dear Sir,
    You are of course perfectly right. These stress tests fail to capture what they really are meant to disclose; the downside risk of the bank’s portfolio of debt, securities and placements.

    When you are setting the values of the bank’s exogenous variables at fixed probability levels, and then compute the scenario – will you get the same level of probability on your answer?

    The answer is NO. The only case when this will happen is when all systematic risk is perfectly correlated. The effect of this assumption is that the risk model becomes a perverted one, with only one exogenous stochastic variable. All the rest can be calculated from the outcomes from this variable.

    Add to this, a valuation analysis with a large number of: both correlated and auto correlated stochastic variables, complex calculations (nonlinear) and simultaneous equations, and there is no way of finding out where you are on the bank’s probability distribution – unless you do a complete Monte Carlo simulation. It is like being out in the woods at night without a map and compass – you know you are in the woods but not where.

    For a simple proof see:

    Tore Olafsen

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: