Right at the bottom of the FSA’s rather mealy Statement on Regulatory Approach to Bank Capital this morning (most of which is a thoroughly anodyne recap on existing policy) is this:
We have also been working with these banks to seek to ensure that the application of the current International Basel Accord, which is implemented through the Capital Requirements Directive, does not create any unnecessary or unintended pro-cyclical effects. In particular, we are amending the variable scalar method of converting internal credit risk models from point in time to through the cycle.
Which would be easily overlooked were it not for the following:
These changes will significantly reduce the requirement for additional capital resulting from the procyclical effect.
The FSA is authorising a modelling trick that will allow UK banks to significantly reduce their risk-weighted asset requirements stipulated under the Basel-II accord.
________
Under Basel-II, banks are required to calculate risk weightings – capital cushions – against their portfolios of assets. In simple terms, riskier assets require more capital to be set aside. Basel stipulates exactly how much must be set aside, and exactly how to calculate the risk of an asset. Institutions can either calculate their risk requirements based on the external ratings of assets (given by the rating agencies) or on internal models.
Internal modelling is the most popular. And of course, in any such internal risk model (just as with any ratings-based model), something like the portfolio’s projected default rate will be of critical importance. Here’s where the FSA’s rule change comes in: the difference between judging risk weightings on a “point-in-time” basis and a “through-the-cycle” basis.
Suppose Institution A has a portfolio of 10 assets, all of which have been modelled and shown to be not very risky. As such, they require only a small risk-weighting under Basel – 20 per cent of their value, lets say. Unfortunately, the economy now takes a downturn. Obviously, the risk models take this into account and as such, the risk of the portfolio has increased. The internal model has effectively “downgraded” the assets, and they now need a higher risk-weighting of 50 per cent. Six months later, the economy recovers. The internal model effectively “upgrades” the assets and so the risk weighting decreases back to its original level. Basel in action. Fortunately, the economy then stays benign for the remaining lifespan of the portfolio (another two years, let’s say).
This is a point-in-time approach to modelling risk-weightings. The risk weighting changes depending on which point in the economic cycle we’re at (downturn or upturn). The disadvantage of this approach is a degree of risk-weighting volatility.
Enter, through-the-cycle modelling. Under a through the cycle model, rather than measuring the risk weighting at every point in the cycle, a risk weighting would be calculated at a flat rate for the whole cycle or in fact, the life of the assets. Thus in our example above, the risk weighting would be smoothed. The idea being that cyclical effects are removed from the portfolio: banks’ capital ratios don’t suffer quite so much from a cyclical downturn.
Fortunately, the FSA itself has left a presentation lingering around the web on probability of default (PD) modelling, from which we reprise this handy illustrative graph (note, the “hybrid” approach is probably close to that used by most banks, who don’t necessarily use a pure PiT model):
All very well and good.
Except there are some major problems with using a through-the-cycle approach. The FSA was itself very sceptical of it. Read their position paper on it from October 2006. First off, is this:
A firm must be able to overcome the considerable conceptual and technical challenges involved in order to carry out these adjustments in a way that properly takes account of their current levels of default risk and changes over time… Based on our experience to date, we are not convinced that all firms will be able to meet the required standard.
In other words, the modelling is too difficult. Given the rather dramatic unmasking of most banks’ supposedly failsafe risk models since the FSA made that statement, you’d have thought the regulator would be less willing to countenance higher modelling risk, not more. Apparently not so.
Indeed, the through-the-cycle approach has one particularly egregious flaw. Again, the FSA know it:
In particular there is a danger of permanently tying their capital requirements to measures of their default risk in the past.
In order to model a through-the-cycle portfolio probability of default, you have to rely heavily on a historical model of asset performance through previous default-cycles. Which means the model will only really capture severity on the scale of past crises. Given the paucity of severe banking collapses in the data, we’d suggest that looking into the current cycle, this is a problem. One could say that it…
…amounts to a quasi standardised approach which is calibrated on the basis of a firm’s historic default experience, and not sensitive to the riskiness of its current portfolio or future changes.
The approach smooths over tail risks – even exacerbates them, since by lowering capital weightings, when such risks do hit, institutions will have lower capital buffers set aside.
The risk is greatest in portfolios like mortgage assets, or for structured finance assets, where the historical default data is even slimmer than that available for corporates.
And then there’s the further issue of correlation in portfolios. Highly correlated situations – such as those that occur following banking crises – can completely throw historical portfolio default models. And through-the-cycle scalar approaches minimise correlation effects.
________
The FSA’s rationale is clear: loosening the RWA modelling rules is another weapon fired in today’s volley of bank capital protection measures. Using through-the-cycle modelling should “significantly reduce” capital requirements under Basel II.
The problem is that the very principle of it has been proved thoroughly wrong by the current crisis. It’s exactly the kind of modelling technique that invites disaster in extreme scenarios. It’s exactly the kind of modelling technique that caused the crisis in the first place.
The FSA should know better. A month ago it published its opus on banks’ liquidity standards, in which it was noted:
…models have only a limited role to play in liquidity regulation, as liquidity stresses are heterogeneous events that make it difficult to construct meaningful probability distributions.
What’s happened to that stance?
Obviously no-one wants to countenance a historically severe default cycle going into 2009/2010, but economic indicators suggest one is very much on the cards. With that in mind, the FSA’s through-the-cycle revisions might just end up creating Zombie banks; organisations who’s statistics look good, but who are actually just as risky – if not riskier – than they were before.
Rather then better equip banks with the capital they require to withstand the crisis, it may be that the revisions simply increase the shock and impact of large corporate defaults as and when they occur.
RBS’ latest writedowns today are a case in point. Look at the huge £1bn hit it took against the LyondellBasell failure. With more charges like that to come for banks, capital ratios need to be more than just superficially bolstered by modelling tricks.
Related link:
Capitalising on cycles - FT Alphaville