Basel isn’t trying to fix bank models, it wants them gone
The endgame for the Basel Committee’s new credit risk rules is to get rid of internal models, but it just can’t get there yet.
The Basel Committee has a habit of sending out vastly important updates to its capital framework before western European holidays — presumably out of an urgent desire to finish overhauling international rules for big banks by the end of 2016, rather than actual malice towards junior lawyers, consultants, and investment bankers.
This Easter was no exception, with a new paper on credit risk that proposed to overhaul the capital rules accounting for around 70% of bank assets at big, sophisticated banks.
It’s only a consultation, but it doesn’t look good for banks and their ability to model and assess their own risks, under either the Foundation Internal Ratings Based Approach or the Advanced Internal Ratings Based Approach (IRB).
This ability is at the core of Basel II, which, in essence, said that banks could and should develop increasingly sophisticated techniques to measure and manage credit and market risk, while their supervisors should spot-check models — but otherwise leave banks to get on with it.
The models in question were not especially baroque. They took the basic inputs of probability of default, loss given default, and exposure at default — the art of Basel II relied on calibrating these assumptions, using historical data, and slicing up the portfolios in question just right.
Generally received wisdom is that this did not work out terribly well. Basel II encouraged banks to pull the wool over regulators' eyes, and make assets disappear using their own models (the banks still had to report total accounting assets and tangible equity in their financial reporting, but for some reason investors didn’t seem interested in these figures).
Hence the decision after the crisis to bring in a regulatory leverage ratio. This had been a feature of Basel I, and was abandoned because it encouraged riskier lending. A risk-weighted capital ratio should encourage banks to look at risk-adjusted returns, not absolute returns. Instead, though, it encouraged banks to hide risk in triple-A rated wrappers, leading, at least in part, to the financial crisis.
However, combine a risk-based capital ratio and a leverage ratio, and the simple leverage ratio acts as a back-up and a sense check for the bank’s models. If too many assets end up with low risk-weights, the leverage ratio will start to act as a binding constraint, making sure enough capital remains in the system.
In the new paper though, there are the first signs the dual system is breaking down, and the endgame is also the end of the models.
Banks will no longer be able to use internal models to measure the risk of their financial exposures, their exposure to large corporates, or the capital they hold against changes in counterparty risk.
Where they can use models, the inputs and the outputs have to be “correct” according to the Basel Committee. Probabilities of default or loss given default figures which are too low can’t be allowed — irrespective of the actual figures experienced in an actual portfolio. Banks can calculate anything they like, as long as they calculate what Basel wants them to.
It’s worth remembering that this is a second, more complicated check and balance. Internal models are already subject to supervision — bank supervisors can and do throw out models they do not believe, and in doing so, gain insight about a bank’s assumptions and approach to risk.
It’s hard to see why the Basel Committee, setting universal rules for large banks, would be better placed to signal the constraints and assumptions about modelling than local supervisors.
Finally, there will be a floor on the output of models.
If a bank uses the prescribed inputs, but still comes out with a lower capital requirements than Basel wants, it will have to raise it to a Basel-prescribed peg. Depending on whether this is 90% of the pre-model level or 60%, this could be cripple the use of models at all. Why bother maintaining the infrastructure of advanced capital modelling if the benefits are negligible?
One suspects, actually, that this is what the Committee wants. It says it continues to value banks having different views about the risks of given financial assets. This difference of views, in theory, is healthy for financial stability, as it promotes variations in business models, the reallocation of assets to whoever likes the risk most, and avoids herd behaviour.
But this pious aspiration is at odds with its actions. It already has a hard back-up to stop internal modelling getting silly, in the form of the leverage ratio. Harmonising bank risk assessments at an earlier point can only be an attack on the idea of modelling at all.
At this point though, the Committee’s hands are tied. It cannot tear up bank capital rules and start again from scratch; it has to consider how much capital is already in the banking system, and how much banks can realistically raise.
That limits the possible attack on models. Credit Suisse estimated a capital floor of 70% would mean up to 200bp off its capital ratio — a floor of 90% could be crippling, and an outright ban worse still. But make no mistake, that’s the way the world is going.