Closing on the impossible May 3, 2012 at 7:38 am

A recent post by the Streetwise Professor made me wonder: how risk sensitive do we want capital requirements to be? Or to try to be?

Let me explain.

In the 1980s and 90s, it was a tenet of both regulators and banks that capital requirements should be risk sensitive. More risk requires more capital. That’s a good thing, right?

Well, yes and no. There are (at least) four problems.

  1. Capital rules are always behind, and sometimes far behind, industry progress in risk measurement. The goal cannot be attained.
  2. Relative risk equality is important, by which I mean that the same risk taken in different ways should attract the same capital charge. The Basel capital rules (increasingly) don’t meet this basic goal. Let’s fix this before we try and figure out the harder task of charging appropriately for different risks.
  3. Risk measures are procyclical. That’s an inherent property of a good risk measure. There is a strong argument against procyclical capital rules.
  4. Risk models can turn out to be flawed, or mis-calibrated. Therefore there needs to be some kind of backstop to prevent the massive growth of products which look low risk according to a (as it will turn out) flawed methodology.

Given this, I am tempted to ask (but not – notice – answer) a difficult question.

Is the task of trying to create a risk-sensitive capital framework worth attempting, given that it is not achievable, and failure necessarily leads to arbitrage?

Your thoughts, as always, are welcome.

8 Responses to “Closing on the impossible”

  1. Hi David,
    what about a charge that is based on a set of global risk neutral scenarios shared across instruments and asset classes?
    That would force a bank to look at different risks in a coherent way, the same set of preferences (or risk measure) is shared across the portfolio, preventing arbitrages.
    Because the same model (say an interest rate model) is used across a wide set of instrument, calibration can be carried out using all the information available and therefore limiting calibration/model risk.
    The main problem would be to find a model that is able to calibrate well “globally”, that cannot be achieved within the limited set of models that guarantee some sort of analytical tractability, but because one needs just one set of scenarios and everything is done with montecarlo analytical tractability is not a constraint anymore and one can choose a more rich and realistic process.
    Of course point 3. would still hold, but I think this would be the best effort in the direction of a risk sensitive model.

  2. Giacomo,

    Banks already have a strong incentive to develop models that are as consistent and risk-neutral as possible in order to calculate CVA. Furthermore, the BIS allows them to use a risk-neutral calibration in order to calculate capital charges.

    I think you are underestimating the strength of point 3, which is the crux of the issue. And in any case, it is not necessary to measure the bank’s risks consistently according to a risk-neutral measure, just to measure them consistently.

  3. David,

    I’m pretty sure that Rich Bookstaber would say simpler is better. Our heros, the cockroaches!

    But does it really make sense to have the same leverage limits on Treasurys, mezz tranches, and equities? I know a risk manager at a hedge fund that had started out with a purely long-short equity strategy and promised investors that their leverage limit would be something like 2:1 or 2.5:1 (I forget the details.) They noticed some yield curve anomalies but were unable to profit from them because of this limit. But a 10:1 repo position in Canadas would have been less risky than some of their equity positions.

    There’s no slam dunk answer.

  4. David

    I tend to be of the view that whilst risk-based is good in theory, in practice it just drives arms races between the regulator and the regulated, with the former always chasing the latter. For me it is interesting how the incidence of bank blow-ups seems to have increased since the 1970s, when regulators first started shifting from a prohibitive approach (“you cannot do that unless we say so”) to an accomodative approach (“you can do that unless we see a key reason not to”). It is a subtle shift but an important one. Then, with the move to AIRB approaches etc, the regulators gave the kids the keys to the candy shop and the arms race began (to mix metaphors horribly). Moreover, imagine if there was a simple non-risk-based framework – a lot of rocket scientists and the like would be released to build rockets (and other more useful stuff).

  5. Thank you all. I will mull on this some more, and write again on it shortly.

  6. I, too, was interested in that post, so much so, I posted it on twitter. In terms of risk sensitivity, as you may recall I am focused on corporate credit risk and in particular its affect on B2B credit risk.
    The benefit of granular, high quality & timely information that records daily observable events has tremendous benefits IF you can also pull the levers that reduce default probability. For example, if a counterparty begins paying slower, you as corporate treasury begin limiting new sales; insurers may restrict insured limits; those actions have immediate effects on the customers business model which – if they are well governed – will elicit corrective actions (unless, of course, they are defaulting.
    Therefore there is a case that risk sensitivity has value if those measuring it can take mitigating actions as a driver does if he sees the speedo showing he is approaching a 30mph speed camera at 35mph!
    It’s a fascinating topic and like you I don’t have the answer, but another important question is for whom exactly is the risk sensitivity deeemd important. Again in terms of a comparison we call all relate to, the taxi driver, the lorry-driver’s employer, the schoolbus driver, the joyrider, the passenger, the police, other motorists?
    In finance, I’m happy for “sophisticated” gamblers to take risks; I’m just not keen for them to tell me they’re doing 29mph in a 30mph zone when they are doing 50mph.

  7. David,
    I’ve meant to right this for a while. You ask many good questions.

    For background, I was a senior bank supervisor and was involved in negotiating and implementing Basel II, both internationally and domestically.

    Some history is in order: Basel I was fundamentally put in place to insure that Japanese banks held sufficient capital – at the time they were fundamentally undercapitalized. By the late 1990s, in particular along side the development of credit derivatives, it became clear that the Basel I regime was flawed. Banks, responding to incentives, optimized their return on the regulators flawed capitalization construct. This is most easily seen in the prevelence of ever-green 364 day loan commitments that received a zero risk weight. If you review regulatory pronouncements during the late 1990s, you will see them struggling to fit credit derivatives into the construct, especially after JPM execute their innovative BISTRO trade to hedge their loan book (ironic given the losses recently in London, isn’t it). Eventually, the Fed was reduced to issuing SR99-17, which fundamentally says “it is your job, bankers, to identify your risks and hold an appropriate amount of capital against these risks.” This, in essence, was implicit endorsement of economic capital modelling since the regulators in the US realized that banks could make their regulatory capital appear however they wanted.

    Basel II embraced “risk-sensitivity” and economic capital modelling because…what else was there at the time? In addition, at that time, globally, most banks at best were nascent in their measurement and modelling of risk (think about this – most banks, even the largest ones, realistically couldn’t tell you quantitatively what their risks were, and how much they could lose with any degree of certainty). In addition, there were developed country markets where credit was not fundamentally “priced for risk” – pretty much all borrowers below the largest firms paid essentially the same rate, and bankers were in the binary decision making position of choosing “good” credits and avoiding the “bad.”

    Over the subsequent decade, firms down to intermediate size all got the “risk sensitivity” religion, and invested in credit portfolio modeling, market risk management (VaR and stress testing) and (when not openly distainful of the concept) tried to apply these techniques to operational risk as well. The Basel process was fundamentally important in creating the incentives to implement these techniques, and for fundamentally empowering the culture of risk management within banking. As a result, arguably, credit is allocated much more efficiently today than it was a decade ago.

    Now to the essence of your question: should REGULATORY capital measures be risk sensitive?

    In the absence of regulation, one would think that banks would of their own accord hold capital based on their risk. And we now know that bankers have recognized their own incentive to do so.

    It would seem to me that the right regulatory response would be to encourage the internal development of risk-sensitive capital allocation, and CONTROL for errors in that process.

    What are the errors in the process? Over-rating high-quality, low-PD assets and thereby encouraging extremely high levels of leverage. Now there is nothing wrong with a highly-leverage highly-rated portfolio PROVIDED it doesn’t also involve a material degree of maturity transformation. Perhaps these portfolios should be required to be funded with long-term liabilities.

    Another way to control for this risk is, I hate to say, a leverage ratio regime. This, however, runs the risk of being distortionary, with credit only flowing economically to firms where the potential risk-based leverage is equal to or less than the regulatory leverage ratio.

    The other failure of the capital regime has been addressed in the CCaR process. Economic capital practitioners made the fatal flaw of equating the required amount of capital with the confidence interval of a distribution. In reality, firms fail considerably before they run out of capital. The CCaR “floor” of 5% conceeds this, and says, essentially, use EC as a risk measure and figure out the capital you need to keep to minimize the potential for breaching the regulatory intervention constraint.

  8. Carter

    Thank you for your detailed and insightful comment. I agree with your account of the development of the Accords, and with the utility of leverage and stress backstops.

    Having lived and worked through roughly the same evolution of capital rules as you – from the 1996 Market Risk Amendment to Basel 3 and the FRTB – I am increasingly of the view that while somewhat risk sensitive capital rules are good, and extensive set of backstops are needed to close off arbitrages and mis-modelling. The cost of crude backstops is lower leverage than is safe; the cost of not having them seems to be the risk of systemic crises. On that basis, most people would chose to have the backstop…