The Bear’s Lair: Bring out the random number generators!

According to the Financial Times, the major international investment banks are cutting their spending on “quants” and their mathematical risk management models by about 40% from the current average of $100-150 million annually per bank. The regulators no longer give a risk capital allocation benefit for banks with “sophisticated” truly incomprehensible modeling systems, so there is felt to be little benefit any longer in going beyond standard models. Since the models used by banks didn’t work, as demonstrated by the 2008 crash and J.P. Morgan’s “London Whale” loss, this is probably a worthwhile economy. Of course, they could try doing the modeling properly, but this might tell them things they don’t want to know.

The new change originated in a typically dozy new regulation from the Basel Committee on Banking Supervision promulgated in March. Under the new rule, banks would no longer be allowed to calculate their capital requirement for interbank loans, loans to large corporations and stocks, but would still be allowed to calculate their own capital requirement for corporate, mortgage, credit card and retail exposures. The Committee is still considering how to treat sovereign debt holdings.

There are two distinct problems with all this: the regulations and the models. Banking regulation has become impossibly more complex, which means in turn that banks have devised ever more devious ways of gaming the regulations. Complexity in regulation is not a virtue. Second, by allowing banks to use their own models to assess risk, and not correcting for the known fallacies in commonly used models, regulators are allowing the banks to game their internal models as well as the external regulations, deceiving both the regulators and their own top management about the risks they are undertaking.

The principal flaw in bank regulation is the use of “risk weighted assets” whereby government debt is given a zero risk weight and mortgages are treated more favorably than ordinary loans. This system reflects the political prejudices of the regulators, most of which are thoroughly economically misguided.

Government debt is the least productive asset to be held on a bank’s balance sheet, because the government should be a better credit risk than its domestic banks. Thus a bank holding government debt is lending at a negative credit margin, even though it may make money by taking interest rate gapping risk. Incentivizing banks to hold government debt also makes it far too easy for governments to borrow and waste vast amounts of money that must be repaid by taxpayers. Similarly, home mortgage debt encourages people to invest excessively in unproductive housing and pushes up prices.

Conversely, small business lending, while the most difficult function of banking, is also the most valuable. It also has the useful property of being self-diversifying, except in terms of geographical concentration – any given small business loan can form only a modest part of a large bank’s risk exposure. It should thus be placed at no disadvantage to other forms of lending, as it is currently, but rather regulators should incentivize it.

It would thus make much more sense to weight all assets fully, and impose a robust capital requirement of say 15% against them all. “Off balance sheet” positions such as swaps and options would be weighted by a highly conservative view of their risk, so for example “short” positions in credit default swaps, in which the entire credit risk is assumed, would be subject to 100% weighting.

Needless to say, this would have prevented the 2008 meltdown. With 15% capital ratios, home mortgages fully weighted and CDS also fully weighted, the shenanigans of 2006-07 would have been impossible. Lehman Brothers would still be with us, as would Wachovia Bank, while AIG and Citigroup would never have had to be bailed out by the unfortunate taxpayers.

Such a regulatory reform would also go a long way to remove the pathological features of today’s market. Governments would have had to rein in their borrowing years ago, since banks would be very restricted on their government debt holdings. That in turn would have forced up interest rates on long-term debt generally, removing the incentives for excess leverage in today’s financial system. Home mortgages would be more expensive and credit standards would be higher, both of which would improve the health of the housing market and consumer savings patterns generally.

With banks requiring higher margins on all their loans to pay for their cost of a 15% capital requirement on all assets, they would no longer be interested in the very skinny margins available on government debt or government guaranteed home mortgages (if the capital requirements were extended to them, as they should be, Fannie Mae and Freddie Mac would soon be out of business.) Instead, banks would be attracted by the higher margins available on credit card lending, non-government guaranteed lending and above all small business lending.

With this simple reform, we would greatly reduce the cost of regulation, putting thousands of regulators out of a job, and would eliminate the banks’ ability to game the current over-complex, lobbyist-ridden regulatory system. We would also need to reform risk management systems, eliminating banks’ ability to game these, fooling their own management as well as the regulators.

The principal problem with current risk models is that they are almost exclusively Gaussian, working on the assumption that financial market prices move in a “random walk” as if they were motes of dust being stirred by a Brownian motion. However, as I have previously shown, and much more distinguished mathematicians, notably Benoit Mandelbrot have shown before me, securities prices do not move in a random motion, but are instead subject to laws of motion that either produce very “fat” tails, in which the probability of extreme movements is much higher than a Gaussian would expect or very “long” tails, in which the probabilities are Gaussian but the deviations from the mean are far more extreme than a Gaussian would expect.

For example, a portfolio of subprime mortgages, perhaps packaged in a collateralized debt obligation, has much fatter tails than expected. If credit conditions tighten or house prices fall, a much larger percentage of such subprime mortgages will default than you would expect. Being subprime, they are subject to a much higher level of non-random fraud and non-random mortgage broker incompetence than a similar portfolio of prime mortgages, and those elements are not properly accounted for in a Gaussian model. Furthermore, subprime mortgages are highly correlated, so a housing market downturn pushes them all over the edge.

The best modelling software for such a portfolio is not Gaussian but a fuzzy logic-based model, which reflects the true reality that if subprime mortgages start failing, a large proportion of the portfolio will do so.

On the other hand, a portfolio of credit default swaps, with the very long risk “tails” on short positions of 100% of the credits insured is best represented by a much longer-tailed alternative to the Gaussian, the Cauchy distribution. With this distribution, the peculiar likelihood of a short CDS portfolio to lead to truly catastrophic losses is highlighted.

It seems likely that a combination of Cauchy and Fuzzy distributions can cover all possible risk profiles that can be devised by even the most devious trading desk quants, so a risk management system that included all three methods and flagged any risk profiles that were out of line on any of them would be well suited to manage the activities of a modern and inventive investment bank trading group.

We are bound to run into another financial crisis fairly shortly; the follies of eight years of negative real interest rates demand it. When it happens, there will doubtless be some “too big to fail” financial institutions that incur staggeringly large losses. The “stress tests” devised by the Fed in conjunction with banking sector lobbyists will prove to be utterly useless – not nearly stressful enough, and missing several large categories of risk that the models had not caught and of which bank top management was only vaguely aware. So even if capital ratios are set at proper levels and models are adjusted to reflect actual risks, it is still more than likely that the banking system will get in trouble because of the world’s very eccentric recent monetary policy.

At which point, the public will doubtless blame the banks and the politicians will demand even more draconian and complex regulation. But the reality is that irresponsible politicians and central bankers will produce irresponsible banks, by the normal operation of market forces. No amount of sophisticated risk management and regulation can change that.

(The Bear’s Lair is a weekly column that is intended to appear each Monday, an appropriately gloomy day of the week. Its rationale is that the proportion of “sell” recommendations put out by Wall Street houses remains far below that of “buy” recommendations. Accordingly, investors have an excess of positive information and very little negative information. The column thus takes the ursine view of life and the market, in the hope that it may be usefully different from what investors see elsewhere.)