A Call for Counterrevolution
Imperfect Basel I Was So Much Better Let’s Ban the Unacceptable
Einhorn versus Brown But, Will We Suffer?
Would VaR have been enthusiastically adopted by financiers and politicians if it weren’t wrapped up in sophisticated-looking mathematical symbols and analytics? I have my doubts. I quite strongly believe that VaR’s quantitative cred, which as we know was particularly acute in the early days of the model, decisively contributed to its embracement as the risk guide that would solve all problems.
The math helped convince many people that the new methodology was imbued with unlimited rigor, a wise conduit to financial precision, an end to bothersome uncertainty. In some cases, those conclusions would have been reached after thorough examination of the technical documents. In others, I suspect, the conclusions would have been arrived at rather unconditionally, the total acquiescence with the model not demanding an actual investigation of its analytical insides: VaR’s high-tech outer appearance would be more than enough credential, no further introspection required.
This speaks of the powerful status that quantitative concoctions have reached in financeland in the modern age; not only those who truthfully abide by the symbols are brought on board, but also those (a vast number perhaps) whose mathematical knowledge of the model is limited to the fact that the model is mathematical. If you want a device to infiltrate the markets, it surely helps if it is analytically clothed.
The reasons for this are probably varied, including a general human infatuation with scientific- seeming accomplishments, a reluctance to challenge apparent sophistication, or a lack of trust in the
“softer” sapience of personal intuition. Whatever the actual factors behind the imposition of a financial model, it seems clear that on way too many occasions common sense is forced to take a back seat, if at all, when it comes to some of the most consequential financial decisions. The crowning of the model as supreme ruler implicates, almost by definition, the excreting of human intuition, which many proponents of the analytical way consider not just a competitor but the enemy. VaR is the most relevant example of this phenomenon, but certainly not the only one. Complaisance toward equations- adorned gadgets has convinced people to put their arms around plenty of silly notions, such that it is possible to know a priori the future risks and returns of a security, markets are perfectly liquid and continuous, crashes and bubbles don’t take place, or it is possible to measure a priori the future correlation between defaults on mortgage loans. None of these assertions, I believe, would have ever become accepted wisdom had the source not been quantitative. Had a, say, innumerate cab driver, not an MIT professor or a JP Morgan quant, uttered such notions we would have immediately dismissed
him as a hopeless crank. And yet, once the very same dictums emanate not from a smelly taxi, but from the hallowed ivory tower or the imposing bank we puzzlingly nod in agreement, endow the authors with the genius label, and shower Nobel prizes on them. Whether a financial tenet is dressed up mathematically or not can be the difference between ignominious rejection and getting a medal from the King of Sweden.
This is not a good state of affairs. Bad theories should be as quickly discarded as bad cabdriver advice. No amount of technical wizardry can justify the embracement of beliefs that would be deemed absurd absent the theorems. If the math becomes the Kool-Aid that makes us accept silly principles, then the math becomes a dangerous thing. If the math forces us to betray our most pure intuitions, then the math must be resisted.
Cab drivers would not have entrusted market risk management and bank capital regulation to VaR.
But they, on the other hand, would not have found themselves in total disagreement with what was going on before VaR. Basel I would have seemed quite reasonable, quite acceptable, at the very least a decently sound starting point. Measuring risks to the third decimal through the use of suspect statistical trickeries and unreliable past data would appear to our no-nonsense taxi-driving friends much more unreasonable, unacceptable, and unsound than ranking financial assets by their obviously intrinsic nature (even if crudely done). Even a financially ignorant individual can see the wisdom of, first, doing no harm: Make sure that the nasty stuff is treated accordingly. Many financial mathematicians and theoreticians may want to convince us that a subprime CDO should be given the chance to appear less risky than a Treasury Bond, but that does not negate the utter silliness of the idea. The concept would not pass the cabby’s test, and thus should be rejected. If the common man thinks it nuts, so should regulators and bankers. Stop flawed models from shoving insultingly unacceptable results down our throats. We can die from it.
It is paradoxical that an attempt to imbue rigorousness and sophistication into something may end up delivering outcomes that deviate from truthfulness even more dramatically than the supposedly plebeian system that had been replaced by the new high-tech ways. The pioneer Basel I international regulatory standards have been ruthlessly lambasted for their perceived lack of attunement to real- world realities. True risk, the critiques posited, is not captured by such rustic architecture, we need something much more accurate and fancy. Third-generation mathematical models that drink from actual market signals will get risk right, the thinking went (and still goes in many quarters). Such erroneous mode of thinking, it turned out, was founded on the idea that financial risk can be implied in some magic way from past behavior and statistical hypothesis. Rather, financial risk can at best be guessed and ranked. We can’t imply the future riskiness of a trade, but we can try to discriminate between different trades, and rank them in buckets. This, the Basel I regime got absolutely correct.
Free from the analytical shackles (the no-holds-barred quantification of finance had not yet conquered completely in the mid- to late-1980s), financial mandarins arrived at a commonsensical solution. The real value of a Basel I-type exercise is not so much that the risk buckets will be perfectly designed or organized (in fact, they were far from perfectly designed or organized), but that the discriminatory approach based on asset fundamentals is bound to guarantee that the most naturally risky stuff will be placed in the worst buckets (i.e., those demanding more regulatory capital and careful steering).
Rather than trying to measure risk, particularly through very inappropriate means, we should focus on making the worst kinds of risk unacceptable. While VaR and other metrics subliminally fail at that,
something like the much derided and denigrated Basel I showed the right path to follow (of course, Basel I dealt with credit, not market, risk but what’s being proposed here is that Basel I–style intuitional bucketing of risk categories be applied to both trading book and banking book assets).
Thus, we should engage in counterrevolution: restore into power the old quant-less monarchy and, Napoléon-like, exile the defeated models to a faraway location. The 2007 crisis was VaR’s Waterloo; we should find a remote St. Helena where the dethroned emperor can spend the rest of its life, terminally incapacitated to incite any more mayhem.
But restoring the old ways would not be enough. A healthy dose of reformation would be in order, not just to procure a more robust regime but also to limit the potential for a second quant revolt down the road. The Basel I monarchs must understand that while their system was superior at what truly matters, it can be greatly improved on.
Basel I was accused of three major sins. First, it can easily lead to higher risk by naively dumping together in the same buckets assets of widely different nature, making it for instance as costly capital- wise to lend to IBM or to the corner shop (and thus, in principle, encouraging more lending to the corner shop than to IBM, as a higher interest can be charged on the former; more interest measured against the same capital charge generates better returns on equity). Second, it could result in unnecessarily taxing capital levies by not taking into account the risk-reducing diversification benefits of owning a portfolio of purportedly uncorrelated assets. Finally, it didn’t cover market exposures, focusing only on a bank’s banking book. Only the third complaint carries real merit. It’s not that the other two charges would be completely off-the-mark, because the old rules could indeed favor lending to weaker credits and since diversification can certainly result in offsetting positions. But the remedies to both shortcomings made things much worse, potentially and indeed in practice: VaR and credit ratings can provide much more dysfunctional risk signals than Basel I’s less-than-perfect bucketing, encouraging punting on extremely dubious assets if the latter happen, possibly for sheer coincidence or reflecting a bubble, to have enjoyed a recent calm market period (much worse than making it relatively economical to lend to the corner shop is to make it almost free to lend to someone who doesn’t have a job or savings or income, let alone own a shop); and allowing more leverage on account of the supposed benefits of diversification can boomerang on you, especially when the diversifying factor is estimated via the statistical concept of correlation (the system can be gamed by scouring the historical data universe for assets that happened to have been uncorrelated of late, yielding very low capital charges for a portfolio of assets that, when things turn sour, can very well tank down in value all together at the same time, rapidly eating away at the diminished equity cushion;
that which was assumed to lower risks becomes a dramatic risk enhancer). Relying on correlation, just like relying on volatility, can lead to bigger and bigger portfolios backed by smaller and smaller amounts of capital.
It turns out that Basel I was a superior architecture precisely because it did not incorporate those things that its critics found inexcusably missing. By not rewarding portfolio “diversification” with lower risk estimates and capital charges, and by not drinking from “market signals” implied by past data Basel I made itself into a more robust system than its later siblings Basel II and Basel III. By not confusing statistical correlation with true codependence and by not confusing risk with volatility, Basel I won the day. Is the use of correlation and volatility always a bad thing? No, of course not.
Shouldn’t asset diversification and actual market data be taken into account when appraising a portfolio’s risk? Yes, of course they should. So, why are we praising Basel I on account of its neglect
of both factors? Because drinking from those sources, while possibly useful at times, can make unacceptable answers possible and embraceable. Denying them center stage, instead ceding it entirely to experience-honed fundamentals-based decision making, makes (or ought to make) unacceptable results impossible. It is feasible that a lot of the time, not giving a starring role to the statistical counsel may reduce the accuracy of our risk analysis. But, I believe, that would be an agreeable price to pay in return for avoiding the emergence of the utterly diabolic. I’d rather settle for the exclusion of some potentially useful bit of information from the risk appraisal than for the possibility of a banking book or a trading book or both leveraged 100 to 1 or even 1,000 to 1 on lethal assets. It’s healthier to potentially err on the riskiness of a conservative or semi-conservative portfolio while making it essentially not possible for a big toxic position to be built. While the financial and economic systems could put up with the former scenario, their very survival would be threatened by the latter’s.
So Basel I was more wholesome than its quant successors. The “improvements” that were required upon it from analytical corners should not have been taken on board. However, and retaking the key point introducer earlier, a number of other tweaks certainly were and would be required to correct for some obvious imperfections. How should this perfected, superior regulatory structure (let’s call it Basel I.5) look like?
One inescapable flaw of Basel I was that it allowed unlimited leverage on developed country government bonds-loans, by forcing a regulatory capital requirement of 0 percent on such positions (technically, on debt obligations by members of the Organization for Economic Cooperation and Development, or OECD, a Paris-based assemblage of rich and quasi-rich nations; 24 members when Basel I was put together in 1988, 34 at the time of writing). Toxic leverage is very bad, but excessive vanilla leverage should be equally avoided. Government-issued securities, even if issued by the most robust of nations, are not riskless, neither from a credit nor from a market point of view. The chance that an OECD country would default on its debt obligations is not zero, and certainly those assets can suffer from the volatile whims of global investors and tumble in value at no notice. Granted, such debacles would almost certainly never mirror those shouldered by more daring securities (while a subprime CDO can go to zero market value, an OECD bond is unlikely to sink nearly as much even under dire government financing circumstances), but they can potentially be significant nonetheless.
Thus, for a bank to bet the house on government-issued securities could lead to losses significant enough to drive it out of business and to ignite widespread economic despair. Regulators, therefore, should not enable free gearing on such plays. Accumulating Italian government bonds or U.S.
Treasuries should cost a little bit more than nothing.
It is often said that regulators decided to treat public sector debt so generously capital-wise as a way to guarantee that developed countries would find it easy to raise the funds they needed at any point; clearly, making that debt very economical for banks to hold is a powerful incentive for banks to lend to governments. So the OECD-originated mandarins in charge of Basel I decided to help their countries by helping global banks accumulate OECD debt very cheaply. Basel I may have placed OECD debt in the right risk bucket (in principle, that asset category should be placed among the safest) but got the risk weight wrong. Future regulatory regimes didn’t exactly correct the problem, as the humble capital charges afforded by Basel II to securities with the highest credit ratings made sure that large leverage on developed nations’ debt (which tend to be endowed with top ratings) continued to be affordable. That, combined with the reign of VaR on the trading side, potentially gave rise to a
particularly dangerous combination of very low capital requirements for both esoteric and government assets. Making leverage on the latter very economical is especially worrisome when leverage on the former is too inexpensive, given that the financial, economic, and social mess that would be triggered by the more-than-likely blowup of the toxic plays would normally lead to shocks in public finances deriving from costly banking bailouts and stimulus policies; the end result could very well be greatly enhanced volatility and price declines in the government securities sphere, inflicting severe setbacks on those institutions that had accumulated sizeable amounts of those assets on the back of a very generous capital treatment (the banking industry, in essence, would be exposed to facing a fatal double blow: first, massive write-downs on the exotic stuff, then more massive losses on the vanilla stuff). So the urgency to correct for Basel I’s lenient attitude toward OECD debt, highly advisable in itself, would be even more pressing under a system where VaR still roams around.
The meltdown that began in 2007 attested to all that. As the dust settled on the mortgage market massacre, an additional crisis was unleashed in certain corners of the sovereign securities arena, particularly in the Eurozone. By mid-2009, the headlines were no longer dominated by CDOs, massive losses on subprime loans, or rescue packages for Wall Street, but by the humongous difficulties faced by countries like Greece, Ireland, Portugal, Spain, France, Italy, or Belgium to deal with ever more unbearable fiscal deficits and levels of indebtedness. Of course, those difficulties had been accentuated by the earlier private-sector financial crisis, as governments had to rush in expensive rescue packages for banks and other firms and as tax receipts suffered from the abrupt decrease in economic activity and the abrupt increase in unemployment. The real plus the perceived risks of sovereign defaults collided to condemn those governments’ bonds to a sharp decline in price, hurting anyone who had dared accumulate them in bulk. On December 31, 2010, the FTSE Global Government Bond Indices indicated the following miserly 12-month returns for some of the above mentioned sorry cases:
Greece −20 percent Ireland −12.5 percent Portugal − 7.3 percent Spain − 3.9 percent Italy − 0.8 percent
Don’t tell me that developed country–issued securities aren’t risky, or that they should deserve a 0 percent regulatory capital charge.
Another obvious flaw of Basel I was that the maximum minimum capital requirements were set at a way-too-low level. The max min capital charge was capped at 8 percent of risk-weighted assets, which effectively implied a cap of 8 percent of total assets on those asset families deemed riskiest (and thus deserving of the top 100 percent risk weight). An 8 percent total capital charge, implying leverage above 10 to 1, can be too lenient if the asset is too daring. The top capital charge should be set at 100 percent, limiting gearing to a 1-to-1 ratio. This naturally implies that those assets placed in the most lethal risk bucket would be assumed capable of losing their entire value in a downturn. Such assumption may be seen by some as a tad excessive: Even highly illiquid stuff may be liquidated into something more valuable than nothing. However, slightly unseemly as they might appear, very steep top capital charges would serve us much better in our efforts to ban the unacceptable than an 8 percent max charge ever could. Again, the main goal is not to get risk metrics precisely right, or to design a risk system that is so fair and just that no asset family is ever demanded more capital than it
should. Many times, the steep top charge would seem unfair and uncalled for. Too bad. What truly matters is to fence hellish trades so stringently that they can’t be accumulated massively, or if they are accumulated massively never without a correspondingly massive equity shield. If a bank wants to lose $100 billion in mayhem-destined positions, it should back that wish up with a $100 billion capital commitment. That way, losses on the bad stuff won’t consume equity raised to support the good stuff. Every dollar of lethality should have its own equity cushion. An eye for an eye, as they say.
The key idea here is discouragement. A 50 percent or a 100 percent capital charge may turn out to be an inappropriately untruthful characterization of some of those assets unsound enough to qualify for the worst risk buckets, but it would always be appropriately discouraging, turning banks away from those, now taxingly expensive, punts. The markets and the economy at large become more resilient, as the possibility that the banking industry may finance a toxic orgy with but a tiny capital slice is made unfeasible. In essence, regulatory capital’s main role becomes the de facto banning of the obviously unacceptable; short of legally banning certain plays, the best weapon against the fragility of finance.
Quite prominent people would cast their vote for such initiative. Famed hedge fund manager David Einhorn is a case in point. Following the fall of Bear Stearns in March 2008, but before the crash of Lehman Brothers (which Einhorn famously shorted) the following September, the successful money manager saw it only natural that as a result of the malaise financial authorities would force banks to accumulate much more capital going forward. His most revolutionary recommendation? 50 percent to 100 percent charge for “no ready market,” that is, dangerously illiquid plays. Einhorn had no doubt that very low regulatory capital requirements, on the back on very low VaR figures, had sunk Wall Street, and that the unavoidable remedy would be to force banks to delever and to make trading on suspect assets much more costly. Capital should also be only of the highest quality, Einhorn offered.1 Anything other than core equity should not be allowed to call itself capital. The president of Greenlight Capital, in other words, presented himself as an indefatigable defender of down-to-earth, dogmatism-proof, common sense: Too much bad leverage is bad, and should not be condoned. The sad irony is that it took a “contrarian” (Einhorn has been portrayed as a quixotic figure, a roguish anti-system maverick recklessly betting on the end of the financial order) to point what should have seemed only natural to anyone all along. When those tagged as rebellious contrarians are the ones lonely carrying the flag of commonsensical decision making, that’s when you realize how maddeningly fragile the VaR-dominated financial universe had become.
There are also prominent individuals on the other side of the debate, financial risk grandees that would find a return to something resembling Basel I impossibly allergic. To them, there’s no turning back from metrics-based analysis. Give me historical data and quantitative models or give me nothing, seems to be the chant of those bent on protecting the status quo (notice that those fellows would now be firmly part of the “traditionalist” camp, given how entrenched analytical risk management has become; those proposing commonsensical, intuitional risk management would nowadays be the “revolutionaries,” inexcusably daring to challenge the supreme authority of the mathematical emperor).
Some of those opposing change would be untameable reactionaries, quantitative Torquemadas for whom law is only what the dogmatic book says. Others would be much more enlightened, much more tolerant, yet still enthusiastically quant-oriented folk. People like successful real-world risk manager