Paul Wilmott, eminence grise of the quant world has posted a manifesto which among other things points to the wrongs of the past, but neglects the fundamental truth of models and risk.
In applied systems the rules determine the outcome.
The very concept of a "truth" in modeling or risk is flawed. I am ashamed to use the term, but will. Markets are post modern. In that I mean that sustainable markets rely on multiple truths: perceptions of risk, reward and timing. If everyone has the same truth, the system fails as it teeters toward an extreme. A uni-model system is less stable than multi modeled one. When everyone looks over the edge of the boat on the same side at the same time it tips over. Everyone working in a Basel 2 or VaR type of framework will share very similar opinions. We need radical dissent and it to hereticaly ignore the pursuit of collective models of risk.
If all groups pursue or use the similar models with slight tweaks they will reach similar conclusions, at the same time and like everyone on a life boat looking over the same side, tip it over. The unified presentation of "correct" or "true" models and unified risk is what caused the problem with CDOs etc. in the first place. The ratings monopoly on Truth combined with similar models and trader laziness among other things tipped the mortgage boat over. AAA all the way.
Bubbles and manias are caused by everyone sharing similar models and horizons of risk. Whether it is gut feeling "mental model" that says Tulip Bulbs can only go up or a CDO valuation cranked through the latest hot quant model, shared truths become false truths. Lets stop chasing them or forcing banks to pursue them. Wilmott's proclamation is nothing but a request for the red queen to run faster, individually that is fine, but a collective pursuit is dangerous.
Basel 2 with its self similar approaches to modeling is crying out for reformation if not outright elimination. Forced use of mean variance type models will induce nasty problems as everyone seeks to game the systems capital weighting in the same way. SIV 2.0 will beget Credit Crunch 2.0, in a bigger more sophisticated way. Our faith will be emboldened by faster computers and brighter quants. "modern" finance always comes up with new ways to blow up. Speed and greed translate into leverage in the financial world. It may take 10 years, but we will get there and we will be sorry. The first group to use a model will make money and the next few less, by the time it becomes a "risk" management tool, the boat tips over. The model bears the seeds of its own destruction in a Popperian sense.
I prefer my inevitable banking failures small if you please. Lets minimize systemic risk, by decoupling the system. De-coupling means getting everyone to form their own opinion of risk and reward. Naive inter bank trading "caveat beware" should return. Yes this will lead to a less efficient system, but it will be a more stable one. It is a Tobin type tax if you will, that acknowledges the unknown and the unknowable inherent in any economic transaction. Get me there slower but safer at the over all system level.
In every other engineering business the rule is safety first! Safety should come before optimization or transactional efficiency. A few basis points for the bankers bonus isn't worth the collective risk. Slow down.
When the models and regulations rule, actor's risk horizons and opinions converge. This convergence on a truth put the system in a dire and unstable state. Does anyone find it interesting that the only entity apparently capable of bailing out the current crisis is the FED which doesn't have a Tier 1 capital requirement? God help us if/when the bond market wakes up to the shenanigans going on at the FED.
As Churchill said," where all think alike, none think much." Dissent and ignorance of counter party risk counter intuitively provides stability as actors tread lightly upon the stage of finance.
My credo is lets all use models mental or mathematical, but let none be beholden to a collective "truth" of them a la Basel 2 or regulatory regimes.
