Rebonato does not mince words, pointing out the weaknesses of mathematical models, and noting that most quants, although mathematically sophisticated, often lacked deep knowledge about markets and banking (I assume he does not include himself in this group). In my experience many quants never questioned the basic efficient market assumptions underlying their models, although some certainly did -- in particular, those with trading experience.
Rebonato is polite, even urbane, but disagrees with Econtalk interviewer Russ Roberts on many important issues. The most important question, which Rebonato addresses immediately, is whether enlightened, self-interested managers of financial institutions can be relied on to properly manage risk. Regulators accepted, on faith, the self-regulating abilities and properties of a system managed by such people. Thus, one of the main ingredients in the crisis was the ideological (as opposed to political or financial) capture of regulators by efficient market proponents.
Other interesting topics covered are the divergent risk tolerances and interests of bond holders vs equity holders vs regulators of banks [1], and whether moral hazard (anticipation of a bailout) played a role in the crisis -- Russ, the anti-government libertarian, says yes. Rebonato says no, the story only makes sense if told at the institutional level, whereas individual incentives were different. I think Rebonato's logic is impeccable. It's more persuasive to me that incentive schemes which allowed huge compensation based on short term (ultimately illusory) gains were much more of a factor. (See Clawbacks, fake alpha and tail risk.)
Dr. Riccardo Rebonato
Riccardo is Global Head of Market Risk and Global Head of Quantitative Research and Quantitative Analysis for Royal Bank of Scotland based in London. Prior to joining the Royal Bank of Scotland, he was Head of Complex Derivatives Trading Europe desk and Head of Derivatives Research at Barclays Capital, where he worked for nine years.
Riccardo is a Visiting Lecturer at Oxford University in Mathematical Finance and Adjunct Professor at the Tanaka Business School, Imperial College, London.
Before joining the financial world, Riccardo was a Research Fellow in Physics at Oxford University (Corpus Christi College) and, before that, Visiting Scientist at Brookhaven National Laboratory.
Riccardo is the author of the books Plight of the Fortune Tellers ('07), The Perfect Hedger and the Fox (Wiley ’04), Modern Pricing of Interest-Rate Derivatives (Princeton University Press ’02), Interest-Rate Option Models (Wiley ’96,’98), Volatility and Correlation in Option Pricing (Wiley ’99). He has published several papers on finance (option modelling, computational techniques, risk management) in academic journals. He is a regular speaker at conferences worldwide.
[1] Footnote: see my earlier post on the vacuous Modigliani-Miller theorem. I recently learned from Vernon Smith's memoir (see pages 230-231 and 276) that he has similar opinions. Google books link; also search under "MM".
What I read from Riccardo Rebonato made a lot of sense.
ReplyDeleteHe knows his instruments enough so that he gets what their limits are.
The critic made on quants apply equally well to the trading side : Most traders do not know enough the underlying assumptions of their scientific tools.
Yes, no one would claim that typical non-quant traders understand the math.
ReplyDeleteWhat I meant by that comment is that people with quant backgrounds who trade typically learn right away that many assumptions in the models are not obeyed in the real world.
Thanks for the interesting post. Your posts about economics/finance and careers from a physics viewpoint are the reason I've been reading this blog for over a year now.
ReplyDelete"many assumptions in the models are not obeyed in the real world."
ReplyDeleteI think this is difficult for some people with a science or engineering degree to grasp. In these classes, even the most difficult problems are fraught with assumptions, such as no heat loss or no friction or no 3-d effects. The engineering student, having worked through 4 years of problems that all come down to one solvable differential equation, thinks he can apply these rules in far more complex systems.
This probably explains their initial failure to formulate predictive models.