Future of Financial Mathematics? 301
An anonymous reader writes "Nassim Nicholas Taleb, a famous 'Quant,' has long been a strong critic of the use of mathematics and statistics in the financial markets. He has been very vocal in his books The Black Swan and Fooled by Randomness. In his article on edge.org, he says 'My outrage is aimed at the scientist-charlatan putting society at risk using statistical methods. This is similar to iatrogenics, the study of the doctor putting the patient at risk.' After the recent financial crisis, wired.com ran an article titled 'Recipe for Disaster: The Formula That Killed Wall Street' in which the quant David Li and his Gaussian Copula were crucified — we discussed it at the time. Now, I've recently been admitted to a graduate program of good repute in Computational & Applied Mathematics. There is a wide range of subjects in which you can pursue your PhD, one of them being Financial Mathematics. I had a passing interest in it for quite some time. In the current scenario, how advisable it is to pursue a PhD in this topic? What would my options be five years down the line? Will the so-called 'quants' still be wanted by the banks and other financial institutions, or will they turn to more 'non-math' approaches? Would I be better off specializing in less volatile areas of Applied Mathematics? In short, what is the future of Financial Mathematics in light of the current financial crisis?"
Monte Carlo, we should have known. (Score:2, Informative)
I always found it troubling that one of the computational algorithms [wikipedia.org] they relied on was also the name of where James Bond liked to gamble.
Re:A bit self-defeating (Score:5, Informative)
But his thesis is that such events are fundamentally unpredictable. If he made a fortune, it means _he_ was able to predict it, well enough to profit for it.
No. His "system" is indeed based on the assumption that such events are unpredictable. That's why he for years bet
simultaneously on a sharp increase and a sharp decrease in stock values - and ran slight daily losses, in the
anticipation and expectation that once such an event inevitably happens, the profits will more than make up for those
losses.
It worked.
He basically had no idea - and didn't care too much - when this (and what this "this" would be) would happen though.
Re:If you enjoy it ... (Score:5, Informative)
Re:Advice from a PhD student (Score:3, Informative)
I generally agree with this. I would like to add a few more points from personal experience:
Read people.. (Score:1, Informative)
I'm not even sure I agree with him or don't but get his argument more or less right before you start making random comments. Also his success or failure isn't an indication he's right otherwise all those guys who made millions then lost it all were "right" at least up until they weren't. I tend to think there is more structured thinking behind Mr. Taleb's work and I respect it but success in one instance is not vindication...that's called luck as often as not.
Re:A bit self-defeating (Score:4, Informative)
It wasn't so much the 5% failure rate that was the problem. That and probably more was expected. It was the fact that when those loans failed, they could not be recovered because house values had actually slipped in a lot of places - that had never happened before on any kind of wide scale. But you're right that it could be predicted, and I'd expect your formula for predicting a market collapse of some kind is probably pretty acurrate in a general "rule of thumb" sense.
This housing market collapse was predicted over a decade ago. The recipe for disaster was there, the only question was exactly how long it would take.
Heavy-handed incetives to take risky loans that were first implemented 20 some-odd years ago, but greatly ramped up by the Clinton administration, created a climate where it was quite profitable to take on bad loans, and then shift, move, and otherwise hide that they were bad. The housing market itself was able to hide the risk of taking on so much bad debt, for as long as house values went up faster than interest rates, even extremely high risk loans (like fixed payment loans, and second and third mortgages) would almost always be covered when a house was sold. Homeowners who can't pay their bill just sell, banks get their money back, and nobody loses. Even when forclosure was necessary banks could reliably recoup most of their money, though forclosure is less than ideal.
However, as soon as housing prices stagnated the model became tenuous, and when it slipped, even a little, the model collapsed. You ended up with thousands of loans that could not be covered by the sale of the home, and forclosure was the only option.
This is BAD. Forclosures are expensive anyway, and only cover the balance left on the house for the initial bank, the secondary balance if there is one, or the homeowner if there isn't gets what's left. Forclosures that can't even get the primary loan balance back are a financial nightmare. What's more, the homeowners that tended to default were also more likely to get second and third mortgages to stave off forclosure. Houses went from being "guaranteed income" for banks to a massive liability. Companies like CountryWide - which embraced the Government's high-risk low income loans and who either Fannie May or Freddie Mac, I don't remember which, touted as the "model" for lending - was the first company to fail (no bailout for you! too bad so sad!). AIG, by the way, profited by shifting the risk of these loans via a special insurance type. Obviously that worked out well for them.
I personally think the healthiest thing to do would have been to let the market collapse and allow other companies to fill the voids. Definitely more painful in the short term, but in the long term I think it would be better. We'll be suffering for a long time with the current bailout plan. Though of course, I'm no economist, but they haven't done so hot anyway so...
Re:Monte Carlo, we should have known. (Score:2, Informative)
Re:A perfect subject.. (Score:1, Informative)
I have worked in this field, for almost a decade, as both a fixed income and equity derivatives quant. I never worked in structured credit, btw. Too unsavory.
Times are grim for new grads as there is a glut of talent walking the street. A masters of finance, plus three bucks, gets you a coffee at starbucks.
When I am hiring a phd I would sooner get a physics phd who self taught himself finance, than a guy who went the finance phd route. Every time.
Re:A bit self-defeating (Score:5, Informative)
His system is basically arbitrage. We have an algorithm for pricing options (Black-Scholes) that makes an invalid assumption (it uses Gaussian statistics where it shouldn't). This fault was recognized almost as soon as it was published, but people continue to use it anyway, which means they're mis-pricing their options. Black Swan makes money in the long haul because they know big price swings occur more often than the options are priced for, and they buy based on this knowledge. Exploiting market mis-pricings like this one is the essence of arbitrage.
Like classic arbitrage, this only works because a) there is a mismatch between price and real cost [risk] and b) there aren't a lot of players making the same purchase. Change either A or B and Black Swan's strategy will become a money-loser, or at least a break-even.
-JS
Re:A bit self-defeating (Score:1, Informative)
After 10 years at 10% growth you have over saturated the market by 100% and every company that got there with unstable books and balances collapses.
Um, 10 years of 10% growth is 159% total growth. You've failed at the maths, so I stopped reading there.
Re:A bit self-defeating (Score:1, Informative)
He is "buying insurance". He pays a small cost each year and hopefully gets a big payoff when things go tits up.
In contrast many quant based strategies are accidently selling insurance - their payoff is a small regular sum, but every so often they make a huge payout (loss)
Both strategies can be modelled very reasonably as a gaussian (bell curve), however, the trick is to realise that you have two (or more) types of scenario embedded, so you have the main gaussian distribution, plus another distribution on the outlier.
This models the situation that the world tends to behave one way for long periods of time, and then suddenly all the correlations change in times of distress and a new distribution dominates
Of course that's a good model of the world, the problem is that it's near impossible to parameterise it until after the event..