"The State of Financial Engineering"

An email suggested this as a follow-up to Quants Did It?:

The State of Financial Engineering, by Sylvain Raynes: ...All over the world, it has become fashionable for Universities and Colleges to offer Masters degree programs in quantitative finance or financial engineering (FE), a code word meaning the solution of the Black-Scholes option pricing differential equation in as many ways as possible. To do so, students are taught to use basic techniques in numerical analysis whenever the equation is either non-linear or does not lend itself to the standard analytical solution. As a precursor to this main task, the program usually includes a course in stochastic calculus during which Ito's celebrated lemma is discussed, proved and used.

In general, the cost and length of such programs are remarkably similar... Even Ivy League schools like Princeton University, who swore up and down they would never play this game, are now happily teaching finance and deriving significant incremental income from a fully depreciated curriculum.

The techniques taught in quantitative finance are completely standard in other fields. In most cases, the only exciting thing about the curriculum is that one day these methods might be applied on Wall Street to the calculation of cash flows. If they were instead applied to the making of widgets or the collection of tomatoes, it is a fair bet that nobody would be interested in them, and certainly no university would be able to charge $35K to learn them. ...

It is a plain fact that the field of quantitative finance has not made a single fundamental step forward over the past twenty years, not to mention that Black himself, by his own admission, had nothing to do with the equation that now bears his illustrious name. The BS equation was first formulated and solved by Casey Sprenkle some ten years before Black's famous 1973 paper in the Journal of Political Economy. Regrettably, it is still politically incorrect to give due credit to someone who made a real contribution to finance. Unlike those of some of his associates, Black's reputation hardly hangs on one paper.

Statistics and numerical analysis have nothing to do with finance per se but are merely tools of financial analysis, just like accounting statements and legal opinions. Finance is quantitative by definition; there is thus no need to add an adolescent adjective to the word. This is like saying aerial flight or wet swimming. Although people employed as aerospace engineers use computers on a daily basis, none would describe him- or herself as a computer programmer.

But if this were about mere semantics, it would not be worth mentioning. Unfortunately, FE programs are also drifting farther and farther away from their purported subject matter. In effect, quantitative finance has entered the scholastic stage whereby numerical techniques are taught completely out of context as if a deal were somehow a differential equation that could be solved for the right solution. In fact, there is no solution to a deal as there is to a differential equation. ...

Students thinking themselves financial experts simply because they can solve the BS equation in a few minutes (there is apparently no other one around) are being misled by their own mentors and teachers into the naïve belief that this amounts to finance. ...

A deal only happens when various constituencies (lawyers, investors, bankers, rating agencies sometimes, regulators, accountants, etc.) are able to come together under a unified framework. This is not the time to deliberate or lecture on cross-correlation and conditional VAR. On the contrary, any such talk is completely counter-productive...

Too often, students who would otherwise have something valuable to contribute are being led down the primrose path... The promise of high-flying jobs in New York is all that has propped enrollment at its current levels, since once on the job, graduates of financial engineering programs quickly become aware that it is MBA graduates, and not themselves, who are destined to breathe the rarefied air of boardroom deal-making. In fact, the label quant has now become quasi-pejorative, the practical equivalent of geek or inconsequential number-cruncher. Deal-makers do not want such people in front of their clients, if only for fear of hearing naïve prognostications of hetero-scedasticity and Gaussian copula bandied about before befuddled investors (and even before lunch).

Not surprisingly, business school professors often warn their students about not being labeled a quant if they ever want a career in the community of finance. Given the current state of affairs, we could not agree more. ...

What is obvious and regrettable is that no effort is ever made to teach numerical analysis as a proper and rigorous discipline. Instead, students literally learn numerical recipes and are no more equipped to handle reality than someone equipped with a driver's license when their car breaks down. One also gets the disturbing feeling that the majority of teachers involved in quantitative finance have limited knowledge of either finance or of the elements of numerical analysis. Unfortunately, too many professors in that field are there for the ... reason that ... they can earn much more for teaching the same material in finance than they would in the original field. It's hard to turn down $150,000 for teaching either statistics or the numerical solution of partial differential equations on a full-time basis when your equally competent buddies from graduate school are doing the same thing elsewhere for $50,000.

Although we freely admit that we have not performed the monumental task of a complete inventory and comparative analysis of financial engineering curricula, a cursory review of the most popular ones reveals what seems to be their central dilemma, which is how to fill 18 months of teaching with a single topic: stochastic calculus and its applications. The general answer appears to be to fill the remaining 14 months or so with subsidiary material peripheral to finance but available for much less elsewhere. ... One meanders across finance discussing swaps, default swaps and various options (instruments that hardly require Herculean intellectual prowess to grasp), engages in endless and meaningless debates on the eigenvalues of correlation matrices, and then dabbles in numerical analysis by learning basic methods applied to the BS equation taken as the last word in finance for the remainder of mankind's existence. At no time, as far as we can tell, are students taught how to construct a numerical method from scratch or how to tell if it will work or fail.

Throughout our Internet search, the following topics were absent from the syllabi of the numerical analysis courses within the financial engineering curricula of the academic institutions we reviewed:

1. Z-transforms and Laplace transforms
2. Banach and Sobolev spaces
3. Fourier series and transforms (one exception)
4. Lax equivalence theorem (same exception)
5. Von Neumann stability analysis
6. Courant-Friedrichs-Loewy (CFL) condition
7. The Nyquist sampling theorem (useful in Fourier analysis)
8. Convergence analysis
9. Error propagation analysis
10. The Weierstrass approximation theorem
11. The interplay between truncation and discretization error

It is simply not possible to claim expertise in numerical analysis if one does not have at least a passing acquaintance with the above foundational elements. However, learning these things takes time. On the other hand, if the goal is not to become knowledgeable in numerical analysis but simply to learn a sundry assortment of basic methods, there are much cheaper ways to do this, for instance in the mathematics or computer science department of the same school. Numerical analysis is a well-formed discipline that does not need finance to give it credibility.

The consequence of all this is that today, and through no fault of their own, students with degrees in financial engineering are ill-equipped to face the rapidly changing face of finance. Once ensconced in their jobs, they are quickly marginalized and relegated to the role of glorified programmer until being eliminated in the next headcount reduction because (with unfortunate justification) they are not considered producers.

It would be a different matter if financial engineering were just a code word for numerical analysis with finance used as a marketing mechanism to attract people to the field. It would be equally acceptable if financial engineering were devoted to the actual practice of finance instead of being largely an obsession with one equation... Someone who has never done an actual deal can hardly be expected to know how deals are done, let alone teach how to do them. Likewise, a manager who used to supervise twenty-five Ph.D.s in some research department on Wall Street has as much knowledge about deal making as an usher at Yankee Stadium has about baseball. On the contrary, it has become painfully obvious that these managers, if the term can be used at all to describe this level of incompetence, are precisely the people who truly need supervision instead of underlings who, at bottom, never make a single decision that could take their firm down.

Financial engineering never grew up within finance; it was taken over by physics. ... Unless the field re-invents itself pronto and starts becoming relevant to what people actually do out there, graduates with newly minted financial engineering degrees hoping to see a decent return on their own or their parents' sizable investment will continue to be sorely disappointed by their actual career prospects, and will keep wondering where in God's name they went wrong. Regrettably, the answer is: nowhere.

Comments

[Tom Hartman]

Maybe the quants don't understand the deal - but that isn't what brought down the companies. Fundamentally, the managers and heads of companies didn't understand the risk, or ignored the risk. How many of these individuals understood at all the FE that was involved in the deals they were doing? Gun is pointing in the wrong direction on this one.

[Massimo GIANNINI - M.G.]

I wonder how and why Europeans got involved in this and were caught into the "trap". We do not have Nobel Prizes or financial engeneers of that level...

[ani]

Isn't the same, to a large extent, true for Economics? Stochastic calculus porn?

[Dave]

The author is right, and the lack of depth in many of these courses is, frankly, scary. I remember being exposed to a great deal of the underlying mathematics discussed above when I studied electrical engineering. I have been amazed at the sheer number of people leaving undergraduate and specialist courses in finance / economics, destined for roles in financial modelling and (in some cases) quant analysis, who lack even a cursory understanding of the mathematical principles they seek to apply.

It is quite amazing that Z/s transforms, fourier analysis (not to mention wavelet theory or more complex tools), basics of nyquist sampling theory (aliasing, anyone?) and the rest could be omitted from these programmes. Never mind complex analysis, non-linear and adaptive methods, Lyapunov stability criteria, Kalman and information filter theory, solving AREs, H-inf and many of the more complex control theoretic techniques that could (and have) found application beyond the physical sciences and engineering in the financial world.

And of course, how could any course in quant analysis be complete without a detailed deep-dive into Mandelbrot's work on fractal aspects of finance? That these topics are omitted to me suggests that these degrees in so-called "financial engineering" are a toy. You would be better off taking an undegraduate degree in electrical engineering (with a control or signals emphasis) or a bachelor/master degree in statistical methods or applied mathematics.

[Valuethinker]

Hi

AFAIK the syllabi *do* include courses in numerical analysis. They also include electives in the likes of corporate finance, which focus on institutional structures of deals, not just the risk mechanics.

The criticism seems overstated.

The problem described is a generic one of professional teaching: the divorce from real world practice.

School can't teach you work: work (career) is a political process with political factors. The 'risk managers' at the banks both failed to appreciate the risks, but also to communicate that at the highest levels of the firm. That is analogous to what happened with Arthur Andersen at Enron. Many professional services firms have been sucked down that path.

I believe the top hedge funds (Citadel etc.) tend to hire Phds in engineering, physics, economics etc. in preference to Msc in quant finance.

[Beezer]

Taleb's book Black Swan seems to be talking to the same thing: The mathematics being used are often mis-applied and, therefore, the conclusions often very wrong.

He admires Benoit Mandelbrot, inventor of the Mandelbrot set and of the word "fractal."

I'm not qualified to discuss this on a mathematics basis, but from my personal experience in finance, Taleb's observations appear to be more right than wrong.

Is Taleb's intuition something missing from the curriculum?

[ken melvin]

Afore my time, anyone who understood autos was considered to be a genius; in my time, some (including the electronics types themselves) considered anyone who understood electronics to be geniuses; while back, anyone who claimed to understand computers was considered to be a genius; now anyone ...

Today, and since at least 1980, finance is king of beast.

[MP]

Ironically, Taleb teaches (or used to teach) a course at Courant's Math Finance program at NYU. He also lectures (or used to lecture) at Wilmott's CQF (which is not even a master's degree)

[Turbo]

Well the same criticisms can be made of graduate education in Economics - emphasis on false mathematical rigour over a true understanding of economics. There's nothing wrong with being taught financial math as long as one realizes that it's a framework for understanding financial problems and not necessarily the solution to them. That distinction has been pretty much absent from the profession.

[Tim Payne]

I 100% agree with this; I worked as a Quant pricing asset-backed securities over two internships and have transitioned over to front office sales in a financial institution; the skills you will use as a Quant with the masters degree is marginal at best -- most Quants are glorified VBA & MS Access programmers. Sorry.

Deal makers are actually the jocks and football players from college -- most deals are based on relationships and drinking... certainly pricing matters but most firms are on an equal footing for the most part on pricing -- it is the relationships that matter.

BUT.......

If you get a financial engineering degree, you are not equipped for deal making; however, you have a lot of doors open to go work for a hedge fund who can use your numerical skills. Don't go to an IBank (and don't worry the market will take care of those soon enough...) but look into prop. trading groups to harness that degree.

[Barkley Rosser]

Sitting in a lounge at Dulles Airport about to board to fly across the Pacific.

Ito's Lemma is based on stochastic processes characterized only by variance. Hence the BS equation does not handle the empirically relevant kurtosis of fat tails. Practitioners figured this out back in 1987, and in that regard, a takeover by the physicists is not all that bad, in that they are perfectly ready to deal with such matters.

I do not know what is taught at some of these courses, but if the description is correct, they are seriously deficient.

Regarding Taleb, ultimately he says that no probability distribution will suffice. In that regard he pushes Keynesian-Knightian uncertainty, although he likes to push various forms of tail insurance through his firm. In his book he describes the approaches that allow for fat tails as "grey swans." Thus, his strategies are themselves really grey swan strategies. For true black swans, one must simply suck it up and be philosophical, which he likes to be some of the time, when he is not pushing his firm's products.

[robertdfeinman]

The criticism is probably fair, but the critic is someone promoting his own brand of snake oil - check out his consulting firm.

I pimped my white paper comparing econometrics to signal processing yesterday, so I'll skip the explicit link again.

[Organic George]

I had a mentor who told me the more complex the deal, the more someone was trying to hide the truth.

He would receive dozens of packages each week from brokers, insurance companies, etc. He would look at each one for about 10 minutes, but he only read the footnotes. 99% of the packages were tossed out and of the 1% he had his team review thoroughly maybe 0.001% would actually be considered for further discussion.

He never drank with the people he did business with, and he fucked most of them at the signing table. But that had more to do with the fact that the first pair of long pants he wore were his great uncles civil war trousers. He came from abject poverty.

He died about 20 years ago but I have followed his advice on researching the baseline of the proposed deal. All the math is crap and all the deal makers are street hustlers with expensive suits.

I do have hope for this country now that Wall St will no longer be the center of the universe. Let's hope all the McMansions are turned into halfway houses for drug offenders and petty criminals working themselves back into society.

To paraphrase Lee Iaccoca: At some point we have to stop selling pieces of paper to each other, we have to manufacture products that people want to buy.

[donna]

Lots of my physics friends were hired to work on Wall Street as the business majors "couldn't handle the math". I think that was the first sign to me that we were in big, big trouble.

Finance isn't meant to be rocket science, guys. Really.

[kthomas]

Interesting article.

[cm]

Nice essay, but I hope I'm not supposed to be surprised at the revelation of:

"The promise of high-flying jobs in New York is all that has propped enrollment at its current levels, [...] Unfortunately, too many professors in that field are there for the ... reason that ... they can earn much more for teaching the same material in finance than they would in the original field."

After all, hasn't that been dangled in front of us all the time, education for "highly skilled" (as in well paid) jobs?

[Mark]

The argument would be more persuasive if it were accompanied by some evidence of an association between performance and the numerical analytic capabilities of individuals and/or firms.

[kthomas]

At the height of the dot.com bubble, many companies sold what is often called vaporware or crapware. This article reminds me of just that.

[Jrossi]

Not knowing advanced mathematics, I can put no stock in it. I can, however, trust in herd psychology, which teaches that securities markets will go nuts from time to time. This might just turn out of be one of those times.

[OhNoNotAgain]

"anyone who claimed to understand computers was considered to be a genius; now anyone ..."

You'd be surprised how many people in the IT field don't really quite understand how everything works. They know enough to get by, but that's it.

And, with no offense to our esteemed host, but I see this a lot with university research that is published in my particular area of expertise in software development - database software. Many of the papers start out with a good idea, but only provide a sketchy outline of the actual implementation, and the end result is often something that is great in theory but would never work in the real world due to its performance, usability, or concurrency implications. I attribute this to the fact that many papers simply don't progress far enough into actual coding to understand that they don't/won't work. Some of them read like that South Park underwear gnones joke:

Phase 1: Collect Underpants
Phase 2: ?
Phase 3: Profit!

[OhNoNotAgain]

Duh, that's gnomes, not gnones.

[Richard H. Serlin]

I have worked with this kind of mathematics in finance (for example). I think it can be very valuable, and eventually will provide a great deal of value, but the process of learning how to use it properly, and utilizing it in general, will be long because the mathematics is so complex, so few finance and economics academics know it, and it takes a long time to learn (at least relative to how much pressure finance and economics academics are under not to undertake even very high NPV projects if the payoffs are long term).

The big problems with "quantitative finance" (I know the author complains that all finance is quantitative, but this finance is far more quantitative, or mathematical, than average, so it's not that bad a way to distinguish it. You could be more precise with a longer name like "extraordinarily quantitative finance", but people really like short names.) are problems you find throughout finance and economics academia:

1) People often grossly misinterpret models, usually because they take their assumptions literally (act as though the model is exactly reality, without thinking how things will differ due to how much the assumptions differ from reality), or too literally.

2) Finance and economics academia is so specialized. An economist may understand his narrow (or super narrow) area well, but little else in economics very well, yet he thinks he does because he has tenure at Harvard, and much of the public thinks he does too for the same reason. A lot of these quant guys have gotten all kinds of accolades and rewards and high level positions because of their understanding of advanced mathematics and workaholism, but they understand very little of anything else in finance and economics.

3) Intuition and high level thinking is grossly under-rewarded in finance and economics academia. You can get tenure at Harvard and a Nobel Prize just from strong mechanical skills and workaholism from age 5. You get to the frontier quickly and be the first to do valued mechanical work, in a formal way, with a pedigree so that it gets published in a top journal and you don't suffer the fate of Casey Sprenkle, who as the article notes, formulated and solved the Black-Schoales equation "some ten years before Black's famous 1973 paper in the Journal of Political Economy. Regrettably, it is still politically incorrect to give due credit to someone who made a real contribution to finance."

However, as I've noted many times, academia may have serious flaws -- very largely due to ginormouns asymmetric information -- but it's still a fantastically high return investment. It's like modern scientific medicine. There's a lot of room for improvement, but it's still a tremendous investment, extremely well worth a lot of money, and we certainly would not want to be without it.

[Karl]

The article's thesis seems to be that quants are not well prepared, by education alone, to do deals in the real world. However, I'm not sure quants are trained to "do" deals as opposed to evaluate them. Someone has to perform this function, however crudely, with the tools at hand. Even if the quants have adequate educational training, they may not have enough independence or clout to do this properly. Organizations will always need the quants to keep score. The failure of good score-keeping is really, ultimately, a failure of the organization's governance, control and reward structures. Management often determine whether the key assumptions are realistic. Then the results, even if valid, need to be communicated--without political noise--up the management chain. Just read the sad story of CitiGroup to understand the limitations of Risk Management in today's corporate world--it has little to do with the Quants and their models, but how management used, abused or ignored them. Fortunately for Citigroup, Rubin, Geithner and Paulson saved the day for them.

[kthomas]

Richard, what good is this knowledge when nobody is buying? Seems the people who aquire this knowledge really add nothing of value to our economy. (Yes, THEIR wages have been high, of late....of late.) Please prove me wrong, my good sir. What I'm trying to say is, the Party is just about over.

OhNoNotAgain, that Phase 1, 2, line was hylari0uso. And don't worry, the Spel1ing Police Troll$ are not usually trolling Prophesor Thoma's cite.

[evagrius]

I like it- "BS equations" is perfectly descriptive.

[winstongator]

I wonder how many companies would admit to using the BS method to price options if they knew it had as an underlying assumption that the stock price fluctuates as a RANDOM WALK. What is the use of giving an employee an option if you are admitting that you are assuming that any change in your company's stock price is RANDOM!

Why don't companies account for options as potential liabilities and take automatic hits to their balance sheet for the difference between strike price and current stock price.

I looked up a paper on modeling the risk of mortgage backed securities, and the way they modeled for default risk (in this paper) - THEY DIDN'T. They assumed that most mortgages were insured against default! Too few uninsured mortgages would actually default that it isn't even worth considering.

No amount of math can make up for stupid assumptions.

[OhNoNotAgain]

kthomas,

"OhNoNotAgain, that Phase 1, 2, line was hylari0uso. And don't worry, the Spel1ing Police Troll$ are not usually trolling Prophesor Thoma's cite."

Yeah, I'm not a big South Park fan, but that bit was pure genius on their part, mainly because it captures so much that is wrong with our financial markets in such a succinct fashion.

[save_the_rustbelt]

When I was in grad school in the mid 70s, operations research was the new, new thing.

Put all of the variables in a matrix algebra equation and the magic math would make all of your important business decisions -- make 20 red cars, 10 blue cars, 5 green cars, blah blah.

(One of the courses had a bunch of engineers working thoward MBAs, and they blew all of the accounting and management types out of the water with their damned slide rule math skills - yuck.)

We had to do all of this crap without the benefit of camputers (unless one wanted to do punch cards and wait 24 hours for the answer).

Math will never substitute for good judgment applied to the best facts available at the time.

[Walt]

Quant is a specialized job, like accounting. Quantitative finance allows you to get a handle on the moving parts of a complex portfolio, so it requires people with specialized expertise. Accountants don't do deals, either. Honestly, I'm beginning to subscribe to Michael Lewis view that dealmakers are a bunch of dumb jocks that get ahead on connections and charisma until they get over their head.

winstongater, one paper does not a literature review make. (It sounds like you found a pretty terrible paper, though.) There are lots of papers on how to price mortgages, and the key variables that they try to value are the default risk and the prepayment risk. It turns out hard to do -- if you try to derive it from economic first principles, then you predict that people default less often than you'd expect, and they prepay less often than you'd expect.

[Dwight]

Nice antidote to the constant refrain from Wall Street that the current problems were created by the best and the brightests, the most brilliant mathematicians on the planet, and lunchmeat intellects like Dick Fuld, Chuck Prince or Vikram Pandit should get a free pass (sorry, VP, but if you're used to being the smartest guy in the room, that doesn't say much for the company you keep).

That was always BS (not Black Scholes).

The mathematicians who gravitated in numbers to Wall Street were uniformly B List. The A List crowd stayed in Princeton at the Institute for Advanced Studies, remained in Cambridge (whether Anglia or MA), is hiding out in Leningrad, or otherwise occupied in any of a handful havens in California (where they may very well be trapped in overpriced real estate, for all I know).

But those guys never keep score by how big the bonus or nice the wristwatch (well, maybe how nice the wristwatch). For an insight into the mindset, read G.H. Hardy.

[Richard H. Serlin]

kthomas asks, "Richard, what good is this knowledge when nobody is buying? Seems the people who acquire this knowledge really add nothing of value to our economy."

The good of this knowledge is like the good of knowledge in nanotechnology or quantum computers. The good is long term. The knowledge won't do much to help a short term liquidity crunch and demand decrease chain reaction (recession or depression); for that you need a great increase in government spending on high return investments (the kind the free market will grossly underprovide due to well established market problems like externalities, especially the pink elephant of economics, positional/context/prestige externalities, and you need a monetary expansion by the Fed, in this case a huge one.

But quantitative (continuous time, to use the jargon) finance and economics research, like nanotechnology research, has the potential to increase productivity substantially in future years, with advancements in understanding and application. Currently, though, it's been applied horribly, and its application needs to be modified greatly.

[Richard H. Serlin]

By the way, Paul Krugman agrees with me. Writing on the recent death of Kyoshi Ito, one of the fathers of continuous time mathematics (the core tool of quantitative finance and economics), Krugman states:

Ito studied random motion, and his work has played a key role in finance theory — and even in some of my own work. I’m not much of a mathematician, tending to pick up no more technique than I need — Avinash Dixit liked to make fun of the way I’d write down the basic Wiener process, then say “whatever that means” — but this was really useful stuff.

[kthomas]

Mr Serlin, thank you.

[some investor guy]

I have a physics degree from a well-respected university. I got an MBA from a highly ranked program. Things which gave the nonquant MBA students fits were easy for myself and a few others from math, stat, and engineering backgrounds.

In 20 years since I graduated, I have done a massive amount of stats and simulations. I have done exactly one convulution integral, and maybe three partial differential equations for work. Things like fat tails, skewness, and biased data samples are problems that crop us. However, those aren't what really bother me.

What has left me aghast about modeling was most often that something was modeled at all, or was assumed to be only the baseline value. Before the credit crisis, I was taking people to task for models which assumed that real estate prices could not go down. I showed them Shiller's price histories, and even reports from Southern California from the 1990s real estate bust, and the Japanese real estate bust. This made them nervous, but they didn't think it would happen, at least not nationally in the US.

I worked on a very large deal where there were interest rate simulations all over the place, and some asset return simulations, but the project was sitting on a well-known major quake fault and no one had anything in their models about it. Why? Because their prebuilt models were made for interest rates and asset returns. We beat out 4 other investment banks partly because of our modeling. Fabulous insights, client loved the work. We got the deal, but got paid exactly what other firms would have gotten paid for the same deal.

Serious quant work is usually being done by people betting their own money, not by people trying to do a deal.

[some investor guy]

I meant to say "What has left me aghast about modeling was most often that something was NOT modeled at all"

[cm]

OhNoNotAgain:

"You'd be surprised how many people in the IT field don't really quite understand how everything works. They know enough to get by, but that's it."

That's true for any field of business, and indeed for any field of human activity/interest. But OTOH the "getting by" skills are often practically good enough. I used to have much higher (and probably unrealistic) professional standards, but with "experience" I have reconsidered.

[kthomas]

Some Investor Guy, thanks.

May I rephrase? "What has left me aghast about modeling was most often that something was NOT honestly modeled at all"

Oh, the virtues of honesty.

[wml]

Do you have any evidence that, say, even academic economics has had any positive impact on real aggregate wealth in the last 30 years?

When I was an undergrad, I once asked my advisor "Has anyone ever done a good study to quantify the benefit of academic economics to the economy?" His response was "I don't know of any, and we probably wouldn't want to, as I suspect the result would come back negative."

[John Emerson]

Scholes really must be a prince. LTCM goes belly up, he's implicated in eight-digit tax fraud, he then marries Jan Blaustein (who was more deeply implicated in the fraud than he was), and now people using sophisticated quant tools he devised have just brought on a depression. Good work, Myron!

Economists who resent the things people say about their profession should keep people like Scholes and Gary Becker in mind. They're not merely members in good standing, but highly honored Nobel Prize winners.

There may be jerks in the other Nobel categories, but by and large jerkish novelists and chemists don't contribute to destroying the world economy or involve themselves in fraudulent get-rich-quick schemes.

But, as Summers says, they have very high IQs and know tons of meth.

[some investor guy]

Kthomas asks, May I rephrase? "What has left me aghast about modeling was most often that something was NOT honestly modeled at all"

Well, you do see some dishonest modeling, but it usually smells like junk even to the moderately sophisticated nonquant. Those are people who ask questions, but don't know how to build models. I was at a conference where a speaker had a financial model showing how much money they would make on an investment. Even the nonquants picked up on the problem: why is someone giving you above-market returns?

There is a good deal of conveniently-selected modeling. This sometimes is quite innocent. For example, people think that 2007 is a low inflation point in time, recent history was low inflation, and they leave out both high inflation and deflation from their simulations. This isn't a bad idea for building a baseline, but it's a terrible idea for a stress test or risk management.

Convenient selection also dogs calculations of returns. The start and finish points are very important, and salespeople will tend toward timeframes that make things look good.

The absence of quake modeling I referenced above arises from people not knowing how to model things, and choosing to ignore those things they don't understand. This is incredibly dangerous, and for a great many reasons. These are not six sigma events, where a variable in your model goes wildly out of its historic distribution. These are cases where very significant things happen which weren't even in your model. Remarkably often, either the analyst knows such things are possible and doesn't know how to model them, or they recognize the problem as soon as someone mentions it and still don't know what to do about it.

In my quake model, very different risks were present at each stage of the project, and they had very different interactions. A quake during early phase construction, in a booming construction environment, with high interest rates would have been immensely painful. A similar quake after completion during a construction recession with low interest rates wouldn't be bad at all. It was not so important that we assigned exactly the right probabilities to these events. We built an approximate model and looked at the problem cases from the simulation runs. Under what circumstances did they run out of money? What could we do about it, either in the financing plan, the construction plan, or somewhere else?

[John Emerson]

Math, not necessarily meth. Except Erdos.

[kthomas]

"Math, not necessarily meth. Except Erdos." LOL

Nice follow-up there.

[kthomas]

six sigma....grrrr.

[wcw]

Hey, nice line. Good old Erdos. I had some Cal math profs with reasonably low Erdos numbers.

In re Black-Scholes: see Rubinstein 2006 for a defense. The argument is that it isn't the formula for which they're known, but the derivation.

In re modeling, these are sins I have seen everywhere I have ever seen anything modeled. Yes, most of my career has been in the investing business, but I've done a bunch of wacky contract work that has ended up in interesting places, and I can assure you that bad math and worse modeling underlies more things in the real world than you would care to know.

In re MFEs, like MBAs and CFAs, they are not about the content, but about getting the good-housekeeping seal. You should hear my father (a chemical engineer who got his MBA to move to the coroporate side) dismiss the MBA curriculum. Heck, the CFA curriculum (yeah, I got mine) is shockingly simple. I believe this practice of semiuseless designations goes back to Confucian exams.

[paine]

diffy q's do wall street ???

baaahhhhd

hey you can say bad things
about the use of colt .45s

[Richard H. Serlin]

wml says...

Do you have any evidence that, say, even academic economics has had any positive impact on real aggregate wealth in the last 30 years?

When I was an undergrad, I once asked my advisor "Has anyone ever done a good study to quantify the benefit of academic economics to the economy?" His response was "I don't know of any, and we probably wouldn't want to, as I suspect the result would come back negative."

Smart economic policy is absolutely crucial to national wealth. The Soviet Union had similar intellectual talent and culture to the U.S., but was far poorer due to it's economic policies. Argentina was as wealthy as the U.S. 100 years ago, but today is far poorer due very largely to its economic policies.

Academic economics not only advances understanding of economics, it trains good (or at least far better than otherwise) advisers, executives, and voters throughout the economy to make better decisions. The economy clearly has problems, but it would be immensely worse if there were no economic training at all. And soon we'll have an administration that believes in, and highly values, thinking, intelligence, and learning, then you'll really see the difference that good economics learning can make.

[Alex Tolley]

Dave: "It is quite amazing that Z/s transforms, fourier analysis (not to mention wavelet theory or more complex tools), basics of nyquist sampling theory (aliasing, anyone?) and the rest could be omitted from these programmes. Never mind complex analysis, non-linear and adaptive methods, Lyapunov stability criteria, Kalman and information filter theory, solving AREs, H-inf and many of the more complex control theoretic techniques that could (and have) found application beyond the physical sciences and engineering in the financial world."

Easy to argue that more expertise is needed. But in what financial modeling situations would these techniques be used?

Donna: "Lots of my physics friends were hired to work on Wall Street as the business majors "couldn't handle the math". I think that was the first sign to me that we were in big, big trouble."

Well I had a math friend who went to Wall Street to find arbitrage opportunities. He did fine. He job was to use math and computers to find trading opportunities, not to construct financial instruments. Seems reasonable to me.

In my limited experience, Organic George is spot on with his comment: "All the math is crap and all the deal makers are street hustlers with expensive suits.". The math they used was basically a stage prop to make their pitch sound as if it had gravitas. Of course if you you are selling to people who have no math skills...

[CBBB]

This conversation seems to have given some people the opportunity to take financial engineering bashing as an excuse to do some general mathematics hating. Since I majored in math I feel I should come to its defense.

Of course using math isn't appropriate in all fields but it can - when used properly - help make arguments much more precise and clear. The problem always arises from the fact that the math must be applied on some kind of model and the model itself might be unrealistic.

Save_the_Rustbelt was talking about Operations Research above, and I think he's wrong. The problems that OR tends to be used on are very specific; how can we lay out this fiber optic network to minimize the cost, how should we schedule this sports tournament to avoid conflicts, how can this major international shipping company best route its airplanes?

STR says that you could just use "good judgment" - but many problems like this become VERY BIG, you have a large number of sports teams, you have a large number of houses that need to be connected, a large fleet of aircraft. One simply can't use "good judgment" to come up with good solutions to very big problems with lots of inputs. You need a methodology - you need math. Since many of these OR type problems are very specific, not subject to much outside/hard-to-quantify influence, you can probably come up with an acceptable model that the optimization algorithms can go to work on.

The same thing is even more true for a field like physics, where the phenomenon trying to be modeled is very specific so good models can be developed.

So the problem is then "can a good model be developed?". I know little about finance but it seems reasonable to say that pricing assets is subject to a large amount of hard-to-quantify influence: psychology, greed, irrationality. This means the fin. engineering stuff is probably garbage.

On a final note it seems like these MSc Financial Engineering people are mostly ex-physicist, engineers, or mathy-type economists. Perhaps the banks would have done better to hire actual mathematicians who spend a lot more time working with theorems and hence know the conditions under which the various techniques work.

Well that's my rant.

[Laurent GUERBY]

"The BS equation was first formulated and solved by Casey Sprenkle some ten years before Black's famous 1973 paper in the Journal of Political Economy. Regrettably, it is still politically incorrect to give due credit to someone who made a real contribution to finance."

Black 1973 paper does credit and discuss Sprenkle formula and solution. The 1973 paper "real contribution to finance" was to bring on board the short term risk free rate by an arbitrage argument, which is not about math but finance.

http://nobelprize.org/nobel_prizes/economics/laureates/1997/press.html

(hat tip to Julien :).

[Richard H. Serlin]

Ok, I just read the whole article. He appears to be only talking about quantitative finance masters programs. Clearly there's a problem with trying to teach that in a 2 year masters program. To learn the enormous amount of math well, plus the finance, as you should, takes a 5+ year Ph.D. program and years of learning afterward as a professor or working in an investment firm. If you try to learn all of that in just 2 years you're not going to learn the math well and you're not going to learn the finance well, and it will show on the job.

But at the level of professors who have trained and worked in this area for years, the material is in general, as Paul Krugman said, "really useful stuff". It's lead to important insights and valuable tools, and will lead to more. It just takes a while to understand and apply quantitative finance well, given how advanced it is, and certainly, as expected, mistakes have been made along the way. You can make most great advances without it, but you can make additional ones with it.

[CreditSpectrum]

Sylvain Raynes' follow-up post, "The Answer is the Question, the Question is the Deal" can be viewed on The Spectrum, the official blog for R&R Consulting and the source of this article.

The Spectrum welcomes your thoughts and opinions, and eagerly awaits all of your debate.

[Ann Rutledge]

CBBB, we like your rant.

But, regarding your skepticism about the applicability of mathematics to pricing assets, check out cash flow ABS and (now a bad word) RMBS. Unlike commodities, rates and synthetic credits, these pure cash flow-backed investments are much more amenable to mathematical analysis, so long as the payment instructions are followed. Of course, if people are too willing to rely on the rating instead of doing the analysis, and if the rating agencies don't do the math, then no one can guarantee that the payment instructions are being followed. Or even that the ratings are close to intrinsic value. -AR

[Alberto]

Doesn't Economics have the same problem? Full of Stochastic General Equilibrium Models, representative agent models, etc., aren't these as wrong and misleading as the material tought in Quant finance? Models that predict nothing and explain even less...they are good only for the accademic career.