Theory versus Statistics, Financial Economics edition

Uh-oh. Here comes an economist, to tell finance what’s up. The very smart and talented Barry Eichengreen:

We thought that financial institutions and markets had come to be self-regulating—that investors could be left largely if not wholly to their own devices. Above all we thought that we had learned how to prevent the kind of financial calamity that struck the world in 1929…

Soon after, business schools jumped to supply graduates to write those reports. Value at Risk, as that number and the process for calculating it came to be known, quickly gained a place in the business-school curriculum. The desire for up-to-date information on the risks of doing business was admirable. Less admirable was the belief that those risks could be reduced to a single number which could then be estimated on the basis of a set of mathematical equations fitted to a few data points…

Because it was not that economic theory had nothing to say about the kinds of structural weaknesses and conflicts of interest that paved the way to our current catastrophe. In fact, large swaths of modern economic theory focus squarely on the kind of generic problems that created our current mess…What got us into this mess, in other words, were not the limits of scholarly imagination. It was not the failure or inability of economists to model conflicts of interest, incentives to take excessive risk and information problems that can give rise to bubbles, panics and crises…

Maybe so. But amid the pervading sense of gloom and doom, there is at least one reason for hope. The last ten years have seen a quiet revolution in the practice of economics. For years theorists held the intellectual high ground…Now every graduate student has a laptop computer with more memory than that decades-old university computing center. And she knows what to do with it…Their next step, of course, is to download securities prices from Bloomberg and see how blue skies and rain affect the behavior of financial markets. Finding that stock markets are more likely to rise on sunny days is not exactly reassuring for believers in the efficient-markets hypothesis….

The top young economists are, increasingly, empirically oriented. They are concerned not with theoretical flights of fancy but with the facts on the ground. To the extent that their work is rooted concretely in observation of the real world, it is less likely to sway with the latest fad and fashion. Or so one hopes.

It’s a good piece, worth your time. He draws an line between statistical and theoretical economics and finance. He also believes the future of economics and finance belongs to the empiricists, doing statistics, not theory, and that this might mitigate future calamities. I think this is wrong for a number of reasons.

1) Value at Risk is a statistical concept, not theoretical. It’s a confidence interval. You take the historical data, project it into the future like any other time series regression, and report back the interval. The 95% (or 99%, or 99.9%) interval is your value-at-risk. There is no theory behind it, in the sense that he means it. It is one of the longer running empirical projects in current quant finance.

2) Finance has been leading the way in empirics and statistics since the late 1980s. Finance in particular with massive data sets that one needs to worry about data quality issues, things like survivorship bias, and over analysis, things like data mining, has been leading the economic sub-field pack with its intensity since the 1990s, with labor economics running side-by-side with the bells and whistles. The options pricing, and crazy math stuff in the academy was mostly done by the late 1970s.

In fact, if I grab one of the more mathy, but quant standard, financial economics textbooks off my shelf, Duffie’s Dynamic Asset Pricing Theory, I get this from the preface:

To someone who came out of graduate school in the mid-eighties, the decade spanning roughly 1969-1979 seems like a golden age of dynamic asset pricing theory…[history of the 1970s]…Theoretical developments in the period since 1979, with relatively few exceptions, have been a mopping-up operations. (p.2-3)

quant finance, like hip-hop, better in the 70s

You can almost here the lament of this quant that the real math theory has been dead since 1980, and that it has all been applied and statistics ever since. It’s like Fischer Black was Kool Herc and Myron Scholes was Afrika Bambaataa, and they’d all go plug in their computers into lamp posts and do martingale representations in the streets and at house parties. And, of course, it was all ruined in 1979 when it went commercial.

2.a) And looking at finance in the 1990s, all the major projects are one where statistics leads the theory. The Fama French Three Factor Model, from 1992, was done with very little theory in the sense mentioned here. There were abnormal returns in the cross sectional data of stock returns, and those returns must be attributable to risk factors unidentified. Or is it that people are impatient, or have poor behavioral quirks? We don’t know – we’ll need to do some theory arguments later to figure it out. And that’s from Eugene Fama, the father of Efficient Markets Theory. This is a case where the statistics are leading the theory, and it continues through the whole wave of abnormal pricing research that has been going on (momentum, earnings-diffusion, etc.).

Also the paper he mentioned about weather returns, Good Day Sunshine, was published in the leading Journal of Finance in 2003. It didn’t stop the financial crisis, nor did it impact people’s thoughts about financial markets working. So a call for “more statistics, more empirical work” doesn’t move us an inch further in the stopping financial crisis, or re-thinking markets, game as far as I see the field.

3) Coming to finance and economics with no background except math and modeling (sadly not the male kind), when I first heard the justification for Efficient Markets Hypothesis I was so surprised I laughed and was confused. The theory is duck-taped together because sophisticated arbitrageurs will simply do the exact opposite of noise traders. Since they are the exact opposite, what is left over is identically, independently distributed, hence efficient markets.

It’s pretty obvious to me that this won’t happen in theory, won’t happen in practice, and hasn’t happened in practice. That leaves a gaping hole in the theory, and it isn’t clear what to fill it with. His arguments about information already exists as the bid-ask spread in the micro-market finance literature. Where can we go for new theory?

4) The best line I have read on the problems currently facing finance and economics comes from an english graduate student, writing about The Wire: “Thus, both [Marlo’s] quasi-aesceticism and his amoral approach to corporate entities: more than any other character, he succeeds because he comes closest to the capitalist ideal of existing within institutional frameworks whose limitations only affect others, ruling structures whose rules only he can ignore.” (his two parts on ‘suction’ are must reads if you are Wire fans.) Replace limitations with the economic idea of ‘externality’ and you got it. Scaling businesses past their market limitations can exist if one can make sure the externalities all don’t come back to your bonus until it is too late.

Somewhere, 5 years ago, in a pig field in Mexico, there’s a risk guy who said “we run the risk of generating a nasty strain of influenza. If that happens our value-at-risk is losing the whole farm.” Management goes “Given how much we are making, that is acceptable.” The risk guy follows “but if it spreads, and even 1,000 people across the country get infected, the costs are…” And Management says “Stop. Not our problem.”

Somewhere, 5 years ago, on Wall Street, there’s a risk guy who said “if we have a nation-wide housing downturn, we are probably going to lose most of the book.” Management goes “given what we are making, that is acceptable.” The risk guy follows “but the costs to our shareholders, our counterparties, the taxpayers if we are judged too big to fail…” And Management says “Stop. Not our problem.”

A general economic model of “money goes uphill, (pig) shit goes downhill” would be great to see, especially in the context of increasing/decreasing costs, scales, and externalities. This is where government has to step in, not in order to handle quirky behavioral cues but to make sure the costs get assigned correctly. One can get very large indeed if someone else is holding the bag for you…

Update: Felix Salmon asks “who’s the Grandmaster Flash of the quant world?” Clearly, Robert Merton is the Grandmaster Flash. In the same way that DJs no longer went back to trying to simply guess on where to drop the needle but instead had to develop a system of cueing, quants no longer used the complicated CAPM derived proof and argument that Black and Scholes used for the Black-Scholes equation, but instead the dynamic replication proof of Merton. It’s easier intuitively, and it makes the MBAs and investors a lot more eager to party with your models.

I think, though this is controversial, that Rubinstein’s Binomial Model is the Bomb Squad of the quant world. One takes the box off the beats and gets deep in, recording specifically for sampling, sampling for just seconds, etc, in the same way one can math it up in a pinpoint matter with the discrete modeling of Black-Scholes.