Ted Harlan emails about this NYT article about Fannie Mae (and about this discussion of it on the Mises.org blog). The NYT article is quite good in pointing out the extent to which FM is a financial black hole in which business-as-usual has consisted of sweeping interest-rate risk under the rug and hoping for the best. This isn’t news, but the enormous extent of the risk, illuminated in the NYT article by an ex-FM employee, and earlier over a long period by the WSJ’s editorial page, has only recently become a mainstream issue.
Essentially, what Fannie Mae is doing, from a risk standpoint, is not much different from what Long-Term Capital Management did (though FM probably uses less leverage). In both these and numerous other cases following a similar pattern, an institution develops sophisticated mathematical models to exploit apparent relationships between financial instruments. The success of such models tends to depend critically on the accuracy of the underlying statistical assumptions about markets. The people who develop the models think they know what market normality — in both the conventional and statistical senses of the word — means. They are often wrong, though their main error is in designing trading systems that depend too much on the accuracy of these assumptions. In plain language: they are fitting their models to a limited data history without taking adequate account of unlikely events, the financial equivalent of hurricanes. Things go well for a while, maybe for years, much money is made, and then the unexpected happens and the firm loses in a brief period more than it made during its entire history. (Better traders design more accurate models, or design models whose success is not so closely tied to the accuracy of their statistical assumptions about markets, or both.)
The recent huge break in the U.S. government bond market (see chart below) may reflect in part a discounting of the risk that Fannie Mae will default on significant obligations and have to be bailed out by Uncle Sam. (An alternative hypothesis, the possibility that the interest-rate increase represents mainly a return of inflationary expectations, is apparently belied by the fact that the price of gold in U.S. dollars has been stable. How much of the bond market’s fall was due to new economic-growth expectations, how much to the fact that the bull market in bonds was overdone, and how much to Fannie Mae’s problems, is the question. Whatever the answer, the FM situation didn’t help.)
Fannie Mae and other big holders of mortgages and mortgage-backed securities chronically underestimate the odds of a big move in interest rates that could devastate the value of their portfolios, said Nassim Nicholas Taleb, a hedge fund manager and the author of two books about risk and the financial markets. In general, he said, Fannie Mae and other companies rely too much on computer models that do not account for rare but devastating breaks in markets.
“The fact that they have not blown up in the past doesn’t mean that they’re not going to blow up in the future,” said Mr. Taleb, who is also an adjunct professor of mathematics at the Courant Institute of New York University. “The math is bogus.”
Taleb is mainly right. Perfect storms do occur every once in a while. When it happens, finely tuned trading models that are optimized to exploit marginal relationships (and which almost by definition don’t predict extreme outliers) tend to fail catastrophically. Fannie Mae may not fail, but the risk is there and we shouldn’t be complacent about it.
UPDATE: I should have linked to my June 25 post on the bond market (press F11 key if post doesn’t display correctly). I had the market’s direction right but, in hindsight, was excessively concerned about inflation. The fact that interest rates have gone up while gold hasn’t rallied much from the US$350/oz level suggests that inflation is not a significant concern for the markets now.
“Fannie Mae and other companies rely too much on computer models that do not account for rare but devastating breaks in markets.”
There’s another aspect here, which may be relevant. I don’t know the exact legal structure of Fannie Mae, but since it has some kind of Government guarantee, I’m sure that the managers face some kinds of restrictions on how they can run the thing. They may be required to use models which are “industry standard.” In other words, the incentives facing the management are not to reduce the chance of catastrophe by innovative or thoughtful modeling. Rather, their incentive is to be able to say, post-disaster, that they did what was considered normal and safe and hence non-negligent. Others will know better than I do whether this factor is in play.
The industry standard in this case goes by some buzz phrase like “value at risk.” The latest and greatest software calculates the value of financial portfolio X based on variables A through n. The model usually works great if next Tuesday is like last Thursday. But in the long run the reliability of the model is critically dependent on assumptions about the shape of the statistical distribution of market events. This isn’t a problem if you are selling life insurance, because the distribution of death rates by age for a given population is well understood and unlikely to change. But Fanny Mae, which functions like an insurance underwriter in the interest-rate markets, can easily be a net loser in the long run if its traders underestimate, even by a little, the number of outlier moves in interest rates. And that’s easy to do because distributions of price/yield moves in financial markets are imperfectly understood. Taleb’s central point is that your model isn’t valid if you underestimate the outliers. It doesn’t matter if you were making money for years: if you give it all back, and more, during a brief market crisis (see: Lloyd’s of London), it means you were probably doing something wrong all along.
Fannie Mae may be in this situation, so it’s not enough for them to say, “Trust us because we’ve been making money.” They deserve the closest scrutiny from hard-nosed auditors who will treat with skepticism assertions of expertise by FM’s quant modelers.
The term “financial engineering” has come into wide use but is misleading. Relationships between financial instruments are nowhere near as well understood as are physical relationships between construction materials in real engineering. Calling it “engineering” no more makes it precise and predictable than calling psychology “human engineering” makes interpersonal relationships precise and predictable. Illusions of precision (economist Reuven Brenner’s excellent term) merely add risk.
Lex,
Fannie’s guarantee is not codified. In other words, Congress has never said that they would bail owners of Fannie paper out, should Fannie default. So people usually refer to it as an “implicit guarantee” or a “moral obligation”. This explains, in part, why FNMA paper trades cheaper (or richer, in terms of yield) than Treasury paper– though I should note that individual investors get state tax exemption on Treasuries that they do not get on FNMA or FHLMC paper. Additionally, for what it’s worth, I believe that TVA and the FHLB issue paper guaranteed by the “Full Faith and Credit” of the government.
Jonathan,
I think you mischaracterize Value At Risx.
The theory behind VAR is that it incorporates “fat tails” and shock-type events. But the structure of a VAR estimate is, “There is an X% probability that your portfolio will lose no more than Y dollars over the next Z days.”
VAR doesn’t pretend to tell you what happens in the events that are 100 minus X– or what hapens when you get the one in twenty or the one in a hundred days…
Ted, you may be correct that I mischaracterized “value at risk” analysis. However, I think that my points about risk itself are valid. If your trading methodology is such that a once-in-10-years market shock can put you out of business, you are doing something fundamentally wrong, in my opinion. The risk of what seem to be highly unlikely events can’t be ignored, because if there’s anything that trading teaches it’s that extreme events happen much more frequently in financial markets than ordinary experience tends to predict. As long as such events, the “fat tails” of the returns distribution, are incompletely understood, it is prudent not to trade in a way that makes perfect risk-analysis into a do-or-die proposition. Yet that is how Fannie Mae appears to be operating. The issue is more how they trade rather than how they analyze risk. (And “extreme events” encompasses not just bond-price moves but also such events as unexpected widescale mortgage prepayments.)
Jonathan– I agree with you completely.
By the way, technically there is no government guarantee. These corporations are public corporations with shares on the NYSE. However, they have a line of credit at the Treasury. This, and their role and size, imply that they are effectively too big to fail. A standard case of moral hazard. While the odd Senator has worried about their exposure and risk management, little is done by either side since your political opponents will accuse you of messing with American home ownership. And both Fannie and Freddie spend huge money on campaign contributions and lobbying (I recall they were among the biggest donors to both parties in the country).
Which limits government meddling. But when and if we get to a point where it must happen, the consequences could be rather costly.
Sure. Fannie Mae is very much a political organization, which explains the choice of an ex-Clintonite political operator, rather than a financial person, as CEO. This fact alone is enough to predict moral-hazard problems. I don’t know what should be done about it, but it’s an especially dangerous situation because the organization uses its money irresponsibly, to buy political hedges, when it should be cutting financial risk. I don’t think full privatization would change anything as long as politicians see homeowners as a major constituency. Maybe the best that can be hoped for, at least in the short and medium term, is increased public scrutiny.
Just a note on TVA securities. They are not backed by the full faith and credit of the US Govt. TVA securites are guaranteed first pledge of payment (are backed by) the TVA power system. The AAA rating on TVA bonds comes from an implicit government guarantee and TVA’s monopoly market structure.
Unlike the GSE’s however, TVA is wholly-owned by the US Govt and it appears on the federal budget. Hope this helps…