One year after the beginning of the modern credit crisis, Randall Forsyth , writing in Barrons, provides an fine round-up of what caused this mess to begin with. His theory is that the problems began when statistical modeling of mortgages replaced old-fashioned credit reports when it came to valuing mortgage loans.
“With statistical modeling, no longer was it necessary to collect and verify information about borrowers. Put everybody in the pool and everything would even out, statistically speaking…. From that, Wall Street made the leap to taking those loans and structuring them into securities. Since financiers knew, with statistical certainty, how many loans in the pool would default, they sliced the loans into tranches.”
He blames the advent of modeling for creating the illusion that loans were just another form of security, and he correctly dates the beginning of the problem from “somewhere in the 1970s,” but he seems to miss the critical ingredient that made all of this possible. The final elimination of the dollar’s tie to gold is what unleashed the illusion that we never again had to worry about whether there was enough money. The Fed cast a spell over the financial markets that risk was a thing of the past, having been reduced to one slight factor among many that is used to assess the market price of load/securities. The element of risk itself was considered to be separate from an actual borrower’s ability to pay and became merely an aggregated statistical construct divorced from human action.
He urges us to keep markets functioning as the best means to solve the problem but he seems to miss the critical point–and this is not just his mistake but a near universal issue–that an essential part of free markets is the sound money that emerges from market exchange. The money we use has been distorted and destroyed in order to make possible a fiat money world in which scarcity and risk no longer play a role — or at least this is what we believed. By all means, let markets clear but the core problems will never go away until money resumes its role not as an infinitely available grease for economic expansion but as a scarce and finite good that merely facilitates sound economic development.
Two new books are crucial here. First, there is Hayek’s amazing writing on the business cycle, which links macroeconomic phenomena to the quality of money. Second is George Selgin’s demonstration of the historic capacity of markets alone to make and guard money.
{ 22 comments }
Jeffery,
Good job in connecting this to Nixon taking us off of gold. There seems to be a fear among the economic pundits about mentioning how so much of our current problems grew from this.
The market will always win in the end but government seems to always find innovative ways of hiding the symptoms through market manipulations (monetary included) which simply prolong and worsen the problem increasing the ultimate pain.
People love to hate statistics, but Hayek had the proper perspective in MONETARY THEORY AND THE TRADE CYCLE (beginning on page 35):
“Thus it is not by enriching or by checking theoretical
analysis that economic statistics gain their
real importance. This lies elsewhere. The proper
task of statistics is to give us accurate information
about the events which fall within the province
of theory, and so to enable us not only to connect
two consecutive events as cause and effect, a posteriori,
but to grasp existing conditions completely
enough for forecasts of the future and, eventually,
appropriate action, to become possible. It is
only through this possibility of forecasts of systematic
action that theory gains practical impor-
tance.* A theory might, for instance, enable us to
infer from the comparative movements of certain
prices and quantities an imminent change in the
direction of those movements: but we should have
little use for such a theory if we were unable to
ascertain the actual movements of the phenomena
in question. With regard to certain phenomena
• It should be noted that the idea of forecasting is by no means a
new one, although it is often regarded as such. Every economic
theory, and indeed all theory of whatever sort, aims exclusively at
foretelling the necessary consequences of a given situation, event or
measure. The subject-matter of trade cycle theory being what it is,
it follows that ideally it should result in a collective forecast showing
the total development resulting from a given situation under given
conditions. In practice, such forecasts are attempted in too unconditional
a form, and on an inadmissibly over-simplified basis; and,
consequently, the very possibility of scientific judgments about future
economic trends to-day appears problematical, and cautious thinkers
are apt to disparage any attempt at such forecasting. In contrast to
this view, we have to emphasize very strongly that statistical research
in this field is meaningless except in so far as it leads to a forecast,
however much that forecast may have to be hedged about with
qualifications. In particular any measures aimed at alleviating the
Trade Cycle (and necessarily based on statistical research) must be
conceived in the light of certain assumptions as to the future trend to
be expected in the absence of such measures. Statistical research,
therefore, serves only to furnish the bases for the utilization of existing
theoretical principles. Dr. O. Morgenstern’s recent categorical denial
(Wirtschqftsprognose, Untersuchung ihrer Voranssetzungen und MGglichkeiten,
Wien, 1928) of the possibility of forecasting seems to be
due only to the fact that he demands more from forecasting than
is justifiable. Even the ability to forecast a hailstorm would not
be useless – but, on the contrary, very valuable – if the latter could
thereupon be averted by firing rockets at the clouds
having an important bearing on the Trade Cycle,
our position is a peculiar one. We can deduce
from general insight how the majority of people
will behave under certain conditions; but the
actual behaviour of these masses at a given
moment, and therefore the conditions to which our
theoretical conclusions must be applied, can only
be ascertained by the use of complicated statistical
methods. This is especially true when a phenomenon
is influenced by a number of partly known
circumstances, such as, e.g., seasonal changes.
Here very complicated statistical investigations
are needed to ascertain whether these circumstances
whose presence indicates the applicability
of theoretical conclusions were in fact operative.
Often statistical analysis may detect phenomena
which have, as yet, no theoretical explanation,
and which therefore necessitate either an extension
of theoretical speculation or a search for new
determining conditions. But the explanation of
the phenomena thus detected, if it is to serve as a
basis for forecasts of the future, must in every case
utilize other methods than statistically observed
regularities; and the observed phenomena will
have to be deduced from the theoretical system,
independently of empirical detection.â€
The mortgage crisis didn’t result from the use of statistics, but as Hayek wrote in his footnote “In practice, such forecasts are attempted in too unconditional a form, and on an inadmissibly over-simplified basis.†In other words, statistical techniques are necessary, but the soundness of models depends upon the insight of the modeler into economic theory. If the modelers were following mainstream econ, they would have no way of including in their model the effects of the business cycle caused by the monetary pumping of the feds.
Two things contributed to the crisis. 1) Poor lending practices by mortgage companies made loans to people who were bad credit risks, and 2) Fed monetary policy caused malinvestments in the housing market. It seems to me that the Feds contributed more to the depth of the problem than did poor lending practices because the Feds ignited the housing boom with low interest rates. Rapidly rising house prices (the price inflation that naturally accompanies a loose monetary policy) encouraged people to buy houses they couldn’t afford because the value of the house was rising. In turn, the rapid rise in housing prices is what made the packaged securities of loans so attractive, not the interest rates on the mortgages, which were low. Many people who were barely making their payments were pushed over the edge when they lost their jobs due to the slow down in the economy when the Feds quit lowering interest rates. Also, the Feds low interest rates encouraged investment banks to goose their earnings by borrowing heavily to finance the purchases of the packages of loans. Finally, rising house prices encouraged mortgage companies to loosen credit standards because they figured the increase in house prices would cover loses due to foreclosure.
Yes, defaults in sub prime lending have increased, but not enough to explain the huge collapse in the industry. For that you have to look at monetary policy and the Austrian Business Cycle.
I think one reason for going with a statistical method like FICO sores was to comply with Federal Fair Lending Standards. In the past, bankers would lend on the “know your customer rule”. This raised complaints about “red-lining” and discrimination. Banks used to have their own underwriting guidelines and would actually use credit scorecards to evaluate loans. Again, directly or indirectly, this led to a pattern that could be construed as discriminatory.
Going to an external criterion such as FICO removed the discrimination complaint, since FICO scores do not take race, age, gender, zip code as input. (They also do not take into account income or assets). Also S&P and Moody’s are not going to do detailed underwriting on every loan in a pool to be securitized. They just want an unbiased 3rd party criterion.
There is nothing wrong per se in using a statistical approach. Credit risk can be abated by either underwriting or diversification. Even with loans that have been carefully underwritten, there will still be a residual number of bad loans due to unforeseen circumstances, such as job loss, illness, accidents, and divorce. The problem arises when the actual number of defaults greatly exceed what was predicted – what Nassim Taleb refers to as a Black Swan.
Mike D “The problem arises when the actual number of defaults greatly exceed what was predicted – what Nassim Taleb refers to as a Black Swan.”
I’ve read Taleb and he makes some interesting points, but I think he is too quick to label events as black swans, by which he means random events. Anything mainstream econ can’t explain it calls random. But what seems random to mainstream economists finds a good explanation in Austrian econ. I’m fairly confident that had rating agencies built their statistical models on Austrian theory instead of no theory or on mainstream macro theory, they would have been able to predict the meltdown. After all, Austrian econ would have made the qualitative prediction that the Fed’s printing presses would cause malinvestment, and Austrian economists could see that the frenzie in the housing market was due to that printing. All a good statistical model would add to that insight would be come numbers and timing.
Bernanke: “A second critical development was an even broader credit boom, in which lenders and investors aggressively sought out new opportunities to take credit risk even as market risk premiums contracted,”
This is precisely why the Fed doesn’t understand the simultaneous massive pockets of inflation and deflation, which will possibly result in a path of destruction ultimately in one direction or another depending upon which poison the Fed picks, or which will also result in a path of destruction anyway if the Fed does nothing.
Inflation of the fiat dollar was and continues to be so bad that even high risks at low interest rates were and continue to be preferable to holding dollars as savings. If your money is going to be lit on fire by the government and burned into oblivion, might as well bet on the long shot horses at the racetrack in the meantime. This is not, however, a “malinvestment” proper. Saving dollars has become a guaranteed loss of purchasing power. We may now be on the cusp of real hyperinflation because big ticket items are not purchased through borrowing except with real negative interest rates.
Those high risk investments and RE schemes and scams have blown up, instantly evaporating trillions of dollars of subjective value in the credit and bond markets. This has created the simultaneous massive pocket of deflation. Interest rates are so artificially low, and extremely negative in real terms that all possible private savings are being crowded out of the market.
This is precisely why a mania of people chasing what they were scammed into believing was a limited housing supply market would happen. To this day the market for houses is still flooded with “investors” and “flippers”. These are by definition people who do not want those houses as end subjectively valued goods, but literally as “means of exchange”. Houses are a substitute big unit of money, a synthetic pile of gold, or a synthetic dump truck of dollars. And to the extent money is a pure “means of exchange” (by definition nobody wants it as an end good), it is incredibly volatile in value, is an inherent bubble. It’s not valued for what you personally get from it, but from you believe greater fool suckers will offer you for it. That’s a classic Mania. And whoops, that line of B.S. which was fed by national realtors about land and houses being scarce was smashed by market reality, which is a very good thing as it will get massive quantities of investors and flippers out of the way of people looking to live in a house without artificial bidding competition.
“Give me a dollar today, and I’ll give you 90 cents back next year.”
“Uh, no thanks.”
Moving rates higher will blow up more mortgage bonds and credit (putting a decent default/walk away dent into the highest rated prime mortgages), as many of those instruments are adjustable to changing interest rates, and/or related to a suddenly highly volatile underlying price of housing asset. Unchecked inflation is simultaneously destroying the value of the fiat dollar, of pretend “money”, and blowing up the commodities markets. But the Fed central bank has no problem lending one dollar today for 90 cents in a year because it can just keep on printing as many dollars as it wants. It’ll still gain 90 cents for every dime it loses from negative real interest rates from each additional dollar it prints.
Hayek: “the Trade Cycle”
Lol. All of the most esteemed economists of the 20th century failed to understand trade. The Austrians weren’t excepted. Mises just misunderstood trade the least. There are no cycles in trade. They literally believed that the value of things exchanged was equal, at least when money was added to the “equation”.
The careful reader should be aware that people talking about economic “cycles” are like people talking about solar “cycles”, aka the theory that the sun revolves around the earth. Economics is not mathematics, and nor is it physics either.
How does such a tangled delusion of cycles evolve? It’s been ingrained in society and culture for thousands of years, from day and night, from the astrological calendar, to seasons, to ritual stages through life to death. People just accept that it’s as real as Zeus and Apollo.
Would Mises have ever called Human Action “cyclical”? Humans only act to go to a state of lesser dissatisfaction from a state of greater dissatisfaction. Would it then make sense to say, “this must be cyclically balanced by humans only acting in the future to go to a state of greater dissatisfaction from a state of lesser dissatisfaction”. What an anthropomorphic deterministic lolMyth!
We need a new term other than “business cycle”, such as “Counterfeit Money Flood Escape Disaster”, because “business cycle” is being used and representing phantom phenomena explanations, all over the place. There’s nothing at all a priori or a posteriori cyclical.
“Trade Cycle” is worse than “Business Cycle” is worse than “Boom Bust Cycle” is worse than “Production Cycle” is worse than “Cycle Cycle” is worse than “Cycle Cycle Cycle”…
It is an epistemological fact that there is not any more business calculation errors or “malinvestments” from changes in money supply than there is from any change in supply (or demands) of any other good, no matter whether they are shorter or longer term production processes. There is no boom, but only a hot potato musical chairs race to get rid of fiat money as quickly as possible, to trade it as fast as one can for other goods before its loss in value is generally recognized.
That people are reluctant to want or to save fiat money is a synthetic distortion monkey wrench thrown in the gears of economic trading activity. But that is a problem fundamentally with fiat money itself, not any particular quantity supply of fiat money. That is a phenomenon that could, would, and does occur, even if the Fed does nothing to manipulate the money supply. The problem is that the world is forced to use relatively infinite grains of sand as its money.
It is absolutely appalling how many awful errors continue to flourish in the field of economics due to a faulty understanding and belief in monetary theory.
fundamentalist
I see your point. When Taleb refers to Black Swan’s he is referring to the occurrence of events that have never been seen before. Jimmy Carter’s 21% interest rates would qualify as a Black Swan. $140 oil is a Black Swan. However, 5.25% interest rates in 2007 would not qualify. You are correct that Austrians predicted problems in 2002 when the Fed dropped rates below core inflation (whatever that might be). So the sub-prime meltdown was not only predicable but also predicted. However, the Fannie and Freddie meltdown occurred as a result of problems with prime loans. The error, again easily predictable with Austrian Economics, is to assume that historic default rates are good predictors of future default rates. Since future default rates will be determined by economic events that have not yet occurred that are influenced by human action, which is not known in advance, then the possibility that future default rates grossly exceed current or historic default rates can not be dismissed.
Your first two paragraphs were spot on and made very good points.
Unfortunately the third paragraph wherein an attempt is made to tie “elimination of the dollar’s tie to gold” to failure to verify loan application data is utter shite.
A shame you ruined one perfectly good argument by tainting it with a flawed second argument. Perhaps next time you might separate the two. At least the odds of providing a worthwhile argument would double.
“With statistical modeling, no longer was it necessary to collect and verify information about borrowers. Put everybody in the pool and everything would even out, statistically speaking…. From that, Wall Street made the leap to taking those loans and structuring them into securities.”
The problem isn’t statistical modelling, that’s nothing but good old technological change. The problem lies with the monopolies who pioneered MBS and CDOs and later introduced statistical modelling; Fannie Mae and Freddie Mac.
Because F&F so dominated mortgage markets by the early 90s, any statistical model they adopted would “shape” the entire market. It’s one thing to have competitive institutions carrying out the process of technological change, but to have two government guaranteed institutions behind the process is a recipe for disaster. No doubt their statistical models were tweaked to accomodate the “houses for all” lobby, so that by the end people with 65% debt to income ratios were getting loans.
There’s never been such a huge string of credit card ABS defaults, which use the same technological innovations as their MBS cousins like FICO, but that’s because there’s no F&F equivalent for credit cards.
to fundamentalist:
i think taleb’s point is that markets can’t be perfectly accurately modelled on the back of normal curves. most of the complex derivative models presuppose normal distribution, and most times this is fine.
but in the real world, the probability curve has fat tails, and chance of massive market moves is significantly higher than would be the case were it a gaussian distribution.
besides, most of the derivative models depend on high and continuous liquidity in order to be trimmed dynamically.
warren buffett was right to be very wary of these weapons of mass financial destruction, and has spent years ridding swiss re of them.
fundamentalist:
That was a very appropriate Hayekian selection. But its appropriateness lies not in its description of what’s going on but its exposure of Hayek’s own ambiguity.
He grasps and propounds the Misesian (and, for all of us, somewhat prosaic) placement of theory as being that which permits us to make certain general deductions about a course of events set in motion by a cause. But, even in his denial of the value of statistics, he yet overvalues their use: it is apparent that, in statistical results, he sees a reading of mass phenomena, such “reading” having value for current assessment of the future simply by virtue of their being “mass” phenomena in the first place.
Everyone, in daily life, is prone to the same transference, essentially of volition, from the individuals comprising the mass–to the mass itself. The chief recommendation for making this transference is that it seems to convey a certain useful information, i.e., the approximate moment that the mass resultant may be approaching an unacceptable outcome (for which, it is alleged, countermeasures must be in order). The delusion here is that the volitions and values of individuals are equated to the mid-way causes known to be operative in the events studied by the natural sciences.
In the natural sciences, the fact that we cannot know the ultimate cause or even all of the actual partial causes–causes relatively little problem. We have, so to speak, a “handle” on our ignorance and can subsume it (or them) in a tiny minority of cases to which we can apply the “fudge factor” of statistics and still come up with results that serve the needs. This is an outcome of the fact of regularity that defines our inclusion of the event in the realm of “natural science” in the first place: the assumption that what we don’t know behaves with the same predictable regularity as what we do–at least to a degree falling within the limits we’ve defined and to which the “fudge factors” have been assigned.
But humans don’t behave as though driven by forces of unvarying regularity (except in certain senses, such as that, up to now, all must obtain a certain amount and approximate distribution of “foodstuffs”). Nothing is quite as illustrative of the utter impracticality of drawing conclusions from observation of mass phenomena in human activity than Mises’ observation about “bank runs.” The idea of the “crackup boom” could be divined at the first appearance of fiduciary media–but it’s not. And at each reiterated injection of additional quantities made evident by rises in the level of prices–but it’s not, even as people respond by restricting savings, etc. But, at the point the end approaches, it’s already too late for most.
If statistics had even a fraction of the utility attributed to them–even by Hayek–the outcomes of “wars” between soft drink manufacturers and political candidates would become far more certain. Nor would there need be any embarrassed (or broke) as the result of the demise of LTCM (and the inadequacy of the much-revered Black-Scholes number-crunching).
MikeD: “When Taleb refers to Black Swan’s he is referring to the occurrence of events that have never been seen before. Jimmy Carter’s 21% interest rates would qualify as a Black Swan. $140 oil is a Black Swan.”
That’s true, and I don’t think anyone could predict exactly those amounts, but I think with the Austrian theory behind the modeling they would have come very close to predicting those black swans and certainly wouldn’t have been totally surprised as mainstream economists were. Take oil, for example. It’s very sensitive to the value of the dollar. We should have been able to predict the rise in oil prices by the massive inflation of the 90′s. It was masked by the rise in productivity and output during the same period, but that can be modeled, too. As soon as production slowed down because of low prices, the effects of monetary inflation were uncovered.
newson: “i think taleb’s point is that markets can’t be perfectly accurately modelled on the back of normal curves.”
You’re right, but I never understood his fixation on the normal distribution. In the first place, statisticians have found that statistical techniques, such as regression, are very robust to large deviations of the data from the normal distribution, so his concern is misplaced unless you’re doing simple probability problems. But if you’re still concerned about the normal distribution, there are many techniques for adjusting the distribution of your data to make it more normally distributed. I have used them quite a bit in the past and they work well. But the best modeling methods are those called machine learning programs, such as neural networks. These methods don’t care what the distribution is. Support vector machines, decision trees, neural networks and several other techniques that combine traditional statistics with iterative search algorithms can run circles around traditional statistical methods and almost always outperform them in forecasting. And the distribution of the data is totally unimportant. I don’t want to make these methods sound too good, but my point is that Taleb is way behind technology if his concern is with real life fitting the normal distribution. It doesn’t matter.
By the way, if anyone wants to learn to use some of these new techniques, there is a free data mining program that is absolutely amazing called RapidMiner. It’s available at RapidMiner.com.
to gene berman:
the point is that all option-pricing models all rely on the normal distribution price movements (as you say, probability problems).
most often the normal distribution is ok, but the occurrence of exceptional market moves doesn’t follow a normal distribution (in real life the massive moves occur far more often, hence the fat tails).
besides, even if the normal curve were correct, the presumption of constant liquidity is fatal. market liquidity dries up just when the model requires the position to be rebalanced.
derivatives markets would never have assumed their current size without the massive volatility contribution from the central banks’ fiat money, but i think it’s reasonable to assume that options (and extended family) will always suffer from this problem of modelling something that is not completely knowable. all models require a volatility input; guess that accurately and you’re going to be seriously rich.
oops, cancel berman and make that fundamentalist.
to mike d:
actually $140 oil isn’t a black swan, what happened to bear stearns is a black swan. it’s not the absolute price, nor the price-trend, but the volatility that counts. some of the moves on the sub-prime lenders and insurers paper have standard deviation moves right off the chart.
ltcm qualifies as a black swan, as gene berman implies.
tndal, what is the consequence of not being adequately discriminatory in giving out loans? Potential money loss. What happens when a CB can bail out banks that give out bad loans with credit it can create? Well, a moral hazard. Hence the gold standard allusion (which is 100% reserves banking), under which such credit expansion would be much more difficult. Severing the link of the dollar to gold was just the final step in a long series of events that allowed the Fed to have a (theoretically) unlimited money supply. It helps to take these things in context…
gene berman: “If statistics had even a fraction of the utility attributed to them–even by Hayek–the outcomes of “wars” between soft drink manufacturers and political candidates would become far more certain. Nor would there need be any embarrassed (or broke) as the result of the demise of LTCM (and the inadequacy of the much-revered Black-Scholes number-crunching).”
You should reach “Super Crunchers.” It’s a good description of where statistical and machine learning techniques have improved processes, forecasts and businesses. I work in statistics and constantly deal with two extremes (there seems to be very few people who are ambivalent about statistics): one extreme thinks statistics can accomplish miracles; the other thinks all statistical techniques are worthless. The truth is in the middle.
I don’t understand what in Hayek’s statement offended you. He clearly limits the role of statistics to the practical application of theory and gives it no role in developing theory. And I don’t see that he makes any exagerrated claims about statistics. He only thinks that they are necessary for business and applied economics. He clearly acknowledges that people can be bad at using statistics when he writes: “In practice, such forecasts are attempted in too unconditional a form, and on an inadmissibly over-simplified basis; and, consequently, the very possibility of scientific judgments about future
economic trends to-day appears problematical,…”
But you seem to fall into the group he warns about when he writes: “… and cautious thinkers are apt to disparage any attempt at such forecasting.”
newson: “the point is that all option-pricing models all rely on the normal distribution price movements (as you say, probability problems).”
I don’t understand why people still use the various versions of Black-Scholes. It was great in its day, I’m sure, but so many advances in statistics and machine learning have come along since then. And from what I remember of Black-Scholes, it only uses the normal distribution to figure the probability of an event. But there are many other types of distributions that you can use to do the same thing. The poisson distribution is good for figuring the probability of rare events. I know of a data mining software maker that has made millions using a statistical technique developed in medical research to forecast early payment of mortgages.
No statistical or machine learning technique if perfect. In fact they’re quite bad in terms of accuracy, but they can still make you a lot of money if you know how to use them. I had a friend who was good at data mining and he created a fairly simple statistical model to forecast horse races. It was only about 60% accurate, but that was more than enough to make him a lot of money. If He didn’t bet all of his money at once, but spread it out over say a dozen bets. He won 60% of the time and lost 40% of the time, which meant he made good money.
I worked with a stock broker one time who was developing a stock forecasting model. He claimed his model needed to be just 51% accurate in order for him to make a lot of money. The key is not so much accuracy in forecasting as good money management.
fundamentalist/newson:
Actually, Black-Scholes uses a lognormal distribution. In the stock market it assumes that the logarithm of stock price returns is normally distributed. (This is needed because stock prices can not be negative, and also because an investor holding 200 shares of $50 stock expects the same return as 100 shares of $100 stock). The reason people still use Black-Scholes is that it forms the basis for no-arbitrage pricing theory.
Also, fundamentalist, I think you mean Pareto distribution when you refer to pricing extremes – Poisson is uses in queuing theory, for instance for forecasting how many customers arrive in a given time interval.
to mike d:
yes, quite right. black-scholes and other option-pricing-models work ok for most price movements with liquid underlying securities, but extreme moves are much more common than the lognormal distribution would suggest. some of the us banking and finance stocks recently have experienced standard deviation moves far in excess than predicted by lognormal distribution (i seem to recall moves of 20 s.d. in the latest washout).
people who made very good profits for many years writing naked equity puts found themselves suddenly wiped out by the 1987 crash. ditto for ltcm, which started out making decent money before going dramatically sour.
this is a bit like fundamentalist’s broker pal and his black-box winner, they can work for a while (trends obviously continue!) but fail, often spectacularly, when the trend turns or volatility spikes.
I think one of Taleb’s main points is to stay diversified, or at least don’t bet all of your money on one deal. No statistical or machine learning model will be perfect, so when it fails you want to make sure your losses don’t sink you and you have enough back up to get back in. The book “Super Crunchers” does a good job of showing the benefits of statistical modeling without going overboard and making it seem like a silver bullet.
MikeD: “When Taleb refers to Black Swan’s he is referring to the occurrence of events that have never been seen before. Jimmy Carter’s 21% interest rates would qualify as a Black Swan. $140 oil is a Black Swan.”
That’s not really correct. What Taleb refers to as a Black Swan is not an event so improbable that nobody could predict it. A Black Swan is an event whose probability is not properly predicted using a normal distribution – some variation of which is used for almost all financial models. A normal distribution does predict the occurance of Black Swans, what it doesn’t predict correctly is the odds of occurance.
Taleb’s criticisms of statistics goes beyond this category but to a fundamental misunderstanding of what statistics really tell you and don’t tell you. Statisitics cannot discern cause and effect period. The best that statistics can ever do is identify events that have occurred at the same time in the past and the degree to which they have occurred at the same time in the past. The accuracy of predictions made with statistics are based on the accuracy of the theory used to identify the cause and effect relationship and the degree to which past events will be repeated in the future. Both are somewhat subjective and only knowable to the degree that the future itself can be predicted.
“Risk Management” as it has been practiced on Wall Street glosses over these issues and does not deal with them adequately. 99.9997% of the time that may not matter. However, given the sheer volume of transactions that allows for a lot of unpredictable outcomes of significant magnitde.
DS
Here is Nassim Taleb’s own definition.
author interview
The Black Swan is an intriguing title — can you give us an overview of what a black swan looks like?
The Black Swan is about these unexpected events that end up controlling our lives, the world, the economy, history, everything. Before they happen we consider them close to impossible; after they happen we think that they were predictable and partake of a larger scheme. They are rare, but their impact is monstrous. My main problem is: Why we don’t know that these events play such a large role. Why are we blind to them?
What is your favorite example of a black swan?
I had to write one for the book, with the fictional character of Yevgenia Krasnova. She starts as an unsuccessful writer, becomes famous and creates a new style –later people found her so obviously talented and believed that the emergence of such style was unavoidable. Most successful stories in the arts and letters are Black Swans. Outside the arts, my favorite one is the emergence of the computer and the internet. Also, almost all drug discoveries are Black Swans. The existence of the universe is a Black Swan…
Comments on this entry are closed.