Bailouts, Burnouts and Non-Linear Innovation

By Kurt Cagle
December 10, 2008 | Comments: 7

Recently, Charlie Rose had a very insightful interview with Nassim Taleb, author of the Black Swan(2006) and noted financial mathematician. Taleb has emerged as one of the more prescient economists in the wake of the collapse on Wall Street, largely on the strength of his argument that too many analysts and economists were trying to "linearize" the market, to attempt to reduce risk by distributing it ever deeper into the economic foundations and then working on the assumptions that economics has entered an era of "new stability" that guaranteed the ability to make money at comparatively little risk.

The slide of the Dow Jones and other financial markets by 35 to 45% in the last couple of months attests to the danger of that philosophy. The idea inherent in Taleb's thesis was brought out nicely by the example of the turkey. Every day for a thousand days, the butcher feeds the turkey, makes sure that the turkey has everything it needs, and after a while, the turkey begins assuming that there is no risk in his life, that he can extrapolate his future investments on the basis of his easy ones of the days before. On the thousand and first day, the day before Thanksgiving, say, the assumption proves rather catastrophically wrong - so catastrophically that it can be said to be very non-linear and disruptive ... especially for the turkey.


Significantly, one of Nassim Taleb's most important mentors was Benoit Mandelbrot, the French economist and mathematician who had been attempting to create a model for cotton futures. As he dug in deeper to the problem, he began to realize that most "random" walks aren't truly random - they actually reflect localized changes that can cause seemingly local linear effects at one scale, and can be self-similar at a larger scale ... but that this self-similarity tended to manifest itself not in overall linear behavior but rather in periods of seemingly linear behavior punctuated by significant disruption.

In Mandelbrot's ground-breaking book The Fractal Geometry of Nature, Mandelbrot laid out a comprehensive mathematical underpinning for the study of "fractional dimensions", or fractals, in which pointed out both the fact that fractal, self-similar systems occur regularly in nature, but also making the emphatic point that fractals are extremely dependent upon initial conditions. He more recently applied these to the markets in his book The (Mis)behavior of Markets.

Are you ready for the future? This conference is a must for anyone interested in where Wall Street meets Web 2.0, raising the level of conversation around technology advances in the investment industry. Money:Tech 2009 is happening February 4-6, in New York City, NY . Register today!

The Danger of Straight Lines

In linear regimes, models about the future tend to be fairly tolerant of initial conditions. If you start out with a situation that is similar to what another "successful" person started out with, then you'll end up in roughly the same place. The degree of uncertainty involved in those initial conditions can be considered one definition of risk. Put another way, if you consider the landscape of "initial conditions" to be a plane, then you could in theory draw a region on this plane (which can be simplified as being a circle) where the size of the radiation is inversely proportional to the risk involved - its a map where you can do the same things and get the same result.

Risk management assumes in general that this radius tends to grow or shrink as a continuous function - that in times of high uncertainty, the radius of the "safe" zone will shrink, but it will do so in a way that is still predictable. However, what Mandelbrot stated (and Taleb has so eloquently expanded upon) is that this is much like rolling a ball on a carpeted stairwell. The ball moves in a very linear fashion with just minor bumps due to the weave of the carpet ... up until the point where it rolls over the edge to the next stair. This transitional point is as much a part of economics as the smooth rolling parts, but the model, in assuming that the world is linear because it was linear, would fail rather spectacular as the ball goes into free-fall.

Nassim Taleb laid this out effectively in his book, describing what he called the platonic fallacy, a fallacy that had three distinct parts:

Narative Fallacy. The tendency that people have to create a story after the facts in order to use that story as a means to determine the cause of those effects. You see this in financial websites, where any movement at all in the stock market is ascribed to one or two events that can be summed up in a ten word title sentence.

Ludic Fallacy. Structured randomness that games (and economic models) use is in fact similar to the unstructured randomness of the real world. The randomness of the world is due to the large number of potential interactions and the unpredictable breadth of actors for any given interaction, and this in turn means that the randomness is usually far higher (has a higher fractal dimension) than the Monte-Carlo methods used for random walks in most economic models. In other words, a model by nature is a simplification, and that simplification may gloss over a major factor that wasn't obvious at the time, and hence skew results dramatically.

Statistical Regression Fallacy. This is the belief that the structure of probability can be derived from a set of data.

Stochastic methods (which encompasses much of probability theory) is built upon the ideas that narratives are a necessary part of understanding the variables involved in a simulation, that a sufficiently sophisticated model is capable of emulating natural randomness and that there are nice orderly envelopes of functions that can in fact properly be used to create such models - they are in fact the very inputs to those models. Yet if these assertions are in fact fallacies, then the whole of financial modelling is in trouble.

Today's World Is Non-Linear

By all indications (and as unpalatable as his conclusion is to many on Wall Street), financial modelling is in trouble, and with it the financial industry as a whole and perhaps most of the modern capitalist system. Probability theory generally works well in the domain where there are comparatively few variables at work, where the speed of transaction is low enough that information moving through the system can generally propagate fast enough for traders to make reasonably well informed decisions, yet slow enough that such traders can react while that regime of stability still holds.

Computers and high speed networks have changed that irrevocably. A significant amount of the world's wealth is currently shuttling back and forth between millions of moderately sized to large computer systems, trading according to preset matrices of rules that were set up in an ad hoc manner to try to cover as many potential tradition variables as possible - these rule sets were, by definition, economic models, employing the same principles disparaged above, and because the transactions occurred at several millions of times the speed of human thought, there was no more way that human being could possible oversee it than they could oversee a major earthquake as it was happening.

Certainly greed and cupidity played a part - too much money was skimmed off the top with each transaction ... the broker, playing the part of the middle man, was able to turn a blind eye to the increasingly radioactive securities that he was peddling because he was making profits without taking much risk - he was paid for selling the risky instruments, not to take on even more risk as a financial institution. Yet because they were making money, the models all assumed that the radius of risk was far larger than it was, and worse, that circle was shrinking far more quickly than nearly anyone were able to predict - because the regime had ceased being linear. Initial conditions mattered ... and this is where things went awry.

Nassim Taleb made an interesting prediction on Charlie Rose. This prediction was that just as people used an insufficiently non-linear model for the rising market, they are similarly making too many assumptions that the process in decline will resemble that of the last major crisis of this type, the period after 1929, when the US (and much of the world) underwent a protracted period of deflation. However, just as this is the build up was partially a "technological" expansion, so too will the deflation be "technological" - where financial information is still moving far faster than human beings can realistically respond to (or even relate to).

Indeed, there's increasing evidence that most of the real "crash" has already happened, and what we are seeing now is the expanding wave as it makes it's way through the economy.

Several years ago, I had an older model Chevy (a Citation, I believe) in which the timing belt snapped. A timing belt is in and of itself a fairly simple device - it is in fact just a belt, like so many others in an engine, but its role is to make sure that the explosions happening in the pistons occur at precisely the wrong time. When the timing belt breaks, the pistons get out of synch, running faster and independently, causing rocker arms to warp and cooling systems to fail. The engine gets hotter and hotter, eventually warping, usually causing the gaskets that hold the top and bottom of the engine to split apart, and sometimes causing enough damage to crack the engine altogether.

Typically, by the time that the car finally rolls to a stop, the engine is toast, the electrical system is usually fried due to heat and wildly fluctuating electrical current, any electronic system has been turned into so much silicon, and various pipes and hoses have melted into noxious piles of rubber. If you're lucky, you can get out of the car before the engine itself ignites. The single proximate cause, the timing belt, could be replaced for maybe thirty dollars in parts and a couple of hundred in labor, but what has happened is a system failure ... a non-linear catastrophe ... you're probably better off just getting a new car.

Of course, its worth noting that eighty years ago, should something like this have happened, the system itself was simple enough that the car would have been fairly simple to fix ... painful at the time, of course, but doable. Now, we're reaching a point where you cannot solve what amounts to a fractal catastrophe with linear solutions.

There's a saying that I think applies well to economics theory - insanity is the process of doing the same action repeatedly and expecting different results. Admittedly, in a non-linear regime that's pretty much what does happen, but what emerges is usually not the result you expect to happen. The current situation, with car company bailouts and bank bailouts, is increasingly looking like it is a non-starter. The infusion of capital may temporarily ease credit problems, but the timing belt is broken, and the car will soon stall again.

Award Fractal Thinking, Not Linear Thinking

Significantly, the demise of the investment banks and the demise of the automobile industry both come about due to the same factor - rather than concentrating on their core business (the safekeeping of money and the production of competitive vehicles respectively) these two industries had (like the insurance and real-estate industries) become far more focused on making money by moving money around.

It is easy (and definitely appropriate) to point at many CEOs with their outsized incomes that were paid far in excess to their actual contribution, but its worth understanding that this had become so attractive because innovation ... actual invention ... is hard. It's difficult and risky and usually costly, it requires the type of mindset that has become increasingly rare not just in the US but globally. It's usually safer just to take your paycheck and pocket the stock options, because you're sending your kids to college and you need to save for your retirement because that's the way that things are done.

The Obama adminstration hasn't even officially taken power yet, but already the president elect is challenging the status quo in interesting ways - chief among them the proposal to rebuild the US infrastructure for the twenty-first century. While this has free market capitalists gnashing their teeth, it's worth noting that it will in fact be companies that do most of the actual work - the change being that rather than essentially rewarding those who have caused the problem in the first place, what Obama is proposing is to restore what has been lost in this country - honest competitive bidding for US contracts, based not upon connection (for the most part - there will always be a certain amount of nepotism) but upon competence.

It is easy for the "free market" to question the role of government intervention, but most people do not realize that the government typically tends to be one of the largest employers of public and private companies. The aerospace industry effectively exists because of governments, because the cost of producing and innovating airframes would be prohibitive if it was not at least partially subsidized by US development. The same can be said for many industries (including the automotive industry, ironically). However, this role needs to be rethought in the face of the collapse of twentieth century capitalism.

One of the roles of government is to provide incentives for innovation. Businesses tend, by their nature, to be conservative - it is typically better, despite all the marketing hype, to find successful products and push them as far as possible than it is to spend massively in research and development - which is of course a very risky proposition. Usually, beyond a certain size, a company innovates by buying up smaller, more agile companies and purchasing their intellectual property ... usually at the cost of forestalling any future innovation.

Suppose, however, that rather than continuing to feed the bloating and inefficient automobile industry (funny how that used to be applied to government), perhaps what the Obama administration needs to do is establish prizes of $8-$10 billion dollars apiece to design and build new vehicles that work in a carbon-sensitive world, open not only to the established companies but to any corporation capable of pulling it off. The prizes would be awarded on the basis of clearly delineated guidelines, with secondary prizes for alternative technologies that look promising.

Indeed, one possibility here is to open up these prizes so that some (perhaps even much) of this money would come from private investors, who would in turn be rewarded voting shares of stock proportional to their investments, and who would get significant tax benefits for doing so. The automotive industry has long been very competitive in maintaining control by its oligarchy; this would serve not only to break these oligopolies but would also give engineers and lineworkers alternatives to their present work that would position them well going forward.

Yet, it is also naive to look at specific industries as being the source of such solutions. Bailouts, whether of banks or of automobiles, presuppose that the industries are in fact the ones best positioned to solve the problems. However, the solution that a car magnate is going to come up with for the fast-approaching transportation crisis will be to design a better car, rather than to rethink our approach to transportation in general. We've entered a regime, as James Howard Kunstler points out, of oil volatility, where oil prices will swing wildly between extremes of price, which will in turn wreak havoc to oil producers and consumers alike.

Perhaps the best new truck is in fact a dirigible, the best new car a bicycle. Perhaps the financial system needs its own Manhattan Project, one designed specifically to rethink the whole notion of business, finance and commerce.

Successful inventions build on the past, but they do so for the most part by asking the question "X was designed to do this, but what if you combined X and Y to do this other thing instead?" It's a non-linear way of thinking - disruptive and game changing, but often the only way to solve a problem that can't be solved by successive small adaptions. Such inventions are seldom "safe" ... they are risky to the inventor, they may have unintended consequences that can't be modelled, and they almost invariably represent a break from the status quo (which pisses the status quo off mightily) ... but such inventions are also typically the foundation of a new way of seeing problems that lead to solutions, rather than simply satisfying the "Powers That Be".

Think fractally, think non-linear, and help those of your linear friends, neighbors and political representatives who can't conceive that tomorrow will not be like today to understand that linear thinking is a dangerous, deceptive illusion.

Kurt Cagle is an author, programmer and yes, journalist, working as an editor for O'Reilly Media. Feel free to subscribe to his newsfeed or twitter feed.

You might also be interested in:


If the economy is as non-linear as the weather, how can it ever be made rational for anyone to absorb the high risks and costs inherent in innovation and investment (and basic research and development)? For that matter, how can anyone estimate any economic risk in a non-linear economy?

When some private entity wins the prize awarded by government, what will stop it from taking the money and then not producing anything? What ensures that the Big Idea will be profitable to produce at all (there are plenty of energy alternatives, just not at the necessary price points)? Will the economy become a centrally-controlled zero-sum game in which the Innovators dominate through government financial support?

@Kurt, I do agree that fractal issues must be taken into effect in these industrial failures. However, I think you've ignored some of the fundamental factors.

Today's Wall Street Journal has a succinct editorial about the Detriot bailout: "The Bailout that Won't" ( ). Detroit's failure is that the CAFE rules require not only that they must produce a certain average fuel economy, but that all of those cars for sale in the US must be produced in the US. The Big Three are capable of producing small cars and making a profit on each -- but they cannot do this with cars they make in the US.

These seemingly-small constraints have a nonlinear effect on these three companies. Because of the fixed costs per car, it becomes problematic for these manufacturers to make money on their smaller cars. Other car manufacturers don't have this onerous restriction on their manufacturing. To no surprise, those other manufacturers are not in much better shape than the Big Three.

The editorial writer summarizes this succinctly:

Under a law of politics, such truths were unmentionable in last week's televised circus because legislators are unwilling to do anything about them. They won't repeal CAFE because they fear the greens. They won't repeal CAFE's "two fleets" rule (which effectively requires the Big Three to make small cars in domestic factories) because they fear the UAW. They won't hike gas prices because they fear voters.

Oops. An error in the last posting. I meant to say:

To no surprise, those other manufacturers are in far better shape than the Big Three.

It is also noteworthy that our language is steeped in the language of linear thinking. We work to solidify our ideas, which has them be linear building blocks. We talk about finding a point of leverage or seeking a fulcrum -- two concepts that are rooted in the world of tightly-coupled systems built on top of permutations of the simple machines.

"Building blocks" itself is an example of how one thinks in a compression-based (or stacking-based) structure. Robert Hooke talked about the linear stress/strain response; systems that behave that way are called Hookean. Most of nature's structures are non-Hookean/nonlinear.

It takes real thought to realize how pervasive our linear thinking is. Fortunately, that thinking to realize how linear we are oriented is itself non-linear.

@Kurt, thanks for the pointer to The Black Swan. It just got bumped up my reading list.


Actually the mention of weather here makes a great deal of sense. Weather is nonlinear and chaotic, but not unpredictable. In most cases, even a nonlinear system acts like a linear system so long as you are outside of the inflection points in the system. For instance, while I cannot predict with any accuracy the precise conditions at 2am on the 18th of December in Victoria, BC, I can say with a reasonable degree of certainty that it is likely to be both within a few degrees of freezing and that it will (with less certainty) be raining or snowing. If I can see weather satellite images (the big picture) then I can actually predict the weather almost perfectly within the next hours, but with a level of certain that drops off very quickly thereafter.

These are stochastic measurements, representing the likelihood will be in any one given state. The problem with such stochastic measures is that they are, by definition, based upon history, and history is not uniform. For instance, consider a typical Victorian summer, which usually tends to be dry and sunny (albeit not all that terribly hot) from June 15 to about August 15.

Now, you have an outdoor event that you're planning for August 20, and you need to decide by no later than August 1st whether in fact its doable. So you keep a graph of temperatures and cloud conditions, and each day except perhaps one or two are perfectly clear, and those two are partly cloudy. Based upon this information, you're pretty likely to conclude by the first of August that its practically certain that August 20th will be clear.

However, as it turns out, typically a high pressure ridge sits off the coast of the Pacific ocean, largely due to heat from the coast off of California. By mid-to-late August, that particular ridge has begun to weaken, just as a similar pattern in Japan kicks in and starts to throw colder, moisture laden air across the pacific.

What this means is that by about August 20th, out of a seemingly clear blue sky it isn't at all uncommon for a storm to blue up, and because of pent up pressures, this storm has a lot of energy to it. Thus, August 20th is actually a pretty lousy time for an outdoor event, even after two months of seemingly perfect weather.

You can argue that you didn't have enough information here, and had you known about the pressures, then you would have planned accordingly. That is, however, precisely the point - you can never have complete knowledge of the system, and because of that, any model that you make will have an intrinsic risk factor associated with it. Put it another way, there is no such thing as a completely risk free investment.

Now, there's another factor that plays into this. Nothing ever occurs without a reason, even if that reason does not seem obvious at the time. Usually systems go through a period of instability as energy enters into them, reach a period of stability as the system reaches para-equilibrium, then becomes unstable again as new energy either enters or leaves the system. In weather, the source of that energy is obvious - the sun moving above or below the ecliptic.

In economics, you have similar inputs - an energy differential, but in this case, the energy comes from the development and exploitation of sources of energy, with oil being, far and away, the most important (I'll cover why in a future article).

Normally, when a market is growing, this means that it becomes increasing resistent to external risks - but those risks don't go away. Rather they build up, adding to the overall instability of the structure. A period of equilibrium usually means that the forces that forced the initial move to the equilibrium state have become balanced to the risks. Eventually, however, the risks exceed the capacity for growth - the system cannot grow fast enough to make up for the decay due to risk. Once this happens, the system degenerates fairly quickly, as risks tend to be interdependent, until it reaches a new point of equilibrium.

Before wrapping this up, a final note. Insurance is a cheat. It in essence provides a way to let you pass the risk off to others. It does not, however, make the risk go away ... and in the end we will all pay for it anyway.

So what does that mean as an investor? First, if you're going to invest, keep yourself educated, beyond simply what is happening in any one given market. The price of oil affects currency moves affects equities affects bonds affects commodities affects the price of oil. The price of oil affects geopolitics affects economies affects legislation affects the value of the dollar affects the price of oil. The more that you know about the big forces, the easier it is to anticipate the more disruptive changes.

Keep your portfolio diversified, understanding that even then that you will sometime lose money. Every so often, you will throw snake-eyes, no matter how well you guard yourself. For all that people believe that the economy is cyclical, it isn't - its very much energy dependent, and barring a significant change in our underlying ability to generate energy, the economy will be in decline long term for quite some time to come.

As to your second question - phase payments of any award over fifteen years, with review every three years, and during that time, a government representative has a voting seat on the board of directors for that company. In effect, the government becomes a shareholder in that company during that period. Additionally, such a representative would need to sign off on any contracts which the company entered into with the government. Finally, this governor can not hold more than two terms of three years each in any one company, in order to minimize the possibility of corruption.

There's no guarantee that anyone Big Idea will in fact work. All this does is provide an alternative simply to writing a blank check to companies that are "too big to fail", and it gives the government an incentive as well in making sure that no one company will have too strong a competitive advantage.

As to your final question - tell me how that differs from the situation now?

Floating Bones,

I'd question the comments on CAFE standards for several reasons. First, the green lobby has in general been fairly ineffective in staving off efforts on far more dire environmental fronts than CAFE standards given the very laissez faire attitude that the administration and Congress has had on just about every other environmental issue in the last decade plus.

Second, most cars sold for import come from countries that have generally had far more stringent environmental restrictions on their automobile industries. This is definitely true of Japan and Korea, and I believe it's true for most EU companies.

The two fleets rule issue may be true - though I'd question the degree to which this is a bad thing. The CAFE rules do not in general place restrictions on GM or Ford selling in foreign markets, only that it doesn't give GM or Ford the right to take US subsidies in order to promote job growth then have them turn around and have these jobs done in Korea or China at a tenth the cost.

Keep in mind that both GM and Ford exist primarily in the twilight world of being banks that happen to sell cars in the front lobby. Thus when the credit crunch hit, they were faced with an inefficient production and sales channel coupled with a highly profitable financing arm suddenly having that financing arm stopped cold.

GM in particular is in panic mode right now because they likely used more than a bit of their pension funds both as collateral for their own financial operations and had invested much of it risky portfolios that collapsed when the market did. Now, not only do they not have any money to facilitate additional lending (causing their sales channels to seize up) but they're on the hook for billions of dollars in pension obligations that have mysteriously evaporated.

I'm not very sympathetic to GM right now, and only slightly more sympathetic to Ford as I think Bill Ford may actually be able to pull off the revolution that the company so sorely needs right now. I don't think the unions are completely blameless in this, but overall senior management salaries have been rising at an exponential rate in comparison to linemen and engineers at a time when senior management was doing everything it could to halt meaningful innovation.

As to the linear bias of language - oh, absolutely. Linearity is a survival skill, based largely upon the ability to see patterns (something the brain does very well). Non-linearity is unexpected, and literally causes the brain to go into a temporary (albeit short) seizure, according to some of the most recent research on brain/mind issues. That's why people talk about being stunned when something unexpected happens.

This is actually a pre-human facility, and is also a survival trait. In all likelihood, early arboreal apes would "freeze" in the presence of a snake or leopard or similar predator, probably because such predators usually tend to react most strongly to motion. (It's the same reason that deer get hit so often by cars ... they see cars as predators, and rather than attempting to run out of the way, they freeze, so as to confuse the lion into thinking that they are part of the background. Unfortunately, cars are not predators (normally anyway).

Most thinking about non-linearity is very contemporary; we have had the ability to numerically solve non-linear differential equations for about the last fifty years, fractals date from the late 1960s, and chaos theory is only about thirty five years old. Not surprisingly, people tend to use it in a very linear, predictable fashion, which is why quants, people who should be absolutely brilliant at the stock market, are so often so outrageously wrong.

You raise some very thought provoking comments. Thanks.


Taking the weather idea a bit further, it seems to me our fundamental mistake was not preparing for a sudden and unexpected change in conditions. I'm not even sure our big problem was in not "seeing" events like the oil spike or financial meltdown beforehand - it was our false sense of security and our lack of understanding about how fragile the underpinnings of our economy were.

In IT, we've learned how to plan for failure. In general, we don't try and anticipate and head off every type of failure, or say "this is our main server and it absolutely, positively, cannot fail"; instead, we simply ask ourselves "how can we mitigate the damage were failure to occur?" and also "is this system sustainable in the face of a failure?" In short, we make sure that our systems can gracefully handle failure. There are many other fields as well where it's common or necessary to plan for failure and identify contingency plans in case of such a failure.

So why not do the same with our "economic network"? Why was there a lack of people out there saying "hey, if this fails, look at the mess it will cause. How will we respond to that?" Or perhaps even "can we afford for this company to fail, and if not, what can we do to mitigate the damage?" Finding the answers to those questions won't be easy, but it's a lot better to ask them and think about them when you're not in the midst of a crisis. You may even be able to take preventative measures (like factoring such risks into a merger discussion) to keep an industry or the economy from becoming too reliant on a few large entities, which could take a sizable portion of the economy with them if they fail.

Of course, it's possible they were seriously considered, but it certainly seems like most of the people involved are thinking on their toes right now...!

News Topics

Recommended for You

Got a Question?