Can Economic Risk Be Tamed And Markets 'Know'? Increasingly: No.

by Stuart Kauffman

In this post, I want to show that we often do not know what can happen in the technological evolution of the economy, hence in the most fundamental sense, in general, risk can neither be tamed, nor can markets accurately factor that risk in. As technological innovation accelerates, this problem is likely to become ever worse. This challenges foundational ideas in economics, and gives the lie to the noisome cant of some professional Conservative think tanks and noisy talking heads that we must leave everything to unregulated free markets that somehow always "know".

In my previous blog, Breaking the Galilean Spell, I described Darwinian "exaptations" in biological evolution. As I describe below, the same ideas apply in technological evolution. Exaptations are Darwin's idea that a causal property of an organism, like heart sounds, that was not the selected function of the heart, (to pump blood), and of no selective use in the current environment, could become of selective significance in a different environment, so be selected. He, and other biologists since him, including myself, note that typically a new function arises in the biosphere.

I told of the evolution of the swim bladder, that adjusts neutral buoyancy in the water column by the ratio of air to water in the bladder, by exaptation from the lungs of lung fish. Water got into the lungs of some fish, and this organ was poised to evolve into a swim bladder. I asked: 1) Did a new function come to exist in the biosphere? Yes, neutral buoyancy. 2) Did this new function alter the future course of evolution? Yes, new species, proteins, and niches. So the becoming of the world was changed. 3). Can we prestate all possible exaptations, just for humans? We all seem to agree the answer is "NO". I've asked thousands of people. And I noted that parts of why the answer is no, were: How would we prestate all possible selective conditions? How would we know we had listed them all? How would we prestate the features of one or many organisms that might become preadaptations? There seems no way to do these things.

Then I defined the"Actual" and the "Adjacent Possible" of a litre of 1000 initial kinds of small molecules. Call these 1000 the "Actual". The Adjacent Possible are those new kinds of molecules that could form in single reaction steps from the Actual. Then, by the paragraph above, we cannot prestate all the possibilities of the Adjacent Possible evolution of the biosphere. It follows that we do not know all of what can happen in evolution.

Further we cannot even make probability statements about the evolution of the biosphere by such exaptations, for we do not know the set of all the possibilities, called the "sample space", of the evolutionary process. Not knowing the sample space, we cannot construct a probability "measure".

Most startlingly, if a natural law is a compact description of the regularities of a process, we cannot have a "sufficient" natural law for the emergence of swim bladders. The becoming of the universe is not fully describable by natural law - thus "breaking the Galilean Spell" since Galileo and Newton that all that unfolds in the universe is describable by natural law.

Then what about the technological evolution of the economy? The same ideas apply. Once again, we often do not know what can happen, so, in general, risk cannot be tamed, and, in general, markets cannot always "know". More, the problem is getting worse.


I tell a story I am told is true, of exaptations in the econosphere. Some engineers were trying to invent the tractor. They knew they needed a very large engine, hence a very large engine block. They mounted the block on successively larger chassis. All broke in turn. Finally one engineer said, "You know, the engine block is so big and rigid, we can hang everything off the engine block and use it as the chassis. It worked! And that is how tractors are made. So too were formula racing cars for some time.

Now the use of the rigidity of the engine block as the chassis by the engineers is a Darwinian exaptation - it is the use of an unused causal feature of a system for a novel function. Did a new function arise? Yes, tractors. Could we prestate all uses of an engine block, or screw driver for that matter? No. Who knows what novel uses an engine block or screw driver might find? How would we list them all, know we had listed them all, or what causal features of the engine block or screw driver might be of use for some purpose? Again, there seems to be no way to do this.

Technological evolution is full of Darwinian exaptations. Most inventions are not used for their initial purpose. A blog ago,my colleagues and I discussed the evolution of the early computer, invented to calculate shell trajectories in WWII, that enabled the invention of the Apple personal computer, which in turn afforded the opportunities to successively invent: word processing, hence Microsoft, files, sharing files between buildings at CERN, the world wide web, eBay sales on the web, Google search engines on the web, and, at last, Facebook. Thomas Watson Sr. foresaw use for three computers in the 1940s at IBM. Well, no, Thomas Watson. Did any of us foresee Facebook or Google 25 years ago? No.

Once again, like the biosphere's evolution, we cannot prestate technological evolution. Once again, not only do we not know what will happen, often we really do not know what can happen.

I tell a funny and true personal story. A number of years ago, a baby Bell sought my advice about investing $2 billion for fiber optic cables. They would pay me $5000.00. I reasoned closely: "I don't know anything about fiber optic cables...But $5000.00 is $5000.00." I agreed and learned a lot in a day about fiber optic cables.

At the end of the day, from nowhere, an image came to me and I said, "How do you know some kid won't learn to keep empty tin cans in the air around the globe with beebee guns and bounce signals around the world? Your fiber optic cables would be worthless!" Seven pairs of eyes glimmered at my stupidity. The company invested the $2,000,000,000 in fiber optic cables. Six years or so later, satellites were launched that bounced telephone signals around the globe, rendering those cables useless for a long time.

I am, of course, unreasonably proud of my tin cans, but the question is: Was the baby Bell company stupid? NO. They could not know what would happen.

Now economists often think they can tame "risk" and that markets typically can and do correctly factor in risk. Suppose the baby Bell had issued bonds backed by expected revenues from the fiber optic cables, in order to buy the cables. Suppose Moody rated them, (at the same time Moody was payed by the baby Bell to do the rating, with the obvious conflict of interest), and rated the bonds AA. Maybe hundreds of thousands of the proverbial little old ladies invested in the bonds and they proved worthless.

Were the baby Bell and Moodys able to tame risk? No. Did the market "know"? No.

There are at least two different reasons risk might not be tamed. The first concerns what are called "power law distributions" with no means or variances. Power laws are distributions plotting the logarithm of, say the number of Nile floods of a certain size on the X axis, and the logarithm of the frequency of floods of each size on the Y axis. It is mathematically true that if the slope of this line is flatter than -1.0, the distribution has no average, or mean, nor higher moments like variance. So no amount of data can tell you what might happen. Others have made this point, for example Taleb in "The Black Swan".

What I am talking about is much more radical. In the case above, at least we knew what variable to measure: the size distribution of Nile floods. But in the case I am considering, the baby Bell, fiber optic cables and the unforeseen invention of satellites to bounce telephone signals around the globe, this lethal risk was not previsible, and we did not even know what variables to measure. This is the "unknown unknown".

So for Taleb's reason, and I think more deeply, for my reason, we cannot, in general, tame risk.

Economists want to believe that fancy trading algorithms, or the market itself, can price in risk accurately. In general, it cannot. We do not know before hand what variables to measure. Economists will treat such innovations as "exogenous shocks", but they are not exogenous. The invention of satellites to bounce signals around the globe grew organically out of technological evolution which, as Brian Arthur says in "The Nature of Technology" always grows out of existing technology. Very often this growth into the Adjacent Possible of the econosphere cannot be prestated.

Our pace of entry into the technological adjacent possible is accelerating, and with it the frequency of our encounters with the unknown unknown. With this accelerating pace, taming risk is ever more beyond reach. Free markets cannot always "know" and are likely to know ever less adequately as we explode into the Adjacent Possible technologically.