How close are we to predicting earthquakes or the next bear market?

“The future is an excellent topic for any author. By the time you realize I was wrong about everything I predicted, I will be dead.”—Scott Adams, creator of the Dilbert comic strip

Actually there are plenty of predictions – just none that you can count on.

Earthquakes and sudden stock market crashes do not fit nicely into Newtonian models (where Sir Issac could predict quite accurately over 300 years ago where Saturn would appear in the sky in a few days). Even when we put our sharpest mathematicians on the job of predicting these things, they don’t do much better than a 12 year old using a dart board. Sure, the experts can tell us that there will be another big earthquake in Los Angeles and that the stock market will drop by at least 10%. But today they can’t get the timing or the magnitude right. And they can’t identify the single straw that breaks the camel’s back. In the meantime, we may have smooth sailing in Los Angles for the next 20 years and the stock market might last for several years without experiencing a major hit – although I would bet against it.

These are complex systems that involve countless forces (some understood and many not) that react with differing exponential forces. In other words when factor A accelerates at a rate of X then factor B might respond by a factor of X squared or X^6.8 . And getting exponential factors wrong throws off the accuracy of the forecast much more dramatically than getting linear factors wrong. For example the difference of 5 X 3 versus 5 X 4 is 15 versus 20.  But the difference between 5³ versus 5⁴ is 125 versus 600.  A small error in an exponential dimension is much more significant than a small error in a linear factor.

For both earthquakes and financial markets, we don’t even know all the aspects we need to watch let alone having the ability to actually track them all. And we are not even close to understanding exactly how each variable impacts all the others. Consider how a butterfly flapping its wings might change the results.

The “Butterfly Effect” whose name was coined by Edward Lorenz derives from the idea that a hurricane might result from whether or not a butterfly flapped its wings thousands of miles away.  It is similar to the notion of a single snow flake being just enough to start an avalanche. The single snow flake results in an avalanche which knocks down a forest. The forest was the home to a large butterfly population. Eventually the ensuing chain of events results in the ocean warming by a degree or two and before you know it a hurricane is headed for the Florida Keys. It’s like the Direct TV commercial: Don’t Fall Into a Dinner Party: “When you pay too much for cable, you feel powerless. When you feel powerless, you want to take the power back. When you want to take the power back, you take Karate. When you take Karate, you want to use your Karate. When you want to use your Karate, you become the fist of goodness. When you become the fist of goodness, you run along rooftops. And when you run along rooftops, you fall into a dinner party.  Don’t fall into a dinner party. Get rid of cable and upgrade to DirectTV.”

These complex systems involve so many interrelationships that we just don’t understand them nearly well enough to be able to make useful predictions. Granted, once a hurricane is in the neighborhood, the models are reasonably good about extrapolating to where it will end up in a few hours. But forget about knowing where and when the next hurricane will originate.

Back to earthquakes. The earth’s crust is on about 62 miles thick and made up of 7 primary tectonic plates (and numerous secondary and tertiary ones) that float on the molten liquids inside the earth. These plates tend to snuggle up to one another, move in opposite directions and resist each other’s movement via friction forces. For example, the Pacific Plate (the largest plate at about 40 million square miles and covers about 20% of the earth) creeps slowly northwest. The 810 mile San Andreas Fault separates the Pacific Plate from the adjacent North American Plate which moves in the opposite direction (southeast). Earthquakes happen when the tectonic plates suddenly shift - releasing energy like a massive spring that has been sprung. Most of the time this movement is small - for example the average relative movement between the Pacific and the North American plates is only about 1.3 to 1.5 inches a year. But earthquakes are like trying to move heavy furniture by sliding it across an uneven floor. You push and you shove and the darn thing won’t budge until all of a sudden it lurches. This is the essence of an earthquake, a sudden staggering movement between two tectonic plates.

So why can’t we predict when the next big quake will happen? First, imagine the difficulty in understanding the friction forces where one plate meets another. There are plenty of other smaller faults that connect to and also influence the action. To model this system would consider the stress building all along the fault, and adapting the model to account for different kinds of rock, shapes and surfaces of the edge of the plates, temperature and then maybe – maybe, maybe one could simulate and predict when a violent movement would suddenly happen. We ain’t close.

And if you think predicting the next big quake is tough, imagine the process of forecasting when the stock market will tumble next (I will go out on a limb and guarantee that it will tumble sometime). It is a function of economic growth, company earnings, unemployment rates, changes in tax rates, wars, attitudes about work and innovation, natural catastrophes, the tendency of the government to enforce anti-trust laws, changes in technology, global economic shifts, heat waves, droughts, minimum wage laws, immigration policies, new import tariffs by other countries, and changes in demographics (to name just a few). Each of these factors influences other factors and when you throw the results all into a pot, mix them together, it is tough to understand their relative importance. The words of nineteenth-century writer William Gilmore Simms apply tenfold: “I believe that economists put decimal points in their forecasts to show they have a sense of humor.”

So far we are nowhere near having models that come anywhere close to predicting the what, where and when.

Back in 1984, scientists predicted with what they proclaimed as a 90 to 95% confidence level the next earthquake of about 6.0 (Richter scale) would happen in Parkfield, California (about half way between San Francisco and Los Angeles, in Monterey County) sometime before 1993. They didn’t use the kind of detailed model envisioned above.  Instead they simply reviewed the frequency of the large earthquakes (over 5.5 magnitude) which had happened in 1857, 1881, 1901, 1922, 1934, and 1966 and extended the historical pattern. But 1993 came and went and this bold prediction turned out wrong.  The next 6.0 quake didn’t arrive until 2004.

The 1984 earthquake prediction is similar to the kind of predictions we hear all the time about the stock market.  For example, pundits will state that we are due for another bull or bear market based on the time since the last one, assuming that there is some kind of order to the market. These predictions are typically based on small sample sizes and they usually sound quite plausible.  And they sometimes even get lucky.  One might also get lucky picking the winning numbers in the next lottery. But in the long run this kind of forecast has limited usefulness.

Some things are simply far more difficult to predict that others. That is why I a am a little skeptical about the precise forecasts we hear about global warming. I accept that more CO₂ in the air is a warming factor. But this is far from the only influence that affects the average temperature around the world.  There is simply no way to fully understand all the chain reactions and secondary and tertiary factors and how they all impact each other.