A head and shoulders wrapped in a triple top?

Photographer: Stephen Voss/ Bloomberg

Seeing Economic Patterns That Don't Exist

Noah Smith is a Bloomberg View columnist. He was an assistant professor of finance at Stony Brook University, and he blogs at Noahpinion.
Read More.
a | A

What if history moves in big, slow waves? What if patterns of growth and decline, or equality and inequality unfold over decades or centuries? Can we observe these waves and make history into a science? The answer is probably “no.”

I’m not saying that no such patterns exist. They might! But if they do, it will be very difficult to prove. Instead, we may end up getting fooled by random movements in the stock market, the economy or whatever it is we’re looking at.

For example, take the theory of Kondratiev waves. Back in the 1920s, a Soviet economist named Nikolai Kondratiev (sometimes spelled Kondratieff) theorized that the economy moved in long cycles lasting 40 years to 60 years. Expansion would be followed by stagnation, then recession, followed by a new expansion. Economist Joseph Schumpeter liked the idea, and speculated that these cycles follow technological revolutions. Others have claimed that Kondratiev waves are due to demographic or financial cycles.

But Kondratiev waves, if they exist, are almost impossible to identify in the data. For a periodic phenomenon like a wave to be identified, you have to see it happen many times. Otherwise you can be fooled by randomness that looks like a wave. Take a random walk generator and create some totally fake data. A lot of time it will go up and then down, like a wave. But because you made the fake data yourself, you know there’s actually no wave there.

Humans, you see, have a condition called apophenia, or false-pattern recognition. It’s what causes us to see shapes in clouds, or the face of Jesus in the folds of tree bark. A one-dimensional wave -- something that rises and then falls and then rises -- is pretty much the simplest, most common pattern in existence, and so it’s probably one of the easiest patterns to imagine where it doesn’t.

A better way of identifying waves is to use statistics. If a phenomenon repeats itself, you can uncover it by determining how often it repeats itself over your sample, and you look for peaks. Naturally, people have attempted to do this for Kondratiev waves. The problem with this approach is the sample size. Since reliable economic data goes back at most about 150 years (and probably less), we will have observed at most three of the 50-year waves. Any statistical test will therefore have very low power -- it won’t be able to distinguish between waves and randomness.

A similar result was famously demonstrated by economist Larry Summers in 1982. He showed that it’s possible that even if the stock market has bubbles that appear and disappear over the course of about six years, it’s statistically impossible to identify these bubbles in a 50-year data sample. Instead, you’d need more than 1,000 years to have a hope of getting a definitive answer. That was a blow to efficient markets theory, because it showed that statistics couldn’t tell an efficient market from one with bubbles. When the length of a cycle is so long that history only contains a few of them, you’ll never pick the cycle out from random noise.

Another example of a slow cycle theory is the idea of Kuznets waves, recently introduced by economist Branko Milanovic and named after Simon Kuznets, an economist who studied inequality and helped develop ways to measure the size of an economy. The theory is that as countries grow, they alternate between rising inequality and falling inequality. Milanovic shows that inequality fell in today’s developed countries between the 1800s and the mid-1900s, then began rising again. He hypothesizes that we’re in the middle of a second Kuznets wave, and predicts that at some point -- it’s hard to say when -- restorative forces will push the economy back toward greater equality:

The pro-inequality trends will be very hard to overturn during the next generation, but eventually they may be – through a combination of political change, pro-unskilled labour technological innovations…dissipation of rents acquired during the current bout of technological efflorescence, and possibly greater attempts to equalise ownership of assets…But history teaches us too that there are malign factors, notably wars, in turn caused by domestic maldistribution of income and power of the elites (as was the case in the World War I), that can also do the job of income levelling. But they do it at the cost of millions of human lives.

Milanovic might be right about this. But if there has only been one such Kuznets wave in modern history, there is just no way to tell. It may be that we should expect inequality to fall as the result of natural restorative forces, or it may be that it will linger for centuries. We just can’t know. Milanovic, recognizing this problem, points to some research showing rises and falls in inequality in pre-modern Europe, but this data isn’t likely to be reliable, and the forces causing the rises and falls -- wars, famines and plagues -- are certain to be very different from what we see today.

So beware of seeing patterns in history. They might be there, or you might be suffering from false-pattern recognition. Maybe it’s really true, as the historian Arnold Toynbee is reputed to have said, that “history is just one damned thing after another.”

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Noah Smith at nsmith150@bloomberg.net

To contact the editor responsible for this story:
James Greiff at jgreiff@bloomberg.net