Tuesday, February 6, 2018

Arnold Kling's Moonshot

I'm a great Arnold Kling fan--I've favorably reviewed two of his books on this blog (here & here). I follow his blog daily--it's one of my favorite reads. Mostly I agree with him. But I have to take issue with his post from a few days ago, entitled A Moonshot to Overthrow Neoclassical Economics, along with the similarly titled article, here.

A moonshot is the Big Thing that a person wants to accomplish in life. In other contexts it might be called a vision statement, or a passion. Though a moonshot is typically much less practical than those--a goal for striving rather than achieving. Mr. Kling's moonshot is "to be a leader in overthrowing neoclassical economics. A better approach would focus on mental-cultural factors and rapid evolution."

I think he gives neoclassical economics a bum rap.

In his account (and I don't disagree) neoclassics begins with two "essential propositions":
1) Production is a process that employs two primary factors--labor and capital. 
2) The distribution of returns to labor and capital reflects their respective contributions to the production process.
He notes that Marxists disagree with the second axiom--they believe capital steals a disproportionate share of the proceeds--and he acknowledges they might be right. But that, in his view, is quibbling over a detail. Mr. Kling's critique is much more damning--he denies the validity of the first premise.

Instead a third axiom must be included:
[Y]ou have to pay attention to what people believe. What they believe affects what they value, what they think constitutes a good investment, what conduct they believe is appropriate when buying and selling, how they interact in their work environment, and so on.
Mr. Kling says that culture matters. In a slightly different context others insist that institutions matter. And of course they do--does anybody deny that? Indeed, Paul Romer took the idea to the nth degree and attempted to build a city de novo in the Honduran jungle, endowed from the beginning with good institutions. I don't think that worked out very well.

My old professor in general chemistry (Norman Nachtrieb--now deceased) told us that "science is the art of successful approximation." Ideal gas law is valid as long as intermolecular forces are assumed to be negligible. A step in an engine cycle is adiabatic as long as no appreciable heat escapes into the environment. The words "negligible" and "appreciable" are purposely left vague, and depend on how precise an answer one actually needs in the result.

Economists have a similar concept--ceteris paribus--which means all else equal. Put that phrase in front of anything an economist says and you've got the economic equivalent to "negligible" and "appreciable." Of course, what with economics involving humans rather than molecules, the approximations are much rougher and the degree of precision much less. Still, at least in this sense economics is just like chemistry--it is the art of successful approximation.

One common approximation is to limit the timeframe during which a result is expected to hold. At picosecond scales (light-matter interactions) almost everything is adiabatic. For long-lived radioactive decay processes (the uranium half-life is billions of years), nothing is adiabatic. Completely different approximations will apply in the two cases--indeed, they can't even talk to each other.

In that sense economists have it easy--economic timescales don't vary by 30 orders of magnitude--maybe only by four or five (from a few weeks to a few centuries). Still, timescales matter, and this is where Mr. Kling's analysis breaks down.

Implicit in neoclassical economics are those two, essential, prefatory words: ceteris paribus. The axioms are not divine writ, but only hold as long as everything else remains equal. Thus we'd expect the approximation to be successful only for a relatively short period of time. How long is that time? Do many people make economic forecasts beyond a year? Even the Fed doesn't dare estimate anything more than four or five years out--and they're roundly mocked when they do even that. Ten years is beyond anybody's ken--ceteris is no longer paribus (pardon my Latin).

And Mr. Kling knows that. He says that culture changes, and furthermore, evolves. He further claims that it's changing more quickly than before. And therefore the all else equal assumption is invalid at nearly any timescale. Of course that's wrong--it is certainly possible to predict many economic phenomena (interest rates, corporate profits, unemployment) on a quarterly or even annual basis. Those predictions (especially the short term ones) are right more often than not. Culture does not change that fast.

Contrary to Mr. Kling, I think culture changes are on a generational timescale--roughly 30 years. My children grew up in a different part of the country than I did, and their mother (my wife) is Filipina. So obviously they see the world differently from me (though there is also considerable continuity). But now that they're 30 years old (give or take), those attitudes are mostly baked into the cake. For the next 50 years their cultural expectations will not significantly change.

I think Mr. Kling is mistaken in his claim that cultural evolution is speeding up. It's not. We're talking 30-year timescales here, minimum. It may even be longer since life expectancy has gotten longer. Within that timeframe, for the few-years horizon of neoclassical economics the ceteris paribus approximation makes sense.

What is more rapidly changing is technology, often known as structural change. That's the way mainstream economists refer to something not being equal. And like physical scientists they treat it with fudge factors, as a perturbation from the original result. Obviously if the structural change is huge (the engine explodes), that's not a fair approximation. But in the real world structural change happens slowly--driverless cars aren't anticipated to be mainstream for another 30 years.

So, over the two or three year window during ceteris paribus, neoclassical economics is a good approximation. Or, as scientists working at a blackboard might put it, a good first approximation, for which more refined iteration may be indicated. Mr. Kling's criticism is valid, but only over longer timescales when the two axioms no longer hold.

Mr. Kling does have a legitimate complaint. Approximate models will never give you exact results. Calculating theoretical implications out to four significant digits is a waste of time. It's computer gaming. Many econometric models seem to me to fall under this category (though I'm not expert).

But economists are not alone in falling in love with their computers. Scientists suffer from the same disease--in our current day so-called "climate scientists" are extending models far beyond where they can reasonably be expected to hold.

Further Reading:


  1. What is it that is approximately true about treating the economy as having two factors of production--capital and labor? --Arnold Kling

    1. Because it leaves out all the other things that my impression is you want to include, e.g, institutions, new technology, a change in buying habits, bad weather. All of those will affect the economy and are not obviously a function of K and L.

      I'm actually confused by your question. It sounds to me like I'm missing something basic. Please explain.