I continue to be fascinated by the similarity between climate science and macro-economics. Both study unbelievably complex multi-variable systems where we would really like to isolate the effect of one variable. Because we only have one each of climates and economies (we can define smaller subsets, but they are always going to be subject to boundary effects from the larger system) it is really hard to define good controlled experiments to isolate single variables. And all of this is done in a highly charged political environment where certain groups are predisposed to believe their variable is the key element.
In this post by Russ Roberts, one could easily substitute "climate" for "economy" and "temperature" for "unemployment."
Suppose the economy does well this year–growth is robust and unemployment falls. What is the reason for the improvement? Will it be because of the natural rebound of an economy after a downturn that has lasted longer than people thought? The impact of the stimulus finally kicking in? The psychological or real impact of extending the Bush tax cuts? The psychological or real impact of the November election results? The steady hand of Obama at the tiller? All of the above? Can any model of the economy pass the test and answer these questions?
The reason macroeconomics is not a science and not even scientific is that the question I pose above is not answerable. If the economy improves, there will be much talk about the reason. Data and evidence will be trotted out in support of the speaker’s viewpoint. But that is not science. We don’t have a way of distinguishing between those different theories or of giving them weights to measure their independent contribution.
I’m with Arnold Kling. This is a time for humility. It should be at the heart of our discipline. The people who yell the loudest and with the most certainty are the least trustworthy. And the reason for that goes back to Hayek. We can’t measure many of the things we would have to measure to have any reasonable amount of certainty about the chains of connection and causation.
I have heard it said that the only way nowadays to advance pure science is to be working on arcana like the first microsecond of the universe or behavior of the 9th dimension in string theory. There is still room for a ton of useful work on the analysis, solution, and forecasting of complex multi-variable systems, even if it is just a Goedel-like proof of where the boundaries of our potential understanding can be drawn.
By the way, I wrote my own piece about the limits of macroeconomics here.