Just in time for the annual meetings of the American Economic Association, Russ Roberts writes,
modeling economic behavior using the tools of the physical sciences in hopes of attaining the holy grail of a full-blown, accurate, model that can track the ups and downs of a complex modern economy is not just a fool’s quest, but a dangerous one.
Read the whole thing. No single excerpt can capture what he has to say. My own thoughts:
1. As you proceed deeper into economics courses, the mathematical difficulty increases at what seems to me a steeper rate than the extent of insight.
2. The project of turning economics into a mathematical, quantitative-empirical discipline is difficult, if not impossible, to stop. My email today included the table of contents of the AEA macroeconomics journal. I looked at some of the articles, including the lead article by he eminent Michael Woodford, and I thought they were pretty worthless. But there is no way for the emperor to be overthrown, even if he has no clothes.
3. When I worry about the impact that government funding has on scientific research, I have economics in mind. Government funding serves to entrench the incumbents in any market, and the impact in economics is clearly visible. Maybe “real” science is less susceptible to being captured by cliques, because failure is easier to demonstrate. I don’t know. But what I see happening in economics does not make me want to sign any petitions for more government funding of science.
4. I can sometimes see cases of inverse relationships between scientific pretension and good economics. Exhibit A would be the folks who think that regression analysis tells them whether or not government workers are overpaid, rather than considering simpler behavioral evidence.
5. With all that said, I think that enough useful insights have been developed over the past fifty years that I am not ready to say that we should blow up the current graduate economics curriculum and replace it with one that just studies the classics. What I would like to see is less homogeneity among the top graduate schools, so that top students are not all allocated to programs that worship at the mathematical altar. The top grad schools produce a monoculture, and I think we would be better off if there were diversity.
READER COMMENTS
OneEyedMan
Jan 5 2011 at 7:54pm
There are still some heterodox programs. I wonder if any country has been successful deliberately following their insights.
I’m imagining something like a heterodox Chicago Boys.
With economics blogs popular outside of the profession, I would think that less-mathematical economics would be experiencing a policy heyday if it were generating useful policy ideas. I don’t know how to reconcile that with your grim assessment. Do academics have a stranglehold on policy makers? I can’t reconcile that with the political positions taken on free trade, minimum wages, Tobin taxes, agricultural subsidies, and so on.
RGV
Jan 5 2011 at 8:27pm
Guess the genius of the markets can’t even produce diversity in economics education. Wonder how it solves more pressing problems.
Chris Koresko
Jan 5 2011 at 8:48pm
I’m not an expert in economics, but I’ve done my share of mathematical modeling in other fields. My impression is that your opinion of mathematical economic modeling may be more unfavorable than the approach really deserves.
My guess (and I admit it’s no more than that) is that the problems you’re seeing actually come from a lack of mathematical sophistication among modelers, which leads them to fail to recognize the nature of the uncertainties in their results. One common mistake is to assume that random variables have well-behaved Gaussian statistics, which is provably true in certain circumstances and tends to make the math easy, but is badly incorrect in many situations. This can lead to large overestimates of the significance of correlations, for example.
But if one approaches modeling with a suitable humility, it has the potential to teach you a lot. I’ve often had the experience of throwing down a few equations which describe self-evident relationships, doing some mathematical crank-turning, and being rewarded with an aha! moment when a new expression comes out that reveals a truth that is obvious only in retrospect.
I’ve had a recurring fantasy lately in which I spend a weekend putting together a first-cut model of PSST and show that it makes qualitatively correct predictions about how the economy behaves in a recession. Wouldn’t that be sweet?
Phil Rothman
Jan 5 2011 at 9:36pm
Many of commenters on Russ Roberts’s post seem to think he said something deep. In contrast, I can’t join them. Sure, ‘scientism’ is chicanery, and economics journals have published far too many unenlightening applications of mathematical modeling in economics. Similarly, lots of silly economic analysis has been done sans a mathematical approach. For an example of what I think is far more substantive thinking of this subject, I would suggest, e.g., the following (in which there appears not a single equation):
http://www.minneapolisfed.org/publications_papers/pub_display.cfm?id=4526
david
Jan 6 2011 at 4:22am
But… but.. your “simple behavioral evidence” implicitly presumes an underlying mathematical relationship! You can’t pretend the assumption doesn’t exist by merely stating it in words!
Arrrrgh.
fundamentalist
Jan 6 2011 at 9:29am
I agree with Chris. The problem with mainstream econ isn’t the math; it’s the theory behind it. As Mises used to say, equations are just prose translated into math symbols. The prose is wrong and that’s why the equations don’t work.
As for changing economics, I see more and more Austrian economists at state schools. I don’t think mainstream is ever going to go away, but I see more heterodoxy developing.
RGV:
Where do you find most economists working? In government. That is not a free market. And the state demands economists who give it credit for all things good in the economy, which is why we will always have interventionist economists.
How can we promote better economics? Try this:
from http://www.theglobeandmail.com/report-on-business/economy/friedrich-august-von-hayeks-time-has-come/article1857752/
Today, the only people who listen to economists are politicians. Investors and investment advisers utterly ignore economics. The way to get people’s attention is to make money. If we can show that our economics creates wealth for individuals by providing good investment advice, people will follow it and start demanding better economics in schools. I can envision a future in which economists who want to work for the government go to one group of schools while economists who want to work in the private sector go to a separate group of schools.
BTW, two good pieces of empirical evidence for Austrian econ are unemployment and cyclical investing. I just taught a class in econ about unemployment in which the text had a graph of unemployment by sector. Unemployment in capital goods industries in the recent depression was twice that of other sectors. That is the Austrian business cycle theory. And that has always been the case in every depression.
As for investing, listen to the talking heads: they’re all advocating investing in cyclical stocks. What are cyclicals? For the most part they are capital goods industries. Cyclicals follow the business cycle. They are the high beta stocks. They boom when the economy is booming and crash when it crashes. “Defensive” stocks are consumer goods makers and retail. Their swings are much smaller. Again, that is the ABCT in a nutshell.
Daublin
Jan 6 2011 at 1:55pm
Like others, I don’t see how rejecting modelling in general helps anything. If you aren’t making models, then you’re not even a science. If you aren’t a science, you are just talking about the economy and not providing anything interesting.
The problems you describe, Arnold, can be perfectly summed up by saying the models are bad. So find better ones. Perhaps best of all, find better variables to bother to model at all. Find a way to, instead of elaborating epicycles in ever more detail, develop a general theory of gravity, motion, and force.
Chris T
Jan 6 2011 at 2:30pm
Modeling has it’s place, but must be mediated by experiments and real world data. Otherwise you’re just a subset of mathematics (the same problem afflicts string theory).
The opposite also applies, if you’re not going to attempt to derive or test general principles; there is little point to conducting experiments.
yes
Jan 6 2011 at 8:52pm
[Comment removed for supplying false email address. Email the webmaster@econlib.org to request restoring this comment. A valid email address is required to post comments on EconLog.–Econlib Ed.]
Comments are closed.