The Frontal Cortex

Heterodox Economists, Kuhn and Prediction

Posted by Jonah Lehrer on May 30, 2007
(10)
More »

So there’s an interesting debate over at TPM Cafe about this article in the Nation, which argues that neoclassical economics (the mainstream) suppresses its heterodox alternatives. If true, this would be a classic case of a Kuhnian paradigm, in which the entrenched dogma resists any alternative explanation. The anomalies are ignored, until there are just too many anomalies, and then the whole edifice comes crashing down. That, at least, is how “normal” science is supposed to work.

But my problem with economics isn’t that it ignores its heterodox alternatives, which is what the article tries to suggest. I think it’s pretty clear that neoclassical economics has consistently incorporated heterodox challenges into its dogma. Just look at Kahneman and Tversky, or much of the work done by behavioral economists. Nobody really questions loss aversion anymore. It’s been neatly woven into the mainstream equations.

The real epistemic problem with economics, at least as I see it, is that we’re just not very good at it. We can’t predict recessions, or growth rates, or really any variable of macroeconomic interest.* Mainstream economists come to completely contradictory conclusions on just about every controversial issue, from raising the minimum wage to the effects of illegal immigrants to cutting taxes.

Now imagine an equivalent situation in any other science. It would be like biologists disagreeing about Darwin, or physicists being unable to predict the movement of the planets. The very reductionist foundations of the science would remain in flux. This is the problem faced by economists: it’s tremendously difficult to know what paradigms are broken and what paradigms are still valid, because no paradigm is ever particularly successful, at least when it comes to explaining the real world. I’m not saying that economics is useless, or that modern economics hasn’t made a tremendous amount of progress. Far from it. It’s just that when every model is so bad at predicting things, it’s hard to distinguish between models, regardless of whether they are mainstream or heterodox.

*In March 2001 – the very month that the American economy began sliding into its last recession – The Economist reported that 95 percent of surveyed economists believed that there would not be a recession. The state-of-the-art models used by the Federal Reserve didn’t do any better. In February 2001, they collectively predicted that the American economy would steadily grow at a rate of 2.2 percent in the second quarter of and 3.3 percent in the third quarter. The actual figures were 0.3 percent and negative 0.4 percent.

(10)
More »

Comments

  1. #1 Ted
    May 30, 2007

    We can’t predict recessions, or growth rates, or really any variable of macroeconomic interest.* Mainstream economists come to completely contradictory conclusions on just about every controversial issue, from raising the minimum wage to the effects of illegal immigrants to cutting taxes.

    Now imagine an equivalent situation in any other science. It would be like biologists disagreeing about Darwin, or physicists being unable to predict the movement of the planets.

    I think economists rely too much on the x of human behavior in the equations — so much as to assume predictive behavior based on their definition of rational behavior. Physics and biology does not depend on the capriciousness factor the way our actions do. If you had to insert a random variable into physics or biology every time you ran the equation, the predictive value may be less.

    Social changes tear the formulas apart routinely because they impact our thought processes and activities. We think (and act) different today than we did in 1990, and that was just as different in the 1970s. Access to information is greater, and many of us have freedom to act in ways that are counter to the formulas.

  2. #2 Nic
    May 30, 2007

    “Now imagine an equivalent situation in any other science. It would be like biologists disagreeing about Darwin, or physicists being unable to predict the movement of the planets.”

    Or would it be like biologists unable to infer the development of an organism simply by looking at its genetic code, or physicists being unable to predict the weather more than a few days in advance? Or for that matter, neurologists being unable to predict mind behavior no matter how good their understanding of neurons may be?

    It’s a fatuous comparison to compare a complex and dynamic system like a macroeconomy to a simple and stable system like planetary motion.

  3. #3 jonah
    May 30, 2007

    Nic is right, of course, to point out that an economy is far more complex than planetary motion. My point was that economics still lacks the reductionist framework that defines every other science, not that an economy is a simple construct. Physics works because physicists have atomic theory, gravity, etc. Biology depends upon the grand synthesis of Darwinian evolution and modern genetics. Neuroscience has the neuron doctrine. Economics has…the rational-agent model? Without a reductionist foundation rooted in empirical experiments, it’s tough to know how to improve the mathematical models. David Laibson, an economist at Harvard, said it best: “Natural science has moved ahead by studying progressively smaller units. Physicists started out studying the stars, then they looked at objects, molecules, atoms, subatomic particles, and so on. My sense is that economics is going to follow the same path. Forty years ago, it was mainly about large-scale phenomena, like inflation and unemployment. I think the time has now come to go beyond the individual and look at the inputs to individual decision-making.”

    I hope Laibson is right, and economics does follow that path.

  4. #4 MattXIV
    May 30, 2007

    Nic choses some good examples of other phenomena that we have similar problems modeling.

    Like weather, macroeconomic phenomena can be studied scientifically and models with qualitative and modest quantitative predictive power can be established, but there are simply too many degrees of freedom to make precise quantitative predictions in most cases.

    Also, the minimum wage and immigration controveries are largely over whether certain very small effects are present – for example, on immigration the controversy is pretty much just over how much it impacts the wages of unskilled workers (and even then all the estmates fall in the 0-6% range if I remember correctly), while the effects on other workers and the economy as a whole aren’t very controversial. The technical disagreements are actually small compared to the heat generated because a small difference in the results would have a large impact on the policy enviroment.

  5. #5 Jonah
    May 30, 2007

    But just to be clear on the meteorology/economics analogy: meteorologists have trouble predicting specific weather events (like a thunderstorm or a hot spell) more than 10 days into the future. However, they are very good at predicting weather in the near future. If I remember correctly, it’s something like 90 percent accuracy for the five day forecast.

    What are economists good at predicting? Do they have an equivalent short-term forecast that they are 90 percent accurate at anticipating? I’m not asking macroeconomists to predict the precise rates of GDP growth or unemployment in the third quarter of 2007. I’d be satisfied if they could simply tell me whether or not there would be a recession, which is not exactly a minor economic phenomenon. If you can’t predict a recession while it’s happening (that’s what happened in March 2001), then I have to wonder about your predictive abilities.

  6. #6 Nic
    May 30, 2007

    The Laibson quotation sounds more or less right; I’ve heard intelligent economists compare the current state of economics to pre-Newtonian physics. There are some big and useful ideas in the field, but no grand synthesis to speak of. But then, it’s a young field.

    “What are economists good at predicting? Do they have an equivalent short-term forecast that they are 90 percent accurate at anticipating? I’m not asking macroeconomists to predict the precise rates of GDP growth or unemployment in the third quarter of 2007. I’d be satisfied if they could simply tell me whether or not there would be a recession, which is not exactly a minor economic phenomenon. If you can’t predict a recession while it’s happening (that’s what happened in March 2001), then I have to wonder about your predictive abilities.”

    I don’t think the business cycle will ever be predictable, or at least not without a grand synthesis, because such a large part of macroeconomic activity is recursively expectations-driven. People cause a depression by losing their jobs because nobody is buying goods because they’re hoarding cash because they hear there’s going to be a depression, and so forth. It seems unlikely that an economist could both predict a recession reliably and inform people that one was coming.

    Is there anything economists can predict with 90 percent accuracy? Nothing driven buy an explicit mathematical model comes to mind. Many forecasts can be built around financial futures, credit spreads, and the like, but in many ways that’s just a recognition at how much better a hundred self-interested traders are at such a task than a few PhD’s.

    Economics is trapped in a sense because its data is pretty much entirely from open systems, and empirical work consists largely of hunting for rich data sources that come acceptably close to imitating a controlled study in real life. (Econometricians might be the only people who’d be happy to see the draft lottery come back.) I believe economics is a worthy field of study with a lamentably narrow box of tools. Perhaps forecasting will improve with technological advance; after all, weather forecasting days in advance only became truly possible when gathering huge amounts of data over a large area became feasible.

  7. #7 JW Tan
    May 31, 2007

    What are economists good at predicting? Do they have an equivalent short-term forecast that they are 90 percent accurate at anticipating? I’m not asking macroeconomists to predict the precise rates of GDP growth or unemployment in the third quarter of 2007. I’d be satisfied if they could simply tell me whether or not there would be a recession, which is not exactly a minor economic phenomenon. If you can’t predict a recession while it’s happening (that’s what happened in March 2001), then I have to wonder about your predictive abilities.

    That’s quite a US-centric view. Given the large size and geographical spread of the US, it’s not easy to predict where it’s going at any point in time (although economists have had more success with the current slump), as different states move in different directions. Ditto the world, which is an agglomeration of several large blocs of economies.

    It’s easier to predict the behaviour of smaller, more homogeneous and less spread out economies. An extreme is Singapore – dead easy, you could go out into the street, ask 50 people some questions, and come back with some useful data. In the UK, the Bank of England uses a system of local agents who continuously survey individuals and businesses on expectations and current activity. It works reasonably well.

  8. #8 Ken
    May 31, 2007

    Macroeconomists may be bad at predicting; I don’t know. But microeconomists are sometimes good at predicting. Price goes up ; quantity goes down (and we can reasonably say how much). Besides price/quantity, there is hourly wage/hours worked, education/fertility, etc, etc.

    Macroeconomics is fundamentally different from the other natural phenomena described. There are not currently any agents who can take actions to influence the weather for material gain. Such is not true of the macroeconomy. All that is required for a prediction to be wrong is for a self-interested agent with market power to receive the prediction before the predicted event occurs.

    The author of the post apparently has little knowledge of academic economics. Kahnemann and Tversky have achieved little gain among top economics journals, nobel prize notwithstanding. Economists disagree about normative questions like the optimal minimum wage level because such judgments require a philosophical standard of goodness; physics, neuroscience, and meteorology do not investigate normative questions. The Economist survey is meaningless if they did not survey the right body of economists.

  9. #9 Drowned
    May 31, 2007

    I’m with Ken. Microeconomists are relatively good at prediciting events, Macroeconomists maybe not as good. But then we typically have greater knowledge of those things most likely to affect our results in microeconomic fields. In health economics for example, we have reasonably reliable data from clinical trials etc. and we put the appropriate caveats around our uncertainty about parameters in our results. The predictions we make are often, though not always of course, good ones.

    With Macroeconomic models the starting point is how to make a model small enough to be managaeble but large enough to be meaningful? They often get it wrong and not necessarily because the model was poor but because some individual agent had a large enough impact to change the results. Look into the tale of George Soros and black wednesday in the UK to see how one agent can effect things.

    Macroeconomic predictions are more akin to biologists trying to predict how a creature will evolve in future. You might be able to guess what the important influences are going to be but it will be difficult to say which ones will have the greatest effect at which time.

    The best predictor of a recession as far as I can tell is The Economists recession index. They basically just add up all of the times the word recession appears in certain publications like the Washington Post. The higher the figure, the more likely a recession is to occur.

  10. #10 Steph
    May 31, 2007

    I understand the Modern Synthesis is an intellectual framework, but I think you confuse a framework with predictive power. What exactly does the Modern Synthesis predict?