Steve Chapman on macro

"The problem is, like meteorology, you can't forecast accurately too far into the future, no matter how good the model fits the past."

Again, a model that fits the data but does not survive time is simply an observation.

I'm not discounting the value of making observations -- I'm simply noting that it's very dangerous to draw conclusions from observations that are not static or predictable. I value information, like everybody else -- and I use it to make decisions, like everybody else -- but I use it in a very limited way that advances my personal pursuits. I don't draw any conclusions about how what I do is going to affect the whole.

Tell me what valid conclusions I can draw from macro that are not a part of micro knowledge. Tell me what the science has added to our knowledge other than the fact that it can make predictions carried out to the fourth decimal place that invariably miss the target by integers.

Sheetwise -

Re: "I've yet to hear a good refutation of that charge that included a documented comparison of macroeconomic prediction over astrological prediction."

Are you serious? Macro models fit the data far more accurately than even micro models. This is very well known. Part of this is because the recursivity of macro models substantially improves model fit, but I think it's also likely due to the fact that our macro models are better specified than our micro models. The problem is, like meteorology, you can't forecast accurately too far into the future, no matter how good the model fits the past. I think that's largely due to two things - first, the Lucas critique about replicating past conditions. But I think a lot of the problem with projections is that that's the nature of the limitations of a complex system (again like meteorology). Nobody demands the news channels fire their meteorologists because they can only give us a five day forecast.

I would say that macro is to micro what meteorology is to thermodynamics, or what population biology is to reproductive anatomy. There are different limitations, but that doesn't make it a charade.

"I actually don't think there's a difference between micro and macro evidence in that approximations of controlled experiments are very rare."

Here I claim an exemption -- (If given, received. If not, not received)

"micro and macro evidence in that approximations of controlled experiments are very rare."

Cewrtainly you want to qualify that.

As has been said many times before --

Macroeconomics is to economics what astrology is to astronomy.

I've yet to hear a good refutation of that charge that included a documented comparison of macroeconomic prediction over astrological prediction.

In either case, it would be entertaining to review the math.

Russ -- this makes macro exactly comparable to Darwinian explanation.

You aren't a creationist are you?

Darwinian biology is perhaps the most understanding-altering explanatory machinary every devised by man -- i.e. it is awesome and powerful science.

As I've said. What you are working with here is a false picture of "science" -- you aren't helping us understand the potential scientific standing of macro (esp. macro built from the ground up using sound micro that "science" really doesn't touch, not beyond the mere edges).

>>Russell Roberts, a scholar at George Mason University and the Hoover Institution, told me, "I think some of macro is useful -- what causes inflation, for example. I just don't believe we're very good at testing theories at the macro level in a convincing way." Many times, he says, all macroeconomists can do is concoct stories that explain events after the fact.<<

Russ,

From the article:

"So why should we listen to what macroeconomists say? Some economists think that often, we shouldn't. Russell Roberts, a scholar at George Mason University and the Hoover Institution, told me, "I think some of macro is useful -- what causes inflation, for example. I just don't believe we're very good at testing theories at the macro level in a convincing way." Many times, he says, all macroeconomists can do is concoct stories that explain events after the fact.

Microeconomists, by contrast, make predictions about the effect of policies on individual markets, and those predictions are easier to confirm or refute. They virtually all agree, for example, that price controls lead to shortages, and governments occasionally take the trouble to prove them right."

I am curious, in the 2nd paragraph you mention individual markets. Just how indidivual are you thinking?

Methinks -

Good point on the "depression within a depression" and indeed, that has been used as a mid-stream test by a lot of people. But even a sample size of 2 is hard to convince somebody on :)

John -

In what way has Krugman HELPED Obama? I think Obama probably wishes Krugman would go away! And how does Krugman always find his way into these articles? I was reading a blog post by Posner today too and in the last paragraph he randomly started talking about Krugman too - what's the deal? :)

Methinks, you said:

I recently made a similar argument and I was shut down Pretty quickly. Politicians need to act. Macro-economic models, complete with all of their biases, provide justification for top down action. I can't really come up with reasonable positive argument against this.

This is sadly the case. No one running for high office would be taken seriously if he/she trumpeted an economic policy by first highlighting all of the things that we do not know. It would be honest. It would be true. It would be laughed at.

What does that tell you about the usefulness in such work for designing good economic policy?

Depends on what you mean by "good".

If "good" means consolidating and accumulating power, then I think Krugman is doing a fine job for Obama.

This piece immediately made me think of OTC trades. Since there is no market or a very thin market, the book is marked based on a set of assumptions. Thus both traders on either side of a trade can book a gain. One is correct, but good luck sorting out which bias is less...well...biased. The longer the duration, the bigger the error because there's less certainty around the underlying assumptions.

As Ackerman said: everything looks like a nail to a person with hammer.

I just wish the search for truth in macro was focused on revealing all of the things we don't know, and may never be able to know -- leading us away from top-down policy decisions instead of toward them.

I recently made a similar argument and I was shut down Pretty quickly. Politicians need to act. Macro-economic models, complete with all of their biases, provide justification for top down action. I can't really come up with reasonable positive argument against this. Also, I don't think it matters how static or dynamic models are. They are always subject to bias, as is the interpretation of outcomes.

Dan,

The Great Depression was at least two depressions. There was a depresseion within a depression in 1938.

Bottom line, it seems that macroeconomists do not understand very well how the macro economy works in the short run. This is reflected in the lack of consensus in macro as evidenced by competing schools of thought--Krugman's Keynesian beliefs vs. Barro's real busines cycle view vs. monetarism.

Sam -

Well, singular in the sense that it was one recession. I'm sure it was a perfect storm of causes. But as Chapman and Russ point out, it's very hard to justify counterfactuals when you only have one recession. However, I think it becomes easier when you're comparing lots of different recessions spanning a century and several countries. That makes the counterfactual game significantly more plausible - and then you can take those findings to singular cases to inform them.

"I guess what I'm saying is that I'm not aware of an assumption that models work over time. There are allowances for shifts in parameters over time, or even for whole paradigm shifts in the way the economy operates over time"

If a model doesn't survive time, it can't be used to make predictions -- even if it does provide a pretty good snapshot. If there are no allowances for shifts in parameters over time, then it's static, and will make great predictions (if it's accurate). If the parameters that will shift, and the degree of change can't be accurately predicted, then it will produce silly results (or whatever results the modeler is searching for).

I just wish the search for truth in macro was focused on revealing all of the things we don't know, and may never be able to know -- leading us away from top-down policy decisions instead of toward them.

It is hard to answer questions about singular events - like the Great Depression

Was the Great Depression really a singular event? Or was it a cacophony of events?

"Russ - do you have any sense of what macroeconomic methods he has reservations about?"-DK

I thought Chapman was clear. On opposing views of stimulus spending: "Both can't be right. But neither has to fear being proved wrong. Because there is no persuasive way of determining the effect of implementing the president's policies. No matter what happens, each side can claim vindication."

LowcountryJoe, I think you are exactly right. Everything looks like a nail to a person with a hammer.

I actually don't think there's a difference between micro and macro evidence in that approximations of controlled experiments are very rare.

Sheetwise -

I guess what I'm saying is that I'm not aware of an assumption that models work over time. There are allowances for shifts in parameters over time, or even for whole paradigm shifts in the way the economy operates over time (ie - think of a Harris-Todaro growth model that specifically incorporates the transition from a "traditional" to a "modern" institutional setting).

It is hard to answer questions about singular events - like the Great Depression - but the methods used to estimate impact of fiscal policy or monetary policy on GDP across many countries and across huge periods of time (ie - not explaining a singular event like the GD) seem to me to be exactly the same as the microeconomic methods that Chapman has no problem with. There are additional autocorrelation sorts of concerns with macro that aren't there in micro, but I guess I'm just saying Chapman is claiming there's this difference but the only evidence he presents seems to be the uncertainty of forecasting - and I'm not sure that that's unacknowledged, and other than that I'm not sure what the difference is.

In public policy debates, saying that you do not know is simply political suicide

Macroeconomics in politics

"That forecasts have error I think is a well known criticism - but is he concerned about a specific type of bias in the methods?"

I think the bias is in the macro belief that models work over time. What's the value of a model with no bias that doesn't work?

Daniel,

I think he's thinking about the challenge of distinguishing cause and effect in a complex system. What ended the Great Depression? Good monetary policy? Good fiscal policy? The war? The return to normalcy? Natural economic forces? No one knows. Not only does no one know, no once can provide evidence that convinces someone on the other side who favors a different explanation. What does that tell you about the usefulness in such work for designing good economic policy?

Russ - do you have any sense of what macroeconomic methods he has reservations about? That forecasts have error I think is a well known criticism - but is he concerned about a specific type of bias in the methods?

Paraphrasing Chapman [because I don't know how to get my phone to copy & paste], "Why would you not talk to macroeconomists about the broad economy if that's what you want to be informed about; afterall, they spent their lives thinking and writing about it?"

That's a good question but here's something that should be considered: how much more likely is it that someone who spends all their time thinking about the macro economy will be using normative thinking? And if I am correct, wouldn't the normative thinkers be more predisposed to being sympathetic to central planning thinking rather than decentralized emergent order think? Any thoughts?

Hit the nail right on the head there