"Our greatest responsibility is to be good ancestors."

-Jonas Salk

Friday, June 6, 2008

More Fish, Same Barrel

What I have been calling "linear programming models" in economics apparently have a different name, "Computable General Equilibrium" models. These are the ones that extend the guns vs butter idea you saw in Econ 101 to multiple commodities.

Clark Williams-Derry on Grist points to Peter Dorman calling this whole enterprise into question on purely economic grounds. Never mind the meta stuff. The modeling approach is wrong, he claims. Nobody "with any skin in the game" as Williams-Derry puts it uses them. Yet academics write lots of papers about them and make lots of congressional testimony about them. Is there is any serious evidence of their utility?

On a theoretical level, it is surprising that CGE modeling has become such a vibrant industry, since its underpinnings in general equilibrium theory have been systematically undermined over the past several decades. (1) CGE models use the technique of representative agents—vast numbers of households and firms are treated as if they were a single decision-making entity—when we now know that multiple agents cannot be modeled as if they were just one. (2) In particular, the Debreu-Sonnenschein-Mantel result demonstrates that full knowledge of all supply and demand relationships in an economy is not sufficient to predict the equilibrium the economy will arrive at when it is not there yet. (3) The behavioral assumptions of these models, typically resting on utility maximization or simple modifications of it, have been empirically falsified. (4) Production and utility functions are routinely chosen for their convexity properties, despite the widespread recognition that nonconvexities (that yield multiple equilibria) are rife. In short, if theory should inform practice, we shouldn’t be doing CGE.

Now for the challenge. As far as I know, there has never been a rigorous ex post evaluation of CGE models in practice, one that compares predicted to actual outcomes. Based on performance, is there any evidence that such models add value—that their predictions are any better than those derived from macro or sector-specific models, or even a random walk? Also, are CGE models employed by any private sector players who bet real money on the results, or is it only in academia and the public sector that CGE modeling is taken seriously?
Emphasis added.

If mainstream economists are claiming that all decision making should pass through them shouldn't they make some effort to demonstrate that they know what they are talking about?

On the other hand maybe I should drop this pursuit. It may be too easy. Via Naked Capitalism an article called The Economist Has No Clothes, which in turn quotes an article of the same title in Scientific American by Robert Nadeau as follows:
But what is not widely known is that these now legendary economists—William Stanley Jevons, Léon Walras, Maria Edgeworth and Vilfredo Pareto—developed their theories by adapting equations from 19th-century physics that eventually became obsolete.
The strategy the economists used was as simple as it was absurd—they substituted economic variables for physical ones. Utility (a measure of economic well-being) took the place of energy; the sum of utility and expenditure replaced potential and kinetic energy. A number of well-known mathematicians and physicists told the economists that there was absolutely no basis for making these substitutions. But the economists ignored such criticisms and proceeded to claim that they had transformed their field of study into a rigorously mathematical scientific discipline.
These curious developments explain why the mathematical theories used by mainstream economists are predicated on the following unscientific assumptions:
  • The market system is a closed circular flow between production and consumption, with no inlets or outlets.
  • Natural resources exist in a domain that is separate and distinct from a closed market system, and the economic value of these resources can be determined only by the dynamics that operate within this system.
  • The costs of damage to the external natural environment by economic activities must be treated as costs that lie outside the closed market system or as costs that cannot be included in the pricing mechanisms that operate within the system.
  • The external resources of nature are largely inexhaustible, and those that are not can be replaced by other resources or by technologies that minimize the use of the exhaustible resources or that rely on other resources.
  • There are no biophysical limits to the growth of market systems.
This theory can no longer be regarded as useful even in pragmatic or utilitarian terms.
The comments are fascinating. While there are a few cogent criticisms, many readers claim that Nadeau "knows nothing about economics" yet eschew making any contrary factual statements. Seems familiar somehow.

And thence to yet another critique of economics, this one from a University of Chicago Ph.D. in economics, and an erudite and witty writer as well: The Secret Sins of Economics by Dierdre McCloskey, no less than a masterpiece of essay writing.

I had a lingering suspicion that conventional economics might be useful, but it's out of my system now, and am ready to start looking into alternatives.


Bunty said...

I tend to feel that economics has a certain similarity with astrology: there's certainly a lot of math and science involved, and they both often seem to give convincingly accurate results [not to say that it isn't genuinely accurate and valid in certain special cases and areas].

Unfortunately there's a bit in the middle where they use magic to go from the input to the output. Magic, or to put it another way: an internaly logical, coherent, and indeed thoroughly convincing (unless you take the time to look deeper, doing a Cartesian like thing of asking 'why' at every level on down), framework of theory and principle that just kinda happens to partly founded on some rather absurdly invalid premises.

All Production is Joint Production - A Thermodynamic Analysis (pdf - 120k) is a step in a better direction, takes some of the magic out of the equation.

ac said...

Is there any such thing as skill assessment for economic models - even in a purely academic setting. I'm asking out of ignorance: do these guys run hindcasts, say for periods in late 20C and then look at correlation, reliability etc?

Seems to me that the assumptions made (rationality, utility maximisation, limitless resources) can be attacked, but the more important question is whether these models really can predict the future better than a pack of tarot cards.

John Mashey said...

SO, a puzzling thing about Nadeau's article is that he whacked necoclassical economics and wished for something better ... but didn't reference folks like Charlie Hall or Bob Ayres & Benjamin Warr.
I sent Nadeau email on this, but have gotten no reply.

If you read 3 papers, I'd suggest:

Hall, et all, The need to reintegrate the natural sciences with economics.

Ayres & Warr, Accounting for Growth: the Role of Physical Work.

Ayres on economic growth and cheap oil.

And if you want more, I can point you at some.

This stuff makes more sense to me than Solow Residuals, Total factor Productivity, etc.

BTW: almost everybody's models [IPCC, Stern, NRDC, EPF, Nordhaus] for the costs of climate mitigation assume an indefinite economic growth over next 50-100 years fairly similar to that seen over the last 50. As to why that might possibly not be a good idea, see the last page of the Ayres reference, or go over to Econ Browser.

Dano said...

I have the lingering suspicion that conventional economics might be useful out of my system now, and am ready to start looking into alternatives.

Ahhh. Good for you.

After too many econ classes, in my mind I say any econ conclusion explains at best ~.25 of the variance. Is it worth all the fetishizing? Nah. But it's the best some can do to keep track.



Dano said...

I think the key in the comments is the bit about being 'predictive'.

I may have bored others here with this anecdote, but my UrbEcon prof at UW (not mt's UW) was a big name, and I made the poor guy spend way too much lecture time explaining to me why all the papers we read had such low r^s. We're talking typically .1 - .3 in the required readings.

Basically, his answer boiled down to: it's really hard to predict human behavior.

And that's it. And we know it. Yet we hide it behind the hocus-pocus. It's a social science.



Bunty said...

Dano: not so hard to predict entirely (assuming here it's groups of humans); if you are using the right tools, and if you aren't blinkered by preconceptual assumptions (if you know what I mean by that piece of grammar assassination).

A nice example of using the right tools is this : http://hussonet.free.fr/wealth.pdf

Just to continue the bombardment of this thread with PDF files, to even out the 3 I have to read for the one I posted :D

There's a more easily read overview here : http://www.austms.org.au/Jobs/Library4.html

Bunty said...

Actually, perhaps the old joke says it best:

Who says economists aren't accurate? They've succesfully predicted 13 of the last 5 recessions!


Michael Tobis said...

John, yes I've already been following your lead in that direction. I am concerned with their intellectual legacy to people who misunderstand the second law of thermodynamics, which is not a serious constraint at the economic scale.

I like Charlie Hall's stuff a lot; he seems to have the intuitive sense to steer clear of entropy arguments, though I'm not sure he understands the problem.

I am not so sure about Ayres & Warr; my BS detector is tingling a bit. What is this 'exergy' idea?

Bunty, yes, if economic modeling is possible it will arise from agent-based modeling. That is to say, not from totally untested equilibrium assumptions leading to convenient simplifications that suffer from the flaw of being completely wrong.

I don't think Bouchaud & Mezard are asking some of the questions we need to ask about economics. I think the methodology is sound except in that they fall into the trap of not proposing empirical tests.

I do think agent modeling will be part of a sensible economics properly embedded in a whole system view of the world.

On the other hand, it's my impression that the Faber et al paper in your first comment is completely unconvincing. Replacing a pseudoscience with a slightly more plausible-looking pseudoscience is not progress.

There is no serious second law constraint on human behavior on earth at least until such time as the sun goes out.

Michael Tobis said...

Here's an "economics" paper out of U of Chicago which uses the entropy concept, I believe, correctly, and which makes grandiose claims for itself about language, truth and decision making. I believe this is a classic "in a balloon" analysis; addressing the question correctly in a way that is totally inconsequential in practical situations.

I note that the author is not actually competing for the Netflix prize, although that is his key example.

However, his use of the information theoretic meaning of "entropy" is not confused with thermodynamic constraints. Hurray for that at least.

John Mashey said...

1) Ayres and Warr are quite serious. I've reviewed a (forthcoming book of theirs, "Energy and work as drivers of economic growth"). They didn't invent the term exergy, which goes way back - see Wikipedia on Exergy.

I'd summarize what they're doing as an attribution analysis for economic growth, specifically, to try to turn the Solow residual into measurable components.

2) Put another way, while energy is neither created nor destroyed, exergy (of a fuel) is more or less the heat of combustion (or enthalpy). Exergy actually gets used up.

3) Their models emphasize work = exergy * efficiency, i.e., one can get twice as much work done by doubling efficiency.

(more later got to run).

Anonymous said...

Fascinating. Read many of the links.

By some cosmic coincidence, I just happened to read "economists Have No Clothes" in Scientific American at lunch today.

By the way, the same issue has an interesting article on the neuroscience basis for 'trust'. That is surely somehow related in an account of a society and its natural economy.

David B. Benson said...

Related to all of this is the thesis of "Into The Cool", which explains to role of life in reducing something called 'free energy'. Irrespective of how closely related to 'neo-economics' it is, you'll faind it enlighteningly different.

Anonymous said...


Interesting site you've got here. You are also extremely good at pulling out gems from elsewhere on the internet.

As you said, Dierdrie McCloskey is an essayist par excellence, although I suspect her idea that for a falling body the mass of it makes any difference to it's acceleration might come as a surprise. Unless I completely flunked physics at school, if you were to, for instance, drop two dissimilar weighted objects from the top of the Leaning Tower of Pizza, as Gallileo probably only did as a thought expirament. they'd both hit the dirt at roughly the same time. See here:


Spoilt an otherwise superb article for me, so it did.

Michael Tobis said...

I was reading quickly, else it would have spoiled it for me too.

I am not entirely uanaware of "g"; actually monkeying with the value of "g" is a key thought experiment in my PhD thesis.

ac said...

Michael, what do you mean by:

"There is no serious second law constraint on human behavior on earth at least until such time as the sun goes out."?

Total macroscopic entropy production still has to be positve - having access to a big heat reservoir in the sky doesn't change this.

If we want our energy supplies to consist of mostly solar energy (whether that's photovoltaic, csp or bio), then I can't see how you can sidestep the second law to compute the upper limit on the work you can extract. Or are you saying that having done this calculation, the amount of usable energy is so large as to be

I agree about agent based modelling: it's easy enough to get from atoms to thermodynamics by making assumptions about homogeneity, indistinguishability and so on - I've never found making the same assumptions about human 'atoms' very convincing. In the same way that doing molecular dynamics allows you to avoid worrying about the stat-mech, economic simulations at the agent level could avoid the classical approximations.

Say, surely someone is already working on this kind of approach?

Michael Tobis said...

ac, what I mean first of all is that people confuse information entropy with thermodynamic entropy, and secondly that they neglect the fact that the earth is an open system.

There are plenty of real constraints on our behavior that are far more serious concerns than the second law. This isn't to say that the effective energy content of fossil fuels isn't declining; that is a real issue and one that is related to thermodynamics.

The fact is that the amount of sunlight hitting the earth is both of very high quality and enormously abundant compared to our present energy usage. It is the wrong place to look for a quantitative theory of managing the earth.

ac said...

I agree that there's a lot of sloppy writing out there in attempts to integrate economics and thermodynamics, but I don't think that just avoiding arguments related to entropy is the answer.

You may be right that the upper limit is so large as to be irrelevant, but I need to see some numbers to be convinced. Hopefully I'll get a chance to do the calcuation before this thread dies - of particular interest is how much work I could do with nothing but my backyard.

Anonymous said...

McCloskey is indeed brilliant.

I find it interesting when people cherry pick what suits their interests. I've mentioned before that people may utter advocation of some particular cause completely independant of what their own foundational beliefs indicate.

I'm reminded of Mike Munger's story of a college faculty meeting where the discussion turned to cars. Every teacher around the table said that they drove a Prius. When he asked one of the science teachers if he really thought the Prius had a positive effect, his answer was "No, but I'm faculty at this university. I have to drive a Prius".

McClosky has criticisms of economics, but she is also in the field and criticizes some economists from that internal awareness. I have no doubt scientists in hard-science fields have similar internal debates. This is marginal criticism from people who believe the great majority of their own fields cannon.

If you think that McClosky is against economics, or has little faith in it, or that it is meaningless, you could certainly simply ask her opinion.

Having heard her talk about economics, I can say with a high degree of certainty that she would definately be pro-trade, libertarian, capitalist, free market economics.

Michael Tobis said...

I don't doubt you about McCloskey's market/libertarian instincts. Indeed there is a conservative streak associated with that flavor of wit. I don't have to buy the whole package to appreciate the effort. Also, in some sense I consider myself very conservative, though most other who consider themselves conservative disagree.

Nevertheless, she has expressed some pretty substantial doubts about the rigor of most of what passes for economics.

I don't claim she agrees with me and others like me about how badly and why economics misses the point about the next two centuries.

She doesn't paint a picture, though, of a field that has things so thoroughly figured out that we must defer to them on every maior question, masters of decision-making that they are.

I am interested. Do you think she believes that we should do that?

ac said...

As promised I've somehow found the time, and the envelope, for this rough calculation.

Effective temperature of sunlight is 5700K (or so). Temperature of my shed roof is around 300K. The second law efficiency limit (1 - Tc/Th) is 95%. If I fit the latest model Sadi-matic Solar Engine to my shed, with a surface area of 30m^2, at around 38 degrees S, where average insolation is 200 W/m^2, I can generate 6 KW. That's a not insignificant amount of power.

So yeah, (barring an error in my working) the second law by itself is a pretty insignificant constraint on achieving steady-state solar paradise.

Anonymous said...

the second law constraint that "matters" is the one about "matter"... our economic activities depend on a finite supply of high-grade (aka low entropy) resources (ores, etc.)... our economic activity is fundametally a process of physical transformation, work which is done using energy... the process necessarily generates a net increase in entropy... Both in terms of depletion of low-entropy resources and insertion of high-entropy wastes into the ecosystem... It doesn't matter whether there is additional energy supplied to the system. The matter within the system is finite, and we are *eventually* constrained by the second law.

The claim is not "Hey, our economic activities are constrained by the eventual heat death of the universe!" Whether the second law choke point is going to be from sources or sinks is open to debate. When we run into constraints is an open debate. That economic activity operates under the second law is not.

In any event, it is a mistake to ignore ecological economics based on too-literal implications of entropy. Personally, I think conceptually it is a good organizing principle, although I realize that part of the issue here is about "bad assumptions, principles" in another discipline... So I understand the antipathy...

Michael Tobis said...

My problem is that there are two perfectly good formal definitions of "entropy", one subject to the second law and one not. They are comparable in that their is something fundamentally logarithmic about them but they are not the same.

If people want to use a technical term they are required to make sure it is clear which meaning they are using.

Now ask me about climate and "feedback".

Yes, there is a problem, and yes there is a very real EROEI issue at its core.

I don't think people who don;t understand the issue should be doing long-term economic analysis. Tangling this issue with entropy is possible but not advisable. Let's just talk about the energy net value of the remianing fossil resources.

I must admit that until the last few months I didn't take the peak oil thing seriously. I thought, running out of fossil fuels, good, that will protect the atmosphere. I see now that barring very creative and insightful planning it tends to make our problems all the worse, as we play around with decreasingly efficient carbon resources.

Perhaps we need a EROCE measure: energy return on carbon emitted.

I think calling using entropy language for this class of problem without a powerful and mathematically rigorous motivation is a bad idea in terms of clarity of communication, especially since the concept is so confused already.

This book is largely at fault.

ac said...

It totally matters that there is new energy coming into the system. Earth is more or less a closed system, so the second law is fine with our entropy decreasing as long as it's balanced by the entropy increase in the environment, of which we can ignore everything except the sun, which is producing an insane amount of entropy as it smears its energy across the solar system.

Put another way - we can transform matter into increasingly 'ordered'/low entropy states as long as we have available energy. The high entropy waste doesn't have to be anything nastier than heat.

Any quantitative integration of economic with physical principles is going to deal with entropy, (even if you studiously avoid it by using quantities like Gibbs' free energy it's still there), but I think the point is that most entropy production / inefficiency that's relevant to humans is not second law, it's due to dissipative processes like friction, having to dig oil up, solar panels fouling with algae, and so on.

Anonymous said...

No, I don't think she would say that (we should leave all decision making to economists). Nor would I for that matter.

What I would say is that economists know a few things about human beings and how they behave, and the results of certain patterns that have been repeated throughout history.

I think there's something worth learning there. Of course, it is a science of human actors, which is not the same as a science of discrete molecules. Sociology and Psychology are sciences also, though they rarely use a test-tube.

The past couple of days I've had a good laugh. I've been listening to a lot of econ podcasts and there have been many phrases that are essentially "People tend to think economics is about (insert some MT statement here)..."

The economists that I consider worth their salt talk about much more than GDP, rarely about money, occasionally about parenting and education, how to think in general, not-infrequently about sports, and a great deal about who is still poor in the world and why.

I don't really know how to deal with an insistence in fighting straw men economists. All I can say is, take a walk on the wild side. You might learn something- if you don't start with the assumption that it inherently threatens you or your work.

When advancing as if under threat of attack- we rarely learn anything new.


New topic: I'm almost finishing up James Soruwiecki's book "The Wisdom of Crowds". It's a fascinating account of how aggregated group decisions or even guesses are more likely right than even experts. Too many examples to try to summarize here- but decision markets, a semi-new thing have been better predictors of elections, succesful pharmaceutical products, best selling lines of printers, and all kinds of other information.

If Surowiecki is right (and this is my application, not his) it seems as if we have a decision market of 6 billion people. While some of them say they are concerned about AGW, the actual outcome of the aggregated decision is that we're not really that concerned.

Is the biggest crowd possible very wrong? (I will confess I have the intellectual's prejudice to doubt the hypothesis, but the book is very convincing).

Michael Tobis said...

Steven, I will appreciate if you stay on topic in future rather than free associating. Even if your associations are interesting.

This is my blog. You write enough here to start your own. Effective immediately if you don't stay on topic you won't get posted, because I don't want this to become Steven's blog. If it helps, I'll promise to read for a while and link occasionally.

Regarding your new topic:

The question of who is an expert on what is key. I think finding examples of people who call themselves experts who were wrong proves very little without a very difficult formal analysis. A few anecdotes are essentially meaningless.

So maybe "experts" shipped us "New Coke". I don't include those people in the class of expertise that matters.

Sefl declared experts also advised Al Gore on his campaign strategy (much to everyone's cost) and GW Bush on his military and economic strategy (ditto).

It tells us little to say after the fact that those weren't the real experts. Detecting the real experts in advance is the problem.

I think all we see is a lack of expertise in picking experts among politicians. Ross Perot's whole campaign was based on his claim that he had a good track record of identifying experts. Whether that was true of him is one question, but I have always appreciated his fortrightness about the fact that this is really the key skill in administration.

Crowd wisdom does fail. Easter Island as described in Jared Diamond's Collapse remains a spectacularly interesting example. I think it is fair to say that crowds are more likely to be right when evaluating 1) their own preferences and 2) circumstances easily extrapolated from those they are familiar with. Whether, as in the cartoon, a few genuine experts understood the crowd madness in time to warn but not in time to be heeded (as in the cartoon) is not known, but the presence of a very badly wrong public is clear.

For another example, I would not subject spacecraft navigation contracting decisions to a popular vote. Would you?

Such a generalization is surely overbroad and vapid.

Anonymous said...

My apologies for getting off-track. I'll try to check in occasionally and make a pointer in a different direction if I think alternate opinion isn't being represented.

Regarding spacecraft. That's an interesting example, Surowiecki uses the Columbia reentry explosion as an example of the criteria that must be met for groups to provide otimal outcomes.

The NASA taskforce (studying the damage and preparing a plan for reentry) had some serious flaws. To hyper-summarize, dissent was not encouraged- it was actively discouraged. People died.