July 2014

Sun Mon Tue Wed Thu Fri Sat
    1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30 31    
Blog powered by Typepad

« Acemoglu in Esquire Magazine | Main | A Contested Exchange, or First Round Knock Out in an Intellectual Boxing Contest? »

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83451eb0069e201348765bfe8970c

Listed below are links to weblogs that reference Read and React -- Colander's Testimony on Capitol Hill About Economics:

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

This is a version of the point I've been making for 15 years, and first presented at an HES meeting in 1996.

"First, it failed society because it over-researched a particular version of the dynamic stochastic general equilibrium (DSGE) model that happened to have a tractable formal solution."

The marginal valuation logic of heterogeneous production coordination across time is mathematically untractable -- and the central / explanatory element in economics (entrepreneurial learning in the context of changing prices) is not capturable in any nomological law or statistical formuka or bit of math or formal logic, etc.

I.e. the core of an economy is formally and numerically untractable and not subject to "scientific" measurement and/or "scientific" math function -- the core involves open ended / non-natural kind causal components that operate over an open ended and ever changing set of physically and "law" defined instantiations (compare Hayek, 1952).

For more:

http://www.hayekcenter.org/friedrichhayek/hayekmyth.htm

Colander writes,

"[economics] failed society because it over-researched a particular version of the dynamic stochastic general equilibrium (DSGE) model that happened to have a tractable formal solution."

This Colander piece is just terrific.

I think he's absolutely nailed it.

My point above is simply that you can prove the the central valuation problem in economics is intractable and you can prove that the central causal / explanatory element in economics (just like those in Darwinian biology) cannot be made into a variable that is specified by measurement and can be plugged into a math function that cranks out law-like or stastistical laws, subject to inductive verification.

Indeed, you can show historically how economics has pursued a bogus scientistic conception of "economic science" -- and you can prove by logic and emprical demonstration that the whole effort is a logically and empirically necessary FAIL.

Both Hayek and Mises have made this case -- in a variety of ways.

Advances in mathematical economics and in philosophy have given us the tools to make he same arguments in more sophisticated form.

Following Colander, I believe we would get that if the funding and career incentives existed to encourage this.

Teh takeaway messages: 1) NSF funding corrupts scientific research by rewarding only research that produces the answers government wants, and 2) the economy is a complex system that requires complex system models to understand it. Of course, the NSF isn't going to fund systems work because if one sees the economy as a complex system, you cannot come up with activist policy recommendations, as the nature of complex systems is that one ca never predict the outcome of interventions. Best, then, to leave it to its own devices.

I do find it interesting, though, that after he goes on about the problems of too-simple mathematical models and political interference in research, he then goes on to recommend that "Along with economists on these reviewer panels for economic proposals one might include physicists, mathematicians, statisticians, and individuals with business and governmental real world experience." Isn't the problem with economics as it is now practiced that it is too mathematical, that it too often tries to emulate the simplicity of physics, and that it is too often infected by the concerns of government? A better set of reviewers: anyone involved in complexity or process theories, biologists, anthropolgists, sociologists, psychologists, philosophers, etc. who study complex systems and processes themselves. The first set of theorists I mention will of course include mathematicians and physicists -- and those who study complex systems will actually be able to make good judgments in the field.

It's a very good post - there are parts to agree and disagree with.

One of my biggest concerns is that it's exactly when you see people moving away from the math and the econometrics that you see them making bad mistakes about properly identification, mixing up correlation and causality, etc.

And he doesn't always seem to say "do less math" -- he specifically talks about how we have the mathematical tools to investigate non-linearities, multiple equilibria, and forward looking agents (I thought we already did these things anyway! We certainly worked with them in my macro courses).

So its not clear exactly what the role of math is - but the point is you have to be innovative and think about what your math can't deal with - and that point is very well taken.

I think the problem that a lot of people don't fess up to is that trying to answer a question without math can potentially be riddled with even more problems. Not always, but it certainly can.

For example, in this Mark Thoma post discussing a DSGE model it has three of those points he raised - non-linearities, multiple equilibria, and forward looking agents. I'm not defending DSGEs as the way forward or anything, but I think people vastly overstate issues with them. And this model that Thoma presents is CLEARLY selected because it fits the problem we're dealing with, not because its the kind of math George Evans happens to know.

It's the use of math -- especially the wrong kinds of math (nonlinear, complexity, etc.) -- that has made economics the basket case of a social science it is today. It oversimplifies and cannot actually deal with social reality in its full complexity. Too often those who use math mistake it for reality rather than understanding it to be a precise approximation of reality -- and often a gross oversimplification. Which is fine for modeling a few simple processes, but utterly useless for understanding the true complexities behind human action.

Is it actually true that economists failed to predict the recent financial crises? I seem to recall hearing lot's of warnings from about 2006 onward.

And I, myself started reducing my personal debt and buying gold as early as 1st qtr 2007 because of those warnings.

I'll give him this, in as far that some economists were deep in DSGE theory, maybe they failed to see the obvious warning signs of a classic high credit- easy money bubble..

The first thing that struck me about this Colander testimony is the view of economics expressed through the lamppost joke. I have no problem with the view that the social world is extremely complex and therefore that it is more difficult to find and corroborate eternal truths and causal relationships than in e.g. physics. But what I understand from Colander's use of the joke is that economics seems to be a thoroughly and fundamentally inductive science. After all, to Colander it seems there is no reason for the economist to search in the light of the lamppost other than that it is easier to "see" there. The fact that he knows not at all where the keys are or where they could be clearly indicates that there is no theoretical framework that guides his research.

I didn't read further than the first section. Frankly, the initial discussion made me a little sick to the stomach. Maybe he elaborates on more interesting and important points later in the text. What I got out of reading the first part was that this means we've finally come to the point where economics has given up on its deductive roots (quite predictably, while undesirable). So let's just go out there and collect data blindly and then try to make sense of it all.

Why we would need to search in the light rather than stumble in the dark escapes me, however.

Colander is arguing for a wiser, more thoughtful direction for NSF funding. That's like arguing for wiser, more thoughtful war efforts.

Colander strategically avoids discussing the unwillingness of most economists to search under the truly beaming light: the adverse role of government on the economy under state capitalism. Of course, to do so would have jeopardized his own fund application and upset those congressmen that he was addressing with evidently simple jokes designed for evidently simple minds. And that is why most non-public choice economists steer clear and end up with mostly irrelevant economic analysis.

Charles -
Have public choice theorists been turned down for NSF grants? It seems like a lot of it would be interesting to them. Do you know if any public choice theorists have bothered applying?

I imagine Austrians would have it harder because the NSF economists on staff - for better or worse - might have concerns about their methodological approach. But it's not immediately obvious why the NSF wouldn't be interested in the other two legs of GMU econ: public choice and experimental.

Arnold Kling made the same point recently, citing Zandi as someone who gets government support because he tells the government what they want to hear. Here's the thing - Barro got NSF money for telling the government what it (purportedly) DIDN'T want to hear and Zandi - the one Kling cites - as far as I can tell didn't get any government money for producing his estimates.

For methodological reasons, yes - Austrians might get shut out. Aside from that I'm wondering if people are overstating the bias. What is the rejection rate for public choice economists vs. others? Do you know?

* And I should be careful about even calling the Austrian thing a "bias". NSF would probably say "we do not think that has scientific rigor". Peter and others may think that's a wrong assessment, but that doesn't make it "bias". I want to be careful about what I accuse NSF economists of, because they do a lot of good work.

Might be of interest..

Deficiencies of Austrian Economic thought:

http://conversationwithcrombette.blogspot.com/2010/09/deficiences-of-austrian-economic.html

It's worth noting that Philip Mirowski lays out how money from the military and the NSF played a significant role in funneling economic "science" into a singular concern with "economics under the lamp post".

See Philip Mirowski, _Machine Dreams_.

Colander gets the problem partially correct, but as others have noted he completely misses the problem of limiting macro to inductive reasoning.

I was taught in statistics that you always formulate theory first and then test the theory on the data. Otherwise, you are data mining, which statisticians hate because you come up with all kinds of spurious correlations.

Another problem is that mainstream modelers do in fact use deductive reasoning; they simply deny that they're doing it. Their models are based on Keynesian or neo-classical reasoning, which is the reason their models don't fit with reality. If the models were based on Austrian theory, they would fit well.

But Colander's solution, more empirical data and inductive reasoning is worse than models based on Keynesian theory. Keynes wasn't a complete idiot and he got some things right. But trying harder to derive theory from empirical data will do nothing but cause even greater confusion and multiplication of theories.

I'm exhibiting a bit of hubris here, but indulge me please. I think the way forward is for Austrians to become more interested in math, but not DSGE. Use structural equation modeling, SEM. It is designed to test competing theories and does well in other social sciences. I use it in survey analysis. You collect data, present your models and SEM will tell you which model fits the data better.

Pete,

I guess you liked David's testimony since you gave it so much space. I'd agree with that judgment too. Notice that he does not question the role of the NSF as an oligopolistic supplier of research support, which reflects his purpose in testifying. But in some sense the "real" problem is, once again, institutions. Given the NSF, however, his suggestions seem worth a try. It's just that it's hard to see where they're really going make an enduring difference, particularly given the NSF's identity of supporting "transformative" basic research.

Part of the problem is that you have better chances of getting funding if you promise to predict the economy rather then identify limits to prediction. A few years ago, Robert Axtell got a big NSF complexity grant with the goal of creating “an artificial economy-that can be calibrated to real economies of today.” The hope is “to produce macroeconomic models that have high performance at predicting the near term unfolding of an economy, thus providing more accurate forecasts of economic evolution.” Admittedly, he does not aim past the “near term.” Still, he is saying complexity makes us smarter, not wiser. Sigh. The Austrian view has always been that complexity makes us wiser, not smarter. I don’t see where the NSF is likely ever to start preferring wisdom to smarts.

I happen to agree with Colander pretty much fully, but then I have been a frequent coauthor and co-conspirator with him, so this should not surprise anybody particularly.

fundamentalist,

I think you are misreading Colander. While he says "more induction compared to DSGE" he is not saying "induction only, no deductive theory." That is a misinterpretation. As he notes, the hard core DSGE people, such as Chari, Kehoe, and McGrattan, who are just astoundingly arrogant, as are their followers (see the lunatic nasties populating EJMR), admit that their models do not "fit the data," but it does not deflect them the least from doing the same old same old and shooting off their mouths to all and sundry about how they have "the answer" and are the only people around doing "serious macro." In light of what has gone on recently, this is truly appalling, and Colander is absoluttely right to nail them on this.

He is for more agent-based modeling and other approaches, with expanded views of theoretical underpinnings. Oh, and SEM certainly has its limits.

Barkley, what method has no limits? Do you know of a better method for comparing theories?

The debate between time-series methods, including Bayesian ones, and structural equations modeling methods, is an old and ongoing one.

To continue the joke, an entrepreneur walks up and offers to sell the economist a flashlight and a map, and to rent to him a metal detector. The economist replies, "so sorry, I don't have funding for that".

The section about applied research is important. Models, if they are to be any good, need to be validated with historical data. Otherwise they are worthless. Econometrics, if it is worthy, does a good job of predicting the history. Going forward into the future, it may provide some predictive power just as an umbrella merchant checks the weather forecast for the chance of rain. Marketing people do this all the time.

Forecasting into the future requires confronting uncertainty and not just risk. True uncertainty necessarily includes the possibility of Nicholas Taleb's "Black Swan" events. So, DSGE models, to be more credible, need to define the limits of their predictive scope. No DSGE model should be expected to forecast the full range of possible events in an uncertain world; it's impossible. However, Austrian theory can enlighten economists to these neoclassical blindspots.

Regarding peer review, the addition of 100 randomly selected taxpayers would add a healthy dose of common sense.

Last, let's not forget the approach of Sir John Cowperthwaite, finance minister of Hong Kong in the 1960's. He purposely refused to collect data in order to deprive the British "policymakers" the opportunity to wreck the economic progress in Hong Kong.

Also, I just finished reading "Raising the Colour Bar" by South African economist William Hutt written in 1964. Unless constrainted by iron-clad constitutional limits, policymakers will always end up favoring sectional interests to the detriment of the general interest. The same applies to the NSF. It would be better if economic research was privately funded.

The bogus choice between "induction" and "deduction" is PATHOLOGICAL -- this whole picture is part of the scientism which has led economics into fail.

The induction/deduction picture is a false picture of science.

Example -- you can't fit Darwinian biology into this picture.

In fact, you can't fit ANY science into this picture (see the work of Thomas Kuhn or Norwood Hanson or Karl Popper or F. A. Hayek).

The induction/deduction picture is the false picture of knowledge and science which economists have inherited from a failed tradition among the philosophers.

It's time the economists dumped the the failed project of the philosophers -- and it's time they picked up the successful strategy for doing science exemplified by Darwianian bioloyg, i.e. figure out what the empirical / causal problem is and focus on the empirical / causal mechanism that provides an explanation for that problem.

Barkley writes,

"He is for more agent-based modeling"

Big problem. We can prove that this is just more buzzing around the lamp post, like a mindless moth.

Why?

Because the central causal / explanatory element in economics CANNOT BE MODELED -- entrepreneurial learning / judgment in the context of changing prices and local conditions takes place outside of any nomological or statistical regularity, beyond any measurable physical / natural kinds, outside of the reach of pseudo-scientific formalism or statistical construct.

And then there is the whole problem of applying the logic of marginal valuation to heterogeneous production goods across time and with technological change ....

Barkley, Are you sure you know what SEM is? It's not the same as the structural models used in econ for ages, like the Fair model at Yale, although there are some similarities. SEM is a unique technique that incorporates many other techniques. It's part factor analysis and part linear regression, but it is based on a comparison of the actual vs the modeled covariances. And is solves equations through iterative algorithms, such as the quasi-Newton. It's the use of the covariance matrices that makes comparison of models possible and accurate.

Greg: "The induction/deduction picture is a false picture of science."

This seems to violate Hayek in "Counter-Revolution" and his Nobel speech, as well as Mises' first section in Human Action, unless I misunderstand you.

Re: Daniel Kuehn's "forward-looking agents," i.e. entrepreneurs, Mises says in _Human Action_ that entrepreneurs do not look at the world through the eyes of an historian. I quoted the line in a paper written in college.
Maybe the smartest guys in the room at Enron did, when they were proofing their goofy, that'll-fool-'em, 2001 annual report, but no one in the real world does. There is no such thing as a backward-looking agent. As a friend of mine puts it, "climb down from your high chair."

Collander has made a major mistake if he thinks it is a good idea to separate pure and applied research.
"What I’m arguing is that it is most useful to think of the search for the social science policy keys as a two-part search..."

Roger Koppl might have picked up on that but I don't have time to read all the comments again.

The mistake has a long history and it was enshrined in the success stories about scientific research leading the way in civilisation and economic progress etc. This story has been subjected to devastating critique by Terence Kealey in "The Economic Laws of Scientific Research".

The bottom line is that theory and practice should never be separated, pure and applied research should synergise, as they did in Australian rural research.

http://catallaxyfiles.com/2010/09/16/4-well-spent-kealey-on-civilisation-and-free-trade/

In fairness to Colander, he may be heading in that direction with his final suggestions.

Greg writes,

"It's time the economists dumped the the failed project of the philosophers -- and it's time they picked up the successful strategy for doing science exemplified by Darwianian bioloyg, i.e. figure out what the empirical / causal problem is and focus on the empirical / causal mechanism that provides an explanation for that problem."

Could you share some links where this idea is elaborated in more detail?

Fundamentalist, read Hayek's "The theory of complex phenomena" or his Degrees of explanation" for a contrary view.

Hayek moved a long way from his early 1940s work in his 1950s publication.

Hayek started in the early Wittgenstein / Carnap / Mach / Mill world like many others -- he finished by exploding all that, and empracing folks who participated in the explosion party, e.g. Popper, the later Wittgensteinians, Ryle, Polanyi, the systems theory people, non-linear phenomena people, even recommending Kuhn to Popper (!).

Hayek ends up embracing Bartley and the whole attack in the justification / demostrative knowledge / Euclid ideal tradition.

Greg Ransom, it behooves someone of your caliber & scope in the field of philosophy of science to have F. S. C. Northrop among your repertoire. Would you please put on your reading list at least his _The Logic of the Sciences and the Humanities_ and _The Meeting of East and West_, and get back to us at some point your opinions of how Northrop fits into your scheme of things vis-a-vis, e.g., your two posts above (September 16, 2010 at 03:48 PM & 03:57 PM) - Darwin, induction/deduction, etc.? I trust that after getting Northrop under your belt, you would put him right alongside "Thomas Kuhn or Norwood Hanson or Karl Popper or F. A. Hayek."

By the way, I noticed that the Google page rank of the JSTOR feed to the Austrian economics-related Northrop article I cited elsewhere here (in the Comments to "Some Clarifications," September 01, 2010 at 03:40 PM) went to the very top a week after my citation, and remains there to this day - instead of references to it by other articles previously. I wonder if that Google page rank jump is attributable solely to activity from Coordination Problem readers.

Apologies in advance, Greg, if you're already quite aware of Northrop, but from your reply (September 01, 2010 at 05:15 PM) to my aforementioned Comment, it sounded like you weren't, and indeed by proxy from my characterizations were placing Northrop among the scientistic figures you so rightly criticize.

At least try-on for size that one Israel M. Kirzner-footnoted Northrop article, and see if you don't find yourself nodding your head, "Yes!":
"The Impossibility of a Theoretical Science of Economic Dynamics," Quarterly Journal of Economics, November, 1941, reprinted as ch. XIII in his _The Logic of the Sciences and the Humanities_

fundamentalist,

Sorry, I knew both Sewall Wright, the inventor of SEM back in the 20s, and Arthur Goldberger, its greatest exponent in more recent years, personally. I will simply note that there are a variety of criteria that can be used to evaluate SEMs. There is no single "ah ha, this is it!" criterion. The arguments are very much ongoing.

Greg, But Hayek's Nobel speech, the Pretense of Knowledge, was written in 1974 and he seems to make the deduction/induction argument there. And counter-revolution was written in 1952 and he decries the application of natural science methods to economics.

Barkley, they must have been interesting men. Yes, I use SEM regularly and use several criteria, but no matter which set of criteria you use, you will be able to compare different models attempting to describe the data. Without SEM, you're left with just comparing error in forecasting. There is no method for comparing models that has no problems. SEM is a powerful technique designed expressly for comparing models. I think if Austrians would embrace it it would go a long way toward getting people to take Austrian econ more seriously.

Read Kealy and find how every war, starting from the civil war, was used to push the barrow of centralised state control of research. We are standing in the consequences and all of the bandaid suggestions from David Collander are just trying to keep alive a termianlly ill patient. Abolish the NSF and get with the GMU program of theoretically sophisticated, applied multi-disciplinary fieldwork.

This is not a paid advertisement for the GMU and I have no blood or pecuniary relationship with Peter Boettke or or anyone else associated with him or it.

Daniil, see the work of David Hull, Ernst Mayr, Alex Rosenberg and Larry White on the explanatory strategy, non-reducibility, and explanatory autonomy of Darwinian bilogy and teleological phenomena more generally.

Michael Ruse and many others demonstrated by the failure the absudity of attempting to put Darwinian biology in the nomological deductive model of the "received view".

The last person I'm aware of who really tries to vindicate whatever is left of the "received view" of science tradition of Nagel & Hempel (the induction // deduction model) along with the deep spirit of Mill/Hume empiricism is Alex Rosenberg -- and while Rosenberg's command of the literature and talent in applying it is unquestion and often pathbreaking, it's hard to find anyone in the field who anymore buys into the old project, which suffers from unending counter examples and pathologies.

The Counter-Revolution essays were written in the mid 1940s.

"counter-revolution was written in 1952"

The Pretense address is a work in persuasion using popular language and broad categories, althoughnit contains deep insights and claims.

Read the essays I've mentioned above for meatier stuff.

Hayek says the dominant view is a false view even of NATURAL science.

Here Hayek was influenced by Popper.

Note that Hayek came to reject even Popper's account.

Greg, are those essays available on the internet? I couldn't find them on your site.

Fundamentalist,

Those essays are collected in Hayek's _Studies_ and in his _New Studies_.

Both are out of print -- which is a real shame.

Hayek's account of the relation of tautological formal constructs to the explanation of empirical patterns in our experience can be found in the first few chapters of his _The Pure Theory of Capital_, with supplemental material in his essays, "Economics and Knowledge", "The Use of Knowledge in Society", "Degrees of Explanation", and "The Theory of Complex Phenomena".

Note that this account is specific to economics, but Hayek suggest something similar is applicable to global brain theory and Darwinian biology.

Interesting read. I'm not sure I'm following him on his spotlight metaphor and his explanation of "pure research", but I like it.

I think the second proposition could be a great opportunity for AE, but is rather unlikely. The guys developing these models know too well that anyone would come to the conclusion that these models are irrelevant. No honest macroeconomist would pretend that rocket trajectory math, and in general papers without a single line of text, can really teach us anything about economics. It's about perfecting the models for the sake of perfecting the models. It seems unlikely they would give grants to people to discredit their equations and basically tell them what they already know deep down inside.

Many thanks, Greg!

Daniel:
My impression is that public choice scholars of the Virginia School have not fared well at NSF. I know that Gordon Tullock and Jim Buchanan were unsuccessful there. As Editor of Public Choice for a long period, NSF grant recognition was restricted to public choice scholars more favorable to government:- some from Maryland, some from Harvard, some from Cal-Tech, some from California universities. But those known to be critical of government never cited NSF. Now what I cannot be sure about, other than for Buchanan and Tullock, is whether many of them applied. Mostly, they would favor private monies, if available, although NSF, at least in principle, might have been viewed as more neutral than the various federal government departments.

Actually, answering your question might be a worthwhile public choice research topic.

Some overall comments: (1) Colander's statement is very good,especially as a statement submitted to a Congressional Committee. This needs to be a "politically astute" document; (2) The public choice aspect of NSF funding is quite important. These grants are one way the government keeps the "top" intellectuals happy -- whatever they do is "fine." We (the State) cannot afford an alienated intellectual class.

Consider for a moment that there is no way that the *theoretical* research done by economists has any likelihood of producing "social benefits" and yet it is funded.

My view is that the NSF will fund whatever is done at the top universities -- and their satellites -- just to maintain good relations with the intellectuals with the most influence.

I'd also recommend Nassim Taleb's testimony:

http://democrats.science.house.gov/Media/file/Commdocs/hearings/2009/Oversight/10sep/Taleb_Testimony.pdf

Taleb discusses the pseudo-science of "measured risk" which came out of the Grad Schools and took over Wall Street.

For anyone near Harrisonburg, Taleb is speaking this afternoon on the JMU campus at 3:30 PM in Grafton Stovall theater on "Towards a Black Swan Proof Society."

Mario, Collander seems to think that the system that created the problem could turn around and fix it (essentially by folding up its tent and going away). That is like expecting the Welfare System to reform itself from the inside.

If a project can't get funding from a successful entrepreneur (a market entrepreneur, not a political operator), send it to the maths school and see if they find it interesting.

Collander has a point. Anyways, I have to tell you, I really enjoy this blog and the insight from everyone who participates. I find it to be refreshing and very informative. I wish there were more blogs like it. Anyway, I felt it was about time I posted, I’ve spent most of my time here just lurking and reading, but today for some reason I just felt compelled to say this.

Sure, it's all about the funding; but let's just maintain the funding the way it is (i.e., NSF controlled) with some superficial tweaks an interested party suggests (i.e., Colander). What was the Einstein definition of experimental insanity? It seems to me that economics is clinically as well as experimentally insane. There seems to be an easy first step here: (1) try something else, and/or (2) stop doing the same thing and expecting a different outcome.

The state-funded research industry is too big to fail.

Feel Immediate,prison ensure earth form tax battle difference indeed pull record need broad data previously assume effort alternative push regional concern serious training regulation course require shop store light motion response location afraid no-one strength scheme thank enjoy myself character however procedure component breath normally staff engine fair telephone affair contrast rich factor talk although physical there practical standard report inside relief throughout also violence life step curriculum early animal become same youth amount corporate direct fashion future admit yet journey birth relief judge health error construction

The comments to this entry are closed.