A simple approach to macroeconomic theorizing

by on October 8, 2011 at 4:16 am in Economics, Education, Uncategorized | Permalink

I can’t say it is guaranteed to work, but I give it a high “p”:

1. Take the macroeconomic theory you hold and stick it into a box.

2. Take the major competing macroeconomic theory, the one you dislike, but taking care that you have selected an approach endorsed by high-IQ researchers.  If you dislike them too, that does not disqualify the theory, quite the contrary.

3. Stick theory #2 into the same box.

4. Average the two theories.

5. Pull the average out of the box, and call it your new theory.

How many times should you apply this method?  At least once I say.

I am indebted to Hal Varian for a useful conversation on related topics.

1 dan1111 October 8, 2011 at 4:47 am

If all economists worked in this way, there would eventually be only one theory. And what would it be?

2 david October 8, 2011 at 5:03 am

Whichever school of thought that is most resistant to change!

3 Doc Merlin October 8, 2011 at 9:53 am

And most widely disliked.

4 Claudia Sahm October 8, 2011 at 5:17 am

Varian is no longer the standard micro text in most PhD programs, but I read “baby” (his undergrad text) and “big” (his grad text) Varian along with the assigned “phone book” textbook by other economists. Varian provides a lot of useful intuition. Good intuition and economics common sense (which he teaches very well) are likely necessary, but not sufficient conditions for properly applying the cutting edge techniques…you gotta read the “phonebook” too. This mixing model advice sounds most applicable for people who have already been in economics a while. If you don’t have a “favorite” model, it seems best to learn several models before you start mixing. Still cool advice in the post.

5 Joseph Sanchez October 8, 2011 at 5:19 am

When taking the average, how do you weight the two different theories?

6 Mario Rizzo October 8, 2011 at 5:26 am

If you weight them by your degree of belief the whole “method” falls apart!

7 Harry D October 8, 2011 at 11:45 am

I think I understand Tyler’s point, and I think you’re agreeing that it’s a sensible point, but I’m not at all sure. As I understand it, the idea is that if you take the test and the theory you end up with is the same as the one you started with, then you’ve not invested enough time and effort in understanding rival theories (or approaches). Macroeconomics is about a complex phenomena, and it’s easy to make mistakes in thinking about it, however you think about it, because the chains of reasoning are so involved.

This point applies to Austrian sympathizers no less than to anyone else, of course, if anyone should doubt it. ‘Apodictic’ science (in Mises’s term) is a science in which, if you make no mistakes in reasoning, will lead you to a set of pattern-predictions that’s true always and everywhere. But in applying the general economic framework to a historical case, more than one theory (set of theories?) might reasonably be considered relevant. And at a more fundamental level, the approach based on viewing macroeconomic phenomena as founded on open-ended decision-making and so on(?) might be no more helpful for understanding and no less truthful than an approach that reduces decision-making to maximization, and that reduces spontaneous order theorizing to general-equilibrium theorizing.

But that doesn’t sound quite right. So I might have missed the point.

8 Harry D October 8, 2011 at 12:03 pm

With apologies for making a second comment just after my first one, I wonder if this might be a specific example of the sort of thing we’re talking about, and so it’s something to focus our (my?) attention on. it seems like a conversation worth having.

‘The’ Austrian business cycle theory. It certainly can and often does work as spontaneous order theorizing, and it seemingly at least sometimes best explains a business cycle’s boom and bust. But is it the only spontaneous order theory of the business cycle, and is it the only one that might explain a lot of any particular business cycle? Or perhaps more precisely, what kind of theory based on monetary injection and monetary disorder? I’m not sure, but it seems like it might matter whether the system has polity entangled in the marketplace, for example.

9 Tim October 8, 2011 at 5:26 pm

Generally when I average I don’t use weighting. Because then it’s not really an average anymore. The main flaw/strength of averages is that they’re indiscriminate.

10 dearieme October 8, 2011 at 5:46 am

So much for economics as a “science”.

11 Claudia Sahm October 8, 2011 at 8:14 am

Economics is a “social science” maybe we should average those two words too?

12 Norman October 8, 2011 at 1:55 pm

They may not do this to the same degree as macroeconomics, but if you think the hard sciences don’t follow a similar process you’re kidding yourself.

13 Ricardo October 9, 2011 at 2:58 am

The difference is that theories in the hard sciences that do not make falsifiable predictions get very little respect and eventually get dropped by “high-IQ researchers.”

By contrast, Stephen Williamson declares (almost proudly, it seems) that dynamic stochastic general equilibrium and the efficient markets hypothesis “have no implications” and “cannot be wrong” (shouldn’t we call it the “efficient markets axiom” then?).

The problem, of course, is that if a particular approach “cannot be wrong” then there is no reason to prefer it if you actually want to say anything useful about the real world. Karl Popper pointed out that legitimate science has as its goal good-faith efforts to falsify certain prevailing hypotheses. The worth of a hypothesis is judged partly by whether it makes “risky” claims — a good theory puts its head on the block and dares high-IQ researchers to take a swing.

14 k October 9, 2011 at 10:18 am

but this is what happened to social choice theory within economics – it offered no predictions, and it is a dead field now

15 Sergey Kurdakov October 8, 2011 at 5:57 am

I think the answer could be found in a road to overcome Lucas critique http://en.wikipedia.org/wiki/Lucas_critique – to derive theory from observation of actual processes in economy, not from out of air assumptions.

it is strange to think that highly aggregate theory can properly describe important macro economic effects, just because the interplay of different important parameters are obviously lost.

So, if FuturICT http://www.futurict.eu/sites/default/files/WhatFuturICTWillDo4Media.pdf is a success – then it will most probably allow to have some ‘visualization tools’ which will be illustrative and also more adequate, and maybe ( using ‘clustering’ or something like that ) it will be possible to derive few important parameters which will allow to have new ‘classical’ ( based on few assumptions and equations ) macroeconomic theory, because maybe important features ‘are just not discovered’ ( like air travel on hot balloon, which was possible for millenia, but was invented in 1783 ) as ‘generalizing’ is not a suitable tool for such a discovery, but rather aggregating observed phenomena might be a proper tool – which was never used so far ( as no one tried to approximate real complexity of economy with ‘complex simulation’- and it, maybe, actually, can be approximated – but we will see as a result of FuturICT effort).

16 Andrew' October 8, 2011 at 6:29 am

In my limited experience in biological modeling, there are good and bad models, neither of which are necessarily realistic. A bad model can be more realistic but can neglect an important factor. And the purpose of the model is not necessarily to be realistic, but to be useful to test and develop the next most realistic good model by comparing to experimental data under conditions controlling other potentially important factors.

17 Brandon Berg October 8, 2011 at 6:55 am

What does it mean to do this more than once? Does two iterations mean 25% your theory and 75% the competing theory? Or one third yours, one third competing theory #1, and one third competing theory #2?

18 Peter S October 8, 2011 at 7:26 am

GF Hegel would approve.

19 haragan October 8, 2011 at 8:00 am

This would work if there are some “nonlinearities” in theory averaging or if under the new preferred theory another theory becomes the one you can’t stand, otherwise, if you do it more than once you end up creeping toward the theory you dislike, no? (The average becomes the your preferred one while presumably the loathed one has not changed.)

20 Jeff October 8, 2011 at 8:03 am

Or you could just take two aspirin and lie down until the urge to theorize goes away.

21 Claudia Sahm October 8, 2011 at 11:30 am

Or you could go outside (and offline) to be around people who are more likely to guess that IS-LM is a new strain of the swine flu and not some econ model. If your are at work with economists, read a quality empirical paper with some “surprising” results, that is, ones that are tough to shoehorn in “standard” model. My fav from yesterday: Olivia Mitchell presented a result (among many other more important ones on financial literacy) that women who are impatient and have trouble carrying out plans (among other characteristics) tend to have higher BMIs, but for me being overweight was almost completely explained by marital status. I can’t think of a single “representative agent” model that could handle that regression result. Happy not be a theorist.

22 Claudia Sahm October 8, 2011 at 11:34 am

Woops… I actually checked my post and I still completely messed it up. Repeat key sentence (with lame attempt at humor):

“women who are impatient and have trouble carrying out plans (among other characteristics) tend to have higher BMIs, but for men* being overweight was almost completely explained by marital status.”

if I have to reveal the “me” overweight it’s the result having two kids. I suspect that is one of the “other characteristics” in the female results. So much for being funny. I should have stuck with my “offline” advice.

23 David N October 8, 2011 at 8:17 am

This is great advice. Does anyone know the formula for averaging a flat earth with spherical one?

24 Bill N October 8, 2011 at 8:22 am

Seems like an oversimplified form of Hegelism or a very low “n” wisdom of crowds. Not that it isn’t a useful exercise, but what do you do next?

25 D October 8, 2011 at 9:21 am

Steve Sailer just wrote something similar:

“Along these lines, I would encourage intellectuals to try to subscribe to a form of vulgar Hegelianism in their personal intellectual behavior that I’ve found very useful.

If you hold a thesis for what seem like good reasons, and somebody counters with a well-argued antithesis, you have several options:

– Reject the antithesis (the most common)

– Convert to the antithesis (the most dramatic)

– Look for a synthesis that makes sense of both your thesis and the other guy’s antithesis (usually, the hardest but most profitable)

For example:

Thesis: A racial group is a taxonomical subspecies.

Antithesis: A racial group is a biologically nonexistent social construct!

Synthesis: A racial group is a partly inbred extended family.”


26 Ricardo October 9, 2011 at 9:00 am

My first question would be what are the testable implications of a racial group being a “taxonomical subspecies”, a “biologically nonexistent social construct” or a “partly inbred extended family”? If there are none, scientists should leave the debate to ideologues and scribblers and move on to more productive endeavors. If there are testable implications, then scientists should choose the hypothesis that is best confirmed by the data, not the hypothesis that is the product of some “vulgar Hegelian” dialectic.

27 Becky Hargrove October 8, 2011 at 9:33 am

Albert Einstein would be proud. Although he might be inclined to completely discredit a few variables.

28 Doc Merlin October 8, 2011 at 9:55 am

I think David gets it. This would just make MMTers or Austrians the winners for being the most hard headed and least liked.

29 anonymous... October 8, 2011 at 10:32 am

In doing this, you give your intellectual adversaries every incentive to be patiently intransigent, or indeed to stake out ever more extreme positions in the hope of luring you towards their original starting point.

If the opposite side is not following a similar process of introspection and consensus-seeking and questioning long-held assumptions (*cough* Krugman *cough*) the result is unlikely to be helpful.

“The best lack all conviction, while the worst
Are full of passionate intensity.” — Yeats

30 Doc Merlin October 8, 2011 at 11:06 am

Absolutely. This plan fails the test of most economic plans. Its not robust to other people knowing about the plan.

31 JLA October 8, 2011 at 11:08 am

How can you average reduced form and structural form modelling?

32 Ram October 8, 2011 at 11:30 am

Disagreement in economics is so widespread largely because there do not exist many knockdown arguments for specific models or even for specific theses. Part of the problem is that people in general, and intellectuals in particular, have comparatively rigid priors with respect to economic subjects, which makes it more challenging to persuade one’s opponents. Another is that economic “expertise” carries some weight with powerful interests, not to mention political movements, encouraging economists who seek influence (which is to say, most economists) to focus their research upon confirming the biases, or supporting the interests, of those whom they wish to influence. This discourages individual economists from being persuaded by the arguments of their opponents. A third is that the phenomena under study are so complex that one must, in choosing a model, trade off simplicity against relevance, opening up a further dimension along which economists may disagree.

To some extent, these problems are present in every science, but they’re especially present in economics. (I haven’t even discussed the financing of research, among other issues.) The way other sciences overcome them is by developing knockdown arguments, arguments so persuasive (typically due to their empirical content) that one’s reputation as a scientist hangs in the balance if one rejects them. Economics simply has not developed many (or any) such arguments. Its models make assumptions that fly in the face of everything we know about the relevant subjects, their causal claims have generally not been well-established with experimental evidence (from laboratory or field research), and they do not make reliable, quantitatively precise predictions over a range of phenomena, nor do they permit the engineering of highly successful interventions into the systems under study. The evolution of economic theorizing mostly is based on reasoning about qualitative stylized observations, logical/mathematical consistency, and “plausibility” (which just amounts to coherence with one’s priors). These merely support the notion that economic models describe possible worlds that are not wholly unlike the actual world, but they do little to persuade that they describe worlds which are substantially similar to, and therefore relevant to, the actual world.

Until economists develop theories with quantitative empirical content so compelling that no serious economist could reject them, there will be a great deal of disagreement in economics. And even where there is agreement (free trade rocks!), this may have more to do with the fact that those sympathetic to existing arguments are more likely to become economists than those who are not sympathetic to them. In the meantime, the rest of us should recognize the extent of this disagreement as THE central fact about economics, and moderate our priors about how the economic realm works accordingly.

33 CW October 8, 2011 at 1:01 pm

There is some work along these lines by Brock, Durlauf, and West on model averaging and on decision-making under uncertainty about which of several models is true.


34 Jason October 8, 2011 at 1:49 pm

Rediscovering ensemble models, huh?


If only there were a way to judge the effectiveness of models … some kind of “Chi square test” but for hypotheses and statistical data (smirk) … Or maybe rigorous criteria …


Gaussian distributed points around a straight line does remarkably well for post-war GDP. Each sample requires only a tenth of a bit of information more to reproduce the actual (non-Gaussian) distribution (derived from Kullback–Leibler divergence).

A straight line with a kink in it for slower growth after 1980 does even better, requiring only a half a tenth of a bit more. Post-war GDP only has about 8 bits of dynamic range. That’s a model that is about 1% accurate. And it’s two slopes and an intercept.

35 Matt Flipago October 8, 2011 at 2:29 pm

Well i’m confused, why would you do it more than once. IF you do it three time you end up with something thats abou 7/8 your hated theory and 1/8 the theory you liked in the first place. Who would do that?

36 jpd October 8, 2011 at 4:09 pm

should you weight them by which one is correct?

37 Lord October 8, 2011 at 10:25 pm

How does that work when the alternative is not a theory at all but just an ideology, non predictive, that can only manage to “explain” anything afterwards?

38 Ramagopal October 9, 2011 at 5:11 am

I take the theory I like , Austrian theory of inflation , and put it in a box
I then take the the theory I don’t like, the keynesian theory of inflation, and put it in the same box.

now can anyone tell me exactly how I am to “average the two theories”? Is this a subjective exercise or is there some objective way of doing it? Tyler, if you actually applied this method, can you describe how you carried out the “average the two” part of the exercise?

39 Colin Fraizer October 9, 2011 at 7:11 am

Something about this technique reminded me of both Justin Timberlake and Paul Krugman. Maybe it was this video:


40 Colin Fraizer October 9, 2011 at 7:14 am

Better link that starts with the steps:

1. Cut a hole in a box.

Tip: to link to a particular time offset in a YouTube video, add “;t=XmYs”, where X=the minutes offset and Y=the seconds offset.

41 gc October 9, 2011 at 10:34 am

“taking care that you have selected an approach endorsed by high-IQ researchers”: If you need to know the IQ of the researcher in order to judge the quality of the research then your “judgment” is limited to being able to distinguish bigger numbers from smaller numbers. Why is it that the blogs of Fields Medals winners such as Tao, http://terrytao.wordpress.com/, and Gowers, http://gowers.wordpress.com/ never mention IQ? Perhaps they are more interesting in teaching than self-aggrandizement?

42 Jason Yip October 9, 2011 at 6:50 pm

Is this not an example of balance fallacy? http://rationalwiki.org/wiki/Balance_fallacy

43 Dave Tufte October 17, 2011 at 11:28 pm

This isn’t that dissimilar to “shrinkage estimators”, covered in some detail early in the “blue” Judge et al. text, and then routinely ignored by everyone.

Comments on this entry are closed.

Previous post:

Next post: