The intellectual pulpification in question involves a post by Dr. Williamson that reviews a book, "Zombie Economics," written by the economist John Quiggin. Williamson introduces Quiggin thusly:
What is Quiggin's claim to fame? His early work is an odd mix of agricultural economics and decision theory, but he seems to have distinguished himself mainly in public policy. He writes regularly in the mainstream media, writes a blog, and Zombie Economics appears to have sold well.
Now, what is Quiggin up to in Zombie Economics? Roughly, Quiggin is the Australian farm team in the Krugman/Thoma/DeLong league.Although I would probably give at least one toe to be described as the "Krugman/Thoma/DeLong farm team," I thought this to be mildly disparaging, and potentially a little unfair. So I looked up John Quiggin on Wikipedia and learned the following:
Quiggin is one of the most prolific economists in Australia, illustrated by his output over diverse, high-quality journals and by citation frequencies in the period 1988-2000. He is among the top 500 economists in the world according to IDEAS/RePEc. He is best known for his work on utility theory. Quiggin has frequently been awarded and recognised for his research, including twice receiving Federation Fellowships from the Australian Research Council.And who is Stephen Williamson? I looked him up on Wikipedia as well, and found the following:
Stephen Williamson (28 June 1827 – 16 June 1903) was a founder of the Liverpool shipping company Balfour Williamson & Co. and a Scottish Liberal Party politician.Oh wait, wrong Stephen Williamson. Where's the disambiguation page? Turns out there isn't one, because Stephen Williamson, the Robert S. Brookings Distinguished Professor of Arts and Sciences at Washington University, apparently does not meet Wikipedia's cutoff for significance.
(Don't take it too hard, Steve. I tried to get myself on Wikipedia, and they booted me. Even when I claimed to be the founder of string theory and son of Sean Connery. But I digress...)
Anyway, Williamson's post is intended as a comprehensive takedown of Quiggin's book. Zombie Economics is about five big ideas that Quiggin thinks have outlived their usefulness. These ideas are:
1. The Great Moderation.
2. The Efficient Markets Hypothesis.
3. Dynamic Stochastic General Equilibrium
4. Trickle-down economics
5. PrivatizationWilliamson goes through these one by one, and purports to show that Quiggin doesn't know what he's talking about. So I'll stick to the same format. Quiggin has a partial response on his own blog, upon which I will draw as I go through Williamson's points.
Warning: brutal pulpification is imminent.
Williamson Point #1:
The Great Moderation: Quiggin is a little confused on this one, as the Great Moderation simply characterizes a set of properties of US aggregate time series. From about 1985-2007, inflation was lower and less variable, and real GDP was less variable about trend than had been the case previously. Quiggin is certainly correct, though, in finding fault with those (Ben Bernanke included) who wanted to argue that the Great Moderation was due to a regime change in economic policy. If policy was so great, it should have done a better job over the last four years.So, according to Williamson, Quiggin is "confused" because he uses the term "The Great Moderation" to describe the notion that certain changes in aggregate time series from '85-'07 represented a structural change. Williamson then criticizes everyone who actually did think that it was a structural change. And he uses the phrase "Quiggin is certainly correct." So how is Quiggin the one who is confused, here? Williamson is dissing Quiggin over a semantic point, then substantively agreeing with him. Not much of a takedown...
Williamson Point #2:
The Efficient Markets Hypothesis: For Quiggin this is "the idea that prices generated by financial markets represent the best possible estimate of the value of any investment." Here, Quiggin is badly confused, but maybe the finance practitioners are not helping him out much. Market efficiency is simply an assumption of rationality. As such it has no implications. If it has no implications, it can't be wrong.First, I will unleash Quiggin himself, who points out that a theory with no implications is a worthless waste of time:
That’s right, Williamson not only defends the EMH on the basis that it’s not even wrong, but follows up by making the same claim about the whole of modern macro. Those are, quite literally, his only criticisms of these chapters. His commenters (none of whom seem particularly favorably inclined towards me) try to suggest that isn’t the most effective line of attack, but he comprehensively misses the point.Yes, Stephen Williamson defended a theory by saying that it can't make any predictions about the real world. You read that right.
But it gets worse. The EMH is most certainly NOT an "assumption of rationality." The EMH does not even require rationality in order to be correct; it requires only that the aggregate effect of individual investors' irrationalities not be predictable. And Williamson also seem not to realize that Quiggin's statement of the EMH is perfectly legitimate; it is the "strong form" of the EMH, as stated by Eugene Fama. Conclusion: Stephen Williamson does not know what the Efficient Market Hypothesis is.
(Random note: I like the EMH. I don't think it's completely correct, but I think it's the unique starting point from which any analysis of financial markets should begin. As opposed to, say, DSGE macro, which is just some random thing that doesn't work. Anyway...)
Williamson Point #3:
Dynamic Stochastic General Equilibrium: Quiggin claims that this is "the idea that macroeconomic analysis should not concern itself with economic aggregates like trade balances or debt levels, but should be rigorously derived from macroeconomic models of individual behavior." I can hear you snorting with laughter. Why is "but" in that sentence? Like the "efficient markets hypothesis," DSGE has no implications, and therefore can't be wrong. Indeed DSGE encompasses essentially all of modern macroeconomics. Which of our models is not dynamic, with uncertainty (and therefore stochastic), and with some equilibrium concept.I hesitate to say this, but Stephen Williamson appears also to not understand DSGE. If he thinks it has no implications, I mean. DSGE is not just "some equilibrium concept," but a very specific one: the dynamic, stochastic generalization of the general equilibrium framework of Arrow, Debreu, and McKenzie. This means that DSGE does have implications - for example, that markets clear and that agents are price takers. Those things can be wrong, as can the results that follow from them (e.g. the Law of One Price). So Stephen Williamson's attempted defense of DSGE, which was looney in the first place ("It has no implications! It's utterly useless and therefore unassailable! Gurrrrhhh!"), also happens to be flat-out wrong.
Williamson Point #4:
Trickle-down economics: This one puzzled me. For the previous three zombie ideas, Quiggin is confused, but I could see where the confusion might come from. However, while the words "trickle-down economics" are familiar to me, I have a hard time associating that idea with the mainstream ideas of any academic economists.Um...how about dynamic Laffer effects, as posited by John Cochrane? I'm pretty sure John Cochrane is a mainstream academic economist. He's on Wikipedia, at any rate.
Williamson Point #5:
Privatization: Here, Quiggin offers a litany of government privatization efforts gone awry. If I wanted to, I could take this evidence as supporting the hypothesis that government is really bad - so bad that it can even screw up privatization.Sure, if you wanted to. If I wanted to, I could take the Mongol Conquests as evidence supporting the hypothesis that Radiohead is the Best Band Ever. It's called "intellectual dishonesty."
Anyway, Quiggin also points out that Williamson doesn't offer any evidence to refute the book's claims. This is true. It's as if Williamson thought that he could refute Quiggin's entire book by tossing off a quick and sloppy mix of vague handwaving, semantic nitpicking, utter devaluation of the theories Quiggin is attacking, complete and utter mischaracterization of those same theories, and lame personal jabs.
I score this Quiggin: 5, Williamson: 0. (Or maybe Williamson: -1, given that he thinks "not even wrong" is a feature rather than a bug.) Pure and utter pulpification.
Oh well. At least Stephen Williamson can take comfort in the fortunes he's made off of the Balfour Williamson & Co. shipping venture. Oh wait...that's the wrong Stephen Williamson! Dang you, Wikipedia!!!
Surprised anyone is still paying attention to Williamson. I tried reading him for maybe a month tops about a year or two ago. Total waste of time.ReplyDelete
Are you OK, Noah? What did I do to make you so grumpy?ReplyDelete
"[M]ild and nonconfrontational," eh? Who are you, and what have you done with my brother?ReplyDelete
DSGE seems to be the equivalent of an engineer trying to predict the bending properties of a beam from the quantum mechanical properties of electrons, protons and neutrons - without actually knowing what those properties are.ReplyDelete
It might be possible. No one has done it. It is easier and more reliable to go and make experimental observations.
I am trying to read Cochrane's paper. I think he loses sight of what is really going on in all of the clever mathematics.
"I could take the Mongol Conquests as evidence supporting the hypothesis that Radiohead is the Best Band Ever."ReplyDelete
It was Idioteque which first gave me the idea that the world needed conquering, and Creep was very influential in forming my strategy to overrun the world by slowly, uh, overrunning the world.
And vice versa.
So, I, Ghengis Khan, certainly think Radiohead is The Greatest Band Ever! Without Thomas Yorke's inspiration, I would have been just another Mongolian king, vaguely dissatisfied with the dimensions of his yurt and wondering if there wasn't more to the world than Ulan Bator and fermented mare's milk.
I suspect Quiggin, and by extension you, are wrong in the description of DSGE: the only framework under which it makes sense is building strictly from =micro= foundations and then stipulating that this is, therefore, the best of all possible worlds because the better ones were not possible (since they didn't happen).ReplyDelete
But I'll keep screaming this to my deathbed, which I guess is why I cannot qualify to be a member of the Axis, Bold as Technocrats.
Don't they also assume complete markets? Unfortunately that is impossible in the real world. And in GEI (General Equilibrium with *incomplete* markets) the equilibrium is typically sub-optimal.
And the Robinson Crusoe assumption?
Rational Expectations rules out belief heterogeneity by assumption and also requires all agents be Laplace's demons.
And the claim that microfounded ratex DSGE models allow us to unearth 'deep structural parameters' is testable and in fact is false.
Please do not waste any time with Stephen Williamson. He is a sad clown deeply lacking attention.ReplyDelete
It is not every day that I can follow a link from a blog post at the New York Times to one by a grad student and find said graduate student talking back to an endowed professor with wit and gumption.ReplyDelete
Dr. Williamson. Let me suggest you get a clue. Since you seem incapable, let me give you some free professional advice - writing about graduate students in public in so patronizing a way (I visited your post about said grad student) is totally unbecoming. I would say it rates you a remarkable high jackass quotient. And I've never met either of you.
That said, I apologize to Noah here since he doesn't need others to stick up for him. I just find the Williamson spectacle astonishing.
Noah, you couldn't pass prelims in wash u or anywhere else for that matter. What's up with dsge assumes arrow debreu with competitive markets and law of one price? Have you ever heard of new keynesianm models, whre firms are in monopolistic competition (and hence NOT price takers) and where prices are not the same since they are sticky?ReplyDelete
Cochrane is mainstream. So is thte whole modern dynamic public finance literature that finds motivations for all sort of taxes (including over capital). But I guess unless you pass the prelims you won't have a chance to hear about them
I would love for someone to explain to me how an "assumption of rationality" has no implications on models that depend on it. Seems to me that it would be like assuming there is no friction, and saying that that has no implications on the feasibility your perpetual motion device.ReplyDelete
In addition to Noah's perceptive deconstruction on economic grounds, Stephen Williamson also thinks that St. Louis is a nice place to live.ReplyDelete
Actually, it seems like Dr. Williamson is a mole in the academic establishment who is trying to ensure that economics has no legitimate respectability or consistency as even the semi-science it is.ReplyDelete
Once you do that, all is permitted. Like unicorns, fanciful theories that have no implication or application, and the hordes of grubby politicians who can't wait to bastardize your field with their own anti-intellectual, anti-elitist doggerel.
Hey look over here: there's a minority of economists who think that the tragedy of the commons is a false plot planted by the Russians to promote communism. That means that the whole economic foundation of cap-and-trade and carbon taxes is an elaborate shell game!
And so on.
Noah, the Arrow/Debreu/McKenzie general equilibrium model allows for multiple commodities, securities, and states of nature at all dates, as well as unrestricted heterogeneity in endowments and preferences. It's far more general than any DSGE model that I have ever seen. Far from being a generalization, I would say that DSGE is a caricature of general equilibrium theory.ReplyDelete
Anonymous @12:26 AM wrote, Actually, it seems like Dr. Williamson is a mole in the academic establishment who is trying to ensure that economics has no legitimate respectability or consistency as even the semi-science it is.ReplyDelete
That was already done in spades by the relative lack of economists expressing alarm in the run-up to the recent crash of the financial system.
Poor old Steve wrote "Are you OK, Noah? What did I do to make you so grumpy?"ReplyDelete
He's not very perceptive, our Steve. Not exactly the sharpest tool in the shed.
I don't think Noah was at all grumpy. In fact I got the impression he was that amused by the stoush and the lameness of Steve's 'argument' that he was at risk of a urinary indiscretion. I know I was.
Oh, so funny.
To be fair to Williamson, Radiohead is indeed the best band in the history of rock.ReplyDelete
The phrase you are looking for is: "U mad, bro?"
Personal nastiness does not look good on Krugman or de Long when they do it and it gets in the way of their message. It is, in a word, self indulgent (oops - that was two words). I suggest that you avoid it.
Remember that the Internet is forever and you can never know who is connected to who. You could find yourself being interviewed by Williamson's best friend some day.
I do think some critics of Efficient Markets theory are glossing over some important distinctions. For example, most people do not subscribe to the "strong" version of the EMH. So, if Quiggin is attacking that version, it's kind of a straw man argument.ReplyDelete
But the most important distinction is between "value" or "fundamental" efficiency on the one hand, and "informational" efficiency on the other. It's perfectly legit to doubt whether efficient markets generate the "true" value of stocks. But even if markets aren't "value" efficient, they can still be "informationally" efficient in the sense they process all public information such that one can't beat the market.
So why is an informationally efficient market important? Well, for one thing, securities fraud class actions depend upon the markets' being informationally efficient. Without informational efficiency, it would be impossible to certify a class because everyone would be affected differently by the fraud.
Herman (and all others who are as poorly informed about DSGE models):ReplyDelete
Don't they also assume complete markets? Unfortunately that is impossible in the real world. And in GEI (General Equilibrium with *incomplete* markets) the equilibrium is typically sub-optimal... Rational Expectations rules out belief heterogeneity by assumption and also requires all agents be Laplace's demons."
It really seems like a lot of non-macro people have this weird notion that DSGE is nothing more than Kydland-Prescott and its immediate descendants. This is really not at all true and I'm not sure where you're getting this. DSGE models do not always assume complete markets, representative agents, homogeneous beliefs, etc.
On asset market structure: There are DSGE models with basically anything you can think of. Complete markets, uncontingent bonds, financial autarky, partially complete markets due to limited enforcement/information asymmetry, etc.
On representative agents: Heard of Krusell-Smith (JPE 1998)? Since then we've been able to solve all kinds of models with heterogeneous agents and both idiosyncratic and aggregate uncertainty. For heterogeneous firms see Khan and Thomas (Ecta 2008) and its descendants like Bloom et al. Make no mistake. These are DSGE models. Hell, Aiyagari (QJE 1994) was a DSGE model (it had idiosyncratic but not aggregate uncertainty). Incidentally, all these papers have incomplete markets.
On beliefs: Sargent has a ton of stuff on adaptive expectations and ambiguity aversion (Knightian uncertainty). There's a working paper by Mertens and Hassan called "The Social Cost of Near-Rational Investment" which studies how consistent but small irrationalities can have large consequences. These are all DSGE models. And as an aside, rational expectations does not at all rule out heterogeneous beliefs. As long as some agents see different information they will rationally have different expectations. There are all kinds of DSGE papers with this feature as well (just search for "DSGE + Bayesian learning").
On price-taking: Ever hear of New Keynesian models? They have monpololistic competition. So do lots of papers that incorporate trade. Also DSGEs.
If you want to see what kind of modern models fit into the DSGE paradigm, just take a look at the papers that were presented at Minnesota Macro this year. The list is here:
DSGE is not "A" model. That's all he was saying. It doesn't make much sense to say it's wrong or "not even wrong" in Quiggin's words. It's a class of models, different versions of which can have many different implications. It's like utility theory - you can dispute axioms if you like but in terms of observable implications it's possible to rationalize almost anything. That doesn't make it bad or "not even wrong" -- only points out that you need to add more structure to get implications.ReplyDelete
When you're on the market, don't take a job that will require you quit blogging. Best laugh I've had all day.ReplyDelete
@Anonymous 2:00PM --ReplyDelete
You were right the first time; "self-indulgent" is one word.
See, there's this thing called the hyphen...
Also, too, kindly cite an example of Krugman or DeLong indulging (pun intended) in "personal nastiness". I am not aware of any nastiness on the part of either that would be considered "personal", i.e., ad hominem.
Stephen Williamson said...ReplyDelete
"Are you OK, Noah? What did I do to make you so grumpy?"
Yup - when he's losing he just tries to confuse the issue. Of course, he's pretty much confused already.....
In response to a few comments above, it might be a good idea to read the book, since Williamson's "review" gives readers literally no idea of its contents.ReplyDelete
1. I distinguish between strong, semi-strong and weak forms of EMH and argue that (a) the semi-strong version is widely (often implicitly) believed (b) the evidence doesn't support it
2. I'm well aware that DSGE models contain frictions, heterogeneous agents etc. The problem is that each new paper goes back to the standard model and introduces its own twist, so it's impossible to model an economy that's a long way from general equilibrium. I drew this point from Akerlof and Shiller and I have a lovely quote from Blanchard saying much the same thing (though with a positive view)
John (perhaps I should just write this on your own blog, but oh well):ReplyDelete
Imposing general equilibrium in DSGE models is really just making sure everything adds up, i.e., no goods are magically appearing or disappearing without being accounted for. Disequilibrium is meaningless. The economy can be in a stationary equilibrium or a non-stationary one, but it has to be in equilibrium. A DSGEer would say that the actual world economy is always in general equilibrium, for precisely the fact that everything would add up if you could realistically do the accounting. It's not a very strong definition.
Perhaps you're thinking about studying dynamics outside of stationary equilibria (or small neighborhoods thereof). This is done all the time, although not generally by the "low-tech" DSGErs like New Keynesians who rely heavily on perturbation methods.
If you disagree with this characterization of your view, it would be helpful for you to define your notion of disequilibrium, or at least give some examples.
I guess I should also note that there are lots of small open economy models that are considered DSGE, even though they're not strictly "general equilibrium." The developing country business cycle literature is full of them. I'd say that we should probably just call them "dynamic stochastic equilibrium" models, with the equilibrium part just there to make it clear that there are some agents that are making rational (or consistently irrational) decisions.ReplyDelete
"It really seems like a lot of non-macro people have this weird notion that DSGE is nothing more than Kydland-Prescott and its immediate descendants."
No. People are talking about the supposedly canonical models.The sort discussed in the Walsh textbook for example.
"DSGE models do not always assume complete markets, representative agents, homogeneous beliefs, etc."
Yes, of course. But one or two at a time, not all together. So while technically general equilibrium models, why should their implications be any more reliable than partial equilibrium models?
Also some people have been making policy relevant comments based on the canonical models. Others (eg Sargent) say that DSGE models cannot be used in the current circumstances. Yet others (eg Chari-Kehoe-McGrattan) say that policy relevant DSGE models do not yet exist (even during times of full employment), but will exist one day in the future.
So who is right?
"A DSGEer would say that the actual world economy is always in general equilibrium, for precisely the fact that everything would add up if you could realistically do the accounting. It's not a very strong definition."
It certainly is a very strong assumption, since it is not about everything adding up at a point in time, but everything, including peoples beliefs and beliefs about beliefs and so on adding up at every point in time for eternity.
"... it would be helpful for you to define your notion of disequilibrium, or at least give some examples."
Well, I can't speak for John Quiggin, but how about this from George Evans
Remember that the Internet is forever and you can never know who is connected to who. You could find yourself being interviewed by Williamson's best friend some day."
Keep your opinions to yourself, Noah. This isn't a hedge fund.
". . . who is connected to [whom]."ReplyDelete
"Yes, of course. But one or two at a time, not all together. So while technically general equilibrium models, why should their implications be any more reliable than partial equilibrium models?"ReplyDelete
What partial equilibrium models do you have in mind that you'd put your trust in? Aside from that, DSGE models are popular in part because their purpose isn't just to predict stuff. In fact, they don't predict out-of-sample particularly well at all. Steve W. himself says that he'd rather leave this kind of task to professional forecasters. DSGE models provide qualitative economic intuition that you simply can't get from large-scale econometric models. Impulse responses are popular tools in the macro literature not because they predict anything, but because they help to explain the qualitative insights models provide. Moreover, DSGE models provide intuition about general equilibrium effects (change one real variable, prices endogenously change, this affects other real variables, etc.) that partial equilibrium models simply cannot provide at all.
"Also some people have been making policy relevant comments based on the canonical models. Others (eg Sargent) say that DSGE models cannot be used in the current circumstances. Yet others (eg Chari-Kehoe-McGrattan) say that policy relevant DSGE models do not yet exist (even during times of full employment), but will exist one day in the future.
So who is right?"
I think you're a bit confused here. If I recall correctly, Sargent's view is not that DSGE models in general are not useful in the current circumstances, but that the workhorses like log-linearized New Keynesian models are not suitable for studying dynamics after large deviations from steady state like the one we're in now. And I'm sure that Chari et al have never said that there are no DSGE models that are useful for policy analysis. Their precise statement is that "New Keynesian models are not yet useful for policy analysis." They're talking specifically about New Keynesian models here. All the work that trio has ever done fits well within the DSGE banner (as Chari himself testified to Congress).
"It certainly is a very strong assumption, since it is not about everything adding up at a point in time, but everything, including peoples beliefs and beliefs about beliefs and so on adding up at every point in time for eternity."
What? This doesn't mean anything at all. It's not an assumption at all, it's a tautology. The world is in fact in general equilibrium right this very moment. It's impossible for this not to be true.
Oh, and that video you linked to on Economist's View is most definitely equilibrium analysis as far as I'm concerned. If that's what you call disequilibrium, you're simply not understanding what equilibrium means to macroeconomists to begin with.ReplyDelete
"DSGE models are popular in part because their purpose isn't just to predict stuff. In fact, they don't predict out-of-sample particularly well at all."ReplyDelete
Amazing. You think this is a feature and not a bug!
Yet these models are used for welfare analysis and for studying counterfactuals. Exactly how reliable are those conclusions likely to be?
So you like the story-telling aspect of these models. But why on earth do you trust these stories?
Reasons For Pessimism: Popularity For The Wrong Reasons
• They are better than structural VAR’s as tools for spinning elaborate stories about how the economy works, which may make them more saleable.
• We know these stories aren’t true. There is no K, there is no P, there is no C. There are no infinitesimal monopolistically competitive firms. Individuals do not respond instantly, continuously and rationally to the macroeconomic state. And that these things are all myths matters for how the data behave.
• It may be an advantage of VAR’s that they restrain our impulse to take our storyspinning too seriously. If Bayesian DSGE’s displace methods that try to get by with weak identification and in the process reinforce the excess weight we give to story-spinning, they may set us back."
------End of Sims's quote
"And I'm sure that Chari et al have never said that there are no DSGE models that are useful for policy analysis. Their precise statement is that "New Keynesian models are not yet useful for policy analysis." They're talking specifically about New Keynesian models here. All the work that trio has ever done fits well within the DSGE banner"
Of course C-K-M are DSGE. In the 2009 paper the only DSGE models they talk about are NK models (specifically Smets-Wouters) and then come to the general conclusion that, as you say, 'New Keynesian models are not yet useful for policy analysis'. Then they talk about their conviction that in the future the models will have policy relevant conclusions.
Narayana Kocherlakota said much the same thing in his widely quoted defense of DSGEs.
Oh, and Kussell-Smith: Try and explicitly note the assumptions. How much heterogeneity is there in the Utility function? In capital? In the Prodn Fn? Is unemployment possible? Why? Trade? Are you seriously claiming that the model tells us anything at all about the real world?
Finally, what on earth is tautological about different people's beliefs about the future being mutually consistent? If you have a model in which this is so, then it is an artefact of the model not a property of reality.
As always a discussion about macro with a dyed-in-the-wool freshwater chap has been somewhat surreal. It is as if one of us is on Planet Earth, the other on Planet Rational Choice - and an overly simplified Rational Choice at that.
This is my last comment. I have no intention of monopolizing the comments section.
DSGEs do NOT require that all agents be Laplace's demons. It is only required that the economist be one!
Without prejudice concerning Radiohead, Genghis Khan's strategy derived from itReplyDelete
"my strategy to overrun the world by slowly, uh, overrunning the world."
was quite ineptly executed, considering how fast the Mongol conquests were.
I once made the mistake of posting a comment on Dr. Williamson's blog, to the effect that perhaps how well the less well off did was an appropriate evaluating criterion for an economy in the real world, and the the US economy did not do very well by this criterion. Williamson replied that presumably if the poor did better and became less poor I would be less concerned about them. I haven't been back since.
I do, however, live in St. Louis, and it is a nice place to live between late September and early December and again between the middle of April and the middle of June. I'm sure that Dr. Williamson has some statistical criterion, of a piece with his general intellectual approach, which gives the desired conclusion.
That is a great Sims quote and it comes from a great article. Thanks for that!
If your model doesn't make good out-of-sample predictions, what reason does it give you to trust the impulse responses it generates?
Noah: Since my goal in using a model isn't generally to make out-of-sample predictions at all, I'm not concerned that my models don't perform well in this dimension. Impulse responses are useful for understanding how the model works.ReplyDelete
Frankly, I don't think macroeconomics is at a point where we can do any kind of remotely useful prediction of the future. Virtually all models rely heavily on shocks (TFP shocks, taste shocks, etc.) I share Narayana's distaste for this, but nobody's come up with a better way to do things yet. And you can't back out shocks from data that you haven't observed yet. VAR models are no better. They have "shocks" too (regression residuals) and only perform well when the economy is puttering along without doing anything too weird. Throw in a financial crisis and VARs go haywire.
Where models are more useful than VARS is in gaining a better understanding of past events/trends, especially anomalous things like crises, and performing analysis of policy counterfactuals. If you don't find this particularly interesting, that's fine.
Herman: Yes, of course the models are wrong. I don't think anyone would dispute that (maybe Ed Prescott or John Cochrane, but they're legitimately crazy). Nevertheless, don't think that making some simplifying assumptions means the model is incapable of generating any useful information. As I mentioned above, if the model fits the data in a particular event or time period well (and in a non-trivial way, i.e., it accounts for variables not used to calibrate it), then I think its implications for policy are worth thinking about. So yes, I guess I do think that the model's "stories" can help us think about the real world. But as Bob Lucas said in the essay linked below, reasonable people can disagree about this. Hell you'll probably find the whole essay ridiculous, but it's a pretty honest depiction of why we like models.
How does a model with no predictive power give you an "understanding" of past trends? A) How do you know the model you are using to give you "understanding" is actually a good one?, and B) If the model has no predictive power, doesn't that imply that past trends were dominated by things that the model cannot explain?
Same questions could apply to policy counterfactuals.
be using it for policy.
Joe/Herman, excellent exchange. This is how these sorts of arguments should be conducted. There are none of the ad hominems, sly underhanded jabs, etc., which so often characterize the econ blogosphere, and which Noah (shame on you) has adopted. Hopefully you two can keep debating, am learning quite a bit.ReplyDelete
I think everyone in the blogosphere recognizes lighthearted teasing for what it is, and not as an ad hominem attack... ;-)
OK, the very last comment; And unfortunately the longest. Hence chopped in twoReplyDelete
Glad to be able to repay you at least a little bit for your excellent blog. I forgot to post the link or even identify Sims's paper. Google and Scholar are a real blessing. Here is the link
I take your comment as a light-hearted one. However, just to clarify, the comment about Laplace's demons was about representative-agent rational-expectations equilibrium models, not DSGEs generally. And even then, for stochastic models it is not literally true, but effectively so. Now Joe will argue - with justification - that macro has moved on and at the frontier people work with much more sophisticated models. And he would undoubtedly be correct. Unfortunately a lot of people's intuitions are still based on those very simple models and this leads to bad policy advice. eg. the claim the Ricardian Equivalence rules out any role for fiscal policy or the recent 'Kocherlakota Controversy': Rajiv Sethi, George Evans, Economist(Free Exchange)
... cont from previous commentReplyDelete
Well, I guess we will have to agree to to disagree. I have no problem with that.
"[VARs] only perform well when the economy is puttering along without doing anything too weird. Throw in a financial crisis and VARs go haywire."
Yes. But maybe this is the best that we can do. The is a limit to the usefulness of time-series econometrics. There seems to be an attitude - I don't know whether you share it - that time-series econometrics is all there is to empirical macro. Surely this is unnecessarily restrictive.
"No one could be against time-series econometrics. When we need estimates of parameters, for prediction or policy analysis, there is no good alternative to the specification and estimation of a model. To leave it at that, however, to believe as many American economists do that empirical economics begins and ends with time series analysis, is to ignore a lot of valuable information that can not be put into so convenient a form. I include the sort of information that is encapsulated in the qualitative inferences made by expert observers, as well as direct knowledge of the functioning of economic institutions. Skepticism is always in order, of course. Insiders are sometimes the slaves of silly ideas. But we are not so well off for evidence that we can afford to ignore everything but time series of prices and quantities."
----End Solow's quote
"Where models are more useful than VARS is in gaining a better understanding of past events/trends, especially anomalous things like crises, ..."
Bob Lucas would disagree [via David Laidler, and Justin Wolfers]:
"... the problem that the new theories, the theories embedded in general equilibrium dynamics of the sort that we know how to use pretty well now – there's a residue of things they don't let us think about. They don't let us think about the U.S. experience in the 1930s or about financial crises and their real consequences in Asian and Latin America, they don't let us think very well about Japan in the 1990s "
----End Lucas's quote
And why would you trust the policy counterfactual analysis from a model that is not good at out-of-sample predictions?
"I guess I do think that the model's "stories" can help us think about the real world"
Fair enough. But so can stories told by models from different traditions. While this might be a legitimate argument in favour of DSGEs, it is not an argument for their exclusivity. Repeating myself from CT:
Again, my main point is about methodological pluralism. People should be allowed to work on many different approaches – old fashioned ‘mechanical’ models, agent-based models, various forms of VARs, even partial-equilibrium models. Premature convergence to a particular methodology without overwhelming empirical justification only discredits the practitioners and the field.
"Ed Prescott or John Cochrane, but they're legitimately crazy"
If you do realize this, there might be hope for you yet :).
Doesn't EMH imply a certain equality of knowledge between buyers and sellers?ReplyDelete
Hasn't recent history show that this isn't the case?
"I think everyone in the blogosphere recognizes lighthearted teasing for what it is, and not as an ad hominem"ReplyDelete
Light hearted teasing is something you do face to face with someone you know and respect and who knows and respects you. On the internet, with its lack of nuance, one man's tease can be another man's mortal insult.
I do not see a lot of light hearted teasing in the blogosphere. I see regular full on attacks on people's competence and integrity. Noah, you are too young to play that game and Krugman and Delong would do better to avoid some of their more petty remarks (and understand that I happen to agree with Krugman most of the time and I respect him almost all of the time).
What is it about econ blogging and the civility trolls?ReplyDelete
Who cares if someone is being intentionally offensive or just teasing? Is he right or not?
Also, everyone knows ";-)" is the universal internet sign for teasing. Sheesh.ReplyDelete
Taking my cue from Douglas Adams, who wrote a 'trilogy' of four books that turned into five ...ReplyDelete
Did you read the Lucas paper pointed to by Joe in his last comment?
I read it more than 36 hours ago and my reaction was stunned outrage and I wrote a fairly incendiary comment which I didn't post.
After a day and a half of reflection my views remain unchanged so that comment is appended below.
I just read the Lucas paper pointed to by Joe.
Incredible! Just incredible!
The stupidity, the idiocy, the scientific and epistemological illiteracy is just unbelievable.
Lucas needs to be in a padded cell, not a tenured position at any University.
If this how Chicago and Minnesota macroeconomists think, there is no hope and no future for macro at all.
Jolly Green said...ReplyDelete
" What is it about econ blogging and the civility trolls?
Who cares if someone is being intentionally offensive or just teasing? Is he right or not?"
The second answers the first. They're wrong, so they throw insults.
Barry, you just made another one of the informal reasoning fallacies. Look it up, it's in the list in Wikipedia.ReplyDelete
Fact is Quiggin and Noah are right. Stephen Williamson... well, if I ever interviewed his best friend, I'd find myself asking, "So, when did Williamson decide to stop doing science and become a cargo cult economist? Is he happy in his religion? How can we keep him out of positions of power?"
I am, incidentally, impressed at the failure of "joe" to understand the meaning of the word "general equilibrium", or the nature of science. Cargo cult science is very popular among economists, isn't it?
"What? This doesn't mean anything at all. It's not an assumption at all, it's a tautology. The world is in fact in general equilibrium right this very moment. It's impossible for this not to be true."
Which is simply false, as anyone who knows the meaning of the phrase "general equilibrium" knows -- if the world were in equilbrium it *would not be changing*.
The world is, in fact, according to every real, scientific economic theory, *NEVER* in full equilibrium. Most of the equilibrium theories in scientific economics propose that individual markets will *approach* equilibrium as time passes, given an absence of outside perturbing effects. This is a good first-order model. But (a) there are always perturbing effects, and (b) even when we can't spot the perturbing effects, in reality practically no market ever seems to be at equilibrium.
I find it astounding that someone purporting to know economics would actually claim that it's a "tautology" that the world is "always in general equilibrium". In fact, it's a statement about the nature of the world which is utterly disproven -- the world is never in general equilbrium.
"A DSGEer would say that the actual world economy is always in general equilibrium, for precisely the fact that everything would add up if you could realistically do the accounting. It's not a very strong definition."
(1) That's a redefinition of "general equilibrium". To mean "conservation law", not "general equilibrium".
In economics, equilibria were originally defined as quasi-stable states (stable given that certain outside parameters to not vary). It's easy for the economy to not be in such states. In fact, it's never in such states, because outside parameters are always varying.
The point of equilibrium analysis is that if you are at a non-equilibrium state you will be *naturally pushed* towards an equilbrium state by feedback (even with outside parameters held constant).
If you imagine that you've included all parameters, then technically you'd be in what you imagine to be general equilibrium, but guess what? You haven't included all parameters. Ever. That would require a full model of the entire world, every single science being accurate (so that not just psychology but also technology and environment are fully accounted for). This is impossible given the current state of knowledge.
(2) Even a conservation law is not even true. Frankly, money appears and is destroyed and so are goods (oops, plague of locusts this week!). There is no conservation law for most of what economics studies.
I'm startled and disturbed at the level of cargo-cultism floating around in economics circles, but pleased by the existence of people like Noah, Quiggin, and Herman.
1) Testing EMH requires an asset pricing model. In this sense, it is always a "joint test". I guess this is what Williamson was trying to say.ReplyDelete
2) DSGE vs VAR? Both are good tools and both have their uses. Not sure what one gains by criticizing the tools and not the specific applications or uses of them. That's again Williamson's point.
"Herman can take comfort in the fact he and Noah will be unemployed together."ReplyDelete
Not being an economics grad student, I don't that that is a problem I face. :)
OTOH macro grad students from Minnesota, Rochester or WUSTL seem to be in trouble:
Justin Wolfers :
Formally elegant but empirically irrelevant macroeconomists had a much harder time getting hired this year. Curriculum committees are also paying attention, looking to see classes that speak to real economic issues. Economists beyond the macro tribe are paying greater attention, and they aren’t willing to support the intellectually insolvent.
Economists outside the empirical tribe don't think that OLS constitutes economics, Justin.Delete
"Zombie Economics: A Review"ReplyDelete
John Quiggin in an Australian economist. He made his name in the early 1980s in an esoteric area called decision theory. The Econometrics Society made him a fellow on the basis of this work, a distinguished award. He writes a blog, which has many devoted followers. The book is primarily about macroeconomics, however, which is not his area. Asking Quiggin about macroeconomics is like going to a podiatrist for your headache: it's the wrong end of the body.
The chapter on Dynamic Stochastic General Equilibrium modeling (DSG modeling) is a good example of Quiggin's lack of expertise about modern macroeconomics. He states that one of the oddities about DSG modeling is the representative agent paradigm. This is an abstraction where the decision making of one representative consumer/worker is taken as a stand-in for the millions of people living in an actual economy. This abstraction was employed in a famous 1982 article by Kydland and Prescott. Finn I. Kydland and Edward C. Prescott justly won the Noble prize in 2004. The stand-in consumer was abandoned in 1994 in important work by the late and great economist S. Rao Aiyagari. Every graduate student in macroeconomics today knows the Aiyagari paradigm. This work is not mentioned in Quiggin. Nor is the celebrated work by Mortensen and Pissarides, done during the late 1980s and early 1990s, on modeling unemployment. Dale T. Mortensen and Christopher A. Pissarides won the 2011 Noble prize for Economics. There has been a flurry of work in macroeconomics embedding the Mortensen and Pissarides framework of unemployment into an Aiyagari/Kydland/Prescott style DSG model. An early example is the research by David Andolfatto in 1996. Interestingly, Noble Prize winner Paul R. Krugman's latest research with Gauti B. Eggertsson borrows from Aiyagari (they cite it) and is essentially a dynamic general equilibrium model, albeit with a very Keynesian flavor. Quiggin is really out of touch with modern economics.
The trouble with Quiggin's book is that to the non-economist his little bit of knowledge will sound authoritative. Like an undergraduate's essay, many of the bits and pieces are indeed correct. But, also like many undergraduate essays, it shows little understanding about modern macroeconomic, just a superficial dropping of names and theories. Beloved Albert Einstein, a hero for scientists, didn't like quantum mechanics and argued against it. Perhaps it was because of the escalation of the mathematics required to understand the quantum world. Some people say that Einstein wasn't good at math. The mathematics in his papers is easy for a modern economist or physicist to understand--look them up on the web. Time has advanced mathematical training among scientists. Anyway, this was one battle Einstein lost. When Keynesians displaced the classical economists in the 1940s, 1950s and 1960s the latter cried out about the mathematics (calculus and statistics) the former used. Keynesians, such as the Noble prize winners John R. Hicks, Lawrence R. Klein and Paul A. Samuelson, were at the forefront of technique in their day. And now it is the displaced Keynesian crying about the new math (dynamic programming, numerical analysis, stochastic processes) used by the neoclassical economists ushered in by the Kydland and Prescott revolution. Maybe the table will be reversed tomorrow. Who knows: if you could forecast this you could be a Noble Prize winner. This is the process of science: New ideas don't come easily and old ones are hard to displace. Professor Quiggin: You sound like an old man whining about the young Turks.
Wow, I know it's been like half a year since this post was made, but I was bored and these comments were a good read. Furthermore, I don't like seeing misconceptions go unchallenged on the internet where they can be seen and repeated.Delete
Mr. MIT - while Einstein's early work (special relativity, photoelectric effect, assorted stat. mech.) is pretty light mathematically and more or less understood by anyone with an undergraduate physics degree, GENERAL relativity, is just the opposite. Very few physicists outside of fields where it's relevant have expertise in it and many have very little mathematical grasp of it at all. This is in large part due to its mathematical difficulty. Einstein was pretty good at math, although his math skill was powerfully outshone by his creativity.
In contrast, the math required to explain Einstein's problems with quantum mechanics is easy and something every physics undergrad learns. QM doesn't begin to approach the mathematical difficulty of GR till you get into quantum field theory. Einstein's problem wasn't that the math was too hard, it was that one of the implications of QM (entanglement) was too strange.
Einstein and QM is even more interesting. His difficulties accepting the implications of QM culminated in a famous paper by Einstein, Poldowski (sp?) and Rosen (a paper so famous that it is known by the initials of its authors to all physicists), one of the results of which has been recent intensive work on quantum computing and cryptography. To simply attribute his problems with the implications of QM to a math lack is ignorance.Delete
it requires only that the aggregate effect of individual investors' irrationalities not be predictable.ReplyDelete
Actually you can make even weaker statements than this. You can have the effect of the irrationalities predictable, but various reasons why people can't exploit the arbitrage (e.g. time, capital, etc.)
are New Keynesian macroeconomic models anything more than a real business cycle model plus sticky prices and sticky wagesReplyDelete