I don't do much macro-dissing these days, but this Brad DeLong post makes me unable to resist putting in a quick dig.
DeLong quotes Robert Lucas from a 2011 panel on rational expectations:
[Using behavioral economics for macro] is like saying that we ought to build it up from knowledge of molecules or--no, that won’t do either, because there are a lot of subatomic particles.... We’re not going to build up useful economics in the sense of things that help us think about the policy issues that we should be thinking about starting from individuals and, somehow, building it up from there.DeLong responds:
[When Lucas] builds his models he is aggregating up the behavior of 310 million American individuals, having made certain assumptions about what things cancel out when that aggregation is made...
Lucas does say that economists definitely should not:
Instead Lucas says economists should:
- Model humans as they actually behave.
- Model their economic interactions as they actually happen.
- Aggregate up.
- And look for emergent patterns to fit to the data.
The second, Lucas says, is "science"! The first, Lucas says, is not--even though the first is a strict superset of the second.
- Do what macroeconomists currently do.
- This is, assume one infinitely-lived hyper-rational representative price-taking agent.
- Assume that all equilibrium-selection and coordination problems are automagically solved.
DeLong is exactly right. Also, I like the word "automagically".
Anyway, Lucas' ideas about modeling are pretty typical of modern macro defenders (possibly because he had a big part in inventing those ideas). Modern macro methods commonly face attacks from two other groups of economists:http://newmonetarism.blogspot.com/2015/03/lucas-and-his-critics.html
Attacker Group 1: "Old Keynesian" economists who want to use aggregate-only models.
Attacker Group 2: Decision theorists and other micro theorists who want to make macro use more realistic models of agent behavior.
The modern macro defenders' typical response to Attacker Group 1 is: "No, we need to model agents' decisions, because only then can we be confident that these are structural (and, hence, policy-invariant) models. Simply matching aggregate data is not enough."
The modern macro defenders' typical response to Attacker Group 2 is: "No, we don't need to model agents' decisions realistically, because the only measure of whether a macro model is successful is whether it matches the aggregate data."
This leaves modern macro defenders in the odd position of saying that it's crucially important to model agents' decisions, but totally unimportant to model them in a realistic way.
The natural next question for a macro troll to ask is: "If the agent's behavior isn't being modeled in a realistic way, how can we be confident that the preference parameters are really structural?"
Updates:
As usual, Robert Waldmann beat me to the troll.
Steve Williamson responds. He writes:
I must be missing something, because this doesn't sound like anything I've read in a macro paper or book. It also doesn't really make sense, but that's a topic for another day...
Meanwhile, in the comments, Tore Ellingsen asks me what I would work on if I were doing macro theory. A great question. I gave some responses, including extrapolative expectations and firm-side frictions and constraints (e.g. borrowing constraints). I also think there's a lot out there to mine from the I/O and the banking theory literature, but here my knowledge is just way too inadequate to name specifics. In terms of things people are doing now, I like the learning models of Evans, and the dispersed-information models of Angeletos, though I'm not sure if either of those were actually inspired by micro research.
In the comments, Steve Williamson says:
(Discussion continues in the comments.)
Updates:
As usual, Robert Waldmann beat me to the troll.
Steve Williamson responds. He writes:
[W]hen we say that macroeconomic theory has "microfoundations," what we mean is not that it is built up from theory that explains the behavior of individuals. For a lot of economic behavior, we're not going to do very well in explaining the behavior of an individual. And, as Lucas notes, behavioral economists can't do it either. Rather, what "microfoundations" is about is finding the model elements - optimizing behavior, constraints, information - that explain the behavior of large (that's large in the "larger than one but much smaller than 280 million" sense) groups of economic agents from first principles. Then we can make predictions about the effects of policy on the behavior of really large (i.e. 280 million for example) groups of people.I guess I don't really get this. So there are some intermediate-sized groups of people - much bigger than 1 but much smaller than 280M - who collectively behave like the "agents" that we see in macro models. But these intermediate-sized groups - let's call them "clumps" - don't behave anything like the individuals that comprise them. But we can still treat the parameters that describe the tastes of the clumps as structural. Why? Because of "first principles." Individuals don't have a disutility of labor parameter, or a coefficient of relative risk aversion. But clumps do. And we know that from first principles.
I must be missing something, because this doesn't sound like anything I've read in a macro paper or book. It also doesn't really make sense, but that's a topic for another day...
Meanwhile, in the comments, Tore Ellingsen asks me what I would work on if I were doing macro theory. A great question. I gave some responses, including extrapolative expectations and firm-side frictions and constraints (e.g. borrowing constraints). I also think there's a lot out there to mine from the I/O and the banking theory literature, but here my knowledge is just way too inadequate to name specifics. In terms of things people are doing now, I like the learning models of Evans, and the dispersed-information models of Angeletos, though I'm not sure if either of those were actually inspired by micro research.
In the comments, Steve Williamson says:
Your rewording of what I was trying to say makes no sense to me. That's not what I'm getting at...So, what are microfoundations? That's not building up a macro model from theory that predicts the behavior of individuals, as we don't have theory that does that. It's taking the best available economy (sic) theory we have - that somehow makes sense of average behavior - and we incorporate that in our macro model.I'm a little relieved that I got Steve wrong at first, since the "clumps" thing indeed made no sense. But unfortunately, this explanation still doesn't make sense to me. What is "the best available economy theory"? Best by what criterion? And what does it mean to "make sense of average behavior", if not matching macro facts? What else could it mean to "make sense of average behavior"? So I'm still confused.
(Discussion continues in the comments.)
You should encourage your readers to read the full transcript of the panel from which the quote was taken. In it, they would find a fascinating discussion of the intellectual history of rational expectations dating all the way back to John Muth and they would find that Lucas is not at all the zealot that Delong and Krugman would have us believe he is. For example, here is the full quote you are referencing:
ReplyDelete"One thing economics tries to do is to make predictions about the way large
groups of people, say, 280 million people are going to respond if you change
something in the tax structure, something in the inflation rate, or whatever. Now,
human beings are hugely interesting creatures; so neurophysiology is exciting,
cognitive psychology is interesting – I’m still into Freudian psychology – there are lots
of different ways to look at individual people and lots of aspects of individual people
that are going to be subject to scientific study. Kahnemann and Tversky haven’t even
gotten to two people; they can’t even tell us anything interesting about how a couple
that’s been married for ten years splits or makes decisions about what city to live in –
let alone 250 million. This is like saying that we ought to build it up from knowledge
of molecules or – no, that won’t do either, because there are a lot of subatomic
particles – . . . we’re not going to build up useful economics in the sense of things that
help us think about the policy issues that we should be thinking about starting from
individuals and, somehow, building it up from there. Behavioral economics should be
on the reading list. I agree with Shiller about that. A well trained economist or a well
educated person should know something about different ways of looking at human
beings. If you are going to go back and look at Herb Simon today, go back and read
Models of Man. But to think of it as an alternative to what macroeconomics or public
finance people are doing or trying to do . . . there’s a lot of stuff that we’d like to
improve – it’s not going to come from behavioral economics. . . at least in my lifetime."
So, Lucas says behavioral economics "should be on the reading list", that "there are lots of different ways to look at individual people and lots of aspects of individual people
that are going to be subject to scientific study", and "a well trained economist or a well
educated person should know something about different ways of looking at human
beings." Does this quotation match with Delong's description.
As to your main point: the main benefit of microfoundations is that many phenomena (and data!) can be explained as a consequence of one underlying cause (unification as you called it) and that theories can be checked against many different types of evidence. Do you have a sense that many of the facts about unemployment, consumption, output, etc. can be explained as a consequence of Prospect Theory? Are there any other findings from individual psychology that you think could withstand such scrutiny on the macro scale? If not, perhaps Lucas has grounds for skepticism...
"Does this quotation match with Delong's description."
DeleteYes. This has been the department of simple answers to simple questions.
"As to your main point: the main benefit of microfoundations is that many phenomena (and data!) can be explained as a consequence of one underlying cause (unification as you called it) and that theories can be checked against many different types of evidence."
DeleteAnd they fail dismally. So now what?
Keep trying. Do you have an alternative?
DeleteUm, no. What Lucas says is, someday in the future it may be possible to "model humans as they actually behave" and then aggregate up, but at the moment that is not possible. We can't begin to model even the interactions of two people, much less whole communities. But it is not reasonable to throw up our hands and say, economics needs to wait another century or so until psychology gets good enough to provide the basics for it.
DeleteThe physics analogy is not a bad one. Isaac Newton knew nothing about sub-atomic particles when he developed the laws of motion. He didn't need to aggregate up from basic data about micro behavior in order to make predictions about the behavior of objects in our macro world. And he made non-realistic assumptions (e.g., large objects like planets can be assumed to have all their mass concentrated at their centers).
Newtonian physics is not "correct" in the sense that at the extremes of the natural world - very small, very large, very fast - its predictions are wrong. But for the world of the size and speed that we live in, it works well.
Gary9:47 AM
Delete"Keep trying. Do you have an alternative?"
Read Krugman, and read Lucas/Cochrane for the past several years.
Who has been far, far more correct?
Insofar as Krugman has been correct, it doesn't really undercut my point. Has he been invoking the research on psychology or behavioral science in his analysis? As far as I can tell, he uses old school IS/LM when he wants something quick and dirty, and uses either a Lucas style cash-in-advance model or a Woodford style New Keynesian model with Eggertsson...both filled with intertemporal utility maximization and rational expectations.
DeleteThe notion that Krugman has actually been correct about anything is hilarious. It reinforces my point that the commenters here, and the blog writer as well, know essentially nothing about macroeconomics.
DeleteAnonymous, seems you are a genius. Time that you showed us your work.
DeleteP.S. I think the rational expectations mob are on a false trail, there are no invariant underlying parameters. History matters and equilibrium (at least as they define it) is illusory.
As opposed to you, who has no work at all. Good luck with your sad, angry life.
DeleteI feel like you've just completely left out modern New Keynesian models here, unless I'm missing something? Most good NK models don't have agents that are price makers, and as you know incorporate various 'frictions' such that the economy can't immediately equilibrate (and don't necessarily all assume infinitely long lived agent either). Are you saying that they don't count as truly micro-founded because e.g. Calvo pricing is too much of a fudge/not robustly decision based? Or did you not mean to imply that all modern macro models are of the style strictly advocated by Lucas (in that quote)?
ReplyDeleteThere's also the problem of ensuring the model is actually mathematically solvable, I think a lot of modern macro theorists are aware of the unrealistic assumptions of the agent in their model, but are inclined to stick with this for now because highly complex models of agents can render models completely unsolvable or highly chaotic and far too sensitive to small changes in parameters, so it's all we're stuck with until the next innovation in mathematical economics comes along. Further still there's the idea of it being a representation of economic actors on average, netting out stochastic mistakes of individuals. Further further still, there's issues like whether it's worth advocating a policy that would completely fail if actors have ration expectations; I would say not, regardless of whether actors in the real world always have rational expectations - I wouldn't ever advocate a policy that relies on people being stupid to work.
I meant to say "price taker" there.
DeleteWhat you are saying is let's stay with the theory that the world is flat until something better comes a long - even if we know it is not flat. Mathematical modellling is irrelevant. When it comes to making decisions what matters is the truth. Otherwise what you are doing is completely useless (and when it comes to the people that count - basically that is what they think). Do you really think when it comes to the crunch that a central bank governor or a head of state is going to take any notice at all of a rational expectations model with Calvo Pricing? If he does, I think his career will be very short.
DeleteOne possibility that could save this view is the SMD theorem. Very few of the details about the micro level survive aggregation to the macro level. In that case you can have agents that do pretty much anything yet result in the same model.
ReplyDeleteIn some sense, something like this has to be true ... your micro model has millions of dimensions (each agent having a couple of parameters) and the macro result has maybe 10's of outputs like NGDP or interest rates. Few of the agent details can survive that kind of dimensional reduction.
I have to admit, DeLong is a funny guy. My psychological (non-economic) assessment of his atrociously bad manner is that he must have had, at some time in his youth, an intellectual wedgie applied to him by Lucas and Co. Moreover, I conjecture that the latter did not even realize it, thereby stoking the fire of vengeance several degrees higher in our hot-headed commentator. Yes? No?
ReplyDeleteYou keep using this phrase, "intellectual wedgie". What is this wedgie? How can we define the intellectual underpants and the intellectual butt that it is shoved into? Do I want to know?
DeleteIs that econosplainin? I'd suggest practicing in private or some role playing with your peers before you try to bring that game online, David.
DeleteFunny David. When someone resorts to name-calling it's probably the last stage before intellectual failure. BTW, wasn't Lucas the guy saying that aggregate rigidities can't possibly be true just because micro rigidities do not imply macro rigidities?
DeleteNoah: "You keep using this phrase, "intellectual wedgie". What is this wedgie? How can we define the intellectual underpants and the intellectual butt that it is shoved into? Do I want to know?"
DeleteNoah, it's similar to what the right says about Krugman, that he must have something psychologically wrong, because he points out that people are incorrect. That's bad enough, but then he points out later that they're still incorrect, even with more data contradicting them.
Obviously, such a person is deeply disturbed, getting heated up over minor matters of truth and such.[1]
[1] Does not apply to the right, who are merely Righteous when they get mad.
This is David's theory about a lot of people - they're harboring some grudge, and that's why they act out. David himself is pretty normal, though sometimes I like to give him an intellectual wedgie just to see what he'll do.
DeleteMy guess is that these attitudes are just common currency among the people DeLong was educated with, and by. David and I know Lucas isn't a jerk, but some people seem to think so, apparently. Too bad for them.
David, have you seen his simply brilliant review Right from the Start? What Milton Friedman can teach progressives. J. Bradford DeLong
Deleteat http://www.democracyjournal.org/4/6514.php?page=all
Stephen, I think DeLong's grudge has something to do with years of destructive wrongness on Lucas' part. The fact that you're effectively talking about whether or not Lucas is an agreeable guy in person and it's too bad for them because they're missing out, I guess, is a bit sad.
Delete"...destructive wrongness..."
DeleteThat's a rather bizarre statement - also a bit sad that you think that. Apparently you're in that bubble too.
Excellent post. Much snappier and clearer than my effort to make the same point (to which you kindly linked somet time ago). http://rjwaldmann.blogspot.it/2012/03/modern-macroeconomic-methodology-modern.html
ReplyDeleteOr important but impossible, therefore we will just do what we want.
ReplyDelete"If the agent's behavior isn't being modeled in a realistic way, how can we be confident that the preference parameters are really structural?"
ReplyDeleteDo you have any thoughts on structural models in other fields (e.g. labor, IO etc)? I get why policy-invariant parameters would be great in principle, but some specifications of utility functions just seem way too ad hoc to make any substantive claim that they are fundamental to human behavior
The missing premise of all of these arguments is something to set up the goal of macro.
ReplyDeleteWhat is the goal? The reason macro gets attacked so much is that everyone thinks they know the goal, while I would get that very few people have a common understanding of the goals of macro.
There are many possible goals:
1) Stability
2) Efficiency
3) Highest GDP
4) Trade relations
5) Happiness
6) Peace
There is very little consensus on the goals. If we don't know the goal, how can we decide what model to use?
In politics, unemployment seems to be the main problems addressed, and this is pretty easy to solve in a crisis situation with pretty simple models. It is harder to address the longer term issues, and even harder if we essentially have no common goals.
The modern macro people's typical response to Attacker Group 2 is: "No, we don't need to model agents' decisions realistically, because the only measure of whether a macro model is successful is whether it matches the aggregate data."
ReplyDeleteBut then how do they avoid their own response to Attacker group 1? Because they too are just taking past performance as a guide to future performance. And it's a point reinforced by the consistent failure of DSGEs in general, and RBC models in particular, to remotely fit the data; they can't even predict the past without lots of Ptolemaic auxiliary hypotheses.
Put another way, how do they know that the relevant human behaviour is not actually fully policy-variant? In which case they have made no methodological improvement on the old models while sacrificing an awful lot of empirics and tractability.
Noah: "The natural next question for a macro troll to ask is: "If the agent's behavior isn't being modeled in a realistic way, how can we be confident that the preference parameters are really structural?""
ReplyDeleteUsing the physics analogy, it does look like the Lucasians are saying that one must buildup up the models for bulk matter based on the behavior of individual molecules, but that the models for the behavior of individual molecules don't have to correspond to reality. Also that if the models don't predict the bulk behavior of matter as well as the people who are starting with that, it doesn't matter.
Noah thanks for the link, but you already linked to that flabby old post of mine which is no where near as good as this post (totally aside from the excellent as always illustration).
ReplyDeleteMacroeconomics is not just aggregated microeconomics. There is this problem called fallacy of composition. Something happens on the way to the coliseum. Choices of one agent influence and restrict the choices available to other agents. Keynes knew this, hence paradox of savings, paradox of costs, etc. Economists forgot this circa 1980 with the ratex and monetarist counter revolutions. Economies are complex, adaptive systems. Yes you have to have an understanding of individual choices. But that is not enough. If everyone optimally choses for themselves 100% savings rate out of income flows, their will be zero income flows to save out of. How did we get so brainwashed to forget these fallacies of composition. Note that both of the macro modeling strategies Delong cites above are micro aggregated macro. That does not work - never has, never will - in the world we actually inhabit.
ReplyDeleteAs usual, Noah and his merry band of moron commenters (in particular, Waldmann, reason (surely an ironic name given the absence of any reason in this idiot's comments), and Parenteau) understand essentially nothing of the work they purport to be "destroying".
ReplyDeleteDo we really need such boring and childish trolling here? If you have something constructive to add to the conversation then do it.
DeleteBecause your comments are so incredibly useful?
DeleteGood Post. Short and sharp, if not very sweet.
ReplyDeleteNoah, you and many other commentators seem to agree that microfoundations are weak for the parameters used in current macroeconomic analysis. I wonder, is there any parameter in particular that you object to? Is is the elasticity of intertemporal substitution, or the risk aversion parameter (they need not be tightly related to each other) that is particularly untrustworthy or should be allowed to vary more across agents? Or, in models with many commodities, is there something about the substitution patterns that are assumed that is clearly contradicted by the evidence? Or are there any obvious additional parameters that should be introduced? - Of the many macro models that introduce more "behavioral" features (like habits or inattention) do you think that some might be on the right track?
ReplyDeleteGreat question, Tore!
DeleteI think one of the things that is most problematic in the current models is the Euler equation. It's at the heart of so many things, but it just can't seem to fit the data: http://noahpinionblog.blogspot.com/2014/01/the-equation-at-core-of-modern-macro.html
What can replace the Euler equation? Well, one idea would be to try ditching rational expectations for a subset of consumers, and replacing it with something like extrapolative expectations: http://www.people.hbs.edu/rgreenwood/bgjs9.pdf
That would introduce a couple of new parameters, while keeping the standard ones.
Another thing I think might yield even bigger dividends is to focus on more complex, realistic models of firm behavior. Most macro models have very unsophisticated descriptions of the objectives and constraints of firms. I think it's no coincidence that New Keynesian models, which add a tiny bit of realism to the firm side, are much better at fitting curves than their RBC cousins. Also, I think that the Bernanke-Gertler mechanism, which is all about firm constraints, has yielded some very promising results: http://www.newyorkfed.org/research/staff_reports/sr618.pdf
Industrial organization people have been working on describing firm behavior for quite some time. I'm not very familiar with the I/O literature but my instinct says it should be tapped for insights.
In terms of the recently popular banking-based financial frictions models, I think there are probably a lot of banking theory papers that could be used to improve on the mechanisms people are currently using. But I can't think of anything specific off of the top of my head (I plan to look into this when I have some time).
In terms of behavioral things, in addition to the above-mentioned extrapolative expectations, I think the next obvious one to try is overconfidence/heterogeneous beliefs. It's very tough to model. I would actually probably try it on the firm side first, rather than the consumer side. That's going to be hard, though.
Very interesting recent macro developments that I think are cool also include the dispersed-information models of Angeletos, and the learning models of Evans. Those are both things I would loosely describe as "behavioral", since I think there is no real difference between behavioral econ and information constraints.
I'm not up to speed on inattention-based macro models, though I know the sticky-information literature (and I'm a bit sad that it hasn't seemed to have worked out). Habits are good, and now ubiquitous, but they seem like they will probably end up as part of some larger framework.
How does that sound?
most problematic in the current models is the Euler equation
DeleteFirst: The equation does not seem to incorporate current expectations of future income. In my experience, expected future income has a huge impact on spending decisions.
Second: Using a time stepping model is computationally unstable if there are time scales in the underlying processes which are on the order of or smaller than the time step length.
Sorry. Should say "Using an explicit, first order, stepping model is computationally unstable ..."
DeleteNoah,
DeleteYou can obviously try any and all of those strategies and I'll guarantee you one of them will "work". Still, you are going to fail.
It's not the ontology, the possibilities are endless. Individual behavior is complex. It's the resolution of our telescopes.
Damn Noah: that's quite the research-agenda-to-save-the-world right there!
DeleteI know it's not your 'official' field, but seriously you should consider contributing to macro at some point in your career, howls and cries of 'outsider'/'heterodox'/'commie' notwithstanding. We need to stop letting the Andolfattos and Athreyas and Williamsons of that world keep defining the economic possibilities of our grandchildren.
To have a field one must publish papers. Noah doesn't have a field.
Delete" We can't begin to model even the interactions of two people"
ReplyDeleteThe ECONOMIC interactions of two people? Bet you we could. But notice - the two people would have ideosynchratic features that we need to model. Modeling two random people though is a completely different problem. I wonder if that is what he meant.
I get Williamson completely, but I don't think this is what people used to mean by micro foundations.
ReplyDeleteIt might make sense if you realize that a person acts in different ways in different situations, and that they can actually be part of more than one clump at a time, but acting out a different set of goals with a different set of assets.
"I must be missing something, because this doesn't sound like anything I've read in a macro paper or book. It also doesn't really make sense..."
ReplyDeleteYour rewording of what I was trying to say makes no sense to me. That's not what I'm getting at. Think of it this way. If you're talking to econ 101 students, and explain basic competitive consumer theory to them, you're expecting the ones who are doing any thinking to question it. They'll introspect and tell you: This is not what I do. I don't think about all the consumption bundles I might consume, how I feel about all those consumption bundles relative to each other, and then pick the one that I like best given my budget constraint. Then, your reply might be: Well, it doesn't really matter if this is not actually the way you think about the problem. If we represent the problem of a consumer in this way, and we look at the average behavior of consumers, this will help us understand what they're doing. Nevertheless, I understand that this theory may not help me understand what you, the student, do.
So, what are microfoundations? That's not building up a macro model from theory that predicts the behavior of individuals, as we don't have theory that does that. It's taking the best available economy theory we have - that somehow makes sense of average behavior - and we incorporate that in our macro model. If we do a good job, that should allow us to think about policy in a more serious way.
What's interesting about the Lucas quote from that discussion is that he really doesn't think about the world in the way some people might imagine he does. Pretty interesting, don't you think?
What's interesting is that, if you're right, it means that good predictions about microeconomic aggregates coexist with bad predictions about microeconomic agents, which might lead one to wonder whether good predictions about macroeconomic aggregates can coexist with bad predictions about macroeconomic agents (e.g., representative agents)...
DeleteOK. Good to see that I misunderstood, since what I thought you meant didn't really make sense.
DeleteBut I still don't understand this:
It's taking the best available economy theory we have - that somehow makes sense of average behavior - and we incorporate that in our macro model.
What determines the "best available economy theory we have"? What does it mean to "make sense of average behavior"?
"What determines the "best available economy theory we have"?"
DeleteThat's just the stuff of everyday economic research. It's in the papers we write, what we discuss at conferences, our experience in applying the science to practical policy problems. Impossible to write that down in a few sentences.
Here's another thought. This I think points out what is at the basis of what makes economics so different from natural sciences. We can take what we know about Physics, make some very precise calculations, launch some people off the planet Earth and land them on the moon. But we could have very detailed data on Joe Schmoe, put the best economists in the world in a room, and they're going to do a crappy job of predicting Joe Schmoe's behavior over the next year. That's not the fault of economic science, it's the nature of the problem. It's part of what we mean when we say all economic models are wrong, but some are more useful than others. Macro models in which we proceed as if all economic agents are optimizing given constraints have proved useful. Show me ones that are better, and I'll use them. That the micro behavior is "more realistic" from your point of view doesn't mean the models are useful.
DeleteThis is a good thought, but it should also point out to most people why macro economists have to extra vigilant in determining if their policy recommendations are humane given that they don't like to think about individual people. Do most economists test their recommendations against anything besides GDP? Most do not. And most have no concept of what creates or destroys stability, and many don't even care about the effects very much.
DeleteThis is why it is important to have ethical and morally grounded macro economists. This is also why those who approach economics as a science and resist any attempts to drive it outside of science will generally fail.
If the usefulness of the model is what characterizes it's success, then we've come full circle and have good reason to accept ad-hoc models, even old-fashioned Keynesian models provided that they can explain something.
Delete"This is why it is important to have ethical and morally grounded macro economists."
DeleteHoly cow. Are you thinking your ethics and morals are superior to those of your average macroeconomist?
"even old-fashioned Keynesian models provided that they can explain something."
That's the problem. They can "explain" anything. They place no restrictions on what you should see in the data. In other words, they explain nothing.
Steve -
DeleteI still don't understand. Suppose we have some micro models that fit the data pretty well for markets with 100k people, and those models are based on optimization of a representative agent. But when we look at the behavior of 200 or 500 people in that market, we see them behaving nothing like the representative agent in that model. Wouldn't we then conclude that the 100k-person market exhibits emergent behavior? And if the 100k-person market exhibits emergent behavior, why should we care whether the models we use to describe it are based on optimization? Also, why should we assume that the parameters that describe that representative agent are structural, policy-invariant parameters?
That's what I'm really getting at here.
Oh, and Steve, one random note: We do have some models that explain individual behavior really well in some situations: https://wiki.ece.cmu.edu/ddl/index.php/Introduction_to_random_utility_discrete_choice_models
DeleteThose models aren't especially complex, either. Just Econ 101 stuff with random terms.
The fact that the models are not especially complex and just assume uncertainty about a lot of structure is exactly the key to their success. As is the narrowly defined scope of the problem. Finally, the model predicts an aggregate (ridership) not behavioral properties of anybody in particular.
Delete"What determines the "best available economy theory we have"?"
Delete"That's just the stuff of everyday ... what we discuss at conferences, our experience in applying the science to practical policy problems. Impossible to write that down in a few sentences." ... there you go!
Perfect breeding ground for nepotism, academic cliques, hiring your own, gatekeepers at academic journals, ...if you ever wanted an example of ivory tower research...look no further
"But when we look at the behavior of 200 or 500 people in that market, we see them behaving nothing like the representative agent in that model."
DeleteThen, of course, you have a bad model. Better think hard about what's wrong with it. If you've gone about constructing it in a useful way, the ways in which the model does not fit might tell you how to fix it.
"Wouldn't we then conclude that the 100k-person market exhibits emergent behavior?"
No idea what "emergent behavior" is. So I guess I could not conclude that.
"Also, why should we assume that the parameters that describe that representative agent are structural, policy-invariant parameters?"
If they're not structural, they're not structural. You can follow the book and specify everything from the ground up - preferences, endowments, technology, information structure - but if it's a poor model, it's a poor model.
You said that we should not aim at fitting the behaviour of individuals, but then "you have a bad model" if you do not match the behaviour of 200 or 500 people, even if you match that of 100k people. So, where does the border lie? At which level should we stop?
DeleteThe impression I have is that representative agent micro-founded models are also aggregative models: they try to fit aggregate behaviour, not individual behaviour. They are different from traditional models (IS-LM, AS-AD and so forth) in that they use the representative agent in a more formal and internally consistent way. Indeed, also the IS-LM is a representative agent model. When you say that expansionary monetary policy shifts the LM curve right, reduces interest rates and increase output, you are implicitly asuming a representative agent who observes lower rates and invests more. There is no heterogeneity there.
Oh. "Emergent" just means something you couldn't have figured out from looking at the pieces. I don't know another word for that.
DeleteIf they're not structural, they're not structural. You can follow the book and specify everything from the ground up - preferences, endowments, technology, information structure - but if it's a poor model, it's a poor model.
Makes sense. But then how do you tell if something's structural? Just fitting past data doesn't make stuff structural - just ask Lucas. How can we tell?
"There is no heterogeneity there."
DeleteThe IS-LM model does not tell us anything about the individuals that live in that world. Are they all the same, different, how many are there? We don't know. Seemingly the people who buy the consumption goods aren't communicating with the people holding the money, doing the investing, and supplying the labor - there's a separate little model for each of these things.
"How can we tell?"
You seem to want everything written down as a flow chart. It's case-by-case. Take the Phillips curve (please). Some people were treating the observed Phillips curve correlation as structural. Then the data evolved in a way that made it obvious the Phillips curve was not structural. Then people come up with theories to explain what is going on, and that becomes the new candidate structural model. If it becomes obvious that's not structural, you move on to something else. Sometimes people have a hard time with the "becomes obvious" part as you know. There are always vested interests.
My impression of Real Business Cycle therorists are that they are our modern-day "Geocentric" astronomers of yesteryear. They have really impressive math and theories to explain how their theory could work, but empirical evidence contradicts them.
ReplyDeleteI know this sounds glib but I think it is a good summation of Chicago, Minnesota, et. al.'s ideology.
Please feel free though to slap my snark down above Noah!
(Also, it would be interesting to apply Thomas's Kuhn's theories of scientific progress and revolutions to the last 100 years of econ).
Frank
The problem with geocentric models wasn't that they couldn't fit the data, it was that you had to overfit in order to fit the data. I think we see the same thing with DSGE macro models. Use a basic RBC model, and you can't fit the data. Add a bunch of bells and whistles (more shocks, different utility, Calvo pricing, and what-not), and you can fit the data, but you can't really predict anything except the numbers you added all those bells and whistles in order to fit.
DeleteBut right now there's no "heliocentric" model around to replace that overfitted "geocentric" model. People are looking around for something new, but there hasn't been a Galileo moment yet. And there certainly hasn't been a Newton moment, where we learn how business cycles come from individual behavior.
Noah,
DeleteThanks for the critique of my analogy.
Frank
"So I'm still confused."
ReplyDeleteNope. Williamson is just a Neoclassical crackpot who idolizes theory while giving a crap about data.
If you know what I do, you would understand that I'm actually much further from neclassical than Mike Woodford.
DeleteThe 'micro vs macro' dichotomy seems misplaced to me. Ditto the 'micro' in microfoundations. Prof Williamson has indicated as much. The real difference is between modeling founded on what might be called informed, forward-looking optimization (RBC etc.) by the herd versus one dependent on aggregate behavioral norms that don't necessarily accord with utility maximization. Is that fair?
ReplyDeleteMaybe the problem is with what is being maximized?
DeleteWhat you need to do is to take simple behavioral rules (optimization might do or something else), which, we suspect, are robust enough to apply to everybody and majority of situation and build models of aggregate behavior and behavior of aggregates from those elements. Any attempt to build special behavioral rules, which apply in some contexts but not in others, or are affected by this cultural phenomenon or another are GUARANTEED to fail.
ReplyDeleteBehavioral econ might be useful in very tightly defined circumstances in which the properties of the set-up are very simply related to a specific behavioral trait or regularity.
However, for aggregate behavior, this strategy will, and I say it advisedly, Never, Ever work.
This is why economists cannot seem to find a better, more accurate model:
ReplyDeleteYou need to stop thinking of models in terms of people or groups of people. Rather, you need to group economic activity by role groups. The representative agent in this model doesn't fully represent a person, it doesn't fully represent any group, rather it represents the size of a group acting in a role and the economic activity under its control, combined with its information set.
It is pretty simple.
I don't know why nobody seems to be able to make sense of this. Just look at a person:
ReplyDeleteMost people in our economy act in different roles with a different set of information and motivations:
1) Worker
2) Consumer
3) Investor
4) Debtor
5) Creditor
And the most hideous of the roles:
6) Rentier