Who Needs Microfoundations?
I’m currently reading Justin Fox’s new book, The Myth of the Rational Market, which is recommended. Specifically, immediately before leaving for work I read a brief aside he wrote about the formation of a Kuhnian “paradigm” in economics in the second half of the twentieth century and economists’ pride in having achieved that kind of methodological consensus and thus elevated themselves above the ranks of mere sociologists or some such. That’s what sprung to mind as I read this paragraph from Steven Levitt:
You might think that macro forecasting would be an important part of what academic economists would do, but in practice there is almost nothing of that sort being done. That sort of thing is left for economists at places like the Federal Reserve or private banks to do. You might think that the models that most successfully explain economic patterns would rise to the top, but in the current regime, if they are not meticulously constructed from “micro foundations,” they aren’t allowed to be considered.
From an outside perspective, what seems to be going on is that economists have unearthed an extremely fruitful paradigm for investigation of micro issues. This has been good for them, and enhanced the prestige of the discipline. No such fruitful paradigm has actually emerged for investigation of macro issues. So the decision has been made to somewhat arbitrarily impose the view that macro models must be grounded in micro foundations. Thus, the productive progressive research program of microeconomics can “infect” the more troubled field of macro with its prestige.
Which, as a sociological matter, I think you’d have to say has worked.
But as a methodological matter, it seems deeply unsound. As a general principle for investigating the world, we normally deem it desirable, but not at all necessary, that researchers exploring a particular field of inquiry find ways to “reduce” what they’re doing to a lower level. To make that concrete, in the modern day we have achieved a decent understanding of how principles of chemistry are grounded in physics’ understanding of the behavior of atoms. But it’s just not the case that advances in chemistry were made by demanding that chemists ground all their models in subatomic physics. On the contrary, chemistry moved forward in the first instance by having chemists investigate issues in chemistry and see which models and theories held up. Similarly, though psychology is intertwined with the detailed study of the biology of the brain, it’s not deemed illegitimate to research psychological issues in the absence of a specific neurological theory. Nor, for that matter, do microeconomists generally deem it necessary to explore in detail the psychological foundations of their models. The models are, rather, judged by whether or not they produce fruitful insights about economics. Trying to enhance models with better information about psychology isn’t against the rules, but it’s not required either. What’s required is that the models do useful work.
So why should it be that “in the current regime, if [macro models] are not meticulously constructed from “micro foundations,” they aren’t allowed to be considered”?
65 Responses to “Who Needs Microfoundations?”
Register
RSSLogin
Topic Cloud
ArchivesBlog RollAbout Matt YglesiasContact Me
Donate
June 4th, 2009 at 2:37 pm
but wait a second, matt.
levitt says no one will do macro *forecasting* without being able to ground it in micro foundations.
you seem to be saying that no one will do *macro* without being able to ground it in foundations.
but clearly lots of people do *macro* in academic contexts, and do it with the same independence from micro that chemistry enjoys from physics. krugman, summers, the whole neo-keynsian gang are doing nation-scale, economy scale work without trying to tell rational-actor just-so stories at every stage.
and they do not depend for their prestige on micro, either: they have the prestige of keynes (or of friedman, for what it’s worth).
so i think maybe you are making a far more general, and less plausible, claim than levitt did.
June 4th, 2009 at 2:39 pm
B/c the greatest modern success of macro came from adding a micro foundation: debunking the unemployment-inflation tradeoff curve. Plus, it is difficult to do expts in macro, so models should be built using a foundation that must be true: individual behavior.
June 4th, 2009 at 2:42 pm
The reason is fairly simple Mat. We can predict things very easily, its called a VAR and not much can beat it. But why it works is because macro variables are fairly persistent. Its similar to predicting the sun rise, you can do it but tells you very little of why the sun rises when it does.
June 4th, 2009 at 2:42 pm
So the decision has been made to somewhat arbitrarily impose the view that macro models must be grounded in micro foundations. Thus, the productive progressive research program of microeconomics can “infect” the more troubled field of macro with its prestige.
Matt, I think I agree with you and the author that this is problematic. But wasn’t “arbitrary,” in fact, it was pretty logical. If you have good models that work on the microlevel it would seem to make sense that you could build on them to create a macro model that was based on what you believe to be a firm base, especially because we know that what the economy “is” is the interaction of millions of individuals. Now, I agree that if you can’t do that in a way that reflects reality and you keep needing to use models that aren’t based on microfoundations that means it might not be possible. Even if it is possible, any macro model will need to reflect reality in ways at least as good as our current macro models. But the quest for a macro model based on microfoundations is neither stupid nor arbitrary.
June 4th, 2009 at 2:44 pm
Doug is wrong. There is no debunking of unemployment and inflation. The Taylor rule exists and its used.
June 4th, 2009 at 2:44 pm
It’s particularly a problem given that the microfoundations of microeconomics are probably also fundamentally flawed. We’ve learned a lot about the cognitive biases that affect economic decision-making, but very few microeconomic models incorporate that research. Since Fox is so keen to present economics as physics’ younger brother, it might be apt to point out that Newtwon was, in fact, wrong about a bunch of fairly important things. How successful was microeconomcis in predicting the growth of the subprime housing market? Or the reckless financial behavior of firms like Merrill Lynch?
June 4th, 2009 at 2:48 pm
People (at the Fed, even) use models without microfoundations to forecast sometimes — Levitt is just making up things when he says “they aren’t allowed to be considered.” The reason that economists like to use microfoundations in models for forecasting is because of policy changes.
Say you want question: what will output be in three years if Obama cut’s taxes next year? It wouldn’t be right to just project current output out three years, because agents behavior will change in response to changes in the tax code. One way to account for these changes is construct a model with “microfoundations,” i.e. some paradigm for explaining individual agents make decisions under constraints and uncertainty.
In any event, it’s simply not true that reduced-form models aren’t being used in forecasting. In fact, one the most active research topics in the field is combining DSGE (micro founded) models with VAR/Factor (non microfounded) models.
June 4th, 2009 at 3:00 pm
To make that concrete, in the modern day we have achieved a decent understanding of how principles of chemistry are grounded in physics’ understanding of the behavior of atoms. But it’s just not the case that advances in chemistry were made by demanding that chemists ground all their models in subatomic physics. On the contrary, chemistry moved forward in the first instance by having chemists investigate issues in chemistry and see which models and theories held up.
This is actually completely wrong. Boyle kicked off the movement toward modern chemistry by attempting to develop an atomic theory of matter in The Sceptical Chymist. That movement was greatly accelerated when Lavoisier incorporated his law of the conservation of mass, a reductionist theory which set the basic modern understanding of chemical reactions, allowing quantification and precise predictions. Proust added the law of definite proportions, the idea that particular chemical compounds were made up of fixed whole-number ratios of particular elements. Dalton’s derived from these two foundational notions his atomic theory of chemistry, which in turn underpinned his theory of gasses. In many ways the subsequent 19th and early 20th Century developments in chemistry can be seen as the gradual victory of the atomic theory of chemistry (see, e.g., the work of Avogadro, Arrhenius, and Faraday) as it provided novel, and testable, predictions. And finally, all this was brought together into a more or less complete theory of chemistry by the Rutherford/Bohr model of atomic structure combined with quantum physics, which gave us a theory of chemical bonds, and was used to derive the helical structure of DNA by Watson and Crick.
In short, before incorporating atomic theory and other such reductionist theories, chemistry was little better than what we would call alchemy, and it was transformed into a quantifiable modern science yielding valuable novel predictions and observations precisely by the incorporation of these bottom-up notions.
So, I would suggest a different example if you want economists to stop viewing micro as the preferred path to understanding macro.
June 4th, 2009 at 3:02 pm
Good post. Except we economists have also failed at building a “fruitful paradigm for investigation of micro issues.” Well, I guess it’s a fruitful paradigm for *investigation.* It just doesn’t give correct answers there either.
Sorry. It’s an unfortunate time in the world’s history for the economics profession to be so bereft.
June 4th, 2009 at 3:02 pm
Science deals with the physical world. Economics deals with human behavior. Human behavior in relation to individual and cultural beliefs about abstractions. Abstractions relating to value, profit, money. Economics is not a science.
Economics can use mathematical means but in every case has to set a starting condition which is in fact arbitrary. Then use rules which are impermanent. The rules always change which is why economists are famous for saying ‘on the other hand’.
Economic activity on the macro level, and there was no macro level until the stone age ended, is a sociological phenomenon driven by political interests.
In 500 years, more or less, what are today’s economic orthodoxies will be seen as bizarre superstitions which lead to mega death.
June 4th, 2009 at 3:14 pm
There are two issues here. One concerns mathematical facts, and other is empirical.
On the mathematical fact issue: microeconomic general equilibrium models of the kind needed for a foundation for macroeconomic models do not predict any particular identifiable patterns. There are enough unobservable micro level variables such that individual level optimization and equilibria are consistent with anything youi can see at a macro level. So conceptually, macroeconomics has a problem that chemistry and physics and genetics do not.
If you make enough assumptions, you can hypothesize an ‘representative agent’ model of the economy as if there were just a few types of agents who are maximizing utility or profit (say: a consumer, a retailer, a middle man, a manufacturer, a bank and a government that does fiscal policy -and including all those types would produce a rather complicated model). This type of model will produce observable patterns as a function of micro variables that can be matched against observable aggregates. These models, beloved of conservative free market types (eg, the Real Business Cycle Schoo), in historical fact, did not match observable aggregates very well, which is why the real business cycle school abandoned traditional methods of statistical inference for their models. Lately Keynesians who work from microfoundations have advanced the methods so that more formal statistical tests are possible.
We do know that some macro variables, for example, the shape of a aggregate demand curve, cannot be determined from microlevel variables, but observable macro characteristics can be determined from system level variables, such as the distribution of disposable income.
The insistence upon microfoundations has been a methodological error, and it has held back progress in the field.
For more info, read the introductory article “Whom or what does the representative individual represent?”
Alan P Kirman – The Journal of Economic Perspectives, 1992.
Also read the introductory chapters to
Imperfect Knowledge Economics: Exchange Rates and Risk by Roman Frydman, Michael D. Goldberg, and Edmund S. Phelps
edcon is correct that
“The reason that economists like to use microfoundations in models for forecasting is because of policy changes.”
The problem is that observable aggregate economic variables can only be derived from micro level variables if very special conditions hold at the micro level, but they may not hold, and it is difficult to determine whether they hold.
June 4th, 2009 at 3:15 pm
I’d say the problems are actually made worse by the cross-cutting effects of theoretical vs. empirical investigation. These days, economists start with a theoretical model of how economic transactions work (the micro) and build up from there to get to their model of how the economy (the macro) should work – and then, when the model doesn’t match reality, they don’t really go back and re-think their ground-level assumptions, they add on Ptolemaic circles to get the reality to line up with their theory.
From an empiricist viewpoint, why should it matter that the individual and the group don’t behave the same way – there’s a reason why sociologists don’t all become psychologists, or why quantum physicists aren’t very good at space science. Why not just build the two separately, get them both working, and then look for a unifying principle, rather than insisting on one off the bat?
June 4th, 2009 at 3:18 pm
Ok, so why, if you can’t actually test it through accuracy in predicting outcomes, do you need a micro base again? I don’t see a problem with macroeconomics having the same level of scientific rigor as other social sciences – maybe the problem is the perverse allure of physics in the post-war world that convinced everyone that the only way to make progress was to mathematize everything and find fundamental laws, instead of starting with observation.
June 4th, 2009 at 3:21 pm
immediately before leaving for work I read a brief aside he wrote
So, to clarify, today’s timeline including reading a brief before work, then the gym, then blogging? You must get up awfully early.
June 4th, 2009 at 3:22 pm
“No such fruitful paradigm has actually emerged for investigation of macro issues.” You’ve answered your own question, Matt.
Also, physics investigates something real, chemistry investigates something real. Microeconomics investigates something that is real only because human beings are around to make it real- namely, intentional human interactions. What is the real thing, interesting independent of and yet reducible to microeconomics, that macroeconomics is supposed to be investigating?
June 4th, 2009 at 3:25 pm
Quick notes.
One: the foundations of microeconomics eventually lead down to the axiomatic assumption of rationality. That is a pretty huge assumption! What good is building things up from the idea of individual behavior if our model of individual behavior itself is inherently flawed?
Two: some of the great properties of micro actually do not aggregate well to the macro level. There are many good arguments as to why this doesn’t matter, but there are also many microeconomists who will tell you that macro does not have nearly the sturdy microfoundations its proponents claim it does.
I remember one conference where a particularly bitter microeconomist really tore into a macro guy about such a detail. After all the excuses fell flat, the macro guy eventually just admitted, “well, if we don’t ignore that flaw then we can’t proceed, and in the end, proceeding with a flawed model is better than not proceeding at all.” The microeconomist came back with, “But if you’re going to ignore that crucial detail, why bother with any of the other micro foundations at all then?”
So the question becomes “why indeed?” Especially if one doubts the axiomatic assumptions of micro itself.
I actually do think that micro foundations are a good idea. I just wished people thought a bit more about the weaker links in the chain, and I definitely won’t fault anyone for ignoring micro foundations altogether if the model has good predictive powers.
June 4th, 2009 at 3:26 pm
re: jl
I agree with most of what you’ve written, but note that you can a have (and the field is working towards) a microfounded model without representative agents.
June 4th, 2009 at 3:27 pm
“The reason is fairly simple Mat. We can predict things very easily, its called a VAR and not much can beat it” How sad.
Hey Rob, where have you been for the last eight months? Have you ever heard of Nassim Taleb or Benoît Mandelbrot? You’re so brain dead and reality resistant that you must be an economist by profession.
June 4th, 2009 at 3:31 pm
More fundamental to the problem is the idea that economics is primary and culture a mere superstructure (in both it’s marxian and classical forms). It’s not just wrong, it’s utterly insipid.
June 4th, 2009 at 3:33 pm
Excellent post.
Let’s not forget that one big reason microfounded models gained widespread acceptance is that the only ones that are easy to solve are ones that imply government should never do anything.
June 4th, 2009 at 3:42 pm
While I’m the first to acknowledge that the current macro research paradigm has given us little useful analysis, there is a standard answer to Matt’s question.
You can start by going to Wikipedia and reading about the Lucas Critique. I’ll try to briefly explain here. The idea is that any macroeconomic model that does not include microfoundations must necessarily be based on observation of historical relationships. If you don’t understand what’s going on at the individual level, the best you can do is to make ad hoc models that appear to explain the past dynamics of aggregate macroeconomic variables.
There are a number of problems with this. First, if you’re not constrained by the assumption of individual optimization, you can make pretty much any model you want, and there are a lot of fallacious models with no predictive power that you can force to “fit” the data you’ve observed. (Astronomy before Copernicus, anyone?) Second, if you’re using your models to make policy decisions, then the resulting change in policy may itself change the macroeconomic dynamics you’ve observed. If you have no idea what’s going on at the micro level, how the hell can you know what a policy adjustment will do?
And this isn’t just an abstract argument — there’s a famous economic policymaking episode that accompanies it. In the 50s and 60s, economists noticed a close relationship between unemployment and inflation, the “Phillips curve.” They theorized (without microfoundations, of course) that there was a fundamental tradeoff between the two. And why not? This “model” certainly appeared to fit the data. It also had a clear normative implication: since inflation isn’t as bad as unemployment, it’s reasonable to accept a few extra percentage points of inflation for a significant reduction in unemployment.
Then we actually implemented a loose monetary policy with this tradeoff in mind. What happened? Exactly what any model with microfoundations would have predicted: people built in a certain level of “expected” inflation, and the long-term increase in inflation didn’t lower unemployment at all. But economists who had simply eyeballed unemployment and inflation and estimated a functional relationship between them had no way to know this would happen ahead of time.
Now, microfoundations can become an unhealthy obsession, and it’s clear that modern business cycle macroeconomics hasn’t accomplished very much. But there were deep, logical reasons for the research paradigm the profession chose. Macroeconomists weren’t all dimwits trying to ride off the prestige of micro.
June 4th, 2009 at 3:45 pm
DTM @ 8,
How did a quantum understanding of chemical bonds lead to the discovery of the structure of DNA? The correct structure for DNA was deduced from X-ray diffraction patterns (which had been used half a century earlier for determining the structure of inorganic crystals, a while before Pauling et al used quantum theory to explicate the chemical bond) and from Chargaff’s findings that DNA always had the same percent A as T and C as G, which relied on pre-quantum analytical techniques.
June 4th, 2009 at 3:47 pm
Scientists are well aware of the issue of scale: some formulations are scalable while others are not.
For example, the issue of light as discrete particles(photons) or waves. The scale of the investigation determines which treatment yields the best prediction.
Likewise, an assumption that is valid for individual actions may not be predictive on a larger scale.
June 4th, 2009 at 3:50 pm
The problem is which micro the microfoundations are coming from. Garbage in, garbage out.
June 4th, 2009 at 3:54 pm
I agree with edcon, that there is very interesting new work that does not use the representative agent model, and may provide better understanding about the how micro level parameters and variables and the disaggreageted individual market equilibria that micro loves so much, produce aggregate phenomena.
I think that insistence on microfoundations also came from a desire to say something about individual welfare. But I think that aggregate phenomena such as total employment ratio, might have predictive power about welfare for the aveage dude getting RIF-ed.
Some of the new work does not look like traditional micro or macro. For example some recent work is more in the tradition of econophysics, except applied industrial organization (distribution of firm sizes, time paths of firm birth, growth and death) rather than financial markets.
Alan P. Kirman looks at heterogeonous agent models and in a more traditional way, looks at how macro variables are effected by micro structure. Frydman takes a more aggregate approach and uses statistics from non-stationary time series.
I think it is true that there was a time where most researchers insisted on “rigorous” representative agent microfoundations (that I think are simple minded and very speculative). Being a statistician, I was very suspicious when they abandoned traditional statistical approaches to model evaluation when “too many good models were being rejected” -that is a direct quote from the Frydman book. Their theory and subjective notions of what made a good theory did not hold up well against empirical data, so they just changed the goalpost, IMHO.
The free market Real Business Cycle school’s insistence on microfoundations limited research in more fundamental ways -for example, the cause of business cycles, or determinants of their severity, were pretty much off limits as a research topic. Those things, which most regular people I know assume are important things to understand, where a priori assumed to be exogenous shocks that were beyond the scope of economic research.
June 4th, 2009 at 3:58 pm
Read more Krugman/Delong and less Chicago-based nonsense when it comes to macro and this is a non-issue.
June 4th, 2009 at 3:59 pm
It is perfectly reasonable to use models without micro foundations for forecasting if the actions you take are too small to affect the economy as a whole significantly. However, it is not possible to use reduced-form models to craft public policy. This is known of Goodhart’s law – any time that we try to exploit some regularity between macro variables, it breaks down.
Why? Because people and firms’ rules of thumb reflect in some way expectations of future public policy. If we use the reduced form models to do public policy, then eventually economic agents will figure out that their expectations of the way public policy is conducted is wrong, and they will change their behavior accordingly.
So the solution is to do micro-founded models: try to figure out what people want to achieve and under what constraints and information they make decisions and crucially how they form expectations. It has not been too successful so far, but that’s the only way forward.
June 4th, 2009 at 4:00 pm
And read Godel, Escher, and Bach: an eternal goldenbraid. The author takes this issue head on and talks about conceptual “chunking” where, with very few exceptions, there is at least some amount of abstraction going on.
June 4th, 2009 at 4:02 pm
From an outside perspective, what seems to be going on is that economists have unearthed an extremely fruitful paradigm for investigation of micro issues.
ummm, no they haven’t. Unless you define “fruitful” in terms of disciplinary prestige and $, not successful prediction of human behavior.
The fundamental issue isn’t seeking “micro foundations for macro”, which everyone has always done at least implicitly. The problem is that the entire set of assumptions at the heart of neoclassical microeconomics is wrong.
June 4th, 2009 at 4:03 pm
Matt Rognlie:
See, when you say “based on observation of historical relationships,” I think yes! That’s what we need. Because ultimately, what we need are models that actually explain the dynamic development of economies over time – not the unchanging stasis of individual behavior.
Because when we get into economics as practical science versus as academic inquiry, the only thing that matters in understanding how the economy as a whole works and what should guide policy, not individual behavior, unless that individual behavior can explain what’s going on at a systemic level.
And I would argue that if we want to put the failures of macro vs. micro on the scales, sure the Phillips curve belongs there, but so does Friedmanite monetarism, supply-side, rational choice, and the whole edifice of economic theory that justified the financial bubble, all of which had solid micro foundations but didn’t accurately describe reality.
June 4th, 2009 at 4:03 pm
Academic finance has the same problem:
From Robert Haugen, Emeritus Professor of Finance, University of California, Irvine in his 2004 book, “The New Finance”, 3rd Edition:
For more on this, see my post, “Induction, deduction, and a model is only as good as its interpretation“.
June 4th, 2009 at 4:09 pm
All conservatives confuse macro conclusions with their own microthinking. Of the 15/16 of the time that conservatives are wrong, that specific error underlies at least 3/4.
June 4th, 2009 at 4:18 pm
My opinion about specific issues involved:
1) what we are calling ‘microfoundations’ is historically, really
-insistence on deriving all macromodels from a) individual agent optimization AND b) assuming perfect and complete markets AND c) using representative agents (and representative agent was part of the mix from tractability when this started) and d) all other assumptions needed for general equilibrium on micro markets (including assumption of rational expectations, and additional needed assumptions that everyone had same, or very similar, probability distribution for events).
2) I don not agree with commenters above, if they are saying that assuming item 1) above is necessary for a useful macroeconomics, since we know that some useful macro properties cannot be derived from traditional microeconomic assumptions -for example, the shape of the aggregate demand curve.
3) I think most of the failure of real business cycle style microfoundation theories was due to fact that labor supply was not flexible enough in response to changes in real wage -so it is true that they had a tantalizingly specific issue to work on that promised to fix things for their models. Problem is that as far as I know, only fix that has been found is a technical gimmick that assumes the problem way.
4) One thing all models have in common, whether perfect market real business cycle, or Keynesian microfoundation models that assume imperfect markets is that rational expectations hold. Frydman is interesting follow hear, because his work centers on whether observable aggregates can possible by consistent with any form of the rational expectations hypothesis in its current form.
5) Jospeh Stiglitz’ advances in economics of information has been, as far as I know, almost completely ignored in microfoundations work. If you assume that Stiglizian information effects are important, then many Keyneisan conclusions follow without much trouble from individual optimization and equilibria on micro markets.
June 4th, 2009 at 4:48 pm
Read Steve Keen’s Debunking Economics. That is, if you haven’t already. I can’t express how useful it is.
June 4th, 2009 at 4:54 pm
Shocking that a crypto-Keynesian non-economist would be pushing for macro without micro foundations. Above @5 I believe you mean the Phillips Curve, not the Taylor rule. It may be believed by some, in some limited capacity, but it is really about as alive as that chick in Florida whose case caused the big hubbub in ‘04. The fall of the Phillips Curve, which, along with Lucas and his famous critique, brought forth the return of microfoundations and killed the Keynesian project. No wonder the Keynesians would like to go back to something based in animal spirits and nonsense rather than in economics.
Anyway, the reason it is necessary is pretty simple. “Economies” don’t do anything. Firms, households and individuals do. We have a pretty good understanding of how those three work, so we build from there. Without understanding the actual guts of the economy, we have no way of understanding how macro decisions will affect individual firms, and that is how you end up with a proliferation of unintended negative consequences.
-AM
June 4th, 2009 at 5:11 pm
DTM @ 8,
How did a quantum understanding of chemical bonds lead to the discovery of the structure of DNA?
My understanding is that quantum chemistry was crucial to developing the model of molecular geometry that Watson and Crick in turn used to interpret the relevant X-ray diffraction patterns. I’m not an expert in this area, however, so basically I am just repeating something I was once told.
June 4th, 2009 at 5:16 pm
Brad DeLongs’s Econ 101b Introduction to the Theory of
Economic Growth:
Thought this might help, Matt.
June 4th, 2009 at 5:26 pm
Isn’t the goal in having microfoundations to do something like what statistical mechanics did for thermodynamics? I’m just a dumb physicist, but macroeconomics looks a bit like thermo to me, but without a solid analogue of entropy.
June 4th, 2009 at 6:39 pm
David Yassen: my copy is well-thumbed through, a very useful guide.
A Microeconomist: ““Economies” don’t do anything. Firms, households and individuals do. We have a pretty good understanding of how those three work, so we build from there. Without understanding the actual guts of the economy, we have no way of understanding how macro decisions will affect individual firms, and that is how you end up with a proliferation of unintended negative consequences.”
Couldn’t you just as easily turn that around and say that firms, households, and individuals all “do” within larger institutional frameworks that guide and shape and structure economic behavior, and without understanding how the larger structure works, you end up with micro-based models that have “no way of understanding” how “individual firms” decisions are shaped by the larger macro-economy?
I guess what I’m getting at here is, from my pov, I see a lot of assertion that the micro is the “guts” of the economy without much in the way of argument as to why that is the case.
June 4th, 2009 at 7:05 pm
Regardless of whether you are a Keynesian or a real business cycle macroeconomic theorist, many results of microfoundations macromodels depend on assuming rational expectations. Further, many conclusions common to both Keynesian and real business cycle models depend on rational expectations.
Rational expectations is a system (or ‘economy’) wide restriction placed on agents’ expectations at a micro level. Expectations play an important role in macroeconomics, and I do not know of a good microfoundations theory for any theory of expectations.
The brute mathemical fact of the matter is that the commonly accepted microeconomic assumptions for general equilibrium don’t tell us enough about what the economy looks like in aggregate to build an empirical macroeconomic theory. The added assumptions that have been used so far have not produced a theory that stands up to statistical tests in the same way that theories in other observational sciences have. That is just the way it is, and slogans cannot change that.
June 4th, 2009 at 7:51 pm
The analogy to of macroeconomics to chemistry or thermodynamics doesn’t hold up to scrutiny. It is true that in chemistry and thermodynamics larger-scale truths could be determined without the modeling of all the smaller-scale quantum details. However, that was only possible because these are experimental sciences. That is the key point – you can’t perform experiments in macroeconomics. As a result, for a macroeconomic theory to have a sound foundation, it must be built from sound microeconomic roots. Otherwise, it is nothing but fancy equations.
June 4th, 2009 at 8:01 pm
History is an experiment, albeit uncontrolled. A gross approximation with some basis in real-world experience seems to me more valid than complex mathematics based on convenient assumption or tautology.
As to physics and chemistry, it is also true that quantum mechanics is of little use at a larger scale.
June 4th, 2009 at 10:35 pm
@10 rapier: “In 500 years, more or less, what are today’s economic orthodoxies will be seen as bizarre superstitions which lead to mega death.”
One of the ten funniest sentences I have ever read.
Obviously, I find revealed truth pertaining to possible species extinction super humorous.
June 4th, 2009 at 11:12 pm
DTM, what the hell are you talking about? Atomic chemistry didn’t win out because it relied on the physics of atoms — the physics of atoms were completely unknown in the time period you describe, and in fact physicists were skeptical of the existence of atoms until the discovery of radioactivity.
June 5th, 2009 at 12:10 am
DTM, what the hell are you talking about? Atomic chemistry didn’t win out because it relied on the physics of atoms . . . .
Nor did I claim that it did, if by “the physics of atoms” you mean our modern understanding of atoms. Indeed, I thought I made it quite clear that the Rutherford/Bohr model of atomic structure and quantum physics only came in at the end of the story.
My point was rather that many of the advances leading to modern chemistry were derived from theories about the underlying constituent parts involved in chemistry, all in an explicitly atomist framework. The fact that our understanding of those underlying constituent parts was progressing in parallel to these developments in chemistry in no way undermines that point.
in fact physicists were skeptical of the existence of atoms until the discovery of radioactivity.
Some physicists perhaps, although natural philosophers (as physicists were once called) had been offering atomic theories at least dating back to Ancient India. And as I mentioned, Boyle, Lavoisier, Dalton, Avogadro, Faraday, and Arrhenius were all atomists, and these were the guys doing the important work in chemistry in their days.
By the way, I would suggest that it was Brownian motion (observed by Robert Brown in 1827) that really started convincing the holdout skeptics about atomism, culminating with Einstein’s mathematical explanation in 1905. Radioactivity was more directly relevant to the investigation of subatomic structure.
June 5th, 2009 at 4:12 am
[...] la nécessité (ou non) du réductionnisme Billet intéressant de Matthew Yglesias qui interroge le bien fondé de la norme qui veut dorénavant (enfin, depuis un petit moment [...]
June 5th, 2009 at 7:29 am
“To make that concrete, in the modern day we have achieved a decent understanding of how principles of chemistry are grounded in physics’ understanding of the behavior of atoms. But it’s just not the case that advances in chemistry were made by demanding that chemists ground all their models in subatomic physics. On the contrary, chemistry moved forward in the first instance by having chemists investigate issues in chemistry and see which models and theories held up.”
I believe that Linus Pauling’s application of quantum mechanics to physical chemistry was quite fruitful. I think a better example may be the relationship of thermodynamics to (micro) physical laws describing particles.
June 5th, 2009 at 10:35 am
Theoretically, of course, according to some, macro forecasting, could be reduced to quantum mechanics–or some other (perhaps undiscovered) foundational theory of physics. A very long, complex (and utterly useless) equation involving 3.9 trillion trillion variables and perhaps Eigen values and such and such should — as the story goes — be able to predict the economic growth in Belize–along the way it would even (out of necessity) predict how many bananas I will eat on my vacation there; it will also have to predict hurricanes and earthquakes. Of course, we would never be able to understand any of it.
Most “reductionist” theories have the “gut” appeal because of the belief that the “lower” level science is more rigorous that the upper level, which involves greater abstraction. Thus, neuroscience is more “scientific” than psychology; physics is more “scientific” than chemistry; molecular genetics is more “scientific” than Mendelian genetics, and so forth.
But it is vital to understand and the “laws” of the higher level, less “scientific” endeavors. It’s important to discuss the laws Mendelian traits, even though the action is all way down below in DNA; it’s important to understand the laws of chemistry even though physics is somehow at the bottom of it all.
It’s comforting to know that there are “bridge” laws that can explain Mendelian traits by reference to DNA, but it is far more useful for a farmer (for example) to know that she’s going to get big juicy tomatoes. Mmmmm.
So, it would be great if you could predict macroeconomic events by reference to microeconomic models — or whatever it is you want to do — but don’t count on it. Remember, it is the time-honored role of the economist to observe what is happening in the world and then to go back to see if what is happening is possible in theory.
June 5th, 2009 at 11:00 am
[...] He writes, the decision has been made to somewhat arbitrarily impose the view that macro models must be grounded in micro foundations. Readers of this blog know that I would be the last person to defend economic methods in macro for the last thirty years. But the demand for micro foundations was not originally “arbitrary.” It came about more or less like this: [...]
June 5th, 2009 at 1:14 pm
The micro/macro distinction in Economics is junk. And the metaphor that economics is a science like chemistry or physics is uberjunk. It is a science like sociology or linguistics or history – in fact, Dugald Stewart’s term for The Wealth of Nations, a “conjectural history”, is the best summation of economics.
An economic model is pretty much a poem, made out of mathematics. And the poems are inserted into ideological frameworks, such as that of the “self-correcting” market, or Robert Lucas’s idea that equilibrium is an economic apriori and makes everything intelligible in economics. Which, of course, like most of Lucas’ counter-revolution, is sheer bs.
It is interesting to see how model talk overtakes reality when economists – neo-keynesian, or the fresh water variety – argue. For instance, the recent argument about the effect of federal spending, in which Taylor and Barro came out with the usual line about inflation and government spending “soaking up” private expenditure, and Krugman came out with the Keynesian line about how Barro and Taylor are operating with a model of the economy at full employment – both sides are discussing an economy that, since 1945, has depended on the government for around 15 to 20 percent of all employment. So, in actuality, when these economists are speaking of full employment, they are speaking of some mix – the private sector employing between 70 – 75 percent of the employable population, the government employing between 15-20 percent, and the unemployed ranging between 10-5 percent. So, in reality they are not talking about zero government spending or total, but about the margins that would allow government to soak up the 5 percent of the employable that the private sector can’t at the moment. This is all the argument is really about. But instead, we get a scrim of rhetoric about full employment or “private investment” that was all simply unreal.
In fact, a ’sciencce” that treats one part of the economic mix – “government” – as though it were an intruder in the economy rather than a stable structure that is fully integrated into it, fails the minimal standard for being a conjectural history. Thus, the conversation degenerates into a joke.
As, in fact, is the case with all large scale developed economies, the private public mix will never ever fit the small government model again. It will always range between 20 to 50 percent of the GDP. We know that much about the wealth of nations. And we actually operate that way. We just pretend that we don’t.
June 5th, 2009 at 1:30 pm
Roger, that was one of the smartest things I’ve read all day. And as a historian, I especially approve the anti-dissing of my profession.
June 5th, 2009 at 2:21 pm
Just for the record, the microfoundations of microeconomics are crumbling.
‘
Do people optimize? Well, when they have a choice to make they have to choose from a menu of possibilities, and economists can assert that the choice made is the “optimal” one. But this is circular, for one thing, and also useless without the additional axiom that people’s preferences are stable from one minute to the next. But that axiom fails in the simplest of cases. Why do you buy chocolate one day and vanilla the next? Orthodox micro says that there must be some environmental variable driving your decision. Like what? Or, there’s some other factor in your utility function, such as perhaps reluctance to let the ice cream vendor think you’re in a rut. But that simply gives license to pack the utility function with any bizarre taste, for which there’s no accounting, that you want. Richard Posner once postulated a taste bud in a utility function for shooting at passenger trains. When the big shots can say such things without embarrassment, we might as well declare intellectual bankruptcy.
‘
Do consumer choices lead to a stable demand curve? Well, as the products we buy become more and more complicated, and as we overspecialize ourselves so that we have no idea how to check the quality of what we’re buying, we’ve knocked out one of the essential props of the competitive market model. The markets for almost all modern consumer goods are choked with asymmetric information; the seller knows more about the product than the consumer does (sometimes vice versa). We make do with proxies: seller’s reputation (see under eBay), moneyback guarantees, consumer-protection statutes and tort law.
‘
Competitive markets? Gimme a break. If we’re relying on reputation, etc., only the big boys can survive, because only they come to the notice of, e.g., Consumer Reports or your co-worker Wayne. Little sellers have no reputation that they can stake as a kind of a hostage in the marketplace: “If we sell crummy stuff, you can kill our reputation.” On eBay, the seller may be little, but those sellers who’ve been around long enough to have several hundred happy customers have a huge advantage. And as bigness gets to be biggerness, there’s yet more secrecy and worse asymmetric information. . .
‘
Oh, and on the supply side, there’s economies of scale, which wreck the competitive model right at the starting line.
‘
Does price convey everything you need to know about a particular good? Not under asymmetric information. And we know that even with price information as perfect as can be, in e-retailing, prices don’t converge to a single price the way the “law of one price” says they should.
‘
Expected utility is going the way of T-Rex. It’s been accumulating fun paradoxes since 1948 (in a classic article co-authored by young Milton Friedman, of all people). But no young Kuhnian hero has stepped forward with a new paradigm.
‘
Game theory tried to save micro, and it’s fun to think about, but it relies on those unreliable microeconomic foundations. And anyway, it gives crisp, testable hypotheses almost never. In fact, it’s usually possible to contrive a game-theoretic hypothesis that contradicts common sense. Is Obama’s embrace of Bush theories of executive privilege give evidence that he’s Bush lite? Maybe, maybe not. If he’s playing a deeper GAME, maybe he’s arguing for the Bush version in court in an effort to sabotage it with a judicial ruling. (That happens to be what I think, but smart guys like Glenn Greenwald seem to reject it out of hand. Unless they agree with me, but realize that for the scheme to work, big-league bloggers have to denounce Obama. . . You see the problem. Game theory can explain almost anything, and its opposite.)
‘
‘
It’s frustrating, because as consumer data gets to be more and more detailed (e.g. “fresh values” cards at your friendly neighborhood grocery conglomerate), and computers get to be more and more powerful, and econometric theory gets to be more and more intricate, we have the capacity to test ever-more subtle theories, if we can think of any.
June 5th, 2009 at 4:05 pm
[...] truth of the problem with Macroeconomics. Following Justin FoxSteven Levitt’s summary, Matt asks the next question: So why should it be that “in the current regime, if [macro models] are not meticulously [...]
June 5th, 2009 at 6:45 pm
So why should it be that “in the current regime, if [macro models] are not meticulously constructed from “micro foundations,” they aren’t allowed to be considered”?
Because that rule introduces an intractable right-wing bias into analysis of macro questions. To idiots like Greg Wankiw, a methodology that pre-judges all economic ideas in terms of their fidelity to “free market!!!!1!” principles is preferable to one that actually solves problems and agrees with reality.
June 5th, 2009 at 10:26 pm
Matt Y: So why should it be that “in the current regime, if [macro models] are not meticulously constructed from “micro foundations,” they aren’t allowed to be considered”?
Because every macro model is only compatible with a subset of the space of possible micro foundations. When scientists make their microfoundations explicit, they enumerate some or all of the subset of micro-level ideas compatible with their macro level model. This makes it possible to easily identify macro models which more or less imply micro nonsense.
As a case in point, the best thing about Kydland & Prescott style macro models is that the assumptions are laid bare for all to see, and reject if they are so inclined.
The worst thing about the old school Keynesian macro model is that it’s unclear what micro level beliefs you need to have in order to agree with it. Exactly what do I need to believe on the micro side to accept the Keynesian “animal spirits” story on the macro side without contradiction? This is where microfoundations are helpful.
June 5th, 2009 at 10:44 pm
I thought that the whole microfoundations project was scuttled by the Sonnenschein-Mantel-Debreu theorem back in th 1970s. Because market demand functions are essentially shapeless, building any kind of macro-level theory on microfoundations does not lead to unique results.
June 6th, 2009 at 1:04 am
56. Patrick Earnest:
Yes, that is the famous result in the mid 1970s that initiated lots of interesting research on this topic. Any behavior you can see on aggregate markets, even at lower levels of aggregation than seen in macroeconomics, almost anything you can observe, is consistent with some well behaved microeconomic structure. So, I do not see how comment 55 can be correct. If you makes lots of auxiliary assumptions, then you can decide some micro configurations are ruled out, but the auxiliary assumptions are very difficult to verify. Without those extra assumptions you cannot really rule out any plausible or implausible microeconomic structures. When you add very plausible complications like imperfect information, imperfect competition, constraints that prevent general equilibrium from holding on all markets, then it gets even worse.
The introductory article “Whom or what does the representative individual represent?”
Alan P Kirman – The Journal of Economic Perspectives, 1992.
starts the discussion with Sonnenschein-Mantel-Debreu and then takes the reader through subsequent developments -such as, the extremely strong and unlikely and finicky conditions needed for the current microfoundation-based macromodels to mean anything at all.
Every economists should have learned this stuff, so puzzling to me how the microfoundations macroeconomics has stayed so dominent after 1970s.
June 6th, 2009 at 2:08 pm
Is an insistence upon microfoundations really what’s wrong with macroeconomics? Everybody seems to think macroeconomics is dreadful becuase it failed to inform us of an impending crisis. One way of explaining this is that (business cycle) macreconomists were busy looking the in wrong place, thinking about how shocks are propagated through the economy and so forth. Say they’d been looking in the right place, asking how shocks arise, and paying particular attention to things like deficits, asset prices, balance sheets and the workings and state of the banking sector. Wouldn’t microfoundations play a role here? Say you think the real story has to do with information problems, or incentives facing individual bankers, or deviations from rationality, or all the other plausible ingredients that have been put forward, won’t capturing these things amount to modeling microfoundations?
June 6th, 2009 at 4:25 pm
Note the framing of the following … James, June 5th, 2009 at 10:26 pm:
The premise is that the only allowed support for the General Theory observations about behavior in financial markets is a model of the micro level beliefs that are required.
But they are empirical observations. If there are micro level models that are inconsistent with the observed behavior of individuals in financial markets, that is not a flaw of the observed behavior, its a flaw of the proposed model.
Indeed, one can just as well ask what are the macrofoundations of microeconomic models, since patterns of habitual behavior have to make sense as being able to persist in the macroeconomic environment in which an individual acquires those habits of behavior. Yet there is no presumption at all that the habitual behaviors which can be emulated by a model of incessant decision making under a rule of individual atomistic utility maximizing have to be shown to be compatible with success in the context of the observable macroeconomic environment.
June 6th, 2009 at 7:49 pm
What is the real thing, interesting independent of and yet reducible to microeconomics, that macroeconomics is supposed to be investigating?”
Well, if macro is to mean anything, society’s REproduction of itself.
June 6th, 2009 at 8:53 pm
jl: “Any behavior you can see on aggregate markets, even at lower levels of aggregation than seen in macroeconomics, almost anything you can observe, is consistent with some well behaved microeconomic structure. So, I do not see how comment 55 can be correct.”
Which claim in comment 55 is false?
BruceMcF: What about framing? I like to frame the world in the categories of true and false. For example, it’s false to claim that “The premise is that the only allowed support for the General Theory observations about behavior in financial markets is a model of the micro level beliefs that are required.”
I have no problem with models that derive their support from something besides micro beliefs. But if your macro model is incompatible with my micro beliefs, I want to know that. Then I’ll worry about which set of beliefs to reject.
Prescott, Lucas, Sargent and that whole group make it clear that their results absolutely depend on perfect markets, ratex, and so on so that I can immediately reject them. See the benefit?
June 8th, 2009 at 10:34 pm
[...] likes economists, but he’s got his finger on a real problem in macro-economics – the misguided need to base macro theories on micro foundations: But as a methodological matter, it seems deeply unsound. As a general principle for investigating [...]
June 8th, 2009 at 11:43 pm
[...] Big Dog pointed out to me a recent post by Matthew Yglesias asking the question why Macroeconomic models, to be even be considered in the profession, must be [...]
June 9th, 2009 at 4:54 am
I think the explanation for the overemphasis on microfoundations is a question of scientific prestige: economics, like most other social sciences, desires the cache of being labeled a rigorous science. It occupies a uncomfortable middle ground between the physical sciences and so-called soft disciplines like the humanities. If it is to raise its profile to the point where people will pay attention, it must have the characteristics of an actual science, which in the popular mind probably involves some sort of reduction to microfoundations (i.e. elementary particles in physics or whatever), because any respectful science must have such microfoundations. Of course, this is horrible philosophy of science, a philosophy of science that could probably be traced back to the Logical Positivists. I think Lord Keynes’s dictum about businessmen being in the thrall of a previous generation of economic ideas applies equally well to non-philosophers and the philosophy of science. Much of the conventional wisdom about what a science is is just simple reductionism, which was the dominant paradigm in the philosophy of science until the second half of the 20th century. That sort of picture has since been severely problematized.
The more important point, however, is the special status of disciplines called “sciences.” The success of the physical sciences (physics, chemistry, astronomy, biology, etc), compared to other disciplines, has resulted in their elevation to a special epistemic status. The physical sciences are treated as models for all areas of intellectual inquiry to follow. Anything that looks, acts, and smells like a physical science should also have the special status of a physical science. The attitude seems to be that the way one identifies a discipline making the leap from pre-paradigm status to paradigmatic normal science, to use Kuhn’s terminology, is by identifying the special features that characterize the physical sciences. As I noted before, the conventional wisdom identifies reduction to microfoundations as one of those characteristics. Economics as a discipline is merely following what folk philosophy of science identifies as the necessary steps to assuming the mantle of “science.”
June 10th, 2009 at 6:03 am
[...] criticism floating about regarding how macroeconomics do their thing. Here’s Matt Yglesias wondering about Levitt wondering about microfoundations. And here’s Ezra Klein turning , unwittingly I [...]