The value of microfoundations

Here is Noah Smith on microfoundations, responding to Matt Yglesias (here and here, here is also a Krugman post on the topic).  I am usually pro-microfoundations, though without any particular philosophy of science-derived attachment to the idea.  Here is why:

1. In some very important, simple, and intuitive models, there is nominal stickiness but money is still neutralCaplin and Spulber 1980 is one of my favorite pieces and time spent studying their model will be repaid with value.  I don’t think the model applies to 2007-2009, or say 1929-1933, when we have “out of the ordinary” shocks, but it may apply to many other time periods.

2. Often the data suggest that money is neutral or roughly neutral or at least the data are not inconsistent with neutrality.  I know that one does not hear much about that in today’s economics blogosphere, but I kid you not.  Again, I differ strongly from this literature for “the times which really matter,” such as 2007-2009 or 1929-1933, but still I take the data seriously.  Try this relatively “atheoretical” piece by Harald Uhlig, which does not reject monetary neutrality (nor does it cover 2007-2009).  Nominal stickiness models have trouble explaining why money doesn’t matter more than in fact it does.  Understanding microfoundations will keep you out of the trouble you will get in if you keep trying to use money to expand output.

3. We can try to nudge people into more flexible wages, but again that requires some understanding of microfoundations.  The Fed can prevent any risk of a deflationary downward spiral, and please note it is no coincidence we are in the two percent inflation range.  You don’t have to view this as a high return activity to favor it  (it is funny how the mere mention of wage nudges will cause many people to suddenly turn against the nudge idea, at least temporarily.)

4. Microfoundations don’t have to mean intertemporal maximization with extreme assumptions about rational or well-behaved preferences.  George Akelof has written some of the best papers on microfoundations.  Nor do microfoundations have to mean staking out an extreme position on the Lucas critique.

5. There are plenty of flex-wage professions which still have been seeing high unemployment.  Like real estate agents.  Or what happened to all those Mexicans who used to stand around on street corners?  Were their wages sticky too?  I don’t think so.  Does their return to unemployment or underemployment in Puebla refute the sticky wage model?  I personally don’t think so, but it’s hard to answer that question — an obvious one to a critical observer — without a clear sense of microfoundations.

6. Just how bad is monopoly?  We wish to know when designing a competition policy.  Under some microfoundations models, the existence of market power is essential for ongoing stickiness, in other models not.  This question matters, and we need microfoundations to better resolve it.

7. Should a government subsidize, tax, or be neutral toward contract indexation?  Try answering that question, or even getting started on it, without some sensible microfoundations.

8. If you are trying to end a hyperinflation or high rate of inflation, do you need to get fiscal policy right, in some kind of credible manner, before enacting monetary restraint?  It is hard to imagine answering this question without good microfoundations.

I could go on.  I am worried that people are rejecting microfoundations because they think microfoundations imply objectionable attitudes toward macro policy.  But note: if those attitudes are objectionable and indeed wrong, they won’t be implied!  It is entirely defensible to argue “we should have a more expansionary monetary policy today, even without the microfoundations to support that view.”  It is much tougher to argue “economics should deemphasize microfoundations altogether.”  The broader the range of questions one considers, the more important microfoundations turn out to be.


It seems the main objections to microfoundations is that models don't work well as an empirical matter and often use over-simplified assumptions (often while complaining the macro models use over-simplified assumptions).

I don't know that anyone has an objection to microfoundations as a theoretical matter. It's when models that don't work empirically are used against models that work much better although they lack microfoundation support.

5. Is a really important critique of sticky wages. It seems that the view of sticky wage folks like Sumner is that the market can't clear because of things like minimum wage. Is that really the case? Should we expect high cost of living cities like New York and LA to have higher or lower unemployment than low cost of living cities like Detroit or Birmingham?

I can't remember where I saw it (maybe Krugman's blog) but the point was raised about associates joining law firms and that the salary has remained constant at $160K despite a glut of lawyers and little chance of advancement for the associates. Reason was given that no big firm wanted to drop the salary because they were afraid of getting rejected if they offered a lower amount. Maybe this is the ultimate of sticky wages and it makes absolutely no sense at all.

I would like to understand exactly what you mean by "no sense at all".

If the associates aren't worth $160K, then why would the law firms be concerned about losing a potential hire?

What seems to be happening is that while the firms would like to play a strict supply/demand line relative to their potential employees, they realize that in fact the difference between someone who would accept $160K as a salary verses someone who would accept $150K is >> $10K to the firm. That is, the higher quality of lawyers have settled into a joint minimum wage of $160K which is supported both by the fact that they are worth more to the firm than $160K, and that their quality is enough better than the competition that they can capture the difference.

Of course, partners are going to complain--but you don't see them actually make the lower offer because they don't want to pass on the best candidates.

That's like someone complaining about the cost of wine while ordering from the top of the menu because they know that the restaurant's cellar is full of lower-quality vintages.

Who cares?

If the associates aren’t worth $160K, then why would the law firms be concerned about losing a potential hire?

Because those new associates have to work alongside last year's associates.

People in large corporations often (usually) have no clue what their peers are paid.

Tyler says several useful things that do not seem relevant to arguing against the proposition that macroeconomics MUST be based on explicit microfoundations, which is the way I understand Yeglesias and Krugman.

Two things:

1) It sounds to me like criticizing macro theories based on a lack of microfoundations means you end up criticizing the theories that seem to best describe the world.

Noah Smith seems to me to be on the right track - it`s not that micro founding stuff is in-and-of-itself bad, its that microfoundations as they exist today are generally terrible at describing the world (and wildly inconsistent with their own microfoundations in psychology), and so the models without microfoundations are often the more accurate ones.

2) I think Tyler is after a straw man, and that no one is saying that microfoundations are never worth looking at.

Tyler lays out a dichotomy between a very narrow defense (we should act today without microfoundations) and a very broad argument (we should never even look at microfoundations). I read Ygelsias as making an argument somewhere in between those two - that we should be comfortable with macro theories that match the data but that aren`t micro founded. In his physics example, he`s not arguing we should never look at relativistic theories of gravity, just that there was nothing wrong with Newton`s theory of gravity for most uses.

"you end up criticizing the theories that seem to best describe the world."

You have inputs, a black box, and the outputs. Science is about "what's in the box!?!" We are conflating science with policy.

Here's a recent philosophy of science paper that is relevant:

I think you have missed the point here, a bit. I am pretty sure that Yglesias, Krugman, embrace the idea that microfoundations should be studied, and learned from where when they provide useful insights. What they are objecting to is the idea that the ONLY macroeconomic research that is accepted in many spheres is that macroeconomics that has explicit links to microfoundations.

If someone comes up with a model, or theory, that happens to provide useful insights or empirical predictive value, but that doesn't happen to be microfounded, why should the researchers face barriers to publication?

Obviously, micro adds up to macro, but for policy, people seem to overlook the effect of drags and resistance to adjustment, or they focus just on the drag or resistance that they have an interest in because it supports their world view.

So, if you want lower wages, you talk about wage stickiness....but, you don't talk about transactions costs, say, from price adjustment, or about oligopolies that RAISE prices and restrict quantity during a recession because they have the market power to do so...afterall, they are strong dividend payers during a recession.

I'd just like to see macro claims anchored to something. Anything. With zero predictive power, no repeatable experiments, and no microfoundations, you are getting dangerously close to "because I said so".

That's because the purpose of macro is to prove that governments can do what no individual, household or business could.

I commented on Noah's blog that we have a good case study for this in major league baseball. Lots of hard data starting off in 1901 up to the present day with enough variables to model things in a variety of ways and I've done my share of data analysis over the years (I'm a card carrying member of the Society for American Baseball Research). Just as in economics one runs into the basic problem that no single microfoundation can be extrapolated over the whole data set. The disruptions that one sees have a variety of explanations (changing the ball from 'dead' to 'live', better nutrition resulting in bigger and faster players, lowering the pitcher's mound after the 1968 season, changing the strike zone, use of performance enhancing drugs, etc.). While we all want predictive models, we are confronted by the axiom that we get from every financial prospectus, "past performance is not a prediction of future performance." Perhaps economics is the one area that Occam's Razor doesn't work.

Funny you should mention microfoundations... If you like microfoundations, you really need to look into agent-based simulation. It's been around for a while and had limited impact (although there are a couple of striking results, such as Thomas Schelling's segregation model and Robert Axelrod's finding that tit-for-tat is an evolutionarily stable strategy in repeated prisoner's dilemma games), I think because there tend to be so many degrees of freedom in the way models are specified, plus PhD economists don't usually have the skills to build agent-based simulations. But I'm trying hard to change that. See this 45-minute video update:

Sorry for off-topic, but Tit-for-Tat is definitely not evolutionary stable strategy. For a simple explanation, you can look up for instance Binmore (

In fact, it has been recently shown that the Prisoner's dilemma can be reduced to an ultimatum game.

That's what I love about math--Euclid's proofs are still valid.

Does microfoundations include behavioral foundations? It seems like many of the arguments for including microfoundations in macro models apply to including behavioral foundations in micro models.

+1 If microfoundations DID include behavioural economics, then you would have federal policy going in the opposite direction of, say, the consumer who reacted to an investment panic, ie, going into deficit and stimulating.

But, that's not what this site wants you to hear.

You might want to read Akerlof and Shiller "Animal Spirits"

I'm probably just an ignorant internet denizen, but I don't know how anyone can say money is neutral--whatever it may have been at one time, it is clearly an influence on our thoughts and actions. A post here on this site two days ago agrees with this (, to wit:

"We found that subtle reminders of the concept of money, relative to nonmoney concepts, led participants to endorse more strongly the existing social system in the United States in general (Experiment 1) and free-market capitalism in particular (Experiment 4)".

There is a weight to money, now. Any weight suggests it lacks for neutrality.

Not only that, but money enters the economy at different points and at different times and people assign money differing values. By the time you've gotten your model pared down to "all other things being equal," you're not left with much. Economics is not physics.


Do please keep in mind when you cite Schelling's work to Tyler, that Schelling was his major prof. OTOH, I am one who would like to see Tyler say more nice things about ABM as a potentially useful approach to microfoundations.


Note that Tyler said nice things about Akerlof's work on microfoundations, which is definitely behavioral.

Oooops! Sorry. I confused Nathan and Noah Smith.

This argument is interesting because it relates to each schools views of microeconomics. In the freshwater world there is no microeconomics. It is just a tool for the macro-economist. And I mean tool in the worst possible sense, it just allows them to reach their preconceived macro outcomes. If the freshwater economist was *actually* interested in microeconomics they would approach it much differently. What happens now is they assume something, if it doesn't aggregate into something like reality add free parameters until it can. What they should be doing is actually looking at how people behave at the micro level. Behavioral economists are the ones doing this. The problem is that the results are not easy to aggregate to macro levels using off the shelf tools. If the freshwater macroeconomists were interested or curious this is what they would focus their energy on. The fact that they largely don't reveals their intent. The more actual micro evidence points to extreme irrationality the more foolish the freshwater macro position will become.

"The more actual micro evidence points to extreme irrationality the more foolish the freshwater macro position will become."

I think the opposite is true. If the actors in the microeconomic models do not behave rationally then most of the microeconomic models are broken.

+1 Seth.

Steve, below: ask yourself this question: maybe the question isn't whether micro models are broken, but rather which micro models are broken...the ones presuming rational man, or the ones which measure of humans actually work.

Dear Tyler,

Micro-foundations are not the problem - neo-classical micro foundations are. Let's take the 'representative consumer'. This entity is supposed to behave according to an emergent indifference curve which has the same properties as indifference curves for the neo-classical homo economicus. This curve is therwith not micro at all - it's as macro as it can be. Unlike the national accounts it's not based on micro measurements of individual transactions which are aggregated in an accounting model which spells out the interrelations between sectors as well as the real world accounting restrictions. One can state that the transactions are caused by purposefull behavior of a myriad of individuals - but that's not the same thing as stating that all these individuals can bee seen as bees in a hive, working for the greater good instead of following their own desires (which is the essence of the neo-classical so called 'micro'-foundation of the sector households). I've up to now seen no defense, not any, of the position that such a bee-hive curve indeed exist for human society. The neo-classical so called micro-foundation of the sector households lacks a proper concept, it lacks a proper definition, it lacks a proper operationalisation, it lacks measurement. There is no empirical or logical defense. Don't waste your time with it - it's a silly, ideological game. Not science. By the way - traits of the bee-hive idea of societies is of course rather prevalent in the present, neo-classical inspired micro-funded austerity ideas which wreak such havoc.

(you might want to read about Hjalmar Schacht, by the way)

My starting point on this issue is the Friedman Article mentioned, although I read him as a follower of Pragmatism, not Behaviorism. I find the distinction between Economic Theory & Political Economy very useful. I'm not sure how it relates to Macro-foundations or Micro-foundations, and I really try and avoid any reference to Foundations in my views, as it tends to land me in soggy cereal.

We don't necessarily need microfoundations to understand *how* the world works ... Simple macro models, reduced form micro analysis, and even partial equilibrium analysis might all illuminate a slice of the world. But if we want to know *why* the world works this way there needs to be more structure (microfoundations). Since the how and the why are closely linked, it makes sense to seek microfoundations without getting hung up on them.

This is a terrific post, especially points 6, 7 and 8. More generally, without some sort of microfoundations one pretty much never has any basis for evaluating any policy at all. IS-LM can tell us (perhaps correctly, perhaps not --- that's not the issue here) how to fight recessions; it can't tell us which recessions are worth fighting. I can't imagine addressing that issue without at least implicitly invoking some microfoundations, and it's always better to be explicit.

We see gravity work because of...gravity. Gravity is its own microfoundation. What rationale do we have that macro interventions would work? Is wage stickiness not itself a "microfoundation"?

Without investigation into microfoundations, how do you know what is being held constant? I was nearly rejected (in engineering) for empirical work without theory, and I agreed. Why should that be different in economics? It might should be different because there is more political motivation to block publication in economics. But that's a separable problem.

Let me start off by saying I don't have any objection to studying microfoundations. But unless we have an unambiguously correct micro theory (and we don't), the idea that microfoundations give insight to macro phenomena runs into some problems.

A couple of basic (and related) lessons from complexity theory:

1. Many different micro models can lead to the same macro behavior. The fact that model A leads to macro concept B does not mean model A is correct except inasmuch as model A is correct for other reasons. Model A doesn't even help you understand macro concept B unless A is correct for other reasons. In thermodynamics, the atom model leads to e.g. the ideal gas law. However the power in that is in the fact that the atom model explains many other things (spectral lines, chemical reactions).

2. The process of aggregation to the macro level destroys information (creates ignorance of the microfoundations). This is related to point 1. Some details of e.g. rational economic agents will be destroyed as you aggregate to the macro level -- and it is difficult to figure out what those details are in advance. Ideal gasses with the same number of molecules under a broad set of conditions operate the same way if they are H2 or O2.

3 Emergence. We really have no idea how this works in any regime, so if macroeconomics is a strongly emergent phenomenon then there might not be any way to predict the macro from the micro. Empirical models of macro phenomenon may be the best we can do. E.g. it is currently impossible to predict life from atomic physics; by that I mean without the theoretical possibility doing a full scale simulation. There is currently no way to predict the Mandelbrot set from the algorithm that generates it except by running the algorithm. Macroeconomic laws may not be derivable from microfoudations except as a brute force simulation. As such, this does not appear to be more enlightening than an empirical model based on macro data.

Comments for this post are closed