This post is prompted by a twitter exchange some time ago between Adam Posen, Noah Smith and myself over the ‘merit’ of microfoundations. [Here’s a storify recap]. And that in turn by the fall-out from the events at the Federal Reserve Bank of Minneapolis, where Naranya Kocherlakota [former microfounding whizz] has begun to push out senior advisors and research economists, including Pat Kehoe and Ellen McGratten. But this debate will be familiar to those following econ blogs for longer than that. It’s tangentially related to the controversy in the UK over how economics should be taught, fuelled by the student campaign group at Manchester University, and by the initiative led by Wendy Carlin at UCL to reform the curriculum. And it probably surfaces whenever a few economists have a beer and talk shop.
In this twitter exchange, Adam Posen said ‘microfoundations are without merit’. Noah challenged me to substantiate my claim that they do have merit.
The merit in any economic thinking or knowledge must lie in it at some point producing an insight, a prediction, a prediction of the consequence of a policy action, that helps someone, or a government, or a society to make their lives better.
Microfounded models are models which tell an explicit story about what the people, firms, and large agents in a model do, and why. What do they want to achieve, what constraints do they face in going about it? My own position is that these are the ONLY models that have anything genuinely economic to say about anything. It’s contestable whether they have any merit or not.
The early microfoundations project was about pointing out the unreliability of pre-microfoundations models, sometimes known as ‘Cowles Commission’ models, after one of the research centres that sponsored such a model. These were models that were long lists of equations for economic aggregates built out of stories economists had that linked some of these aggregates together [like ‘people tend to consume something, plus something else times disposable income’ – the consumption function]. This contribution [crystallised in Lucas’ 1976 ‘Critique’] was to note that if policy was based on statistical estimates of these guessed-at relationships, when that policy changed, those relationships may change too, invalidating the original policy choice. So the contribution was negative. It was about warning that another way of doing economics did not have as much merit as first thought, and might in fact entail substantial economic costs.
It seems likely to me that this early contribution ‘had merit’. To me, it seems highly probable that major policy mistakes, informed, for example, by the belief that permanently higher inflation might buy permanently lower unemployment, were avoided. I can’t prove it. But there are dozens of empirical papers exploring this same point, in the light of what Lucas said. The evidence there is not entirely on one side. How could it be? But I would say it was decisively tilted in favour of Lucas/Phelps/Friedman’s warning that higher inflation doesn’t get you lower unemployment forever. Here is an example of recent empirical work on this by Luca Benati that concludes that high inflation doesn’t buy permanently lower unemployment.
Pointing at the Lucas Critique doesn’t establish without doubt that microfounded modelling leads to a better world. It suggests that it might. Revealing that there is some probability that high inflation policies will not work, when previously this probability was discounted, has merit. Lucas showed that if the world behaved according to the postulates of a particular micro-founded model, and you didn’t bother with microfoundations, then you would mistakenly infer that you had found causal and stable statistical connections between your instrument and your goal variable, when these would mutate once you used them to inform policy. You might doubt that a microfounded laboratory is a legitimate tool to discover anything. If you don’t accept microfounded models as an interesting laboratory to test out any thought experiment, then you might not care about the Lucas Critique.
Caring about it requires that you accept that something that is true in a possible world (the model) may be informative about the actual one. I think it’s impossible to refute this, unless you know how the real world works. In which case you would of course not bother with the model. To refute it, you’d need to represent the real world, somehow, and show that acting on something that was true in the false word was useless in the real one.
The discussion about how to do macro often neglects that there are serious people trying to work out the details of how to do policy when you don’t understand how the world works. Or how the world looks when it’s packed full of agents who doubt their own representations of how the world works. Tom Sargent and Lars Hansen, now both Nobel laureates, have spent the last 15 years exploring these topics, sharpening discussions that used to and still go on inside central banks.
I would not try to claim that microfounded modelling is the ONLY way of doing macro that has merit. Sims explained one of the other ways. Form systems of statistical equations linking variables you care about. Where you regress everything on everything else lagged. Then you can forecast things if you like. To get further, you have to start using economics. Further meaning trying to measure the effects of a policy, for example, so that you can work out what good policy would look like. Identifying policy shocks need not use microfounded economic logic. For example, you could study the runes of history and declare that it was self-evident that some policy change was exogenous and not related to the things the policymaker cared about (despite the obvious quandary – discussed in the last blog post, since policy changes should surely only be motivated by care about something). An example of this approach in the past is the use of military spending changes to measure fiscal policy shocks. These being a result of decisions to go to war, or ideological shifts in government, or both, not business cycle policy.
Another kind of non-microfounded thinking I would accept is thinking that has been shown in a microfounded context to work. So, for example, Sims and Zha talked of ‘modest policy interventions’. These are small enough changes to the way policymakers have typically gone about things for it to be conjectured that the Lucas Critique won’t matter. If you don’t mess around with policy too much, the statistical relationships you have estimated won’t change much either. This isn’t a cast-iron guarantee, however. Because to get even this far you need to write down a possible economic world (a microfounded one), and show that the statistical relationships it generates don’t move much for a given type of policy change that you want to contemplate. And in writing down that world you are going to assume some things that are false. And have to keep in the back of your mind that anything you deduce using it might be false too.
Another kind of non-micro-founded thinking I think is valuable is knowledge building exercises like solving non-micro-founded rational expectations models. [The Bank of England had a model like this until just a couple of years ago]. Without going through the motion of solving a model like this, you won’t know how to solve a micro-founded one. And some insights can be got about how forward-looking behaviour manifests itself in time series. Of course, the assumption of rational expectations may often be very unrealistic. But if there are circumstances when it’s useful, and you need to figure out how to go about building a microfounded RE model, then building a non-micro-founded RE model might be a useful learning step.
A final possibility is that there is no alternative but to proceed in non-micro-founded way. Yet some business has to be done – some policy decision, or some investment based on a forecast. In these circumstances, it’s ok to take a stab at the decision rules or laws of motions for aggregates in an economy might look like if you could micro-found what you are concerned with, and move on. Perhaps doing so will shed light on how to do it properly. Or at least give you some insight into how to set policy. Actually many so called microfounded models probably only have this status; guesses at what something would look like if only you could do it properly.
Adam Posen dismissed my many examples of microfounded economic thinking that I thought had changed the face of economics and policy. [See the storify collection for some examples]. He said that the merit in examples of microfounded models lies in the ‘intuition’ behind the ‘one line idea’. In my tweets I explained my position on this. This statement is highly perplexing to me. Economic ideas are claims about what people and firms and governments do, and why, and what unfolds as a consequence. The models are the ideas. ‘Intuition’, the verbal counterpart to the models, are not separate things, the origins of the models. They are utterances to ourselves that arise from us comprehending the logical object of the model, in the same way that our account to ourselves of an equation arises from the model. One could make an argument for the separateness of ‘intuition’ at best, I think, as classifying it in some cases to be a conjecture about what a possible economic world [a microfounded model] would look like. Intuition as story-telling to oneself can sometimes be a good check on whether what we have done is nonsense. But not always. Lots of results are not immediately intuitive. That’s not a reason to dismiss it. (Just like most of modern physics is not intuitive.) Just a reason to have another think and read through your code carefully.
Let’s google Adam and take an example from his written or spoken work. I agree with almost all that Adam says about everything. I think he’s super smart and almost polymathic in his grip on all kinds of different economic issues and problems. But, I vehemently disagree with his claim about microfounded models, and I also think he himself is the great improvising microfounder of our times.
Adam said or was reported to have said [and I agree with him 100% on the substance in this example]:
‘If they [Germany] had resolved and done more transfers to southern Europe in the form of writing off more of the bad loans they gave to southern Europe. If they had pushed for a monetary policy that was more expansionary instead of blocking expansionary monetary policy, and if they had invested at home and in their own public and in their own people, these global balances and the imbalances in Europe would be reduced.’
As someone who thinks microfounded models have merit, but knowing that Adam claims to think they don’t, this is a tricky paragraph to grasp as one that still ‘has merit’. What on earth is he saying? [And marvel at the confidence with which it is said]. I read it as a good account of what would happen in a two-country sticky-price RBC model with rational expectations, and possibly other partially forward-looking models for expectations, of the sort you can find in Obstfeld and Rogoff’s textbook from 1995, modified to encode issues of sovereign default and financial frictions. Or rather, since neither of us (and perhaps nobody, I haven’t checked) has actually worked out a model with all these details, I should say that reads like a conjecture about what such a model would say, based on having seen previous ones like it.
There are other ways to read it [and as something that nevertheless ‘has merit’]. For example it could be read as having meant ‘based on past historical correlations, I predict that if debt had been forgiven, and monetary policy looser, imbalances would have been reduced, and the world would have been a better place.’ But to make such a claim, Adam would have had to develop a sharp econometric procedure to isolate a kind of natural experiment from the historical time series that enables him to replay it now in his head. Or to be referring to others who had. Without isolating these experiments, you can’t confidently claim from past episodes what was due to policy, and what was due to things that were perhaps prompting a particular policy. Worse still, although in principle there are ways to identify these shocks without using microfounded economic theory, they are pretty controversial. The runes of history or military spending approaches are contestable. [How can we know what was in the heads of the policymakers?] And for monetary policy these approaches are very tricky. You can’t help but rely in part on identification schemes that rely on microfounded models. Such schemes might include the following: ‘for any microfounded model I can think of, a monetary policy contraction surely does not boost output, surely raises interest rates, and surely lowers inflation’. Or ‘in the long run monetary policy should be neutral on real variables.’
Aside from these two ways of characterising Adam’s statement, there are no other scientific ways to put it. You might read those words, agree with them, and think to yourself: ‘surely all he’s saying is that looser fiscal policy would put money in German consumers’ pockets, and the looser monetary policy would lower real rates and make German firms want to invest more, and they would buy more foreign goods, and this would provide the missing demand for Greek exports….’. The trouble, is, this claim is empty and baseless without it making reference to one or both of the literatures that I have pointed to above. It can’t be a claim about the real world. On what basis would Adam claim to know about that? [We are excluding econometrics now, remember]. Perhaps if we had surveyed all the households in Germany and outside, and asked them what they would do in certain situations, (and also asked them how happy they would be if certain things happened, since there is an implicit welfare claim in Adam’s statement), then we could interpet Adam’s statement as being useful. [‘With merit’]. Then we could claim that we really knew something about what the world looked like. But we know that no such survey has been done. So this leaves us basing Adam’s utterance either in a conjecture about what a possible microfounded but false economic world would look like (and hedging our statements about it more cautiously than Adam did in that soundbite, a slightly unfair point since it’s possible that the hedging was stripped out of the interview report, or was implicit). Or a statement about the consequences of identified fiscal shocks. If we ground it in the latter, however, we can’t say anything about whether such an outcome would be better. To do that we need to have a model in which we compute how people feel about stuff and in which such a shock can be replicated, to see if they feel better when Adam’s suggested loosening is tried out. In other words, we are stuck with the microfounded but false way of thinking again.
This is not to say that all microfounded thinking is flawless. Much of it is highly dubious. The freshwater lot groan to see all the rigidities built into New Keynesian models partly because they suspect that the micofoundations for them are baseless; that they are put in because they make the models fit aggregate time series better. The particular mechanism for price stickiness widely used is a case in point. One of the most popular devices is Calvo’s: this assumes that prices roll a dice each period and get to change prices with some probability. And they get to never ever change prices until the end of time with some probability too. This makes the model behave very oddly in some circumstances. And is responsible for generating welfare costs of inflation instability that dwarf everything else in New Keynesian models. Calvo price stickiness makes the model tractable and fits certain aspects of the data better than the model without it. [For example, monetary policy shocks have effects on the real economy]. But no-one thinks this is anything like what happens in real firms. But it is an approximation that might well be helpful in lots of circumstances. Example: in such models, rules like Taylor Rules (in which interest rates respond more than one for one with inflation, but also to the output gap) do a pretty good job. And during periods when central banks actually followed them, performance was usually pretty good. I deduce from this that with some probability the false microfounded model has taught us something useful about good monetary policy design. [Almost all policymakers in central banks accept this too, as you can tell from the minutes of their meetings, or the documents explaining their monetary frameworks].
There’s something irksome about defending micro-founded macro from the attack that it is ‘without merit’. A voice inside me says: if they aren’t doing macro, by which I mean, generating new empirical or theoretical work themselves, who are they to go about proclaiming whether something has merit or not, or how macro should be done? [I’m not singling out Adam here. Lots are at it.] And why should anyone care what they say?
Three concerns make it hard to resist attempting a defence, however. First, there is the concern that the macro project at large gets tarred with the same brush as the microfounders who insist almost religiously that prices are flexible and markets are efficient. So it’s worth trying to disentangle the defence of the project as a whole from defending those substantive positions. Second, there’s the concern that someone with some influence might take Adam or others who say these things seriously. Third, there’s the concern that if people inside macro don’t respond to challenge, no matter how high-handed and ill thought through, there’s the risk that we come to seem like a cult bent on disengaging, concerned to interact with those outside the cult only so far as is necessary to squeeze them for the money we need to continue playing with our toys. [Perhaps that day has already come?!]
In my time in central banks one definitely encountered a breed of policymaker that behaved as if they were above actually doing macro, but yet seemed to know all the answers for sure, and know how macro should be done [of course by someone else, not them]. It seemed to many of us who observed them as though they had fallen victim to the illusion that since they had done so well in life, their gut feelings about stuff must really be valuable, and that perhaps that’s where macroeconomic truth lay, in what they as great individuals felt and said. Many can tell stories of attempting to advise them, and being met with the condescending twinkle in the eye that translates as ‘Ah, so that’s what’s true in your silly little toy world, is it, tee hee, how quaint that you think such things worth repeating, well, I can only hope that one day you glimmer the real source of truth, namely, the instinctive knowledge of the chosen’. If the meme that microfounded macro has ‘no merit’ were to gain any more traction, I assert that great danger would lie ahead!: theorising that is incomplete and ‘accidental’ [in the sense meant by Krugman]; policy promises that are unverifiable; discretion untamable; and a search for new economic knowledge that is empty and futile (since the truth is already felt by the great policymakers, and the only way to divine it is to draw the few charts they ask us to plot, and sit around and wait until the charts work their inner magic and they are kind enough to write it down in speeches for us).
[Includes edits for typos and minor stylistic changes not in original].