Simon Wren Lewis’ latest post on Mainlymacro rails against what he sees as an overemphasis on microfoundations in macro models. Paul Krugman picked up this post, and it was reblogged and retweeted by Brad DeLong and probably others, referring to the notion of ‘macro wars’, between those who microfound, and those who don’t. Simon points in particular to the failure to incorporate downward nominal rigidity in wages, and other features that spring from behavioural economics.
To recap for the uninitiated, a model that is ‘microfounded’ is one in which you write down explicitly what it is the firms and consumers in the model are trying to do, and what constraints are placed on them when they try to do it. Microfoundations came to be seen as important for a very good reason. Namely Robert E Lucas Jnr’s work showing that models that didn’t articulate exactly what agents were doing, but instead just relied on correlations between macro time series, could give very unreliable policy advice. The example topical at the time was the apparent tendency for unemployment to fall as inflation rose (the long run Phillips Curve). A tendency that disappears (at least in post Lucas Critique models) if policymakers try to exploit it to lower unemployment permanently. The risk with non-microfounded models is that all you have is a set of difference equations that don’t really have any economic meaning at all.
First point. Some notable microfounders have looked at downward nominal rigidity in otherwise standard and microfounded real business cycle models. For example, this paper by Schmitt-Grohe and Uribe, models the periphery of Europe as suffering from downward nominal wage rigidity; a credit boom pumps up nominal wages, and then when the crash comes, unemployment soars because wages don’t fall. Another example: this paper by Junil Kim, looking a the optimal inflation rate. Or take this paper by the two Maliars, prominent in the literature on numerical methods for real business cycle models, putting downward nominal wage rigidity into a heterogeneous agent model. In finance, the idea is alive and well too. For example this paper by Cohen and coauthors, which goes back to an old 1975 paper by Modigliani and Cohn, arguing that stock prices behave as though investors have nominal illusion. (Probably one of the necessary ingredients for downward nominal wage rigidity). Or this, on a similar theme, by Monica Piazzesi, a prominent finance academic who will have started out life learning Lucas’ microfounded model of asset pricing.
There are many other examples too where authors drop the assumption that agents are maximisers or know all about the workings of the model. Tom Sargent, George Evans and Seppo Honkapohja made their living studying what happens when agents have to learn about the model in the same way that econometricians do. It is common for modern DSGE models to have rule of thumb consumers that make no complicated intertemporal decision, but simply eat what they earn.
The practice of building in a friction into macro models without explaining where it comes from is extremely widespread. The ‘New Monetarists’ that revolve around Randall Wright and Steve Williamson would say that all modern New Keynesian models fall foul of this. The apparently microfounded practice of including (symmetrically) sticky prices and wages is a case in point. Most users of these models acknowledge that this aspect is not seriously microfounded, and is in fact simply a way of getting real business cycle models to fit the data. (Even Nobuhiro Kiyotaki, who co-invented it with Olivier Blanchard, called the sticky price part of the model ‘a fairy story’).
In the heterogeneous agents literature, which began as a direct outgrowth of the old Lucas/Kydland/Prescott business cycle models, a key aspect of the model is how agents can store their wealth, if at all. What kinds of assets can they save into? Many assume, realistically, that there are very few assets that the typical consumer has access to. But without explaining why in the model (without microfounding in other words). In their elegant summary of the state of the art in this literature, Heathcote et al call this a division between ‘model what you see’ and ‘model what you can microfound’. The same issue colours open economy macro, where results differ markedly between environments where markets are complete and where they are assumed, without microfounding, that they are not. The literature on credit frictions which the financial crisis has breathed life into also contains many examples of ad-hocery; the assumption that firms issue standard non-contingent debt, and usually only one period debt. Borrowing constraints that relate what someone can borrow to the value of their collateral. All common sense, but not microfounded in the models themselves.
Simon worries that the influence of these microfounded models is too great in our policy institutions. Well, I’d like to reassure him on that point. In the Bank of England, generous use is made of other models, and judgement overlaid on the main DSGE model. Second, many of those who produce the forecast and many of the policymakers have had no direct exposure to these models in their careers, and consequently don’t really understand them. Many think exactly in terms of older, non microfounded models that Simon and PK would prefer. There may be much less to worry about than Simon thinks! (At least in the UK).
Perhaps a third of macroeconomists either specialise or have a toe-hold in ‘empirical macroeconomics’. Here the bread and butter model is the vector autoregression. (A system where you regress everything on lags of itself and lags of everything else). This tool was explained to macroeconomists by Chris Sims, Nobel laureate. The whole point of it is to avoid making ‘incredible’ assumptions that emerge from particular models. And instead to interpret them using assumptions that would be true of many models, perhaps all models that the researcher would accept as having something sensible to say. Debates rage in this literature about just how safe these assumptions are, but this is the proper place for atheoretical modelling, Sims would say. Not tacking together equations that the researcher thinks roughly describe how firms or consumers go about life.
Simon’s tag-wrestlers in this complaint about microfoundations talk of ‘macro wars’. I think by this they conceive of macro as a battle between those who are content to use IS/LM or similar models, and those who want to use real business cycle models and their descendants. But I don’t see it like this at all. Although the blogging community might manifest this war, and although the IS/LM model is alive and well in introductory macro courses and textbooks, in the community of those publishing in peer-reviewed journals there is no longer any war. The pragmatic microfounders and empirical macro people have won out entirely. (At least if you ignore a few policy institutions!).