Yesterday I had lunch with an old friend who is a top monetary economist. We were discussing a conference on the state of macroeconomics that I attended recently. The conference seemed to me to reveal quite a deep division between those macroeconomists who thought their pre-crisis New Keynesian DSGE models were basically fine, although obviously in need of having a financial sector added to them, and other economists taking a wide range of approaches but pretty sure the pre-crisis mainstream approach is defunct. I’m in the latter category, with the caveat that it’s nearly 25 years since I’ve done any proper macroeconomics.
Alas, my dear friend is in the amended status quo camp. He said the tools a monetary economist needs can be found in the Michael Woodford tome [amazon_link id=”0691010498″ target=”_blank” ]Interest and Prices: Foundations of a Theory of Monetary Policy[/amazon_link]. His argument went like this. Any macro model makes simplifying assumptions. We macroeconomists could have assumed that prices and wages are perfectly flexible, which would make the models easier, but we didn’t. That was a correct call. We did make the simplifying assumption that the financial sector was a veil and we could ignore the incentives on bankers’ behaviour, and that was a misjudgment. But economists understand incentives, it’s straightforward to fix, and we’re putting the financial sector into the models now.
[amazon_image id=”0691010498″ link=”true” target=”_blank” size=”medium” ]Interest and Prices: Foundations of a Theory of Monetary Policy[/amazon_image]
This argument is on the face of it perfectly reasonable, but it reminds me of a conversation I had with another tleading economist in the mid-1990s, when I was working as a journalist reporting on technology and becoming enthused about the likely implications of the internet and World Wide Web for the economy. This economist argued that all the internet did was reduce transactions costs, and economists understand how to model transactions costs, so the new technologies would turn out not to be very interesting analytically. This was also true in a small way, and false in a much bigger way.
The moral is that the judgments reflected in simplifying assumptions are not a small technical matter. Transactions costs are not a side-show in economic analysis, they are the main attraction, driving the fundamental structures and institutions of the economy. So it is with the assumptions behind conventional macro models, which encapsulate a world-view about the nature of economic and political institutions – the failure of pre-crisis DSGE models is not a question of small and easily corrected misjudgements but rather missing the main point.
This non economist is dumbfounded by both the brass neck and complacency of this statement “basically fine, although obviously in need of having a financial sector added to them”
When I was reading Physics, many years ago, we always used to use the pat phrase “an elephant of negligible mass on an infinitely thin string” to mean an unrealistic model of the real world that would lead to incorrect conclusions, but I think I like “the banking sector is a veil” much better now.
Ah yes, we economists are good at invisible elephants on imaginary string….
It depends what the models are for.
If you want a model that’s going to predict crises, or warn us what things to look out for and try to correct in order to prevent crises, you need a model in which things can go wrong in some sense, where imbalances and fragilities build up and then snap. You need endogenous business cycles. Even these might be based on a DSGE framework but with some different behavioural assumptions.
If you want models that help us work out what to do after a crisis hits, you may be able to use a model where economies are hit by exogenous shocks, but which capture how the economy response to these shocks and to policy interventions. DSGE models with the right sorts of frictions, a banking sector and whatever other additions prove necessary, might be fine from that point of view. Mainstream DSGE models were never in the business of explaining where crises came from.
Well, I agree about DSGE models – like any formalised model – not being good at predicting crises. So the question is about the exclusion of all other intellectual approaches, if one wants to have some predictive ability. Old-fashioned ad hoc policy – keeping an eye on the markets and rule of thumb ratios – may still have a place in the policy armoury? I don’t think I agree with you about DSGE models + frictions being useful post-crisis. For example, how can one include in them thinking about the structure of the banking market in considering monetary transmission mechanisms?
I’m not quite sure what you mean by “thinking about the structure of the banking market in considering monetary transmission mechanisms” but if you look in the Handbook of Monetary Economics (ed Woodford and Friedman) you will find a section on the monetary transmission mechanism and within that articles that look at the role of financial intermediation therein (Gertler Kiyotaki, Adrian and Shin). I don’t know if these papers study the “structure of the banking market” in the fashion you desire, but then again I wouldn’t want to claim those models of financial intermediate could not be extended in that direction, if you sat down with Gertler and similar economists and told them what you’d like their models to incorporate.
I agree with you about not wanting to exclude alternative intellectual approaches … which means I don’t want to exclude DSGE either. I’m open to the possibility that other approaches may ultimately prove more useful and displace DSGE, and as I’ve said the whole “assume exogenous shocks” approach is never going to help us understand the source of shocks, but I think in the rush to condemn mainstream macro, people might be underestimating the potential usefulness of DSGE, for certain purposes at least.
Yes, that’s an absolutely fair point.
On the former, I meant the kind of market analysis we do in competition work, looking at what affects margins – ie. interest rate differentials in this case.
I think it’s important not to focus too much on DGSE – because in the future we’ll invent new models with a different four letter acronym and the problem will remain, despite the change of name.
That problem is that any model which appears to be “mathematical” and “complete” is likely a very dangerous beast, because economists invest faith in it and stop looking for the flaws in the model. Even worse, they tend to embark on ideological crusades to use policy to make reality look more like the model.
Thus, for me, the key question is how can we improve the relationship between economists and models. How can we educate them in such a way that the models do not become articles of faith? Or perhaps more immediately, how can we change economic journals so that producing “epicycles” on current models is not over-rewarded relative to exploring alternative models.
Did you see Gary Gorton’s interview with the FT on Oct 25 ? He argues that the key problem is one of measurement. We dont have the data on the things that matter in the modern financial system, and so we cannot correctly calibrate our models. We need a measurement revolution for the 21st century comparable to Kuznets.
Paul Ormerod also sees measurement problems of a different kind – our last-century income accounting is not keeping pace with the knowledge economy so we are likely underestimating economic growth and overestimating inflation
I didn’t see the Gorton article – thanks for that.
Hmmm… and what about the judgement and simplifying assumption that the pre-crisis New Keynesian DSGE models have, in fact, failed? (As opposed to the people we entrusted with the responsibility of interpreting and applying them, i.e. the macroeconomics profession)
It’s hard to argue that those models alerted policymakers to the need to act ahead of the crisis, even if you think they’re working terrifically well now. By definition, equilibrium models, consistent expectations don’t address disequilibrium!
if we’re talking about the sticky price Calvo fairy models, they failed comprehensively because of their superficiality and ruling out coordination failure by fiat. There is actually a much “deeper” Keynesian literature about coordination failure, by the likes of Peter Diamond, Russell Cooper and John Bryant and Roger Farmer. This theory did not fail, in fact it was very prescient.