What place do applied middlebrow models have?

Mainly for economists, although the jargon I cannot help using is not critical to the message

Paul Krugman writes that “the effect of the insistence that everything involve intertemporal optimization has been to drive out middlebrow economic modeling.” I like the term middlebrow. A ‘highbrow’ is ‘One who possesses or affects a high degree of culture or learning’, and I think that has a nice ambiguity to it.

Paul gives a couple of examples where middlebrow research is getting squeezed out. I want to add an applied example of my own which I believe shows clearly that there is a problem, but also why there is no simple answer. The story is a little on the long side, but telling it provides evidence of a pattern which is my main point.

I first started working on the calculation of equilibrium exchange rates with Ray Barrell in the late 1980s. The partial equilibrium model we constructed, looking at the G7 currencies, was based on an approach pioneered by John Williamson, which he called the FEER. This essentially uses trade equations to estimate the exchange rate which would produce an off-model guess at a ‘sustainable’ current account. If you want to think of it in terms of a two dimensional diagram, think about a supply and demand curve in real exchange rate/output space (sometimes called the Swan diagram: see this post for example.) It is a good example of a well used middlebrow model. What FEER analysis effectively does is estimate the demand curve, conditional on off-model assumptions about asset accumulation.

Our paper attracted a lot of interest, and we did send it to a couple of journals. However the response was consistent: the model was rather traditional, partial equilibrium, not microfounded, and therefore of not enough interest for a major journal. All this was true, so we gave up on publication. However the UK was about to enter the ERM, so we used the model as part of a comprehensive analysis of an optimal entry rate. Our analysis, published by the Manchester school as part of a conference volume, suggested that the rate we did join the ERM at - 2.95 DM/£ - was too high and would intensify the recession. Subsequent events did not prove us wrong. [1]

In the late 1990s John Williamson asked me to update the analysis, which I did with Rebecca Driver. It was published as a monograph, and neither of us thought it worth the effort to go for what we knew would be a minor journal publication. 

In 2003 the UK government had to decide whether to join the Euro. It undertook a number of background studies to help it make that decision, but it also needed to know what rate to enter at if they decided to join. Initially they asked me to review existing studies on what the equilibrium Euro/Sterling rate might be. There were no papers (microfounded or not) published in top journals. I focused on three studies: my earlier work with Driver, one available as an IMF working paper, and one published in a policy oriented journal. They then asked me (in 2002) to update my own earlier estimates, which I did with a new model of the yen, dollar, euro and sterling. At the time the exchange rate had been around 1.6 Euro/£ for over 2 years - I calculated that the equilibrium rate was a little below 1.4 E/£. By the time the study was published in 2003 the rate had fallen to around 1.45 E/£, and it stayed near that level until 2007.

There is a lot to say about that work, including a rather amusing incident that happened just after the study was published, but that is for another day. The point I want to make here is that I did not even try to publish this 2002 exercise. I feel rather guilty about this now, because the study is so hard to find. [2] Yet as an academic the incentives for me to publish in a minor policy orientated journal were zero.

John Williamson’s FEER framework anticipated aspects of the New Open Economy approach. In the mid 2000s Obstfeld and Rogoff applied (without as far as I know acknowledging Williamson) a similar approach to an analysis of the US exchange rate and current account, at a time when many people thought the financial crisis would be all about the dollar. The best known example of this work was published as a Brookings paper. Their analysis is microfounded in the sense that it includes deep parameters like demand elasticities, but it remains partial equilibrium and there is nothing intertemporal. Their microfoundations focus is elegant, but its output is more limited: they only calculate the impact of a given US current account change on equilibrium exchange rates, and do not try and estimate what equilibrium rates actually are. However the point I want to make here is that this was analysis of key interest, undertaken by two of the best international macroeconomists in the world, but it was not published in one of the top six journals. [3] [4] [5]

It is not difficult to see why. There is little here to interest either a theorist or an econometrician. The theory is well known, and models are either calibrated or involve very simple estimation. Yet constructing a model (microfounded or middlebrow), and taking it to the data, is a non-trivial task, and there are plenty of ways to get nonsense out. I use a lot of skill and experience in this work. (You could say they are the skills and experience of an engineer or experimental scientist, rather than those of a theorist.) More importantly, the output is of considerable interest to policymakers and others. Every time someone says a currency is under or overvalued they are making a judgement about equilibrium exchange rates.

Now I suspect a lot of academic macroeconomists would say that this is the kind of work that should be done in policymaking institutions, and not by academics. Yet I kept being asked to update my work, and I guess Obstfeld and Rogoff did theirs, because policymaking institutions - with the important exception of the IMF - typically do not have the resources to maintain and develop models of this kind. Indeed increasingly I suspect that because it will not be published in good journals they infer that it is somehow not worth doing. [6]

I cannot help feel that there is some kind of ‘knowledge failure’ here, and that it is fairly specific to macro because macro involves models. This is not frontiers of research stuff, so you can see why you would not find it in the journals that leading theorists or econometricians would routinely read. My work was an empirical application of a middlebrow model, but still in my view the most reliable method we have at calculating equilibrium rates. It is quite consistent with a microfounded approach, but the microfoundations themselves are not terribly interesting. It can be very important (suppose we had joined the Euro in 2003), yet I did not have the incentive to even try and publish what I had done. Isn’t that rather strange?

[1] We published our analysis before we joined the ERM. A well known FT journalist told me at the time that he thought we had won the intellectual argument, but he still felt instinctively that 2.95 was the right rate to enter at. The UK government also followed their instincts, with disastrous results.

[2]  It recently took me 30 minutes to find a ‘snapshot’ from the Treasury website hosted in the national archives where the links worked - thanks HMT!

[3] This update of my Treasury analysis (published as a book chapter) has an appendix which goes through the microfoundations of the FEER approach. I could talk about the relative merits of a microfoundations centred approach and my more data based FEER approach, but that is not the key point I want to make here.

[4] Why not make the model general equilibrium using intertemporal theory to model the current account, for example? Unfortunately the standard intertemporal model is hopeless at describing trends in the data, yet matching trends in the data is vital in calculating equilibrium exchange rates. I essentially make the same point here when discussing the US savings ratio. There have been empirical studies of medium term current accounts (I discuss a recent one here), but these are reduced form and pretty ‘ad hoc’ by today’s standards. Again you will not typically find these studies in the top journals.

[5] Why is publication in top journals important? Because that gives the best economists the incentive to try and improve or develop work, and therefore to get even better answers. Policymakers asking economists like me to do occasional work is fine, but you really want a forum where others can – uninvited – critique and improve on that work, and which provides a memory so that new work does not ignore what has gone before.  

[6] You might be tempted to say that if this analysis was any good, those involved in the FOREX market could make money from it. I have been in a number of meetings with people like that, and they are interested until they ask how long an exchange rate might typically take to get to its equilibrium rate. When I say around 5 years, they nearly always lose interest! For those who think all you need at this horizon is PPP, read this.