Modern macroeconomics: Is it really so bad?

I have to confess that I have long been torn between the intellectual pursuit of academic economics and the uselessness of much of the output. Part of the appeal of economic theory is that it attempts to address problems in a rigorous manner. Of course, that is also its Achilles Heel: the intellectual underpinnings of much that passes for current state of the art thinking are simply bogus. For that reason papers such as the one by Paul Romer, chief economist at the World Bank, entitled ‘The Trouble with Macroeconomics’ (here), always strike a chord. In many ways, this is a subversive read for macro economists and makes a number of serious, but in my view substantiated, allegations regarding the state of economics today.

Romer’s key thesis is that the “identification problem” in economics has essentially taken us round in a circle back to where we started in the 1970s. I hope readers will forgive a little digression at this point so that we can more easily understand the nature of the problem which Romer sets out. The identification problem requires, as Chris Sims noted in a famous 1980 paper, that we must be able to identify “observationally distinct patterns of behaviour” for a given model. This is both a philosophical and empirical argument. Philosophically, it requires us to specify very carefully how our economic system works. In an empirical sense, it means we must construct models in which unique values for each of the model parameters can be derived from other variables in the system. This in turn allows us to clearly identify how economic linkages operate.

As it happens most empirical macro models in use in 1980 were over-identified: It was possible to explain each variable in the model by various different combinations of other variables. Consequently, we were unable to determine precisely how the macroeconomy worked. In his 1980 paper entitled ‘Macroeconomics and Reality’ (a title which, when I first read the paper, seemed to be most inappropriate) Sims noted that such models could only be made to work by applying “incredible” identifying restrictions.

This identification problem is the key to understanding Romer’s critique of much modern macroeconomic theory. In his view, by trying to get away from imposing such “incredible” restrictions, macroeconomists have ended up devising models which themselves are increasingly divorced from reality. Romer starts by taking direct aim at Real Business Cycle (RBC) models which were a direct response to many of the criticisms of the over-identified macro models of the 1970s. He argues that they make a hugely simplifying assumption that cyclical fluctuations in output are solely the result of shocks. The question Romer poses is "what are these imaginary shocks?" Is there really no role for monetary policy, as much of thinking in this field suggests? If that is true, we should all pack up and go home – and the Fed, ECB, BoE et al should abandon attempts to stabilise the economic cycle.

He then aims his blunderbuss at the DSGE models which followed from this, arguing in effect that they are a post-truth way of looking at the world because they rely less on data and more on a series of assumptions about how the world works. (I posted on this topic here). Indeed, such is Romer’s apparent contempt for some of this analysis that he states “the noncommittal relationship with the truth revealed by these methodological evasions [and] dismissal of fact goes so far beyond post-modern irony that it deserves its own label. I suggest ‘post-real’.”

Even worse, in his view, is that many of the proponents of modern macroeconomics have tended to band together and reinforced each other’s views rather than challenging them. This unwillingness to challenge belief systems has, in Romer’s opinion, promoted a stagnant culture which has slowed the advancement of new ideas in economics.

Faced with such a nihilistic view of economics, you may wonder what is the point of it all? I share Romer’s criticisms of many of the ideas which have found their way into the economic mainstream. But I also adhere to the George Box school of thought that whilst all models are wrong, some are useful. There is thus nothing inherently wrong with the idea of going in the direction of RBC or DSGE models – it is just that they have captured the high intellectual ground and have proven difficult to shift. And like it or not, many of the competing theories have not proven up to scratch either.

What is also interesting is that whilst many economists passionately advocate a particular school of thought, few people of my acquaintance argue in public in such terms. So either this debate is a particularly American academic thing or it is confined to those pushing hard to get their material into the journals (and failing). And finally, I have long argued that the financial crisis will act as an efficient way of winnowing out many of the worst ideas in macroeconomics. Just as the Great Depression of the 1930s produced the ideas of Keynes and his acolytes, so the current crisis in the western word may yet lead to a more fruitful approach to many economic issues. So chin up, Mr Romer. The darkest hour comes just before dawn.

The state of macreconomics: A practitioners view

The academic blogosphere is currently abuzz with posts about the state of macroeconomics in the wake of an article by former IMF chief economist Olivier Blanchard on dynamic stochastic general equilibrium (DSGE) models. Blanchard's post, which is well worth reading for anyone interested in a non-technical overview of the models used in cutting edge academic research, highlights that these models are "seriously flawed, but they are eminently improvable and central to the future of macroeconomics."

As Blanchard notes, DSGE models are based on microfoundations, and essentially model the behaviour of various representative optimising agents (firms, households, central banks) whilst making a number of unrealistic assumptions. These are not just simplifying assumptions – they are probably downright wrong. One such example is the assumption of pricing behaviour assumed in such models which is designed to handle nominal rigidities. This so-called Calvo pricing assumption  is elegant but is not observed in the real world, thus rendering it somewhat useless. Blanchard goes on to point out that such models are mathematically dense, and thus are impenetrable even to many economists. As he put it, they are bad communications devices. My major criticism of such models, and one echoed by Blanchard, is that they are designed to fit the theory rather than the data with the result that they are empirically unsound. This probably reflects the way I was taught to do econometrics, but there is a quaint old fashioned notion that models should be congruent with the data.

Paul Krugman argues very strongly that Blanchard is too kind on such models and that far from offering scope for improvement, they have led us up an intellectual cul-de-sac. In his view they have contributed no insight into the workings of the economy and that old-fashioned ISLM models offered more useful predictions in the wake of the financial crisis than did DSGE analysis.

Others, such as the always readable Simon Wren-Lewis are less trenchant in their view but still critical. As SWL put it "DSGE completely dominates academic macroeconomics, and there is no way that all these academics are going to suddenly decide this research programme is a waste of time ... What is at issue is not the existence of DSGE models, but their hegemony." He has long argued that journal editors have routinely turned away good papers on the grounds that they do not offer a microfounded or general equilibrium approach and he offers the hope that "criticism from one of the best macroeconomists in the world might" prompt them to change their view.

Now I am no academic but I have spent enough time building and running economic models to know when a modelling paradigm is going broadly in the right direction. And I am pretty sure that DSGE is not the way for someone like me to go. My job is to understand how the economy fits together, and we know that there are structural changes over time which means that things that once worked no longer do. Estimating structural models allows us to do that, in order to see whether econometric relationships which once held continue to do so. The models I use are structural forecasting models, which would not have looked out of place 20 years ago, and many academic economists dismiss them out of hand. For one thing, they often struggle to model expectations in a way which satisfies academic thinking (though I am quite happy to pretend that the rational expectations revolution never happened). In addition, they are not identified (i.e. it is impossible to derive a specification for each individual variable in the equation which gives a unique specification in terms of others in the system) and the economic theory underpinning such models is often ad hoc.

To a degree, these criticisms are all valid. But these models have not been usurped by a superior paradigm, and certainly not by DSGE. I thus have a lot of sympathy with Krugman's view that macroeconomics has taken a wrong turning. DSGE models do not help most economists to understand how the world really works, and they certainly do not help to inform the general public. They are highly sophisticated mathematical tools which explain the world as many economists think it should be, rather than as it really is. DSGE models were a brave attempt to try and move the economic debate forward but they have failed. If economics wants to be more relevant to the wider policy debate, we need to find better ways of understanding how the world works, but as Wren-Lewis notes, too much intellectual capital has been sunken into the field to hope that it will go away quickly.

However, perhaps macroeconomics is ready for a Popper-style paradigm shift in which empirical falsification will prompt another generation of economists to push the field in a new direction and get us out of the rut into which we appear to have sunk. We could certainly do with a bit more public credibility and in that sense DSGE models are clearly not the way to go.