Recently the FT’s Wolfgang Munchau referred to an article by prominent US economist Paul Romer, The Trouble with Macroeconomics, published in September 2016. Mr Munchau used it as an argument to rethink the conventional wisdom of macroeconomic management, such as independent central banks and inflation targeting. I agree we need a rethink, but Mr Romer’s article is the wrong jumping-off point for that idea. Instead Mr Romer shows that academic macroeconomics has lost touch with the real world.
Mr Romer references the Trouble with Physics an article from 2007 by Lee Smolin. At this time there was a lot of nonsense going on in theoretical physics, in particular with the idea of string theory. String theory attempts to be a theory of everything, and at its core is a lot of hard mathematics. But it makes no verifiable predictions and, indeed, seems to avoid areas where there is any danger that data might challenge it. It is more metaphysics than physics. And yet it commanded a sizeable academic following, including a number of big hitters – or at least it did in 2007.
Mr Romer suggests that something similar is happening in academic macroeconomics. People are creating elaborate models whose complexity runs well ahead of any data that can test their relationship to reality. This is covered up by sophistry and obfuscation. By itself this is not so strange, except that people are reluctant make public criticisms of these models, and the often prominent academics whose names attach to them. And yet that process of criticism is the stuff very of science. This makes it a pseudoscience – something that adopts the outward language of a science, but where a core set of beliefs and people are beyond criticism. (There is, of course, always a core set of beliefs in any system that are beyond challenge, including science, but I am talking about something wider here).
The prime target of Mr Romer’s criticism is a theoretical system referred to as the real business cycle. This was developed in the 1980s, and commanded supporters such as Chicago Nobel laureate Robert Lucas. It suggests that government actions, such as fiscal and monetary policy, have little effect on the real economy, and that the business cycle is almost entirely driven by changes to technology, a macroeconomists’ way of saying “just noise”. The real business cycle was presented to me as an economics undergraduate in 2006 as a curiosity that was so silly that it didn’t need comment or study. I can think of no serious piece of economic policy in recent years, and certainly since the financial crash of 2008, that makes reference to it. So it was a surprise to me to read that it remains the subject of serious academic support in the US – and that other serious academics look the other way rather than criticise it.
While Mr Romer spends most of his paper taking apart models based on the real business cycle, he makes it clear that this is a general problem, affecting Keynesian models too. And in particular the Dynamic Stochastic General Equilibrium (DSGE) models that are used for economic forecasting, and so hard-wired into economic management. Economists like to create huge models with lots of variables, and then pump huge amounts of data through them. But it is mathematically impossible to identify, that is to fix, the variables without building assumptions into the model about how they relate to each other – which the model is then unable to test. The models are therefore more a product of their assumptions than a test of those assumptions against data. Mr Romer complains of a conspiracy of silence not to undermine the fragility of all this. This ressembles string theory and other hobby horses of theoretical physics – though these days I read a lot about constructive work going on in physics, as scientists grapple with the problems of dark energy and dark matter. Also I think theoretical physicists are much more transparent about what they are doing – so far as I can see they aren’t even pretending that what they do is useful, except in some abstract sense of advancing the boundaries of human thought. Macroeconomists are dishonest by comparison.
How has all this come about? I don’t think it helps that macroeconomics has been politicised, and in the highly polarised environment of US politics. Real business cycle models are beloved of the right, as a basis for cutting government down to size; DGSE models are liked by the left, as the basis of fiscal and monetary intervention. Pretty much any important development in macroeconomics is parsed for its political significance. Neutrality does not seem to be an option – and yet cloaking policy prescriptions in academic mumbo-jumbo make them look more authoritative, and so demand for academic economists remains strong. US academics used to knock seven bells out of each other (the famous dispute between “salt water” and “fresh water” institutions) – but no doubt they now realise that this just devalues the whole discipline, and so the different schools ignore each other instead. Besides, they both rely on similar conceits.
How much does this matter? Those involved in the practical business of running economies have long since ignored real business cycle theory. Econometric models are used, but with a great deal of caution. Alternative ways of constructing models might be fruitful (for example Kingston’s Professor Steve Keen suggests the use of non-equilibrium complexity theory, as used in meteorology) – but these are liable to suffer from same identification problem. The future is inherently unpredictable. Practical economists can get on with the job without help from an increasingly irrelevant academia.
And yet there is a clear crisis in economic management. Too many people in developed economies feel left behind, fueling political instability that will not help economic management. The authority of the western democratic and inclusive system of government is waning as a result. If academic macroeconomists coud somehow change the direction of their discipline, rather than resorting to obfuscation to insist they were right all along, and deluding themselves that massive computer models are telling us anything useful, then the outlook would be better.