Rationalising the Macroeconomy
Writing about web page http://www.ft.com/cms/s/0/478de136-762b-11de-9e59-00144feabdc0.html
In The Financial Times on July 21, Paul de Grauwe published the best comment I have read so far about the crisis in macroeconomic policy. If your time is scarce, don't read on; click the link and read him.
De Grauwe makes a fundamental argument, which I will summarize in four steps.
- Today, macroeconomists are distributed along a spectrum from "Keynesian" at one end and "Classical" at the other. They tend to clump at the extremes so there are many passionate Keynesians and passionate Classicals, as well as less passionate scholars in between.
- The Classical macroeconomists expect the macroeconomy to bounce back quickly from a major disturbance (for example, a credit crunch) on its own accord; government intervention is more likely to hinder than help. The Keynesians believe the opposite.
- For practical purposes, both schools model the behaviour of the people in the macroeconomy as follows: their behaviour is based on expectations of the future that are guided by the model, whether Keynesian or Classical. Classical macroeconomists assume that people's behaviour is based on the expectation that the outcome of the Classical model will be fulfilled, and Keynesian macroeconomists similarly.
- In both models, these expectations are self-fulfilling.
De Grauwe's punch line:
So what? Does it matter that economists disagree so much? It does. Take the issue of government deficits. If you want to forecast the long-term interest rate, it matters a great deal which of the two camps you believe. If you believe the first [Classical] one, you will fear future inflation and you will sell long-term government bonds. As a result, bond prices will drop and rates will rise. You will have made a reality of the fears of the first camp. But if you believe the story told by the second [Keynesian] camp, you will happily buy long-term government bonds, allowing the government to spend without a surge in rates, thereby contributing to a recovery that the second camp predicts will follow from high budget deficits.
In short, in a Keynesian model, the agents are assumed to expect that a credit crunch will have lasting adverse consequences. As a result they will rein in consumption (because households expect lower incomes) and investent (because firms expect depressed markets). The economy will stay depressed until government action flips the economy back to normal. But in a Classical model, the agents are assumed to expect that a credit crunch will soon be overcome, provided markets are allowed to work normally. Do nothing, and any damage to confidence will soon be restored. Unnecessary government action, however, by enlarging public spending and debt, will depress long term expectations and so inhibit the restoration of confidence.
This point is not new. I'm not sure who made it originally. It has been around a long time. I checked my notes from 1998/99, the first year I lectured to first year undergraduates at Warwick on this particular topic. I found the following passage:
We’re trying to explain a state of the world in which at least some unemployment is involuntary, money isn’t instantly neutralised by price change, and business cycles last anywhere between 5 and 9 years. The fundamental problem of the RE [rational expectations] approach is that it proves this state of the world can’t exist. Underlying this are some basic conceptual faultlines. Learning from experience may be more difficult than RE theory assumes. Large experiments are rarely if ever repeated under controlled conditions (e.g. joining, then leaving the ERM). Large shocks (e.g. oil shocks, monetary shocks) make it hard to discern the underlying things which remain the same. What is the true model of the macroeconomy? RE theorists tend to assume that most people adhere to a Classical philosophy. But since economists have such difficulty decided how best to model the economy, it’s not clear why rational non-economists should be different. Policy demonstrably does affect the real economy, so why should rational people believe it won’t? This is particularly important since the outcomes of actions based on RE tend to force the world to conform to the model, not the other way round. What is created here is a "guessing the winner" problem: what’s important in forming rational expectations is not "how does the economy work?"; nor even "how does the economy work in my opinion?"; but "how does the economy work in most people’s opinion", bearing in mind that in forming their opinions they are all asking themselves the same question.
I claim absolutely no credit for this; I was not saying anything original. I got the argument from somewhere or someone else. My point is that the basic paradox in rational expectations has been understood for a long time, but the horrendous policy implications are perhaps only now fully apparent.
How bad does that make economists? Ten years ago I told my students that the idea of rational expectations, although not wrong, contained a paradox. I had no idea how to resolve it, however. One route the profession has taken has been to consider that, just as economists learn, so do non-economists. As a result, macroeconomic models have been developed that incorporate heterogeneous expectations -- when different people in the macroeconomy start out with different models of how the economy works and so different forecasts of the future -- and model how they might then learn from experience. A recent review by George W. Evans and Seppo Honkapohja is here.
This takes me well outside my comfort zone. I thought about it, however, when a friend forwarded some lines from an internet discussion including the suggestion:
Until the "science" of economics detaches itself from econometrics and unilateral modelling and realises that humans are "rationalising beings", not "rational beings", then the predictions and opinions stemming from its adherents should be treated with caution.
In the context I took the gap between "rationality" and "rationality" to reflect some falling short of cognition or computation. It wasn't that I disagreed; the suggestion seemed almost trivially true (apart from the reference to econometrics, which seemed silly). What it made me think is this: If humans are "rationalising beings," then so too, being human, are economists. All economic models have cognitive and computational limits. They model reality; they don't and can't reproduce it.
In the often misquoted words of George Box and Norman Draper (from Empirical Model-Building and Response Surfaces, New York: John Wiley 1987, p. 63):
All models are wrong; the question is how wrong do they have to be to not be useful.