Putting up another post all too quickly between conference sessions:
Forecasts vs mechanisms in economics, by Chris Dillow: This discussion between Edmund Conway and Andrew Lilico on the Today programme on the alleged crisis in economics seems to me to rest upon a misunderstanding of what economics is.
Conway says the crisis has been “an earthquake for economic thought” and Lilico says we need “new theories.“ This, though, seems to regard economics as a settled but inadequate body of knowledge and theory. It’s not. It is instead a vast number of diverse insights. What’s more, all of the insights that help explain the current economic crisis were, in truth, well known to economists before 2007, for example:
- Risk cannot be simply described by a bell curve. But we learnt about tail risk on October 19 1987. And we learnt from the collapse of LTCM in 1998 that correlation risk, liquidity risk and counterparty risk are all significant.
- Assets can be mispriced. But we’ve known about bubbles for centuries - since at least 1637. Their existence does not disprove the efficient market hypothesis; as I’ve said, the EMH is not the rational investor hypothesis. Nor, contrary to Conway’s implicit claim, is the EMH inconsistent with the possibility that behaviour can be swayed by emotions; the EMH allows for the possibility of time-varying risk premia*
- Long periods of economic stability can lead to greater risk-taking. We’ve known this since (at least) Hyman Minsky.
- Banks can suffer catastrophic losses - which are correlated across banks. We learnt this - not for the first time - in the Latin American debt crisis of the early 80s and in the crises in Japan and the Nordic countries in the early 90s. Banking crises are a regular feature of even developed economies.
- Institutions, such as banks, can be undermined by badly designed incentives. But there’s a huge literature on the principal-agent problem.
- The current crisis, then, has not thrown up much that economists didn’t know.
Instead, our problem is a different one. It’s that what we have are lots of mechanisms, capable of explaining why things happen and the links between them. What we don’t have are laws which generate predictions. In his book, Nuts and Bolts for the Social Sciences, Jon Elster stressed this distinction. The social sciences, he said:
Can isolate tendencies, propensities and mechanisms and show that they have implications for behaviour that are often surprising and counter-intuitive. What they are more able to do is to state necessary and sufficient conditions under which the various mechanisms are switched on.
This is precisely the problem economists had in 2007. We knew that there were mechanisms capable of generating disaster. What we didn’t know is whether these were switched on. The upshot is that, although we didn’t predict the crisis, we can more or less explain it after the fact. As Elster wrote:
Sometimes we can explain without being able to predict, and sometimes predict without being able to explain. True, in many cases one and the same theory will enable us to do both, but I believe that in the social sciences this is the exception rather than rule.
The interesting question is: will it remain the exception? My hunch is that it will; economists will never be able to produce laws which yield systemically successful forecasts.
What’s more, I am utterly untroubled by this. The desire for such laws is as barmy as the medieval search for the philosopher’s stone. If you need to foresee the future, you are doing something badly wrong.
* The basic insight of efficient market theory is that you cannot out-perform the market except by taking extra risk. I am sick and tired of hearing people who still have to work for a living trying to deny this.
I think the statements on prediction are overly broad. If you raise the price of a good, in all but a few cases such as when price is interpreted as a signal of quality, we can predict what will happen, quantity demanded will fall. By exactly how much will quantity demanded fall? In some microeconomic applications, the bounds can be fairly tight. For example, I suspect Hal Varian at Google - who has access to vast amounts of data and the ability to conduct all else equal type experiments - has a fairly tight estimate of important parameters that indicate how, say, changing the price of an ad will impact Google's revenue stream. He has also been doing some interesting work on prediction, e.g. see Predicting Initial Claims for Unemployment Benefits. But in other cases, particularly in macroeconomics and the prediction of turning points, success has been much more modest (or absent altogether). However, I am not as pessimistic as Chris that we will never be able to do predict the course of the economy, but it will require that we begin to better understand how pressures build within the macroeconomic system, how to measure and monitor these pressures (e.g. measures such global and sectoral imbalances or price rent ratios, but those are hardly sufficient in and of themselves), and ultimately how to relieve the pressures when they begin to build to threatening levels.