Never trust an economic forecast

tauringhuang.

When people discover that I am an economist, they rarely ask me for my views on subjects that economists know a bit about – such as how to respond to climate change or pay less at a supermarket. Instead they ask me what will happen to the economy.

Why is it that people won't take “I don't really know” for an answer? People often chuckle about the forecasting skills of economists, but after the sniggers die down, they keep demanding more forecasts. Is there any reason to believe that economists can deliver?

One answer can be gleaned from previous forecasts. Back in 1995, the economist and FT columnist John Kay examined the record of 34 British forecasters from 1987 to 1994, and he concluded that they were birds of a feather. They tended to make similar forecasts, and then the economy disobligingly did something else, with economic growth usually falling outside the range of all 34 forecasters.

Perhaps forecasting technology has moved on since then, or is the British economy unusually unpredictable? To find out, I repeated Kay's exercise with forecasts for economic growth for the UK, US and Eurozone over the years 2002-2008, diligently collected at the end of each previous year by Consensus Economics.

The results are an eerie echo of Kay's: for 2004, for example, 20 out of 21 non-governmental forecasts made in December 2003 were too pessimistic about economic growth in the UK. The Pollyannas of HM Treasury were more optimistic than almost any commercial forecaster, and closer to getting their forecast right. So one might suspect that systematic pessimism is to blame.

But no, in 2005, the economy grew more slowly than 19 out of 21 forecasters had expected at the end of the previous year. At the Treasury, they were again more optimistic than anyone, and thus more wrong than anyone. A year later, all but one of the forecasters were too pessimistic again. Yet at the end of 2001, three quarters of the forecasters were too optimistic about 2002.

tauringhuang.

An interesting anomaly is 2003: the one year for which the average UK forecast turned out to be close to reality, but also the year where the spread between highest and lowest forecast was widest. The rare occasion when the forecasters couldn't agree happened to be the occasion on which they were (on average) right.

Recent US forecasters have done a little better: the spread of forecasts is tighter and the outcome sometimes falls within that spread. Still, five out of six were too pessimistic about 2003, almost everyone was too pessimistic about 2002; three-quarters were too optimistic about 2005 and nearly nine-tenths too optimistic about 2006. Perversely, the most accurate forecasts were made about 2007, despite the fact that the credit crunch was a surprise to many.

In the Eurozone, forecasting over the past few years has been so wayward that it is kindest to say no more.

The new data seem to confirm Kay's original finding that economic forecasters all tend to be wrong in the same way. Their incentives to flock together are obvious enough. What is less clear is why the flight of the flock is so often thought to augur much – but then, some astrologers are also profitably employed.

The curious thing is that forecasters often have something useful to say, but it is rarely conveyed in the numerical forecast itself on which so much attention is lavished. For instance, in December 2006, British forecasters were warning of the risks of an oil price spike, a sharp rise in the cost of credit, and a dollar crash. Their guesses at economic growth were wrong, and would have been little use had they been right. But the forecasters said something worth hearing – if you had been listening carefully enough.