How accurate are economic forecasts?

1 Apr 2016

3 min read time

At a glance

By Adrian Rollins and David Walker

For most businesses, the most useful thing an economist could do would be to tell them when a downturn was coming. Then they could run off inventory, pull back on investment and new hiring, stop replacing employees who leave, and so on.

The bad news: predicting the economy’s next few years – so-called “macroeconomic forecasting” – is one of the areas in which economists fail miserably.

This was confirmed recently by a study into recessions around the world. Of 62 recessions in 2008 and 2009, economists predicted not one of them. Of 88 recessions from 2008 to 2012, economists picked just 11.

As International Monetary Fund researcher Prakash Loungani put it: “The record of failure to predict recessions is virtually unblemished”.

Central bankers would love to make perfect predictions: they could cut interest rates ahead of downturns, stop some recessions and ease others and save millions from hardship and joblessness.

Yet a 2012 Reserve Bank of Australia (RBA) study concluded candidly that RBA forecasts “explain very little” of the changes in measured economic growth, unemployment and inflation. (Private forecasters weren’t substantially better, it noted.)

Economists themselves mostly understand their poor forecasting record. Many businesses, however, want at least a best guess about the economic future.

Economists, moreover, mostly find that the media won’t report their comments about their forecasts’ weaknesses. In the words of RBA governor Glenn Stevens, people appear to want “the illusion of certainty”.

Swayed by sentiment

On the problems of forecasting, many economists point out that one of the most important inputs to any short-term economic prediction is people’s feelings about the future.

National Australia Bank chief economist Alan Oster, a former IMF and Australian Treasury staffer, describes economics as “applied psychology with a bit of statistics around it”.

Oster and other economists pay close attention to consumer sentiment surveys. Sentiment is particularly hard to predict, as it can be self-reinforcing. If enough people act like there is a recession, there will be one.

Limits of the data

Forecasters also have to grapple with the fact that what is going on in the economy is only ever imperfectly known and understood. Economists rely heavily on data provided by the official statistician – in Australia, the Australian Bureau of Statistics.

This information is only as good as the methods used to collect it, and Oster believes we should be more aware of the limitations of economic statistics, which can rely on data from a small base.

For instance, Australia’s official employment figures use data collected from just 26,000 households, which contain 0.32 per cent of all Australians aged 15 years or older. This small base can sometimes throw up strange results.

The known unknowns

To make forecasts, economists typically create computer models that seek to emulate the interaction between dozens or hundreds of variables in the real economy.

As Commonwealth Bank chief economist Michael Blythe points out, the quality of results that a model produces is only as good as the ideas and assumptions that underpin it.

Don’t expect precision

The RBA’s Stevens said in a November 2015 speech that the bank now tries to emphasise a range of possibilities, of which its “central forecast” is only the most likely.

He also disclosed that he and other RBA staff have considered leaving out the central forecasts altogether and just publishing the likely ranges.

They concluded the media would just report the number in the middle of the range. Sadly, Stevens concludes, “we, as human beings, will be irresistibly drawn to those who claim to be able to forecast the future.”