The forecasts used by the Bank of England to set interest rates are biased and contain little useful information, a Financial Times audit has demonstrated.
Despite having hundreds of economists working in the Bank, and the most sophisticated suite of economic models in the UK, the monetary policy committee’s forecasts since 1997 have achieved no better outcome than if the committee had simply predicted the average level for inflation and growth over the 13-year period.
With concerns rising over the Bank’s failure to foresee recent rises in inflation, these findings suggest the problems are, in fact, endemic. They will cast doubt over the credibility of its predictions for the course of the economy, due to be updated when it unveils its quarterly inflation report on Wednesday.
The Bank insists there is no automatic link between its forecasts and interest rates but its predictions form the basis of monetary policy discussion every quarter. In a sign of how seriously it takes the forecasts, the MPC has increasingly waited for the updates in each quarterly inflation report before taking decisions on interest rates and quantitative easing.
Bias has not always been such a big problem for the Bank. But this has changed since the Bank failed to spot the recession and repeatedly predicted a more rapid recovery than materialised.
Last week, Charles Goodhart, former MPC member and Bank chief economist, laid the forecasting problem at the door of the Bank’s main economic model. “[It] has got no banks in it and no possibility of default, which is perhaps one of the reasons the model got it so wrong over the past few years,” he said.
Danny Gabay, director of Fathom Consulting, said: “Until 2007, they were slightly optimistic on growth but in the last three or four years, they have been wildly optimistic.”
At the two-year forecasting horizon, the Bank’s average error since 1997 has been to predict growth 0.9 percentage points too high. This bias stems from the Bank’s default forecasting practice that assumes good times will last, but that when times are bad they can only get better.
The Bank’s persistent optimistic bias extends to inflation, which has turned out higher on average than the Bank’s forecast, again with most of the problems being of recent vintage.
Beat the MPC at economics
If you have always fancied trying your hand at economic forecasting, but thought you needed access to hundreds of economists, millions of pounds and tens of PhDs to have a hope of beating the Bank of England at its own game – think again, writes Chris Giles.
So bad is the MPC’s judgment, and so flawed the Bank’s extremely sophisticated model, that all you need to get started are some simple data from the Office for National Statistics’ website. Here is how to do it.
For growth, first assume the current rate of year-on-year growth is equal to the figure just published for the previous quarter. Never mind that this figure relates to a different time period and will be revised, it is not a bad predictor of the present. Second, assume the economy will grow by 2 per cent a year. And third, adjust your forecast gradually from your initial point to 2 per cent over the course of that year. Hey presto, you’ve matched the Bank’s forecast.
For inflation, do the same. Forecast a smooth transition from the previous quarter’s inflation rate to 2 per cent over a year and assume the rate stays there. Over the past 13 years this approach – basic to the point of naivety – would have beaten the Bank’s forecast.
Just for safety’s sake, add a large margin of error, wider than the usual movements of growth and inflation. Then it is odds on that your forecasts will always be well within your self-designated margin of error.
Go and put your feet up on the beach with a self-satisfied smirk on your face.
Independent economists say that while these problems are undesirable, they are understandable be-cause, as Simon Kirby of the National Institute of Economic and Social Research says, everyone missed the recession and “forecasting is very, very difficult”.
But he adds that a more problematic aspect of the Bank’s forecasts is that they contain little useful information.
When the Bank is predicting growth more than a year into the future, its forecasting errors would be lower if it ditched its models and just said the year-on-year growth rate would be the average for the series – about 2 per cent.
For inflation, the forecasts since 1997 lose their power to beat the simple average of inflation after only six months.
As such, the models and forecasts published by the Bank in its quarterly inflation report do not tell readers anything they did not already know.
“There is a big question about exactly what the information content of the Bank of England’s forecasts is,” said Mr Kirby.
One possible answer is that because the Bank is not only forecasting inflation but also setting policy to achieve 2 per cent inflation two years hence, it is natural that it cannot beat an average of about 2 per cent.
On this logic, not doing better than the average is a sign of success not failure.
But in unveiling its report, the Bank is unlikely to push that argument because it does not pretend its forecasts are remotely infallible.
Instead, the MPC puts a wide margin of error around them and, in every August inflation report, tells the world that the outcomes are within that margin of error.
Unlike GCSE and A-level results, however, this assessment is a proven example of grade inflation because the Bank has regularly widened the margin. That leeway is now wider than the normal variability of growth and inflation, meaning that, according to the Bank’s own yardsticks, it is almost impossible for the forecasts to be wrong.
However, the downside of consistently widening the margin of error around the forecasts is that it again reduces the information the forecasts contain. If the practice continues, Mr Kirby said, “they’ll never get anything wrong but the forecasts will be of absolutely no use in setting policy because they do not contain any information”.
