The Bank of Canada's failure to forecast the fall 2012 economic downturn is an ongoing mystery. What makes this failure even more puzzling is that the Bank of Canada's forecasts, while overly optimistic, have been relatively accurate since the recession.
Forecasting year-over-year quarterly gross domestic product (GDP) growth is naturally a difficult task. Any unexpected economic event anywhere in the world will affect the Canadian economy. The variable being forecast, GDP growth, is subject to all sorts of measurement error, which is why it is often revised several times after the initial estimate is released. Any criticism of economic growth forecasting must keep in mind the difficulty of the task.
In order to judge the central bank's forecasting ability, I examined the Bank of Canada's forecasts from the beginning of 2008 to the present. Bank of Canada forecasts are available each quarter in its Monetary Policy Report (MPR), with quarterly GDP growth predicted seven or more quarters past the date of each MPR release. I examined the forecasts for year-over-year quarterly GDP growth for 11 periods from the first quarter of 2010 until the third quarter of 2012. The starting date was chosen to avoid examining the recession, in which forecasts all over the year were poor.
Three trends pop out when examining the data:
- Bank of Canada forecasts are overly optimistic on medium-term growth. Between five and seven periods out, the average growth forecast is 0.6 percentage points higher than actual growth. This is not due to one or two observations skewing the sample; 75 per cent of medium-term forecasts by the bank were overly optimistic. The economy has simply not recovered as quickly as the bank had anticipated.
- The shorter the forecast period, the more accurate the forecast. The Bank of Canada has done a terrific job in revising its forecasts as more information is revealed to the bank. Forecasting two quarters out, their average forecast error is zero.
- While short-term forecasts, on average, are more accurate, there is still a great deal of error in those forecasts. In forecasts of one to three quarters in the future, the average absolute error in forecasts is roughly 0.5 percentage points of GDP. This differs only slightly from the 0.65 average absolute error in forecasts from five to seven months in the future. This should not reflect badly on the Bank of Canada, as there is a fair amount of noise in the actual GDP data, making year-over-year quarterly GDP growth difficult to forecast.
Mike Moffatt is an assistant professor in the Business, Economics and Public Policy group at the Richard Ivey School of Business, University of Western Ontario.