An analysis of forecasting errors in business tendency surveys

Business tendency surveys question a panel of companies about short-term changes in economic measurements such as output, employment and selling price. Companies express a qualitative opinion on expected changes (“over the next three months”) in these factors, and on past changes (“over the last three months”). These responses provide a unique window of observation onto their forecasting behaviour. By monitoring companies over time it is possible to analyse their short-term forecasts and identify any errors or surprises, when changes differ from their expectations.

These forecasting errors do not happen by chance. They depend first of all on the economic factors being considered: businesses make fewer forecasting errors over their selling prices and their workforce than over demand for their products or even their own output. In addition, errors have a seasonal component, which may be more or less significant depending on the nature of the economic factor. Businesses are therefore more often surprised during the April surveys and contradict the expectations they expressed three months earlier more often than they do in the other quarters of the year. And finally, these errors appear to be procyclical: businesses are more often surprised by an increase during phases of recovery, and are more often surprised by a decrease during recessions. From one quarter to the next, forecasting errors are therefore correlated with changes in activity the following quarter.

The observation of companies’ forecasting errors raises question about the process by which their expectations are formed. Responses to questions on expected changes in orders in industry, and turnover in services are modelled to reveal that these expectations display a certain inertia. Companies tend to expect the same changes that they have just observed in the recent past, either with an increase or a decrease. However, they take into account the global economic environment and their own mistakes, correcting their expectations when they have been surprised by an increase or a decrease. Thus they use the information available to them to shape their expectations, which appears to be compatible with the rational expectations hypothesis in economic theory. When this hypothesis is formally tested, it is rare for it to be rejected in the services sector. However, it is more difficult to verify in industry, a sector where common short-term shocks at macroeconomic level can take companies by surprise. Forecasting economic turning points thus appears all the more difficult as they are not always anticipated by the companies themselves.