Real gross domestic product growth forecasts by major international organizations frequently fail standard tests of forecast rationality, though differences in accuracy between the organizations are rarely statistically significant, according to a Monday research discussion paper by the Reserve Bank of Australia.
The paper examined forecasts by four international organizations, the International Monetary Fund, the World Bank, the European Commission and the Organisation for Economic Co-operation and Development, alongside private-sector forecasts, noting that such projections influence policy decisions by governments and receive significant attention from markets.
When comparing accuracy between organizations for individual economies, the researchers found that one forecaster almost always outperformed another over the sample period, but those differences were rarely statistically significant, meaning they could have been by chance rather than reflecting superior capability.
By contrast, each forecaster showed statistically significant departures from forecast rationality for the majority of economies, meaning their forecasts did not minimize the chosen measure of forecast error given the information available at the time, the paper added.
Common reasons included optimistic bias, overly large revisions and forecasts that were too extreme, with the researchers noting that accuracy would have improved had forecasts been slightly lower, revisions smaller and extremes avoided, though this does not mean the forecasts were irrational in the everyday sense, per the paper.
The researchers explored two explanations particularly relevant to international organizations, the first being that forecasts are often conditional on assumed future outcomes for a set of explanatory variables, which can make them look irrational under the standard test.
The second explanation is that organizations may produce modal forecasts, reflecting their view of the most likely outcome, while standard rationality tests assume a probability-weighted average forecast, with evidence suggesting the modal forecasts tend to run slightly higher, the paper added.
The researchers said both practices are reasonable, and concluded that while failures of rationality tests may point to ways of improving accuracy, they could also reflect legitimate forecasting practices, making rigorous evaluation essential to interpreting performance and guiding better methods.
