Over the past few years, the growth forecasts made by the International Monetary Fund (IMF) have displayed one consistent pattern: the forecasts are rosy at the start of the year, then revised downwards towards the end of the year, and the actual estimates of real growth turn out to be even lower.

Has predicting the fate of the global economy become more difficult in a volatile post-crisis world, or does the IMF suffer from an inherent optimism bias?

A look at the historical data on forecasts made by the IMF in its World Economic Outlook (WEO) reports seems to suggest that optimism bias may be the bigger culprit.

As the chart below shows, for much of the past quarter century, the IMF has over-estimated global growth. The forecast error for the spring WEO report tends to be higher on average compared to the autumn WEO report.

In the 25-year period between 1990 and 2014, the spring forecast error for real global growth was negative (implying over-prediction) in as many as 17 years, while the autumn forecast error was negative in 13 of those years.

Analysing the data on forecasts between 1990 and 2011, IMF’s Independent Evaluation Office (IMF-IEO) found that forecast errors were greater in low-income economies where reliable data was hard to come by, and in regions where the staff responsible for making the forecasts were inexperienced. The optimism bias of IMF forecasts was in large measure because of the inability of IMF economists to anticipate recessions correctly, as per the report of the IMF-IEO published in 2014. The large negative forecast errors in the recession years skewed the historical averages. The report also pointed out that the IMF tended to over-predict growth on average for the advanced economies. For most G-20 economies, the forecast errors were negative even if the errors were not statistically significant for all of them.

For several emerging markets, such as China and India, the IMF has historically under-predicted growth. In India’s case, the under-prediction bias is not statistically significant for the one-year ahead forecasts but is significant for the five-year ahead forecasts. In China’s case, the bias is significant for both forecasts.

Despite an average tendency to over-predict, the accuracy of IMF forecasts is comparable to that of other multilateral institutions and private sector firms, the IMF-IEO report observed. But that may be of little consolation when all macro-economists are getting their forecasts wrong.

Why has the team of bright people at the fund not found a way to check their inherent exuberance, especially at the start of the year? After all, this problem is not entirely unknown to economists at the fund. As the IMF-IEO report pointed out, the first study to examine the possibility and nature of biases in IMF forecasts was commissioned by the organization as far back as 1988, in response to growing concerns about bias in IMF forecasts. Since then, there have been three other studies which have examined this issue.

One reason why the forecast bias may have persisted is that there has been no systematic attempt at learning from these studies, as the IMF-IEO report pointed out.

“The fund did not put in place a formal process defining what is expected from each successive study; how the results of the study are to be communicated to staff, management, and the board; how staff and management should respond to the recommendations in the study; or how the follow-up should be implemented and documented," the report said.

If the IMF is unwilling to learn from its mistakes, perhaps we should. And next time the IMF releases its outlook for the global economy, maybe we should add a pinch of salt before we dip into it.