Forecast bias for U.S. GDP

How will U.S. gross domestic product change during the next two years? For asset allocators evaluating the state of the global economy and pondering its effect on asset prices, the question of the U.S. outlook remains important. The answer, however, seems to have consistently eluded forecasters from many government, quasi-government (e.g., international authorities), and private organizations for the past five years.

Point-in-time forecasts from some of these organizations since 2011 show an interesting and statistically significant trend. During the past five years, forecasters repeatedly have proffered overly optimistic predictions (by more than 50 basis points) for inflation-adjusted, long-term growth rates but excessively pessimistic predictions (approximately 25 basis points) for the near term. As asset allocators begin planning for 2017 and beyond, they may want to account for this bias when formulating their own long-term outlooks.

The forecasting history of the International Monetary Fund offers a representative case study. The IMF releases analyses of global economic developments through its World Economic Outlook report, published twice a year. The IMF's most recent forecast, from October 2016, predicts the U.S. economy will grow 2.2% next year and 2.1% during 2018.

Historically, the IMF's predictions have proven consistent with the subsequent official GDP estimate from the U.S. Bureau of Economic Analysis. Figure 1 shows the mean ex ante forecast error with 95% confidence bounds for the IMF's prediction relative to the BEA's ex post report. The figure shows the IMF's forecast error did not differ significantly from zero between 1996 and 2010. This statistical result holds true for forecasts 24 months ahead and for forecasts less than six months ahead.

However, since 2011, the IMF's forecasts have exhibited a consistent bias. When making two-year horizon forecasts, the IMF has overestimated GDP by 50 to 100 basis points. The bias shrinks as the forecast horizon shortens.

The IMF does not stand alone in offering too-rosy outlooks during the past five years. A closer look into U.S. GDP forecasts since 2011 from a set of forecasters that also includes the European Commission, U.S. Energy Information Administration, the Survey of Profession Forecasters (Philadelphia Fed), the Wall Street Journal, and the Congressional Budget Office reveals a similar upward bias.

Figure 2 shows the mean forecasting error in annual U.S. GDP growth rate with 95% confidence bounds based on forecasts pooled across the full set during the five-year period through 2015. Similar to the IMF, the average across all forecasters shows a positive bias (approximately 50 basis points) when looking two years ahead. The forecasts become more accurate as the forecast horizon shrinks, indicating that most forecasters tend to revise their estimates downward as data on actual economic conditions materialize. For near-term forecasts (i.e., less than six months), the average bias appears negative.

In short, forecasters have been too optimistic about the long-term economic growth rate but too pessimistic about the near term conditions.

Asset allocators planning for the next few years might still want to consider the views of experts from a range of government, quasi-government and private institutions, but they would do well to remember that even these experts have struggled of late.

In its latest annual real GDP forecast for the U.S. (released in October), the IMF revised down its estimates for 2016 to 1.6% from 2.4%.

Economists rarely earn criticism for excessive optimism, but the data suggest that could change.

Jeffrey N. Saret is head of thematic research, and Subhadeep Mitra, thematic researcher, at Two Sigma Investments LP.

Opinions expressed in this article are the authors' only.