Evaluation of the OECD’s prediction algorithm for the annual GDP growth rate

Abstract

Forecast of gross domestic product (GDP) growth differences between countries serves an important role in the justification of governments fiscal policy measures and in making decisions on future plans. Prior to using predictions in practice, one should evaluate the proposed prediction algorithm and ensure that it produces accurate and reliable predictions. This evaluation includes many aspects but one of the most important ones is the comparison of the predicted values with the actual outcomes. The Organization for Economic Co-operation and Development forecasts twice a year the GDP growth rate of its 36 member countries. In the present article, the actual annually GDP growth rates for all member countries, from 2013 to 2018, are examined simultaneously at separate years, and not in the long term, with respect to their corresponding forecasts given one or two years earlier. The results indicate a weak and in most cases inaccurate used prediction algorithm for the GDP growth rate for the OECD member countries.

KEYWORDS:

• Forecast accuracy
• forecast evaluation
• outlier formal detection
• general linear hypothesis test
• macroeconomic variable prediction