AN EVALUATION OF USA UNEMPLOYMENT RATE FORECASTS IN TERMS OF ACCURACY AND BIAS. EMPIRICAL METHODS TO IMPROVE THE FORECASTS ACCURACY

Abstract

The most accurate forecasts for USA unemployment rate on the horizon 2001-2012, according to U1 Theil’s coefficient and to multi-criteria ranking methods, were provided by International Monetary Fund (IMF), being followed by other institutions as: Organization for Economic Co-operation and Development (OECD), Congressional Budget Office (CBO) and Blue Chips (BC). The multi-criteria ranking methods were applied to solve the divergence in assessing the accuracy, differences observed by computing five chosen measures of accuracy: U1 and U2 statistics of Theil, mean error, mean squared error, root mean squared error. Some strategies of improving the accuracy of the predictions provided by the four institutions, which are biased in all cases, excepting BC, were proposed. However, these methods did not generate unbiased forecasts. The predictions made by IMF and OECD for 2001-2012 can be improved by constructing combined forecasts, the INV approach and the scheme proposed by author providing the most accurate expections. The BC forecasts can be improved by smoothing the predictions using Holt-Winters method and Hodrick - Prescott filter.

Keywords

• forecasts
• predictions
• forecasts accuracy
• combined forecasts
• multi-criteria ranking