Forecasting Performance and M-Competition. Does the accuracy measure matter?

Resumen

Economic and financial time series are widely considered as one of the most challenging applications of modelingand forecasting. The increasing in forecasting availability and the controversial debate about the advantages ofalternative forecasting procedures suggest the need of further research on the forecasting evaluation metrics.In this context, this paper focuses on two information-based accuracy measures: Theil´s U Index and the QuadraticInformation Accuracy Measure (QIAM), and aims to re-examine the empirical results of the M3-Competition byMakridakis and Hibon (2000), and specifically those referred to the subset of macroeconomic and financial series.The computation of the proposed accuracy indicators leads to new rankings of forecasting techniques, showing somesimilarities and disagreements with the main conclusions by Makridakis & Hibon (2000), found on five error basedaccuracy measures. The obtained results also allow a complexity–accuracy analysis.