A data envelopment analysis-based framework for the relative performance evaluation of competing crude oil prices' volatility forecasting models

Forecasts of crude oil prices' volatility are important inputs to many decision making processes in application areas such as macroeconomic policy making, risk management, options pricing, and portfolio management. Despite the fact that a large number of forecasting models have been designed to forecast crude oil prices' volatility, so far the relative performance evaluation of competing forecasting models remains an exercise that is unidimensional in nature. To be more specific, most studies tend to use several criteria and their measures to assess the relative performance of these models, but competing models are always ranked by performance measure; thus, leading in general to different rankings for different criteria and to a situation where one cannot make an informed decision as to which model performs best with respect to all criteria under consideration. The purpose of this paper is to propose a single ranking that takes account of several criteria using a Data Envelopment Analysis framework. Our empirical results reveal that the unidimensional rankings for different criteria might differ significantly and that the multidimensional ranking of some models could be substantially different from their unidimensional rankings, which highlights the importance of the proposed performance evaluation tool.

► We propose a multidimensional framework for the relative performance evaluation of forecasting models. ► The proposed framework provides a single multicriteria-based ranking of competing forecasting models. ► For crude oil prices' volatility forecasting models, SMA20 performs best among all models tested. ► GARCH models, except CGARCH, have an average performance compared to smoothing models such as SMA20. ► Empirical results suggest that the data generation process of crude oil prices' volatility has a relatively long memory.