Proposal of New Forecast Measures: M Indicator for Global Accuracy of Forecast Intervals

Abstract

The objective of this study is to analyse the problem of forecasts accuracy
from two aspects: the point predictions and forecast intervals. For the point
predictions the classical measures of accuracy were used by adding some
new indicators for forecasts of quantitative variables. For forecast intervals,
there are no specific measures of accuracy mentioned in the literature.
Therefore, this paper proposes different methods for constructing the
intervals and some measures of accuracy for this type of forecasts as a
novelty in literature. Classical and proposed accuracy measures of
predictions were computed for the inflation and unemployment rate forecasts
provided for Romania by the European Commission, Institute for Economic
Forecasting (IEF) and National Commission of Prognosis (NCP) on the
horizon 2010-2012, the best forecasts being proposed by IEF. A novelty in
literature is also brought by introducing the methods of building forecast
intervals. To the classical interval based on the root mean squared error
method, we add the intervals based on the standard deviation and those
constructed using bootstrap technique bias-corrected-accelerated (BCA)
bootstrap method. Also M indicator was proposed to evaluate the global
accuracy of forecast intervals.