Accuracy of Forecasting

Abstract

The ultimate test of any forecast is whether or not it is capable of predicting future events accurately. Planners and decision makers have a wide choice of ways to forecast, ranging from purely intuitive or judgemental approaches to highly structured and complex quantitative methods. In between, there are innumerable possibilities that differ in their underlying philo- sophies, their cost, their complexity and their accuracy. Unfortunately, since information about these differences is not usually available, objective selection among forecasting methods is extremely difficult. The major purpose of this paper is to deal with one important aspect of choosing a forecasting methodology: accuracy. Section 1 will survey past research on accuracy and will look into the reasons why the reported accuracies of different studies vary, often significantly. Section 2 will report our own empirical findings on the accuracy of 111 time series and show how the results could be consistent with the conflicting conclusions of three major previous studies on the subject of accuracy. Section 3 will develop regression equations that express accuracy as a function of a number of factors related to the internal characteristics of the time series used. Section 4 will examine the reasons for variations in the relative accuracy of various methods and suggest needs for future research and Section 5 attempts a brief conclusion. The appendices describe the methods and technical details used to make the comparisons.