The purpose of this study is first, to demonstrate how multivariate forecasting models can be effectively used to generate high performance forecasts for typical business applications. Second, this study compares the forecasts generated by a simultaneous transfer function model (STF) model and a white noise regression model with that of a univariate ARIMA model. The accuracy of these forecasting models is judged using their residual variances and forecasting errors in a post-sample period. It is found that ignoring the residual serial correlation can greatly degrade the forecasting performance of a multi-variable model, and in some situations, cause a multivariable model to perform inferior to a univariate ARIMA model. This paper also demonstrates how a forecaster can use an STF model to compute both the multi-step ahead forecasts and their variances easily.
KEY WORDS
Multi-variable models Simultaneous transfer function models ARTMA models Regression analysis Forecasting performance LTF method Over the past two decades, we have witnessed widespread use of forecasting in business and industry. In manufacturing, for instance, firms need to know in advance how much of their goods to produce in order to meet the demand of their customers. Over-estimates of demand can cause over-production leading to excessive inventories and cashflow problems. Under-production, on the other hand, can mean shortages and unsatisfied customers who may turn to other suppliers for their needs. Regardless of the particular industry in question, manufacturing, banking or trucking, accurate forecasting has real value in its ability to reduce costs and maximize profits. In short, the importance of accurate forecasting in industry cannot be over emphasized.
Traditionally, quantitative forecasting methods are often based on a single time series. A wide variety of methods have been employed including moving averages, exponential smoothing, decomposition of a time series, and Box-Jenkins ARIMA models (Box and Jenkins, 1970). These methods are essentially used to extrapolate the past history of data into future projections. Recently, a number of multi-variable methods have been discussed, including the use of vector ARMA