Applying Multivariate Time Series Model on Sales Forecasting of Technological Products

Abstract

Sales forecasting plays a crucial role in conducting marketing and mix strategies in technological industries. However, traditional sales forecasting methods focus only on customer behaviour and other quantitative variables. This paper proposes multivariate time series models, using the vector autoregression (VAR) model and the Litterman Bayesian vector autoregression (LBVAR) model, for sales forecasting in technological industries. In this study, macroeconomic data are considered to be useful leading indicators and are included in the VAR and LBVAR models. The LBVAR model possesses superior Bayesian statistics in small sample forecasting and holds the VAR model dynamic properties. An empirical study of Taiwan`s portable computer industry is used to examine the VAR and LBVAR models to validate the informative effect of macroeconomic data on sales forecasting. As a result, multivariate time series models with macroeconomic data appear to be useful models for technological product sales forecasting.