Key points are not available for this paper at this time.
The logic underlying the Box-Jenkins transfer function analysis is developed and compared with established econometric approaches. The technique is shown to provide a safeguard against spurious effects by eliminating all time dependency in the data before assigning any effects to proposed independent variables, and by the systematic elimination of a wide variety of candidate effect models. The procedural steps in the technique are demonstrated in an application to the advertising-sales relationship, with particular focus on the advertising lag structure. The two best Box-Jenkins models are compared with previously derived advertising effect models and are shown to have greater forecasting accuracy.
Helmer et al. (Sun,) studied this question.