Journal of Marketing Research (JMR)

The description of the nature of the current and cumulative effects of advertising has long been recognized by both managers and public policymakers as an important but difficult problem in marketing. Theory-based normative models are available (Sethi 1979), but must assume some model of the dynamic process of carryover effects. A time series analysis was conducted for six cereal brands. The analysis proceeded on the assumption that feedback was potentially present, but the relationship was found to be extremely weak. No lag structures are evident. In fact, the hypothesis test of no causal relationship cannot be rejected except in the advertising-influencing-sales direction for Corn Flakes and in the sales-influencing-advertising direction for Rice Krispies. The danger of applying a misspecified structural model is shown by estimating the Koyck distributed lag model with the data base. Our conclusion is that in situations where lag structure is important, the a priori assumption of a particular structure--e.g., geometric--in a structural model analysis can produce grossly misleading results. A superior procedure may be to conduct a time series analysis first, to obtain some insight into lag structure, and then to test a reasonable, theory-based structural model.