NONSTATIONARY CONDITIONAL TREND ANALYSIS - AN APPLICATION TO SCANNER PANEL-DATA

Planning and evaluating consumer promotions is facilitated by knowledg e of the types of consumers who contribute to incremental sales. In pa rticular, interest may focus on identifying the contributions of buyer s segmented on the basis of their prior purchase history. When the dis tribution of the number of purchase occasions in a base period can be described by the negative binomial distribution (NBD), conditional tre nd analysis (CTA) is a simple and effective approach for identifying t he sources of incremental sales during a test (promotional) period. As currently implemented, CTA assumes a stationary marketing environment . The authors propose an extension of CTA that explicitly incorporates varying marketing activities. They also show that the often observed underprediction of purchases in the test period by nonbuyers in the ba se period is a consequence of the skewness of the NBD and is not neces sarily due to model misspecification. An illustration with scanner pan el data is provided.