Taxonomy of Television Programs Based on Viewing Behavior

Development of classification systems of television programs has been a major interest to marketing researchers as well as networks. Past attempts at deriving classification systems utilized either viewing behavior data [4, 5] or program preference data [6]. But, these studies had not taken into account the effects of prior knowledge about television viewing as it relates to the audience characteristics or television scheduling. Frank et al. [2] developed a methodology based on regression to correct for this deficiency in previous research and employed it for developing program types with use of preference data from TvQ for a period of three years. This methodology encountered Ehrenberg's criticism of previous research [1]. This article is a replication of the study by Frank et al. using the viewing behavior data rather than preference data (which may be confounded by past behavior of individuals). The study attempts to compare the classification schemes of television programs before and after removing the effect of prior knowledge on viewing behavior. Prior knowledge was measured in this study by two sets of variables: (1) socioeconomic, demographic, and television ownership characteristics of individuals and (2) variables describing the scheduling of the television programs. The method employed for forming program types is the same as that employed in past research, namely, factor analysis.