This dissertation examines spatial competition in the network television industry. The primary challenge posed by this objective is that product (i.e., show) characteristics are difficult to measure, and more suitably treated as latent characteristics to be estimated.
Chapter 1 provides an overview of the broadcast television industry and positions this research in relation to previous studies involving choice modeling or network television.
Chapter 2 presents a structural model of viewer choice with both state dependence and a components of variance structure for the unobserved, random to the econometrician, component of utility. The components of variance structure is specified to have an ideal point structure with utility from each product increasing (or decreasing) in the distance between the individual's ideal-point (or anti -ideal point) and the product's latent attribute levels. The model is estimated using maximum simulated likelihood and panel data detailing the weekday prime time viewing choices of 3286 individuals over one week.
The estimated attribute space spans four dimensions of horizontal differentiation and one vertically differentiated dimension. Interpretations of these dimensions reflect the traditional show labels. For example, one of the dimensions represents the degree of realism in a show. Furthermore, the clustering of shows based on the estimated characteristics corresponds to traditional show labels. The clusters are identified as sitcoms for mature viewers, sitcoms for younger viewers, reality based dramas, and fictional dramas.
Regarding spatial competition and strategic behavior, the model suggests the networks should use counter-programming (i.e., differentiated products) within each time slot and homogeneous programming through each night. The estimated show locations reveal an extensive use of these strategies, as well as a limited degree of branding. Nonetheless, by unilaterally changing their schedules to increase both counter-programming and homogeneity, ABC, CBS, and NBC are able to increase their weekly ratings by 16%, 12%, and 15%, respectively. In a Nash equilibrium of the static scheduling game, these gains are reduced to 15%, 6%, and 12% increases.
In Chapter 3 the impact of audience size and demographic composition on advertisement revenue is investigated. Show-level data reveals a pricing premium for large audiences, homogeneous audiences, audiences with a high proportion of 35-49 year-olds, and audiences with a low proportion of viewers 50 years old and older. Best-response schedules and Nash equilibria are computed using advertisement revenue as the networks payoff function. Interestingly, the resulting schedules (and the implied strategies) are the same as when the payoff function is simply show ratings. Costs of implementing schedule changes are also considered. Such costs are found to reduce, though not eliminate, gains to be reaped from merely altering the networks' schedules.
My Research