Files
Abstract
CHAPTER 1: Consumers' response to mass-media can be difficult to assess because individuals choose for themselves the amount of media they consume and that choice may be correlated with their other consumption decisions. To address this important confound, this paper examines the introduction of television to the U.S. during which some cities gained access to television years before others. This natural experiment makes it possible to estimate the causal impact of television on the decision to start smoking, a consumer behavior with important public-health implications. Difference-in-differences analyses of television's introduction indicate that (1) television did cause people to start smoking, (2) 16-21 year-olds were particularly affected by television, and (3) the response to television occurred within a couple of years of its introduction. Estimates from this analysis indicate that television increased the share of smokers in the population by 3%-21% in cohorts exposed to television through age 25. More broadly, these results offer causal evidence that (1) mass-media can have a large influence on consumers, potentially affecting their health, (2) media exerts an especially strong influence on teens, and (3) mass media can influence consumers more than typical changes in prices.
CHAPTER 2: Many advertisements, such as national television ads, are purchased for large populations, preventing marketers from perfectly targeting all subsets of the population. Additionally, researchers increasingly have the data required to measure which populations have received too much or too little advertising. In this paper, I argue that, together, these events generate a natural experiment which can be used to obtain consistent estimates of the response to advertising. I present a formal model for exploiting this natural experiment and report the results of its application to multiple product categories. Estimates from this “coarseness” strategy are consistent with recent literature, suggesting many standard approaches to estimating the response to advertising may produce misleading results due to unobservables.