Essays On Health Economics
The essays of this dissertation study the effect of alcohol advertising on individual drinking, alcohol firm advertising decisions, and the relationship between education and mortality. The first essay focuses on the possible effects of alcohol advertising on youth drinking. Researchers still disagree about how advertising affects alcohol consumption. This disagreement largely arises because alcohol firms target marketing at people who already drink. Drinkers prefer particular media; firms recognize this and target alcohol advertising at these media. Endogenous targeting of alcohol advertisements presents a challenge for empirically identifying a causal effect of advertising on drinking. In this chapter, I overcome these challenges by leveraging a plausibly exogenous source of variation in advertising exposure, and by utilizing novel data with detailed individual measures of media viewing and alcohol consumption. I adopt three approaches to control for endogeneity bias due to targeting. First, I use average audience characteristics of the media an individual views to capture targeting. Second, I use media fixed effects to directly control for media choice. Third, I exploit variation in advertising exposure due to a 2003 change in an industry-wide rule that governs where firms may advertise. I use the rule change as an instrument for exposure to alcohol advertising. Though the unconditional correlation between advertising and drinking is strong, this relationship is not robust to more rigorous controls for targeting and to the use of an instrumental variables estimator. The results suggest that any effect of alcohol advertising on youth drinking is modest. The second essay studies the effects of the end of the liquor broadcast advertising ban on firm behavior. I study which firms and brands first took advantage of this new medium I study which spirits brands take advantage of the newly available medium of television. I compare the consumer characteristics and market competition of brands that transition to television advertising to those that do not, using two different definitions of television advertising adoption. I model brand-level, yearly television advertising spending and estimate hazard models of the transition to the use of television advertising. I find evidence that competitive pressure correlates with a brand's adoption of the "new" medium. Firms that are dominant in their market are much more likely to adopt television advertising when their competitors possess a larger share of the market. However, I find little evidence that the demographic characteristics (age, gender, race, income, education, magazine reading, and television viewing) and alcohol consumption of a brand's consumers are related to the adoption of television advertising. The results suggest that television advertising in the spirits market may play larger role dividing market shares than growing market size. The third essay revisits the question of whether people live longer if they get more education or if people who get more education have unobservable traits and habits that cause them to live longer. Like previous studies, we use compulsory schooling laws as instruments for education, However, we use better instruments and Panel Study of Income Dynamics data that include each respondent's date and cause of death. We find our compulsory schooling instruments are stronger predictors of education than those used in previous studies. However, relying on within-state variation greatly reduces the predictive power of our instruments, which only weakly predict educational attainment. We model three different measures of mortality: probit models of mortality over 5- and 10-year age spans and continuous-time survival models of the number of months a person lives past forty years of age. We confirm a strong statistical association between education and mortality in all three model types. However, due to the weakness of our instruments, our results are imprecise and provide little useful insight into whether education reduces mortality. We show the relationship between schooling and mortality is strongest for post-secondary education, though there exists little evidence in the literature concerning whether this link is causal.
Alcohol; Advertising; Drinking; Education; Health; Mortality
Kenkel, Donald S.
Mathios, Alan D; Lovenheim, Michael F
Ph. D., Economics
Doctor of Philosophy
dissertation or thesis