Does Exposure to Alcohol Advertising Cause People to Drink and Drink and Drive?

We empirically estimate whether and how exposure to alcohol advertising affects what types of alcohol youth and adults choose to drink and whether they choose to drink and drive. We use unique measures of alcohol advertising exposure using a method developed in Avery et al. (2007) and using the same rich marketing database those authors used to examine how smoking behavior varied with exposure to advertising of smoking cessation products. Those data, the Simmons National Consumer Surveys, Adults 1995-2009 (NCS-A) not only extensively measures each individual’s demographic characteristics but also their drinking behavior and detailed information on the amount of each type of alcohol (down to the brand level) an individual consumes. We develop measures of exposure to magazine and television advertising. The NCS-A asks respondents whether they read each of about 180 magazines and watch each of up to 600 television shows. (We use an NCS teen survey to track readership across a smaller set of 90 magazines). Finally, the information on alcohol consumption is very detailed. All respondents (including 18-20 year olds) report whether and how much alcohol they consumed in the past month and the number of drinks of each of several types of alcohol by brand (e.g. light domestic beer, rum, Miller Lite, Bacardi). No other large nationally representative survey combines detailed media viewing and alcohol consumption data. Our strategy to estimate the alcohol/drinking relationship parallels the strategy used in Avery et al. (2007). Briefly, we will compare differences in drinking behavior of individuals exposed to different amounts of advertising but control for program/magazine specific fixed effects and/or comparing differences in exposure within magazines/program types that plausibly attract similar people (i.e. people whose unobserved characteristics are likely to be similarly distributed).

In addition to estimating how type/brand choices vary with exposure to advertising, we use data from National Longitudinal Survey of Youth 1997 (1997-2008) (NLSY97). We use the nationally representative sample (of approximately 9,000 individuals who were 12 to 16 years old on December 31, 1996) to estimate how predicted exposure to alcohol advertising affects a richer set of behaviors that involve alcohol. In particular, we investigate whether and how exposure to advertising affects not only whether and how many times over the past 30 days that an individual drank but also whether the person drank and drove and how many times he rode with a driver who had been drinking. Because the NLSY97 does not include the rich information on media consumption, we use a two-sample IV estimation method to use characteristics in both the NCS and NLSY97 (including DMA of residence from the restricted use Geocode files) to predict alcohol advertising exposure at each age for individuals living in each place.

The results of our analysis constitute state-of-the-art estimates of the relationship between exposure to alcohol advertising and drinking behavior.