Academic journal article Contemporary Economic Policy

Comment-Alcohol Advertising in Magazines and Youth Readership: Are Youths Disproportionately Exposed?

Academic journal article Contemporary Economic Policy

Comment-Alcohol Advertising in Magazines and Youth Readership: Are Youths Disproportionately Exposed?

Article excerpt


In the July 2006 issue of this journal, Nelson (2006) analyzes data on alcohol advertising in magazines and youth readership and concludes that alcohol advertisements in these magazines are not targeted at underage youths. Nelson's article seeks to improve a similar analysis by Garfield, Chung, and Rathouz (2003), which concludes that the volume of beer and liquor advertisements in magazines is significantly related to the number of youth readers. In an effort to refine Garfield et al.'s methods, Nelson uses an improved data set and econometric methods to address the multicollinearity in the Garfield et al. model. Garfield et al. include the numbers of youth readers, young adult readers, and total readers in their analysis. These variables are highly correlated. Nelson respecifies the model using percentage of youth readers as the main explanatory variable. He finds that this variable plays no role in explaining advertising placements in magazines either overall or for beer and spirits individually. Nelson has presented a similar analysis more recently that uses more magazines and a longer time series of data and focuses specifically on spirits advertising (Nelson, 2007); this analysis also fails to find any significant effect of youth readership.

Addressing the multicollinearity problem in Garfield et al. by using the percentage of youth readers as the main explanatory variable is an important contribution of Nelson's article. His specification poses the analytical question as whether or not advertising counts are predicted by the proportion of a magazine's readership who are underage youths (which he defines as youths aged 19 yr and younger), after controlling for other predictors of advertising placement, including the cost of advertisement placement per reader reached, number of readers per copy, and the demographic characteristics of magazine readers.

There remains a problem with Nelson's approach, however. His analysis introduces two new variables--median age of adult magazine readers and magazine category--that are highly correlated with percentage of youth readership. This results in multicollinearity that threatens the validity of the article's conclusions.

There is a strong correlation between the proportion of youth readers and the median age of adult readers of a magazine in Nelson's data set (r = -.75, p < .0001). The magazine with the highest median age (46.4) has the lowest proportion of youth readers (only 6.9%), while the magazine with the lowest median age (23.7) has the highest proportion of youth readers (45.9%). Thus, median age serves as a proxy variable for percentage of youth readership. If there were a true effect of youth readership on advertising placement, inclusion of median age in the model might be expected to result in the coefficient for median age picking up the effect of youth readership. This multicollinearity makes it impossible to exclude the possibility that youth readership is related to advertising when median age is present in the model.

The magazine specialty categories (automobiles, black, men's style and sports, women's style, and entertainment and music) that Nelson uses are also highly correlated with youth readership. These specialty magazines have on average 27.1% youth readers compared with 17.3% for general magazines. The relationship between magazine categories and youth readership is even stronger for several of the magazine types. Music magazines, for example, have an average youth readership of 30.4%, almost twice the youth readership of the general magazines. Including this variable in the model along with percentage of youth readership is therefore likely to obscure any true relationship between youth readership and advertising since the magazine category coefficients may well pick up any effect of youth readership.

Since the median age of adult readers and the magazine category are so highly correlated with the percentage of youth readers and may act as a proxy for youth readership in the models, allowing any true effect of youth readership to be taken up by the coefficients for these variables, we wonder whether the multicollinearity problem introduced into Nelson's refined model creates precisely the problem that his respecified model attempts to solve. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.