The campaign to restrict the advertising of booze in order to save the public could end up driving us to drink.
Despite a skeptical literature on the relationship between alcohol advertising and drinking initiation and consumption, there remain powerful public health campaigns to restrict or eliminate alcohol ads. Exhibit A: the British charity Alcohol Concern's declaration last week that alcohol should not be advertised on television before the 9pm watershed. According to Alcohol Concern's spokesman, 'Given the evidence with regard to… the influence of alcohol advertising on young people, it is appropriate that the current rules should be tightened'.
Alcohol Concern's pronouncement is the progeny of two books published a half century ago, journalist Vance Packard's million-selling The Hidden Persuaders and French demographer Sully Ledermann's Alcohol, Alcoholism, Alcoholization. These works shaped today's public health establishment consensus about the effects of alcohol advertising.
Packard asserted that advertising exerts a strong manipulative influence on consumption. Ledermann claimed that there is a fixed relationship between total alcohol consumption and the proportion of heavy drinkers; the only difference between heavy drinkers and the rest of the population being the amount of alcohol consumed. Hence, there is a causal relationship between consumption and misuse.
Between them, Packard and Ledermann provided the basis for the public health establishment's demand that alcohol advertising be either tightly regulated or completely banned. According to the public health view, increases in average alcohol consumption increase the number of problem drinkers and thus the amount of alcohol-related harm, including healthcare costs. Given that alcohol advertising both initiates new consumers and increases total consumption, it should be restricted or banned.
At the very least, this view asserts, exposure to advertising causes individuals to drink who might not otherwise drink and causes people to consume more alcohol than they otherwise would. Restricting or eliminating advertising is justifiable since it would reduce total consumption and with it aggregate alcohol harm.
Are the public health community's claims about alcohol advertising effects true, and are its demands for restrictions or complete bans on alcohol advertising, based as they are to a large degree on Packard and Ledermann, justified?
In order to test these claims, we examined the public health model of advertising's effects, experimental studies, studies of alcohol advertising exposure and recall, econometric studies of alcohol advertising, drinking initiation and consumption, and studies of alcohol advertising restrictions and bans.
The public health model is foundational to the view that advertising affects drinking choices. The empirical evidence for this model is weak and even taken on their own terms studies of alcohol advertising consistently fail to demonstrate that the drinking behaviour of an individual is the causal result of an alcohol advertisement.
Even if we were to allow that this effects model is strongly supported by the empirical evidence, which it is not, there are no studies of alcohol advertising that trace the effect of an advert from exposure to purchase behavior across a sample population in such a fashion to demonstrate that the advert actually caused the behavior. Without such a demonstration, it is impossible to conclude legitimately that alcohol advertising caused a behavior.
As for experimental studies, even allowing for the substantial issues around methodology, small sample sizes, and the appropriateness of drawing conclusions based on artificial settings, it is nonetheless clear that there is no support in the experimental literature for the claim that alcohol advertising causes initial alcohol use or increases alcohol consumption.
Some prominent regulation advocates, such as Professor Gerard Hastings, claim that newer studies provide evidence of the link between alcohol advertising and drinking behavior. Our analysis of 11 cross-sectional and longitudinal studies of advertising exposure and recall suggests otherwise.
Three problems undermine the findings of the alcohol advertising exposure and recall studies. First, and most importantly, none of the studies can justify a causal conclusion about the relationship between advertising and drinking initiation or consumption given their cross-sectional or longitudinal design.
Second, all have significant methodological issues, particularly with respect to warranting that they have in fact accurately measured alcohol advertising exposure, and also in terms of their reliance on unverified subject recall. Finally, the studies generally report data that is either not statistically significant or, if significant, is a weak relationship, or in fact contradicts their thesis.
Together, these studies present virtually no support for the claim that alcohol advertising causes young people to begin drinking. Collectively, these studies suggest that alcohol advertising either does not increase total alcohol consumption, or has an impact that is, in the case of beer advertising, so marginal as to be insignificant.
With respect to econometric studies, out of over 30 such studies over the past two decades, only a handful support the claim that alcohol advertising leads to drinking initiation or increases total consumption.
The imposition of alcohol advertising bans represents a reasonably direct way in which to test the Ledermann-derived public health hypothesis about both the effects of alcohol advertising and the corrective of advertising restrictions and bans. Although the evidence is not completely consistent and has, as we have noted, significant limitations in its ability to control for possible confounding factors, there is still very strong evidence that the imposition of bans has not reduced consumption.
Of 17 cross-sectional and longitudinal studies of the effects of advertising restrictions and bans on drinking initiation and consumption, only three find that such measures have a statistically significant effect on either initiation or consumption. There is strong evidence that restrictions have not reduced consumption and the evidence from jurisdictions that have removed bans shows that consumption has not increased when advertising has resumed.
Nor do such studies provide support for the claim that such restrictions on advertising reduce alcohol abuse or alcohol related-harms such as road fatalities or disease. Indeed, one study found that broadcast bans of spirits advertising resulted in both higher consumption levels and increased levels of traffic fatalities.
Moreover, where alcohol advertising bans have been lifted, there is no evidence that consumption has increased. This does not, of course, mean that advertising is ineffective, as many of these studies, both nationally and internationally, have demonstrated the expected advertising outcome of substitution effects and movements between brands and beverage categories.
Based on the empirical evidence, it is clear that the public health establishment's claims about the effects of alcohol advertising are incorrect. Indeed, the weight of the evidence substantially argues against its assertions about alcohol advertising initiating drinking and increasing consumption and alcohol-related harm. Consequently, there is no public policy justification for measures to restrict or completely ban alcohol advertising that is directed to legal consumers.
But what about warning labels on alcohol? Either instead of or in tandem with advertising restrictions, can they not be an effective deterrent against drinking? In short, no, they cannot.
The main evidence against the effectiveness of such warnings comes from the US. According to the proponents of warnings, they serve both to inform the public about the specific risks of drinking and reduce the drinking of specific groups most at risk, such as pregnant women, adolescents, and problem drinkers. But the research evidence suggests otherwise.
If the warning is to be effective it first has to be noticed. But US telephone surveys have found that only between 20 and 25 percent of respondents noticed the labels in the first six months after introduction, with only 16 percent recalling the content of the warning. Effective warnings, according to Kip Viscusi of Harvard University, must also provide new information. Yet an Ipsos-Reid survey in February 2005 found that 99 per cent of Canadian women of childbearing age knew that there were risks with drinking during pregnancy, suggesting that the proposed warnings would not be providing new information.
Equally unimpressive is the evidence for warnings affecting behavioral change. The US Surgeon General's warnings about the risks to pregnant women from drinking have been mandatory since 1989. A study of alcohol consumption by pregnant women reported by the US Centers for Disease Control noted that almost eight years after the implementation of the warning labels the number of women drinking during pregnancy had risen. As Dr Janet Hankin in a review of fetal alcohol prevention discovered, only the lighter drinkers who were less at risk of having children with fetal alcohol syndrome followed the warning. 'Among high-risk drinkers', Dr Hankin noted, 'the label law clearly has not affected drinking behaviour'.
A similar result has been noted with adolescents. David MacKinnon reported in the American Journal of Public Health that in a group of 16,661 high school students followed from 1989 to 1995, 'there was no beneficial change attributable to the warning in beliefs, alcohol consumption, or driving after drinking'. The World Health Organisation's 2003 study on alcohol noted that warnings failed to increase young people's perceptions of alcohol risks and had 'no direct impacts' on consumption. Studies have also found that heavy drinkers, while aware of the warnings (they see them more frequently), are more likely to consider them less believable and to discount them more than other drinkers.
This suggests that warnings may not only be ineffective, they might also be counterproductive in at least three senses. First, warnings appear to reduce at-risk drinkers' acceptance of the risks associated with their behaviour. The very act of warning actually works against itself.
The reasons for this are various. There is a natural tendency to avoid information that has negative self-implications. People are very good at avoiding processing information like warnings that they perceive as threatening. Through a process known as cognitive readjustment people tend to exempt themselves as subjects who ought to be concerned with the warning.
Then, too, there is the fact that large numbers of risk-takers display what psychologists call 'reactance', in which there is a high level of resistance to the demands of outside authority and control. For these individuals, a label represents an unreasonable attempt to shape their behaviour.
Second, as the WHO study and others have noted, warnings highlight risk and for those attracted to risk this serves to make alcohol more attractive than it might otherwise be. Finally, in pregnant women, research by Professor Ernest Abel, director of reproductive toxicology at Wayne State University, suggests that warnings might provoke stress in pregnant women that in turn may result in stress-related physiological changes that compromise the health of the fetus.
Even if the evidence of failure does not convince, perhaps the strong probability that alcohol warnings and advertising restrictions drive some people to drink might bury these pernicious regulatory instruments.