The alcohol advertising bluff

Activists who want to ban booze ads are abusing politicians' trust

The Scottish government has been consulting on whether to ban alcohol advertising (the consultation closes on Wednesday). And not just advertising. The government has proposed banning alcohol merchandise, branded glasses, sponsorship and displays of alcohol in shops (e.g. on the shelves!).

Quite a lot of the consultation document is worryingly detached from reality and seems to have been written by people who do not drink. For example, take the assertion that different brands of whisky and wine are all basically the same.

It repeatedly cites evidence of an association between exposure to alcohol advertising and the propensity to drink alcohol without considering the obvious explanation that people who consume a product are more likely to remember seeing advertisements for it.

Even when talking about people who own alcohol-branded merchandise, it does not occur to the authors of the document that causality might run in the opposite direction to what they assume.

I found two things particularly irksome about the consultation. Firstly, it relies heavily on the opinions of members of the ‘Children’s Parliament’ who are described as ‘investigators’.

I have written about the Sturgeon Youth before. In short, a bunch of 9 to 11 year olds are put in a room with the state-funded temperance group Alcohol Focus Scotland and indoctrinated with anti-alcohol propaganda which they then regurgitate on camera. It is deeply creepy. Votes at 16 is one thing. Policy-making at 10 is another.

Secondly, the consultation document relies heavily on a report from Alcohol Focus Scotland who are a third party but hardly a disinterested one. Whenever the Scottish government is unable to find an academic citation for its claims, which is often, it cites the AFS report.

The AFS’s approach is to simply assert that alcohol advertising makes people drink more alcohol and that banning it would save many lives. According to them, the case is closed. For example, they say…

‘Research has now established a causal connection between children and young people’s exposure to alcohol marketing and drinking’

‘There is a wealth of evidence that exposure to alcohol marketing is causally linked to consumption.’

‘There is conclusive evidence of a small but consistent association of advertising with consumption at a population level.’

‘Significantly, research published since the Network’s first report has now established a causal connection between children and young people’s exposure to alcohol marketing and drinking.’

‘The evidence is clear that exposure to alcohol marketing is a cause of youth drinking. This is the conclusion reached by researchers applying the same methodology that established the causal link between tobacco and cancer.’

These are big claims. They are not only saying that there is strong evidence linking advertising to aggregate - as opposed to brand - consumption, but that the link has been shown to be causal.

Demonstrating causality is hard enough in real science. The idea that causality can be proven based on studies in which people are asked them how much they drink and how many adverts they watch is absurd. It is simply a bluff aimed at scientifically naive policy-makers.

The causality claim comes from this article by two proponents of alcohol advertising bans, James Sargent and Thomas Babor, who claim that the association between alcohol advertising and underage drinking meet Austin Bradford Hill’s criteria for causality.

¹

Hill is best known for his studies conducted with Richard Doll which identified cigarette smoking as a major cause of lung cancer. The association between smoking and lung cancer was initially controversial, with critics claiming that they had only found a statistical correlation, but the correlation was very strong and once all other explanations had been eliminated the association came to be accepted as causal.

In 1965, Hill set out nine criteria that should be assessed ‘before we cry causation’. They are not iron laws but they contain enough common sense for them to have stood the test of time. The evidence for alcohol advertising causing underage or heavy drinking does not meet a single one of them and Sargent and Babor have to jump through all sorts of rhetorical hoops to pretend it does.

Most obviously, the evidence is extremely inconsistent and the associations are very weak. I have included my response to the consultation summarising the evidence at the end of this post, but let us briefly look at Hill’s criteria and see how evidence from the social sciences on booze advertising stand up.

Strength of the association

Hill suggested that the nine- or ten-fold increase in lung cancer risk among smokers was large enough to imply causation whereas a doubling of risk would have been small enough to allow room for doubt. Associations between alcohol advertising and youth drinking, when they exist, rarely achieve a relative risk as high as 2.0 and never come close to 10.0. Even Sargent and Babor admit that such positive associations as exist are ‘modest’. Of the seven longitudinal studies looking at youth drinking initiation in a 2017 meta-analysis, the relative risk associated with alcohol advertising ranged from 1.00 to 1.69 (i.e. between zero and 69 per cent). Only one was statistically significant.

Consistency

As you will see below, the association between alcohol advertising and drinking at any age is anything but consistent. Even Sargent and Babor acknowledge that the evidence is ‘mixed’.

We really need go no further down Hill’s list. The claim of causality has fallen down at the first two hurdles, but let us proceed nonetheless.

Specificity

There are many factors which influence the propensity of young people to drink. Insofar as advertising is one of them, no one would claim that it is necessary or sufficient.

Temporality

For Hill, it was important to establish that the disease came after exposure to the risk factor, rather than the other way round. Substituting ‘drinking’ for ‘disease’, it is often difficult to establish this in some of the advertising studies. Reverse causation is clearly possible when studies rely on individuals recalling how much advertising they have seen and very likely in the case of owning alcohol-related merchandise. Longitudinal studies are more reliable in this respect but they cannot overcome the challenge of establishing whether the propensity to drink alcohol precedes the exposure to the advertising, nor whether alcohol companies target their advertising to people who are most likely to drink (which, of course, they do, but they wouldn’t if their aim was to get people to start drinking).

Biological gradient

In epidemiology it is generally expected that there will be a dose-response relationship, with greater exposure to the risk factor leading to greater risk. Several studies have found an association between alcohol advertising and alcohol consumption that rises as exposure to advertising increases. However, these studies suffer from the usual problem of recall bias which may also be on a gradient, with heavier drinkers more attentive to alcohol advertising. Other studies which do not rely on self-reporting have found a conspicuous lack of a dose-response relationship.

At the national level, an increase in alcohol advertising does not tend to lead to more alcohol being sold. In the USA, alcohol advertising expenditure rose by nearly 400 per cent in real terms between 1971 and 2012, and yet there was little change in per capita alcohol consumption. In Britain, alcohol advertising spend fell by 10.8 per cent in real terms between 1991 and 2001 but alcohol consumption rose by 15.8 per cent. Nearly all the rise in consumption was due to increased sales of wine, which tends to be less advertised than beer and spirits.

Plausibility

The broader economics literature indicates that advertising generally only affects brand share in mature markets and has no impact on aggregate consumption. Since alcohol is a mature market, it is implausible that it would be an exception to this rule. As Ambler notes:

‘The understanding that total advertising does not affect total market size is not particular to alcohol; it is normal for other mature categories’

For some public health and temperance campaigners, however, it seems highly plausible that the promotion of a product will lead to greater consumption of it. They often say of alcohol advertising: ‘If it doesn’t work, why would the industry spend so much money on it?’. Their mistake is to see the alcohol industry as a monolithic entity rather than a group of rival businesses. The alcohol industry does not advertise. Alcohol companies advertise, and it is worth their while to spend money attracting other companies’ customers and keeping hold of their own. The plausibility of the link between alcohol advertising and drinking may come down to subjective opinion, but it is fallacious to argue that the mere existence of advertising proves the association.

Coherence

Hill noted that the epidemiological evidence on smoking and lung cancer emerged against a backdrop of rising lung cancer cases in countries where cigarette smoking had become popular. It was a known problem looking for an explanation. As mentioned above, there is no such temporal relationship between the amount of alcohol advertising and the amount of alcohol consumed in mature markets.

Experiment

Randomised experiments have produced conflicting and unimpressive results, with most studies being consistent with the null hypothesis (see below).

Analogy

Alcohol advertising is analogous to the advertising of other well established products in mature markets. Evidence from the broader economics literature has been summarised in the New Palgrave Dictionary of Economics as follows:

‘At the aggregate level, advertising tends to lag cyclical changes in total consumption slightly, not to lead those changes… Overall, advertising does not emerge from the empirical literature on consumer demand as an important determinant of consumer behaviour.’

This is not the first time that Austin Bradford Hill’s good name has been abused by neo-temperance academics. In 2017, a study was published claiming that the evidence for minimum pricing - a policy that had never been tried - met his criteria for causality. That was absurd at the time since the ‘evidence’ largely came from theoretical models. It looks even more absurd now that we have seen minimum pricing crash and burn in Scotland.

There is something mildly offensive about using the reputation of a legendary scientist to push your quack policies, rather like when insane cranks compare themselves to Galileo. The reality is that the evidence for alcohol advertising bans is about as weak and inconsistent as it gets. Comparing it to the evidence on smoking and lung cancer is ridiculous and these people should be called out for deliberately trying to mislead politicians.

My response to the consultation…

There is no strong case for any further restrictions on alcohol advertising in Scotland. The evidence presented in the consultation document is biased and partial, relying heavily on a report from the anti-alcohol pressure group Alcohol Focus Scotland and the easily manipulated opinions of children in the ‘Children’s Parliament’ (who are referred to, rather creepily, as ‘investigators’).

The consultation document essentially ignores the benefits of advertising to consumers and business, and displays ignorance of the alcohol market when it says:

'Without branding and other marketing strategies, alcohol products in each beverage sub-sector are essentially variations of the same thing.’

This statement is either trivial or wrong. Human beings (and politicians) are also essentially variations of the same thing, but the differences are extremely important. The implication of this statement is that the difference between a blended whisky and a single malt whisky is unimportant and that the differences between individual brands of whisky are little more than a product of marketing. As is often the case in the consultation document, this strange statement cites nothing more than a report from Alcohol Focus Scotland as evidence.

The consultation document states that:

‘The strongest academic evidence underpins the impact alcohol marketing has on children and young people and is set out through the consultation.’

Four references are given to support his claim. The same four studies are cited as evidence for the Scottish Government’s claim that:

‘A number of systematic reviews of these studies assert that there is a strong relationship between children and young people seeing or interacting with alcohol marketing and then starting to drink alcohol, or if they already drink alcohol, drink more.’

These four publications are:

Anderson et al. (2009) which looked at 13 longitudinal studies.

Smith and Foxcroft (2009) which looked at 7 longitudinal studies, all of which were included in Anderson et al. (2009).

The Scientific Opinion of the Science Group of the European Alcohol and Health Forum (2009) - a review which, due to ‘the voluntary nature of the work, and the limited time frame to prepare the opinion’, simply re-used Anderson et al. (2009) and Smith and Foxcroft (2009).

In other words, there is nothing in two of the cited studies that cannot be found in Anderson et al. (2009). The same material is rehashed.

The fourth citation is Jernigan et al. (2016) which reviews 12 longitudinal studies published since 2008, none of which were included in Anderson et al. (2009).

The problem with the evidence cited here and elsewhere in the consultation document is two-fold. Firstly, it is dominated by observational studies, which are the weakest type of evidence in this field. Secondly, it does not address the question of what happens when alcohol advertising is restricted or banned.

Longitudinal studies (otherwise known as cohort studies)

Longitudinal studies usually find a modest association between alcohol advertising and alcohol consumption, but they are plagued by confounding factors and their weaknesses are well recognised by academics. There are major endogeneity issues which are exacerbated by various potential biases, including selection bias and recall bias. Neither the amount of advertising seen nor the amount of alcohol consumed is measured objectively. Both are self-reported by the people being studied and it is easy for participants to guess the purpose of the study. This can lead to problems with demand effects, as Aspara and Tikkanen (2013: 13-14) note:

‘For instance, if teenagers assume it is socially “cool” to remember many ads or to drink beer on weekends, they may give high scores to both questions (giving rise to a correlation), regardless of how they actually behave. Another common version of this bias is that participants may try to guess the purpose or the research question of the study and then deliberately give answers that “confirm” the question by the researchers. (E.g., if teenagers guess that researchers are studying the relationship between advertising and alcohol consumption, they may deliberately say that they have seen numerous ads and consumed a lot of alcohol, regardless of the actual number of ads seen and amount of alcohol consumed).’

Even if respondents answer as honestly as possible, their recollection of how much advertising they have seen is susceptible to the well recognised problem of recall bias. People who consume a product are more likely to notice and remember advertisements for it, and heavy consumers may be especially likely to pay attention to them. Longitudinal studies can partially address this by estimating their exposure to alcohol advertising at first and then following them up months or years later, but it is very difficult to control for a young person’s propensity to drink, as well as other inherent differences between groups. For example, someone who sees little or no television advertising for alcohol either watches very little television or watches very different programmes from someone who sees a lot of alcohol advertising. It is not difficult to imagine such a person having different tastes and character traits that influence their attitude to alcohol. Some studies of young people have found an association between susceptibility to drinking and positive reactions to alcohol advertising. The authors of such studies typically infer that the advertising creates the susceptibility whereas it seems more plausible to assume that young people who not interested in drinking pay less attention to alcohol advertisements and find them less appealing when they see them.

The targeting of advertising introduces a further problem which academics have only recently made serious efforts to address. Companies target their advertising at people who are most likely to buy their product. It should therefore not be surprising if drinkers are more likely to recall seeing more alcohol advertising than non-drinkers. Lillard et al. (2018: 890) found that ‘firms concentrate alcohol advertising in particular magazines and on particular programs consumed by men, especially young men’ and noted that:

‘If alcohol advertising could actually induce people to drink (who were otherwise not so inclined), one would expect a profit-maximising firm to advertise in a wider variety of media read by different consumers than the ones who consume the media firms currently use.’ (ibid.)

Having acknowledged the problem that drinkers ‘may systematically select different television programs and magazines than non-drinkers’, Molloy (2016: 150) attempted to control for targeting and found that ‘alcohol advertising is, at most, a small influence on youth drinking’ (ibid.: 162).

This element of reverse causation is most obvious in the case of alcohol-branded merchandise. The consultation document says:

‘Research demonstrates that actively engaging with alcohol marketing, like owning alcohol-branded merchandise, has a stronger association with alcohol consumption, than seeing an alcohol advert does.’

It would be remarkable if this were not the case. People who own football merchandise are more likely to go to football games and people who ‘actively engage’ with computer game marketing are more likely to play computer games. Alcohol-branded merchandise is not randomly distributed across the population. Those who like drinking alcohol are more likely to seek out, purchase, and (if given away for free) keep merchandise related to alcohol.

Longitudinal studies therefore have inherent limitations which make them likely to find a correlation between alcohol marketing and alcohol consumption. This correlation is then portrayed as being causal by anti-alcohol activists and academics. 

Alcohol advertising expenditure

Economists have tended to examine the question by looking at alcohol advertising expenditure and per capita alcohol consumption. Both measures are more reliable and are not susceptible to recall bias. Studies using this methodology have consistently found no association. There are too many of them to discuss individually, but the following are typical:

In his study of spirits advertising in the USA between 1976 and 1989, Mark Gius (1996: 75) concluded that ‘brand-level spirits advertising results only in brand switching and does not increase the size of the spirits market’.

A study of the US beer market found that sales were affected by price but not by advertising (Lee and Tremblay 1992).

A study using data from Ontario, Canada concluded that ‘advertising is not effective in enlarging markets and this suggests that firms (especially breweries) use advertising to compete in zero-sum market share games’ (Larivière et al. 2000: 147).

Most of this evidence comes from North America, but Duffy (2001) found the same null effect in the UK:

‘Advertising is found to have had no significant effect upon the “product composition” or “level” of total alcoholic drink consumption in the UK over the period from 1964-1996, and this result is robust with respect to variations in the specification of functional form. The consumption of alcoholic drink is affected by relative prices, total consumer budgeted expenditures and, to some extent, by autonomous shifts in tastes. These results imply that manipulation of the aggregate level of alcohol consumption (and its distribution between beer, spirits and wine) is not an easy matter for policy makers to achieve.’ (ibid.: 454)

Randomised Controlled Trials 

Further evidence comes from a number of randomised controlled trials (RCTs) that have been conducted to see whether alcohol advertising makes people drink more in the short term. Although it would be unethical to use children in these experiments, almost all of them have used college students in a quasi-naturalistic setting. Their strength is the randomisation of participants and their weakness is that some participants guess the true purpose of the experiment. They also tend to be small-scale (and therefore have low statistical power) and typically give the alcohol away for free, which may bias them towards finding an association in the experiment which would not happen in real life.

This evidence is mixed. Of nine relevant RCTs, five found no effect on consumption, one found that advertising and consumption were positively associated, two found mixed results, and one found an effect on heavier drinkers but not on moderate drinkers. On balance, the evidence leans towards alcohol advertising not being a significant driver of impulsive drinking.

Impact of advertising bans

Finally, there are studies looking at what happens when alcohol advertising is banned or restricted. This literature is largely overlooked in the consultation document despite it being most directly relevant to the policies being proposed. Again, the evidence is mixed. Three studies from Canada used an interrupted time-series and found little evidence of changes in alcohol consumption when bans were introduced or lifted.

The first looked at a short-lived ban on alcohol advertising in British Columbia introduced in 1971 and found ‘little support for the view that the B.C. advertising ban reduced alcohol consumption’ (Smart and Cutler 1976: 20).

The second looked at restrictions on beer advertising in Manitoba and concluded that there was ‘little evidence that per capita beer consumption has changed in any way since beer advertising ceased to feature in Manitoba media’ (Ogborne and Smart: 294).

The third looked at the lifting of a 58 year ban on alcohol advertising in Saskatchewan in 1983. It found evidence that consumers shifted from spirits to beer but that there was ‘no impact on wine and total alcohol sales from the introduction of alcohol advertising’ (Makowsky and Whitehead 1991: 555):

‘This evaluation suggests that alcohol advertising is not a contributory factor that influences the overall level  of alcohol consumption’ (ibid.)

Calfee and Scheraga (1994) found that Sweden’s ban, which began in 1979, had no impact on the sale of alcohol. By contrast, Rossow (2021) looked at Norway’s ban on alcohol advertising, introduced in 1975, and estimated that it had reduced alcohol consumption by 7.4 per cent.

Several studies looking at OECD countries have produced conflicting results, with Henry Saffer and Dhaval Dave generally finding an effect from alcohol advertising bans while their fellow economists Jon Nelson and D. J. Young generally find no effect.

Overall, it is difficult to argue with the conclusion of a 2014 Cochrane Review titled ‘Does banning or restricting advertising for alcohol result in less drinking of alcohol?’ which concluded:

‘There is currently no robust evidence for or against recommending the implementation of alcohol advertising restrictions.’ (Siegfried et al. 2013: 2)

References

Alcohol Focus Scotland (2022) Realising Our Rights: How to protect people from alcohol marketing. June.

Anderson, P., de Bruijn, A., Angus, K., Gordon, R. and Hastings. G. (2009) Impact of alcohol advertising and media exposure on adolescent alcohol use: a systematic review of longitudinal studies. Alcohol & Alcoholism 44: 229-243.

Aspara, J. and Tikkanen, H. (2013) A Methodological Critique of Alcohol and Addiction Researchers’ Studies on the Effect of Advertising on Adolescent Alcohol Consumption. SSRN http://dx.doi.org/10.2139/ssrn.2205112

Calfee, J. E. and Scheraga, C. (1995) The Influence of Advertising on Alcohol Consumption: A Literature Review and An Econometric Analysis of Four European Nations. International Journal of Advertising 13(4): 287-310.

Duffy, M. (2001) Advertising in consumer allocation models: choice of functional form. Applied Economics 33: 437-56.

Gius, M. (1996) Using panel data to determine the effect of advertising on brand-level distilled spirits sales. Journal of Studies on Alcohol 57(1): 73-6.

Jernigan, D., Noel, J., Landon, J., Thornton, N. and Lobstein, T. (2016) Alcohol marketing and youth alcohol consumption: a systematic review of longitudinal studies published since 2008. Addiction DOI: 10.11111/add.13591.

Larivière, E., Larue, B. and Chalfant, J. (2000) Modeling the demand for alcoholic beverages and advertising specifications. Agricultural Economics 22: 147-62.

Lee, B. and Tremblay, V. J. (1992) Advertising and the US market demand for beer Applied Economics 24: 69-76.

Lillard, D., Molloy, E. and Zan, H. (2018) Television and Magazine Alcohol Advertising: Exposure and Trends by Sex and Age. Journal of Studies on Alcohol and Drugs 79(6): 881-92.

Makowsky, C. and Whitehead, P. (1991) Advertising and alcohol sales: a legal impact study. Journal of Studies on Alcohol 52(6): 555-67.

Molloy, E. (2016) This ad is for you: Targeting and the effect of alcohol advertising on youth drinking. Health Economics 25: 148-64.

Ogborne, A. and Smart, R. (1980) Will Restrictions on Alcohol Advertising Reduce Alcohol Consumption? British Journal of Addiction 75: 293-6.

Rossow, I. (2021) The alcohol advertising ban in Norway: Effects on recorded alcohol sales. Drug and Alcohol Review 40: 1392-95.

Scientific Opinion of the Science Group of the European Alcohol and Health Forum (2009) Does Marketing Communication Impact on the Volume and Patterns of Consumption of Alcoholic Beverages, Especially by Young People? - A review of longitudinal studies.

Siegfried, N., Pienaar, D., Ataguba, J., Volmink, J., Kredo, T., Jere, M. and Parry, C. (2014) Restricting or banning alcohol advertising to reduce alcohol consumption in adults and adolescents. Cochrane Database of Systematic Reviews Issue 11. Art. No.: CD010704. DOI: 10.1002/14651858.CD010704.pub2.

Smart, R. and Cutler, R. (1976) The Alcohol Advertising Ban in British Columbia: Problems and Effects on Beverage Consumption. British Journal of Addiction 71: 13-21.

Smith, L. A., Foxcroft, D. R. (2009) The effect of alcohol advertising, marketing and portrayal on drinking behaviour in young people: systematic review of prospective cohort studies. BMC Public Health 9(51).

1. The study is titled ‘The Relationship Between Exposure to Alcohol Marketing and Underage Drinking Is Causal’. Search engine friendly titles containing a blanket assertion are usually a red flag for activist-driven research.