Does reported alcohol intake, according to population surveys, only account for half of the alcohol sold?

Question

A recent book by the House of Commons Library statistician Georgina Sturge, criticises an over reliance on data collected by self-reporting surveys in areas where there may be an incentive to minimise the reporting of "undesirable" activity.

In one example^* she claims:

The amount of alcohol that people claim to consume doesn’t match up with the amount sold. It has even been estimated that the total amount people admit to consuming is around 40 per cent to 60 per cent lower than the amount being sold. Yes, a lot probably ends up on the pub floor on a Friday night, but what this really shows is that people massively underestimate the amount they drink when answering surveys.

Some of this stems from the way in which surveys ask about drinking, which is usually in terms of `units'. Did you know that in the UK one standard-sized glass of wine (175ml) is 2.1 units? [...] But even when people are asked how much they drink and are given a list of unites per glass, they still under-report.

Do people, on average, under-report their alcohol consumption?

^* Sturge, Georgina. Bad Data: How Governments, Politicians and the Rest of Us Get Misled by Numbers (p. 53). Little, Brown Book Group. Kindle Edition.

Answer 1

Do populations (as a whole) under-report the amount of alcohol they drink on simple surveys?

Oh, yes.

This 1983 experiment with 49 men, Diary gives more accurate information about alcohol consumption than questionnaire compared drinking levels reported on a questionnaire versus a daily diary. The questionnaire only revealed 60% of the mean daily consumption compared to the diary.

Do populations (as a whole) under-report the amount of alcohol they drink on simple surveys compared to national sales figures?

Well, yes... but...

Knibbe, R. A., & Bloomfield, K. (2001). Alcohol consumption estimates in surveys in Europe: Comparability and sensitivity for gender differences. Substance Abuse, 22(1), 23–38. doi:10.1080/08897070109511443

The coverage of sales estimates by surveys varies between 39% (Germany) and 56% (France).

Citing a 1974 paper, it explained it was a:

long-known fact that survey estimates of alcohol consumption in a country are lower than sales estimates for that country

The motivation of this paper seemed to be more "Can we judge/rank the quality of a nation's research by the size of the discrepancy?" rather than judge how much each nation's inhabitants tended to underreport their own drinking. They appeared to give up on that goal:

It is concluded that the methodologic differences between studies and the differences in sales coverage do not allow cross-national comparison of survey estimates of alcohol consumption of different European countries.

I would not take the actual percentages too seriously. The paper lists a number of sources of error:

• the complex (region-based) systems used by researchers to convert between the survey answers given, and actual alcohol consumed on average, with severe limitations in the accuracy. [i.e. a respondent could be honest with the answers, but still not accurately represent their alcohol intake]
• legal and illegal import of alcohol into countries (especially, high-taxed countries like, for example, Finland)
• unregistered alcohol production (e.g. wine-making in France)

I note no mention of alcohol used in cooking, wastage (which is more substantial than Sturge suggests), and limited accounting for seasonal differences (e.g. binge drinking over Christmas and New Year might not be reflected)

Does including a clear explanation of "standard drink" affect the reporting?

Sometimes yes, sometimes no.

• Gilligan, C., Anderson, K.G., Ladd, B.O. et al. Inaccuracies in survey reporting of alcohol consumption. BMC Public Health 19, 1639 (2019). DOI: 10.1186/s12889-019-7987-3

This experiment compared including a guide explaining what a standard drink is versus leaving it out, for university students from both New South Wales, Australia and the US Pacific Northwest.

Australian students did not adjust their responses to questions with and without the standard drink term and pictorial guide. The US students were more likely to adjust their responses based on the detail of the question asked. Those US students who drank more frequently and in greater volume were less likely to adjust/apply a conversion to their consumption.

A number of possibility for the differences between the cohorts is discussed in the paper.

Can we assume that individuals consistently report around half of their drinking?

Oh, no! Absolutely not. Going back to the last paper:

This study supports previous findings of the inaccuracy of alcohol consumption volume in surveys, but also demonstrates that an assumption of underestimation cannot be applied to all individual reports of consumption. Using additional questions to better understand drink types and serving sizes is a potential approach to enable accurate calculation of underestimation in survey data.