To find out how much alcohol people drink we need to ask them, but precisely how we ask may have important implications on the answers we get.

New research suggests we need to be asking about more than just typical drinking habits, if we want to get an accurate picture of how much alcohol the UK is really consuming. Drinkaware looks at how alcohol consumption is currently measured, the problems with these measures, and potential for more accurate measurement in the future.

How is average alcohol consumption calculated?

The most common way of calculating alcohol consumption in national surveys is the quantity-frequency method. This method involves asking people how often over the last year they have drunk each of a number of different types of drink, and the usual amount drunk on any one day.(1) The 'quantity-frequency' method was used by the Health Survey for England from 1991 to 2002, and reinstated again in 2011.(2)

Another technique used in some surveys is the graduated frequency method. Using this technique people are asked to state the number of times they drank at different levels of drinking, for example how many times they had more than 12 drinks, down to how many times they had 3 to 4 drinks, and 1 to 2 drinks.(3) A version of this 'graduated frequency' method was used in the Alcohol Research UK report exploring the alcohol harm paradox.(4) (please visit our Databank for more information).

But the figures don’t add up

Using these survey measures we can estimate the amount of alcohol that people in the UK drink. However, when this is compared with HM Revenue and Customs information on alcohol taxed for sale, a big difference is found. Survey estimates of alcohol consumption have been found to represent just 55 to 60% of the alcohol sales figures.(5) This leaves a huge 40 to 45% of alcohol sales unaccounted for. 

How do we explain this difference?

A number of possible reasons have been suggested for the disparity between reported alcohol consumption and alcohol sales figures.

One factor which has been suggested to contribute to this issue is underestimation of the amount poured into a glass at home. A 2013 study put this to the test, asking people to pour their usual glass of wine or spirit, and then to estimate how many units their glass contained. The authors concluded that underestimation of drinks poured at home did not offer any substantial contribution to explaining the difference between self-reported alcohol consumption and alcohol sales figures.(6)

A further explanation suggested that when people are asked to recall their drinking habits in surveys, they may not mention the drinking they do on special occasions, such as New Year or holidays. Forgetting about special occasions, where drinking is likely to be heavier, is a recognised weakness of the quantity-frequency method.(7)

A new way of calculating average alcohol consumption?

A study published in 2015 by Bellis and colleagues set to unpick the issue of special occasion drinking, asking people about their usual drinking as well as drinking they might only do on special occasions. People were asked whether there were days where they drank more, or less, than usual, for example heavier drinking every Friday. Additionally, people were asked about drinking on special occasions e.g. holidays, or weddings. The authors found that by including questions about drinking on special occasions their estimation of alcohol consumption could account for 78.5% of alcohol sales. Without including questions on special occasion drinking their estimation could only account for 62.7% of alcohol sales. Accounting for special occasion drinking was found to add over 120 million units of alcohol per week to population consumption.(8)

The way forward?

The evidence from Bellis and colleagues suggests the importance of incorporating more questions about special occasion drinking into national surveys of alcohol consumption. This is likely to provide a more accurate reflection of alcohol consumption, and a figure which is closer to that of alcohol sales.

However, consistent measures of alcohol consumption are important for understanding trends in the data (we have further information in our “Alcohol related hospital admissions up by a third article”). Whilst additional measures of special occasion drinking provide important and valuable insights into alcohol consumption, they should be an addition, and not a replacement, for current drinking measures.