2. Lessons for the future of alcohol surveys

Representativeness as derived from statistical properties of sampling would require a) some kind of probability sample, b) that every person in the population has a probability to be drawn, and c) that non-response is not systematic (i.e., missing at random). These requirements can no longer be matched in modern alcohol surveys. DEEP SEAS is just one example of the failure to achieve results which are meaningful for describing countries.

Narrative review of current surveys and their limitations, and potential ways forward.

The literature will be discussed along the following lines:

1) Problems of underreporting and ways forward (e.g., new STATA module to correct for misclassification; triangulation of sources – alcohol per capita consumption as gold standard)

2) Problems of missing populations (e.g., military personnel, homeless, institutionalized, aged) in the sampling frame and ways forward (e.g., modelling populations from multiple surveys and sources of information)

3) Minimal requirements for each survey (e.g., quotas for key characteristics which apply jointly)

The results of the review will be presented, and will be discussed by two alcohol survey specialists, before involving all participants.