Bias due to Non‐response in a Dutch Survey on Alcohol Consumption

Abstract

Summary Increasing non‐response rates force survey researchers to focus on the danger of biased results. In this study two methods were used to determine whether a non‐response rate of 31% in a survey of drinking habits in the Netherlands seriously affects estimates of alcohol consumption. Firstly, several indices of mean consumption were corrected for differences in response probabilities of the distinguished substrata. Secondly, data from a follow‐up among non‐respondents were used to study differences in consumption between the response and non‐response stratum. Both methods led to the general conclusion that the effect of non‐response on estimates of consumption for the entire population appears to be small. Among female non‐respondents, a higher proportion of abstainers and a lower average consumption was found. Although occasional heavy alcohol use in the previous 6 months was more frequent among male non‐respondents than among male respondents, the reverse was true for frequent, weekly heavy use.