Seasonal variation in survey and sales estimates of alcohol consumption.

Time variation of drinking is substantial and has an effect on aggregate estimates of consumption. In this article it is shown that because of a considerable seasonal variation in consumption (+/- 20%) a serious bias in annual consumption estimates can be expected in surveys with a limited time frame. The present study analyzes drinking data collected in the general population of the Netherlands from March 1985 through December 1985 (including Christmas and New Year's Eve). Since it was expected that sensitivity to temporal fluctuations might not be equal for different methods of measurement, several indices of consumption were compared. Although the assessed seasonal effect varies indeed across types of measurement, across male and female subsample and across types of alcoholic beverage, the general tendency is for consumption to be highest in the spring season and lowest in the autumn. Sales figures fluctuate accordingly. It is evident that the risk of biased estimates is larger the shorter the time frame of the survey. Seasonal variation was highest in the frequency domain. Furthermore, exclusion from the time frame of collective holidays, during which people drink more often and more per occasion (viz., Christmas), increases the risk of biased estimates. Even estimates of abstention, but also regular heavy drinking among women, appear to vary considerably over the three seasons in this study. The main conclusion is that results of comparisons of survey data on drinking, particularly those over time, are more or less invalid if the respective time frames of the surveys do not correspond.