Flux and stability: individual fluctuations, regression towards the mean and collective changes in alcohol consumption

Abstract

First, to extend the predictions of the collectivity theory of drinking cultures to contexts of ordinary panel data, thereby integrating collective change and individual fluctuations in consumption. Secondly, to develop a technique for controlling the regression towards the mean effect, in order to estimate the pattern of individual change, net of this effect. The task is, first and foremost, to develop a theoretical model, guided by earlier contributions and empirical results. The applicability of the model is illustrated through re-analyses of panel data on annual intake of alcoholic beverages from the Finnish alcohol reform in 1968-69, and Norwegian panel data from 1975 to 1976. The model offers a reasonable fit to the data. It is demonstrated that a previous attempt at controlling the regression effect has overlooked a methodological complication, and therefore underestimates the real change. Analysis of the Finnish data demonstrates that the large increase in average consumption was due to approximately the same percentage-wise increase in the consumption level of drinkers at all levels, when the regression effect was controlled for. Thus, in absolute terms the increase was largest among the heavier drinkers.