Essays on Societal Cost of Alcohol and Related Issues - A Health Economic Analysis

University dissertation from Department of Clinical Sciences, Lund University

Abstract: The consumption of alcoholic beverages has wide effects, for example, causes premature mortality, prevents certain heart diseases, increases crime rates, and affects quality of life. The main problems with alcohol consumption from an economic point of view are lack of information for consumers when making consumption decisions, externalities, and the addictive nature. It is expected that this results in non-optimal consumption levels, causing higher costs than benefits. Studying the effects of alcohol consumption is thus important in order to increase information and to allow interventions and regulations to be implemented targeting market failures, with the overall purpose of improving societal and individual welfare.

The aim of this thesis is to study the effect of alcohol consumption on society, calculating the societal cost of consumption, and investigating possible improvements with regard to the estimation methods, data materials, and methodological assumptions. The focus of the latter is on issues related to labour market outcomes. Four research papers are included, together covering the aim.

Paper I conservatively estimates the societal cost of alcohol consumption in Sweden, including health and quality of life effects. The costs add up to a net cost of SEK 20.3 billion (0.9% of GDP) in 2002. To this should a partial estimate of reduced quality of life be added, totalling 122,000 QALYs. Sensitivity analyses indicate a sensitivity range of 50% of the net cost. However, even the lowest plausible estimate shows net societal cost of alcohol consumption.

Paper II investigates the effect of low alcohol consumption on health, measured as medical care costs and prevalence of alcohol-attributable diseases. It is found that low alcohol consumption increases medical care costs and episodes, with the exception for individuals above 80 years of age. Thus, the protective effect of low alcohol consumption for some diseases can not fully counter the detrimental effect from those diseases where low alcohol consumption increases the risk. Based on the epidemiological literature, low alcohol consumption should therefore not be considered to improve health.

Paper III studies a methodological issue in connection to the wage equation; whether failure to account for individuals’ drinking histories causes heterogeneity within commonly pooled consumption groups, potentially causing bias in econometric estimations. By applying a multinomial logit model, it is found that pooled drinking groups (current abstainers and light drinkers) are heterogeneous, and that this might implicate estimation bias due to confounding and misclassification. This study thus argues that it is imperative to account for drinking history when studying the effect of alcohol consumption.

Paper IV analyses the effect of women’s alcohol consumption on the likelihood of being long-term absent from work. Drinking history and selection effects are controlled for by applying a Heckman model. Women who are not a long-term light drinker is associated with an increase in the probability of long-term sickness-related absence, except for the insignificant effect of being a current light but former heavy drinker. The strongest effect is found for former drinkers (18%) followed by former abstainers (15%). Surprisingly are both being a long-term abstainer and a long-term heavy drinker associated with an increase of around 10%. Several simulation models were estimated, for example investigating the potential societal gain in productivity if all women were long-term light drinkers. It is found that the effect of alcohol consumption on long-term sickness-related absence is rather small on an individual level, although the added societal effect is substantial.

It is shown in this thesis that alcohol consumption has a large societal impact. The societal cost was estimated in Paper I and Paper II – IV have supplied new information, with focus on the possible wage effect of alcohol consumption, in order to improve future estimations. Paper II rejects, based on the epidemiological literature, the possibility that the positive wage effect of low alcohol consumption is mediated through a protective health effect. According to Paper III, drinking history should be controlled for although this can not explain the commonly found inversed U-shaped relationship between alcohol and wages. Paper IV in turn suggests sickness-related absence as a mediator, potentially explaining (parts of) the alcohol – wage effect. Finally, the thesis has shown that the results of cost estimations are sensitive to what type of data is being used. Compared to a society without alcohol, the current Swedish consumption increases long-term sickness-related absence when using epidemiological data (Paper I), and decreases absence when using microdata and econometric methods (Paper IV).