Abstract
Economic arguments for acting for health are increasingly important for policymakers, yet to date there has been no consideration of the likely economic burden of alcohol on the global level. A review of existing cost estimates was conducted, with each study disaggregated into different cost areas and the methodology of each element evaluated. The range of figures produced from more robust studies was then applied tentatively on the global level. The reviewed studies suggested a range of estimates of 1.3 – 3.3% of total health costs, 6.4 – 14.4% of total public order and safety costs, 0.3 – 1.4‰ of GDP for criminal damage costs, 1.0 – 1.7‰ of GDP for drink-driving costs, and 2.7 – 10.9‰ of GDP for work-place costs (absenteeism, unemployment and premature mortality). On a global level, this suggests costs in the range of $210 – 665 billion in 2002. These figures cannot be understood without considering simultaneously six key problems: (i) the methods used by each study; (ii) who pays these costs; (iii) the ‘economic benefits’ of premature deaths; (iv) establishing causality; (v) omitted costs; and (vi) the applicability of developed country estimates to developing countries. Alcohol exerts a considerable economic burden worldwide, although the exact level of this burden is a matter of debate and further research. Policymakers should consider economic issues alongside evidence of the cost-effectiveness of particular policy options in improving health, such as in the WHO's CHOICE project. [Baumberg B. The global economic burden of alcohol: a review and some suggestions. Drug Alcohol Rev 2006;25:537 – 551]
Key words:
Notes
1 ‘Economic’ is here narrowly defined as arguments involving the money economy, rather than the broader sense that encompasses changes in quality of life (which is therefore much closer to a public health approach); this point is developed below in note 21.
2 A number of methodological considerations are not discussed here for both readability and space. However, readers with unanswered questions should refer to Anderson & Baumberg (2006) or contact the present author.
3 Interested readers can also find further discussion elsewhere Citation[17],Citation[28],Citation[56],Citation[106].
4 The following search terms were used: alcohol*, combined with economic*, cost* or burden*.
5 Data in languages other than English, French, German or Spanish were translated by the relevant APN member, using a standard form to extract relevant information only. In two cases the studies were not publicly available; for transparency purposes, the English summaries have been made available on the APN website.
6 All transfers between individuals—whether deliberate or stolen—were also removed Citation[17],Citation[107], as have the health costs of violent crime (due to the risk of double-counting Citation[28]) and non-market costs such as household work (as these cannot strictly be compared to GDP Citation[17]).
7 Crime costs are expressed as a % of GDP given the lack of available data on ‘public order and safety’ expenditure as presented in .
8 GDP figures taken from the UN Statistical Division 1/2/2006; health spending as percentage of GDP figures taken from the World Health Report 2005, published by the WHO. All figures are stated to the nearest $5 billion to avoid giving a misleading imprecision of precision.
9 For homicide: 31 (developed countries) vs. 32% (developing) of deaths for males and 31% vs. 22% for females. For road traffic accidents: 41% of deaths in men aged 15 – 29 compared to 30% in other regions. ‘Developed countries’ defined as countries with very low child and adult mortality; ‘developing countries’ refers to all other countries. Average figures are obtained by scaling the reported alcohol-attributable fraction for homicide (Rehm et al., 2004) by the numbers of homicides in each region (see http://www.who.int/healthinfo/bod/en/index.html).
10 This assumption is that 20% of all alcohol consumed was drunk by addicted drinkers and should therefore be included in the social cost. This excludes the cost of drinking when the drinker does not have complete information on the risks of alcohol.
11 It should be noted that health costs in the Collins & Lapsley studies are relatively low compared to other studies as they take into account the ‘savings’ of premature deaths; see below.
12 More theoretically, however, there are two reasons why we may not expect individual-level studies to match the social cost results. First, there may well be systematic biases in how individuals react to health care, with some research suggesting that relatively heavy drinkers may have shorter stays in hospital as they cannot carry on with their drinking Citation[36]. Such biases are particularly important for out-patient service use, where variations in care-seeking behaviour are likely to explain at least part of the negative relationship between alcohol use and out-patient service utilisation (see summary in Citation[35]).
13 Ironically, Collins & Lapsley also found that the health benefits of alcohol increase the total health care burden. This is presumably because cardiovascular disease leads to death relatively more than long-term disability, and so preventing cardiovascular disease leads to an increase in health costs.
14 Even this US evidence is not altogether convincing—a wage penalty was found only for men who had ever suffered from alcohol dependence (but not alcohol abuse, and not for women at all). The estimate also assumes that education is a mechanism through which ever being alcohol dependent can affect wages (i.e. it uses a reduced-form model). Harwood et al. justify this using research that suggests those reporting youthful alcohol abuse have less education than would be expected from their background, although more recent research contradicts this Citation[108]. If this assumption is dropped, then Harwood et al. find no significant effect in any group.
15 Although the risk of suffering from an alcohol abuse disorder increases at higher levels (and more detrimental patterns) of consumption, there are more intervening variables (such as genetic vulnerability) that create a more uncertain probability than found for the level of consumption.
16 Many studies use called an ‘instrumental variables’ technique by looking at the association between wages and variables that are correlated with alcohol use but not with wages (e.g. alcohol tax changes).
17 Similar results have also been found for drinking and unemployment, in that alcohol use disorders are associated with higher unemployment, light alcohol consumption is sometimes associated with lower unemployment, and complex methods often produce implausible results Citation[51].
18 For example, alcohol consumption was reported retrospectively at 2-week intervals, and the study did not ask about the quantity of alcohol consumed (or any other features) of the drinking occasion. Such weaknesses were enough for a systematic review to classify the study as of ‘low quality’, although it should be noted that only one study world-wide met their definition of even ‘medium quality’Citation[109].
19 The same study also valued stolen property as a cost to society, but this has been removed from the current comparison as it was felt to count as a ‘transfer’ rather than a ‘loss’.
20 Cook & Clark argue that ‘problem drinking is clearly associated with an increased risk of STDs’[73: 159], given that eight of 11 studies found a significantly increased risk of at least one STD among problem drinkers compared to non-drinkers. However, this conceals the fact that many studies performed multiple tests separately (e.g. for men and women, for different measure of alcohol use, or for different STDs)—if we look instead across the full 22 contrasts among problem drinkers, we find that over half were non-significant, and that two further contrasts were insignificant in multivariate analyses.
21 As an aside, it should be noted that ‘health economics’ as a discipline deals substantially with matters other than production losses, such as quality of life and leisure time. The difference between public health and economic approaches is therefore merely a matter of whether money is used as a metric that enables otherwise incommensurate areas (work impairment, premature mortality, and so on) to be combined in analyses. However, ‘economics’ in popular and policy usage tends to refer to the narrower set of concerns that have here been labelled ‘economic arguments’ in this article, as seen in, for example, the economic impact assessment for the prospective European Commission Communication on alcohol (RAND Europe, to be published later in 2006). Many thanks to an anonymous reviewer for stressing the need to clarify this point.