Organisation de Coopération et de Développement Économiques Organisation for Economic Co-operation and Development

Transcription

1 For Official Use For Official Use Organisation de Coopération et de Développement Économiques Organisation for Economic Co-operation and Development 12-Jun-2013 English - Or. English DIRECTORATE FOR EMPLOYMENT, LABOUR AND SOCIAL AFFAIRS HEALTH COMMITTEE HARMFUL USE OF ALCOHOL: TRENDS AND POLICY CONCLUSIONS OECD Conference Centre, 2 rue André Pascal, Paris, France on Monday 1st July - Tuesday 2nd July 2013 Please note that the annexes mentioned in this document are in a separate document with the code /ANN. For further information, please contact Mr Franco SASSI, Tel (0) ; franco.sassi@oecd.org English - Or. English JT Complete document available on OLIS in its original format This document and any map included herein are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.

2 NOTE BY THE SECRETARIAT 1. As part of the OECD Programme of Work and Budget for the years , the Health Committee undertook to conduct analyses of trends and patterns of alcohol consumption, with a view to exploring the potential impacts of policy options to address the harms associated with such consumption. The project is set in the broader framework of the OECD Economics of Prevention programme, laid out in OECD s Health Working Paper 32 The prevention of lifestyle-related chronic diseases: an economic framework (DELSA/HEA/WD/HWP(2008)2). The project was undertaken on the basis of a work plan discussed at a meeting of the Expert Group on the Economics of Prevention in April The project involved several streams of work and output. Preliminary versions of some of the outputs were presented as background papers and discussed at a meeting of the Expert Group on the Economics of Prevention, in October An early version of this report was submitted to the Health Committee for consideration at its December 2012 meeting. The Expert Group and the Health Committee provided guidance to the Secretariat on a number of dimensions in which they wished to see the work expanded or revised, including: a. broadening the geographical coverage of the analyses of trends and disparities in alcohol consumption; b. extending the analyses of the impacts of alternative policy options to cover both morbidity and mortality from a range of relevant diseases and injuries; c. discussing the policy impacts beyond those on health and health care expenditures, particularly the labour market impacts and possibly selected crime impacts; d. revising the set of policy options assessed in simulation analyses, in line with existing evidence of the effectiveness of alternative policies, and with the current policy debate. 3. The Secretariat produced new analyses and technical documents, which were circulated in batches to the Expert Group in the period January to April Members of the Expert Group provided detailed comments, which the Bureau of the Expert group considered extensively in a meeting held on 26 th April at the OECD Headquarters in Paris. At the same meeting, the Bureau of the Expert Group provided guidance to the Secretariat on how to finalise the analyses (a draft Summary Record of that meeting is appended to this document). 4. The Secretariat met on a number of occasions with representatives of the alcohol industry, pharmaceutical industry and large employers, under the auspices of BIAC, to discuss aspects of the work and receive evidence on the impacts of selected policy options. 5. A final set of outputs from the project will include the following: An OECD Health Working Paper entitled Alcohol consumption in OECD countries: trends and social disparities, reporting the detailed results of a range of analyses based on health survey data from half of OECD countries. 2

3 An OECD Health Working paper entitled Tackling alcohol-related harms: the health and economic impacts of prevention policies, reporting the results of a brief review of policies adopted in OECD countries to control alcohol use and reduce the harms associated with it, as well as the findings of an economic analysis of the impacts of such policies on health, employment, productivity and public expenditure in a selection of OECD countries. A final report for the project, to be published under the authority of the Secretary-General, which will present the findings of the various streams of work in a cohesive form for an audience of policy makers, health professionals and health policy scholars. This report will provide an overall summary of OECD work on the economics of alcohol consumption and on the effectiveness and cost-effectiveness of alcohol policies, and an assessment of the policy relevance of the findings presented. The Secretariat aims at producing this final report by the last quarter of A manuscript will be made available to Health Committee delegates and members of the Expert Group on the Economics of Prevention before publication. 6. The enclosed paper is meant to provide the Health Committee with an overview of the project and key findings so far. 7. The Committee is invited to: NOTE, and COMMENT on, the main findings of the OECD Alcohol project and their policy implications, as set out in the enclosed overview paper; NOTE the proposal for the eventual publication of the main findings from the project, after accounting for delegates comments, under the responsibility of the Secretary-General. 3

4 TABLE OF CONTENTS NOTE BY THE SECRETARIAT... 2 EXECUTIVE SUMMARY... 6 ALCOHOL, A COMMON FIXTURE IN HUMAN SOCIETIES, A LEADING RISK FACTOR FOR HEALTH... 8 SECTION I ALCOHOL TRENDS: A SLOWLY CHANGING CONSUMPTION WITH COMPLEX SOCIAL PATTERNS The challenge of measuring alcohol consumption Aggregate data sources Survey data Trends in alcohol consumption Trends in different age groups How to explain changes in alcohol drinking? Social disparities in drinking patterns Disparities in alcohol consumption Hazardous and heavy episodic drinking by education level and socioeconomic status Time-trends in social inequalities Disparities by ethnicity SECTION II THE IMPACTS OF ALCOHOL CONSUMPTION ON LABOUR MARKETS AND PRODUCTIVITY Alcohol consumption and labour market outcomes Alcohol and employment Alcohol and wages Alcohol and labour productivity The value of lost productivity SECTION III AN ECONOMIC CASE FOR GOVERNMENT INTERVENTION Policy objectives and rationale for government intervention Types of policies The private sector response: how are markets adjusting to the new challenges? SECTION IV WHAT POLICY OPTIONS? ARE THEY COST-EFFECTIVE? Assessing the impacts of policy options: the need for a population-wide modelling approach The CDP-Alcohol model and the rationale for the selection of policy options assessed Main analysis Further analysis Design of the OECD CDP-alcohol model The impacts of alternative policies The effects of preventive interventions on alcohol consumption, longevity and disability Effects on diseases and injuries The cost of preventive interventions The efficiency of prevention policies SECTION V CONCLUSIONS AND POLICY RELEVANCE REFERENCES

5 Tables Table 1. Share of hazardous drinkers in ten selected OECD countries Table 2. List of diseases in CDP-Alcohol and health impact Figures Figure 1. Alcohol consumption in litres per capita among population aged 15 and over, (or nearest year), OECD and key partner countries Figure 2. Evolution of alcohol consumption, (or nearest year), selected OECD countries Figure 3. Age-standardised rates of any drinking in the past 12 months Figure 4. Age-standardised rates of hazardous drinking Figure 5. Trends in HED by age group, France and Germany Figure 6. Adjusted probabilities for any alcohol drinking by education level Figure 7. Concentration index for hazardous drinking by education level, men (left) and women Figure 8. Concentration index for HED by education level, men (left) and women Figure 9. Odds ratios for hazardous drinking by ethnic status, England Figure 10. Simplified structure of the CDP alcohol model Figure 11. Diseases and injuries included in the CDP-Alcohol model Figure 12. Impacts of selected policies on levels and patterns of drinking, Canada, Figure 13. Life years gained through selected policies. Average per year, Canada, Figure 14a. Cumulative DALYs gained over time, main analysis Figure 14b. Cumulative DALYs gained over time, further analysis Figure 15. Decrease in number of new cases of diseases at the population level over the simulation period (average per year) Figure 16. Economic assessment of the interventions at the population level million ($ PPP per year).. 44 Figure 17a. Cumulative impact on health expenditure ($ PPP / capita) over the simulation period (discounted 3%), main analysis Figure 17b. Cumulative impact on health expenditure ($ PPP / capita) over the simulation period (discounted 3%), further analysis Figure 18. Incremental cost-effectiveness ratio ($ PPP / DALY) of the interventions over the simulation period Boxes BOX 1. Alternative sources of alcohol consumption data in the United States

6 EXECUTIVE SUMMARY 8. Alcoholic beverages, and the problems they engender, have been familiar fixtures in human societies since the beginning of recorded history. Worldwide, alcohol is responsible for a larger share of the burden of ill health and premature mortality than tobacco or obesity. It accounts for one in 15 deaths in the European region, and for an even larger proportion of disability, especially among men. Its social costs are estimated to be in excess of 1% of GNP in high- and middle-income countries. 9. A basic principle in the economics of prevention is that Maintaining good health is an important goal for most individuals, but health is by no means the only outcome that individuals value when they choose how to lead their own lives. Individuals wish to engage in activities from which they expect to derive pleasure, satisfaction, or fulfilment, some of which may be conducive to good health, others less or not at all. [ ] An assessment of the role of prevention must not ignore those competing goals (Sassi & Hurst, 2008). Alcohol drinking is a case in point. However, there are strong reasons to believe that the negative consequences of alcohol are not fully reflected, or accounted for, in individual drinking behaviour, which provides a rationale for governments to enact policies to reduce the harms of alcohol consumption by creating conditions that would change some of those elements of the individual behaviour. Governments need to balance individual choices with social responsibility. 10. Drinking is a social phenomenon, with strong cultural connotations. Alcohol use is driven by social norms more than other health-related behaviours. This is reflected in unique patterns of social disparity in drinking, showing the well-to-do often more prone to hazardous use of alcohol, and a polarisation of problem-drinking at the two ends of the social spectrum. Governments can play a major role in tackling the harms associated with alcohol use, and have many tools at their disposal to achieve this goal. Some policy approaches are more effective and efficient than others, depending on their ability to trigger changes in social norms, and on how well they can target the groups that are most at risk. The OECD Alcohol project provides a wide-ranging assessment of the health, social and economic impacts of key policy options for tackling alcohol-related harms. 11. Both aggregate national estimates and individual-level survey data show that trends in overall alcohol consumption remained relatively stable over the past 20 years. However, a closer look at trends and patterns of consumption among specific population groups reveals a more complex picture. Higherrisk drinking behaviours (hazardous drinking and heavy episodic drinking - HED) were found to be on the increase among young people, especially women, in several countries examined. In these countries e.g. England, Korea, Switzerland, United States, Ireland and Germany rates of hazardous drinking were found to increase more, or decrease less, among younger age groups compared with the rest of the population. In Canada, Germany, Italy, New Zealand and United States HED increased fastest among younger age groups. On the other hand, in countries where youth HED is most widespread, i.e. England and Ireland, rates fell over the same period. 12. The trends observed among younger age groups are a source of major public health and social concern because the spread of high-risk drinking behaviours is associated with an increase in morbidity and mortality from accidents and injuries (the most common cause of death in adolescents and young adults), as well as violence and social disturbances. In addition, and most worryingly, early onset of drinking and high-risk drinking at young ages make drinking problem more likely in adult life. 13. In most of the countries examined, highly educated women are more likely than those with less education to engage in HED, while the opposite is true for men. However, social patterns of drinking tend 6

7 to differ across countries, as they are largely the result of social norms, cultural and environmental influences, and of government policies in place in the countries concerned. 14. Government intervention is supported by a strong economic rationale. Policy efforts to reflect individual choice while reducing the negative social impacts of alcohol use have been mainly focused on making alcoholic beverages more expensive through taxation, on limiting their availability, and on controlling their supply through state monopolies. An increased emphasis on the public health consequences of alcohol use in the past few decades has raised the importance of alcohol in policy agendas worldwide and broadened the spectrum of actions countries are prepared to consider and finance to address problems associated with harmful use of alcohol. 15. Governments have seen both public support and opposition to actions that are seen by some as undue interference with individual choices. The involvement of all stakeholders in the fight against the harms caused by alcohol use is desirable, although the contribution of alcohol beverage manufacturers and retailers to tackling alcohol-related harms (e.g. through product reformulation to reduce alcohol content, or through self-regulation of advertising) has so far been limited or ineffective. 16. In order to take effective action, governments need evidence of population-level health and economic impacts of alternative policy options, which can only come from computer-based simulation models like the one developed by the OECD as part of its Alcohol project. The findings presented in this paper show that fiscal measures have the greatest potential for reducing high-risk drinking behaviours and overall mortality, and do so at a very low cost, and with a significant revenue-raising potential. Regulation of advertising may achieve important reductions in high-risk alcohol consumption and related harms if implemented on a relatively large scale, leading to sizable cuts in advertising expenditure. Actions within the health sector, including brief interventions in primary care and pharmacological and psychosocial treatments for alcohol dependence may also significantly reduce alcohol-related morbidity and mortality but the costs of brief interventions and their limited effects on health expenditures suggest that their costeffectiveness profile may be less favourable, at least in the form assessed in this analysis. Minimum pricing policies, if their potential in reducing alcohol consumption detected in existing studies is confirmed, can produce significant health gains and reductions in health care expenditure, at a low implementation cost. 7

8 ALCOHOL, A COMMON FIXTURE IN HUMAN SOCIETIES, A LEADING RISK FACTOR FOR HEALTH 17. Few commodities are as deep-rooted in people s cultures and habits as is alcohol. The discovery of Stone Age beer jugs confirmed that fermented beverages already existed in the Neolithic period, i.e. about 12,000 years ago (Patrick, 1952) and anthropologists claim that people started growing crops with the aim of brewing alcoholic drinks even before they had expertise and technology to bake, making this ancient form of beer precede bread as staple food (Katz & Voigt, 1987). Throughout history, and until the present times, alcohol has played a fundamental role in arts and religion, and even became an important public health tool in the medieval age, when the process of brewing was used to purify water from bacteria. Alcoholic beverages provide many individuals with pleasure, and have always had, and still have, an important social function too. However, the side effects of alcohol, too, have been present throughout the history of its use. Excessive alcohol use has traditionally been associated with disease and loss of physical and social function, with idleness and poverty (although excessive drinking is by no means an exclusive habit of the poor and socially excluded), with crime and social disturbances. Therefore, governments have been trying for millennia to control the consumption of alcohol, through measures such as monopolies and taxes. 18. Since alcohol is causally related to diseases covering more than 60 ICD codes (Rehm et al., 2004)], disease outcomes are among the most important alcohol-related harms. Excessive alcohol consumption contributes to a variety of health problems, including chronic diseases like cancer, alcoholic liver disease, cardiovascular disease and mental health conditions, as well as the sequelae of intoxication and alcohol dependence, injuries (such as falls and road traffic injuries) and suicide. WHO estimates that 4% of the global burden of disease is attributable to alcohol, which accounts for about as much mortality and disability as tobacco and hypertension (Rehm et al., 2004; Rehm et al., 2009). In the US, excessive alcohol consumption is the third-leading cause of death, accounting for 80,000 deaths per year and 2.3 million years of potential life lost (CDC, 2012). Health care costs associated with excessive drinking in the United States are estimated at $25.6 billion (Bouchery et al., 2006). 19. Many studies point to a beneficial effect of moderate alcohol consumption on overall mortality among middle-aged and older men and, for more modest volumes of alcohol drunk, in older women. This is because of lower rates of ischemic heart disease. However, rates for other conditions increase steadily with consumption, which means that drinkers in the above age groups have to trade off a reduced chance of death from, say, myocardial infarction with an increased chance of death from, say, cancer or accidental injury. Interpretation of the evidence is exceptionally difficult, because of factors such as people quitting drinking because of ill health, thus inflating the short-term correlation between abstention and death. 20. The impact of the harmful use of alcohol extends well beyond the direct health-related consequences to drinkers, however. A broader assessment of social impacts, including criminal damage, violence and lost productivity at the workplace, demonstrates the considerable harm imposed on society as a result of alcohol consumption. Social cost studies have been carried out in a number of countries (Rehm et al., 2009). Estimates for the United States, for instance, point to extra health care costs of $25.6 billion, and broader societal costs exceeding $220 billion, almost ten times the health costs alone (Bouchery et al, 2006). 8

9 21. Harmful use of alcohol, therefore, constitutes a very serious public health problem, to which national health authorities and international health organizations have responded with comprehensive control and prevention strategies. Evidence-based preventive measures are available at both the individual and population levels, with alcohol taxes, restrictions on alcohol availability, and drink-driving countermeasures among the most effective policy options (Babor et al., 2003; Anderson et al., 2009; Task Force on Community Preventive Services, 2009). Other measures aim specifically at high-risk or dependent drinkers in order to encourage modifications of behaviour (brief interventions, pharmacological treatments). Since 1979, the World Health Assembly called upon WHO member states to develop and adopt appropriate legislation and measures to tackle excessive alcohol use (WHO, 1979). Such efforts culminated in the endorsement of a Global Strategy on the Harmful Use of Alcohol in 2010 (WHO, 2011), which supports ten target areas for national action including: community actions, drink-driving policies, limitation of the availability of alcohol, and actions affecting the marketing, promotion and pricing of alcohol products. Even though the types of alcohol policies adopted are broadly similar across countries, policies vary widely in how they are implemented and in the legal limits set for law enforcement. For instance, blood alcohol concentration (BAC) limits can range from as low as zero tolerance in Russia for some driver categories (Lerner, 2012) to 0.08% of blood alcohol level in the US, UK and Mexico (WHO, 2011). On the other hand, the US enforces a minimum legal drinking age of 21, one of the highest across OECD countries, while many other countries have minimum ages as low as 16 (see Annex 1 for more details). 22. A better understanding of which measures or strategies represent the best use of society's scarce resources - and by how much they can reduce the harmful consequences of alcohol use - is key to an evidence-based approach to alcohol policy making. Potentially valuable investments in prevention are forgone because of uncertainty and ignorance about the likely returns on those investments and the time frame in which they will materialize. Governments wishing to allocate their resources efficiently need to use appropriate and relevant approaches to assess the full range of costs and consequences of different actions. 23. At the international level, the WHO CHOICE project compared the costs and effectiveness of a number of potential strategies for the prevention and control of alcohol use in different regions of the world (Anderson et al., 2009; Chisholm et al., 2004). Assessed strategies included drink-driving legislation and random breath testing, taxation of alcoholic beverages, reduced hours of sale in retail outlets, advertising, and brief advice in primary care for high-risk drinkers. Mass media or school-based awareness campaigns were not analyzed on the grounds that evidence for effectiveness was weak, both in terms of methodological quality and the estimated effect on consumption. Results showed substantial differences in costs and effects, both between interventions and in different WHO regions. The most costly interventions were random breath-testing and brief advice in primary care. In populations with a relatively high prevalence of hazardous alcohol use (more than 5% of the total adult population), taxation was found to be the most effective and cost-effective strategy. Comparative analyses of the costs and effects of different strategies for the prevention and control of alcohol use have also been carried out at the national level, for instance in Australia (Cobiac et al., 2009) and in the UK (Purshouse et al., 2010), using models to generate estimates of the health effects of interventions. Alcohol and the OECD Economics of Prevention programme 24. The overarching aim of the OECD Economics of Prevention programme is to support governments in developing and evaluating policies to improve population health by reducing the occurrence and the impact of non-communicable diseases. The economic framework in which the programme was developed is centred on the hypothesis that the prevention of chronic diseases may provide the means for increasing social welfare, enhancing health equity, or both, relative to a situation in which chronic diseases are simply treated once they emerge. The framework outlines a process for testing 9

10 whether such a hypothesis holds in relation to the prevention of specific risk factors and chronic diseases. This process entails an assessment of the pathways through which chronic diseases are generated and an exploration of whether such pathways are simply the outcome of efficient market dynamics, or the effect of market and rationality failures preventing individuals from achieving more desirable outcomes. Where failures exist, there may be scope for adopting preventive interventions, if such interventions are shown to have the potential to increase social welfare or improve its distribution. The expected impact of preventive interventions on individual choices should be commensurate to the extent of the above failures and to the severity of the outcomes arising from them. 25. Section I of this paper provides a summary of new work on patterns of alcohol consumption, including the spread of hazardous and harmful consumption among people of different social groups, mainly based on analyses of health survey data from 13 OECD countries (4 additional countries are currently being included in the set). This is complemented, in Section II, by a review of existing evidence of the main impacts of alcohol consumption on labour markets and productivity. 26. Section III provides a logical link between the above analyses and the analyses of government policy options addressed in Section III. In particular, the economic justification for government action on alcohol is reviewed in Section II with the aim of identifying specific factors that may hinder the smooth functioning of market mechanisms in alcohol consumption, leading individual consumers to make choices that do not maximise their own, and society s, welfare. In the case of alcohol, the economic justification for government action on the grounds of market failure is stronger than in some other areas of public health action (e.g. on aspects of diet and physical inactivity). 27. Section IV contains a review and analysis of actual and potential interventions to reduce alcohol consumption and alcohol-related harms. This includes a review of policy measures adopted in OECD countries and an assessment of the health and economic impacts of a selection of policy options, identified on the basis of their prominence in the current policy debate, and of the availability of evidence of their effectiveness. 28. Finally, section V contains a discussion of the policy relevance of the findings produced thus far as part of the OECD Alcohol project, as well as a summary of key messages emerging from the work undertaken. 10

11 SECTION I ALCOHOL TRENDS: A SLOWLY CHANGING CONSUMPTION WITH COMPLEX SOCIAL PATTERNS 1. The challenge of measuring alcohol consumption 29. Multiple data sources exist on alcohol, which may be used to derive complementary estimates of consumption serving different purposes. However, it is virtually impossible to obtain objective populationwide measures of alcohol consumption. Sales data provide relatively straightforward estimates of consumption for countries, but they may not be comprehensive, and mapping purchases to consumption requires difficult assumptions. Individual-level surveys provide detailed information on actual consumption but have to rely on often biased self-reports. All types of data sources may contribute to drawing a picture of alcohol drinking in a country, and all must be taken into account as far as possible. The main data sources are discussed below Aggregate data sources 30. The OECD regularly collects from its member countries data on alcohol consumption defined as annual sales of pure alcohol in litres per person aged 15 years and over. Data sources are mostly national statistical institutes, except for eight countries for which the WHO Global Information System on Alcohol and Health is used (Belgium, Germany, Greece, Israel, Italy, Luxembourg, Portugal and Spain). The methodology to convert alcoholic drinks to a more readily comparable metric of pure alcohol may differ across countries. Typically beer is weighted as having 4-5% alcohol content, wine 11-12% and spirits 40%. In the OECD area, the consumption of alcohol is, on average, around 9.1 litres per capita per year. It should be noted that in some countries (e.g. Luxembourg), national sales do not accurately reflect actual consumption by residents, since purchases by non-residents may create a significant gap between national sales and consumption. Consumption of alcohol per capita over the period , on average, decreased by 9% in the OECD area (Figure 1), although consumption increased in Iceland and Mexico by 70% or more (from relatively low levels). Alcohol consumption increased significantly in some Key Partner countries, including the Russian Federation, Brazil and China, although consumption per capita is still comparatively low in the latter two countries. 31. A closer look at trends in alcohol consumption per capita reveals a steady decrease over the past 30 years in countries where consumption was originally high, such as France, Spain, Italy, and Germany (Figure 2). In countries that originally had lower consumption levels, the data show increasing trends, in some cases followed by a downturn in the recent years (e.g., Ireland, Finland and, the UK). However, it should be noted that aggregate consumption data reflect volumes of alcoholic beverages purchased or consumed, weighted for the typical alcoholic strength of different beverages. Trends over time may not fully account for changes in alcoholic strength within beverage categories. For instance, the alcohol content of wine has gradually increased in recent years, on average, by an estimated 0.2 to 2.0 percent on a base of percent, because of changes in production and consumer demand (Alston et al., 2011). Therefore, declines in consumption over time may be overestimated, and increases underestimated. 11

12 Figure 1. Alcohol consumption in litres per capita among population aged 15 and over, (or nearest year), OECD and key partner countries 2009 Change per capita, Litres per capita (15 years +) Indonesia India Turkey Israel China Mexico Brazil Norw ay South Africa Iceland Japan Sweden Italy Canada Chile United States Korea Slovak Republic OECD Greece New Zealand Netherlands Belgium Germany Finland Spain Australia Denmark Sw itzerland Poland United Kingdom Ireland Russian Fed. Slovenia Hungary Luxembourg Estonia Czech Republic Austria Portugal France n.a n.a n.a % change over period Source: OECD Health at a Glance WHO produces estimates of adult per capita alcohol consumption (APC) based on data from the Global Survey on Alcohol and Health (WHO, 2011), which in turn draws upon various data sources, including sales, tax, international trade, and survey data. Much effort was made to collect and crossvalidate various sources of information, and to estimate the unrecorded alcohol consumption which is likely to account for nearly 30% of total worldwide adult consumption. The WHO APC is indeed an estimate of recorded and unrecorded adult per capita consumption of pure alcohol in 2005 (a 2012 update is currently under way). Unrecorded alcohol is defined as alcohol that is not taxed and is outside the usual system of governmental control, because it is produced, distributed and sold outside formal channels it refers to homemade or illegally produced alcohol, smuggled alcohol, alcohol for industrial and medical use, alcohol obtained through cross-border shopping, and alcohol consumed by tourists. Unrecorded alcohol consumption is estimated based on empirical investigations and expert judgements. The highest unrecorded consumption figures are about 4.0 or more litres per capita in Hungary, Poland, and the Russian Federation. The resulting estimated APC is about 6.1 litres on average worldwide, and about 10.6 litres in high-income countries. Thus, WHO total estimates (recorded and unrecorded) may significantly differ from OECD figures for some countries (e.g., 9 litres in OECD Health Data and 14.4 litres in WHO data for the Slovak Republic, or 9.0 and 15.0 litres, respectively, for Korea). 12

13 Figure 2. Evolution of alcohol consumption, (or nearest year), selected OECD countries 19 Liter of pure alcohol per capita, age Australia Canada Finland France Germany Ireland Japan Spain United Kingdom United States Source: OECD Health Data Existing aggregate sources and estimates of alcohol consumption provide the most reliable information to determine broad national trends and draw country profiles of alcohol consumption. Aggregate sales data are a good indicator for monitoring population-wide measures of drinking in line with population-wide policies on pricing and availability. Also, there is some evidence of existing correlations between total per capita alcohol consumption and excessive consumption, and its related harms (Håkan et al, 2002). 34. However, the aggregate nature does not permit identification and examination of individual patterns of drinking. To design appropriate policies, it is desirable to understand how harmful forms of drinking evolved over time, and to identify which population groups are most likely to engage in, and which are most affected by, problem drinking. Such analyses help policy makers to target population groups for strategies to reduce harmful drinking. Population health and lifestyle survey data providing information on alcohol consumption and individual characteristics are best suited for assessing drinking behaviours across different sub-groups of the population, although these surveys do suffer from important limitations. Hence, for these reasons, the OECD analyses of alcohol consumption are based on information from individual-level surveys Survey data 35. The OECD undertook analyses of alcohol consumption in 20 OECD countries selected on the basis of data availability (Australia, Canada, Chile, Czech Republic, England, Finland, France, Germany, Hungary, Ireland, Italy, Japan, Korea, New Zealand, Portugal, Slovak Republic, Slovenia, Spain, Switzerland, and the USA), in order to explore: (i) trends in different types of alcohol use over time; and, (ii) social disparities by education and socio-economic status, and differences by age and sex, in alcohol drinking. 13

14 36. Individual-level alcohol consumption data were obtained from national health and lifestyle surveys, or alcohol and drug use surveys. These surveys provide the most detailed information currently available on individual socio-demographic characteristics combined with alcohol drinking patterns, assessed either over the week prior to the interview in 9 countries (Australia, Canada, Czech Republic, Finland, Hungary, Japan, Slovak Republic, Slovenia, Switzerland) or based on questions on drinking frequency and quantity on a typical drinking day in the 11 remaining. Moreover, some countries questionnaires use standard scales to measure drinking behaviours such as the Alcohol Use Disorder Identification Test (AUDIT) in Chile, France, Ireland (2007) and Korea (2008). The AUDIT is a questionnaire covering alcohol consumption, alcohol related problems and abnormal drinking behaviour. It was developed by the WHO as a screening tool for health professionals to identify people at risk of developing alcohol problems. 37. Analyses of trends in alcohol consumption over time used multiple waves of survey data available in 12 OECD countries: Australia, Canada, England, Finland, France, Germany, Ireland, Italy, Korea, New Zealand, Switzerland, and the USA, providing a temporal coverage up to 18 years (England). The national surveys used and the number of waves available for each country are listed in Annex The use of different national surveys for several years may be a source of data heterogeneity across countries and over time, although all the variables are constructed in order to get the highest level of comparability. International comparisons in alcohol consumption need to rely on a common measure of alcohol level. Each country has its own definition of a standard drink (how much pure alcohol it contains) and its own recommendation for hazardous drinking limits (defined as number of glasses or amount of pure alcohol per week). No international consensus in drinking guidelines exists. While some countries do not have official recommendations, others set national guidelines for maximum weekly and/or daily consumption, either expressed in standard drinks or grams of pure alcohol (Furtwaengler and de Visser, 2013). 39. Further limitations of survey-based data on alcohol consumption are due to measurement bias, including underreporting by surveyed respondents and selection bias in survey sampling, as discussed extensively in Annex 3A. Drinking levels reported in surveys were shown to account for only 40-60% of alcohol sales (Midanik, 1982; WHO, 2011). OECD analyses based on quantitative approaches for correcting survey estimates for self-report bias (Rehm et al., 2010b), suggest that large discrepancies exist in some countries. The approach used is based on the triangulation of survey data with recorded aggregated per capita consumption data by modelling the upshifted distribution of alcohol consumption. Hence, because under-recording is apparently larger in France than in England (see Annex 3 for a discussion of these discrepancies), the adjustment is larger for France than for England. Table 1 illustrates those discrepancies for a selection of OECD countries. 40. National health survey data provide information on individual quantity and frequency of drinking. As far as possible, comparable outcome measures were derived from the surveys. These include: a. Drinking status, indicating whether people drank in the past 12 months or were abstainers. This variable is available over time in nearly all countries. b. Hazardous drinking, defined as a weekly amount of pure alcohol of 140 grams or more for women, and 210 grams or more for men. This measure refers to the limits above which people are at risk for their health. However, other studies use different thresholds (Rehm et al., 2002). This definition is specific to our study and is in line with the limits observed in most of countries. 14

15 c. HED, commonly called binge drinking, collected in most of national surveys through questions such as: In the past 12 months, how often did you have (n) or more drinks on one occasion? (5 drinks in Canada, Germany, and the US; 6 drinks in Chile, Czech Republic, France, Hungary, Ireland, Italy, Japan, New Zealand, Portugal, Slovak Republic, and Slovenia; 7 drinks for men and 5 drinks for women in Australia; and 8 drinks for men and 6 drinks for women in Switzerland) 1. Analyses focused on regular HED, i.e. at least once a week. Table 1. Share of hazardous drinkers in ten selected OECD countries %hazardous drinkers in total adult population Survey based estimates Corrected estimates Men Women Men Women Canada % 3% 22% 9% Chile % 0% 31% 10% England % 12% 37% 22% France % 2% 41% 23% Germany % 11% 30% 21% Hungary % 1% 47% 3% Ireland % 5% 42% 26% Japan % 5% 30% 5% Spain % 5% 36% 12% USA % 6% 29% 15% Source: OECD estimates, see Annex 3A for the methodology. 2. Trends in alcohol consumption 41. Figure 3 presents age-standardised proportions of people reporting to have drunk (any) alcohol in the past 12 months in the adult population. In all countries, rates are higher among men than women. Rates of male alcohol drinking are around 90% in most countries whereas female drinking rates are more variable. Rates are relatively stable over time, once sampling variation is accounted for. 42. Rates of hazardous and heavy episodic drinking display a larger degree of variation, partly due to differences between surveys, which exist not only across countries, but also across surveys undertaken in the same countries (e.g. see Box 1). In particular, the definitions used for hazardous and heavy episodic drinking vary, and so do interview approaches. Despite efforts to standardise definitions in the analyses presented here, differences remain across surveys, which suggest that the value of these analyses is more in the assessment of trends over time than in the comparison of rates across countries. 1 Further variation across countries derives from differences in the alcohol content of a standard drink. For instance, the 5-drink threshold in Canada corresponds to 68 grams of pure alcohol, while the 6-drink threshold in Ireland corresponds to 60 grams. 15

16 Figure 3. Age-standardised rates of any drinking in the past 12 months Proportion of alcohol drinkers, men age Proportion of alcohol drinkers, women age Note: Age range may differ across countries (Germany (age 25-65) and Finland (age 25-74)). Source: National health surveys mentioned in Annex 2, OECD estimates. 16

17 Figure 4. Age-standardised rates of hazardous drinking 30% AUSTRALIA 20% 10% CANADA ENGLAND FINLAND FRANCE GERMANY IRELAND NEW ZEALAND SWITZERLAND USA 0% % Proportion of hazardous drinkers, men age AUSTRALIA 20% CANADA ENGLAND FINLAND FRANCE GERMANY IRELAND 10% NEW ZEALAND SWITZERLAND USA 0% Proportion of hazardous drinkers, women age Note: Age range may differ across countries (Germany (age 25-65) and Finland (age 25-74). Source: National health surveys mentioned in Annex 2, OECD estimates. 43. As shown in Figure 4, most countries show relatively stable trends of hazardous drinking, although declines were observed in Ireland and Germany, where rates were originally high. These findings are in line with national trends reported elsewhere (e.g. Morgan et al., 2008; Pabst et al., 2010). Most countries present low rates of female HED. In 6 out of 8 countries, less than 8% of female drinkers report 17

18 HED at least once per week, this rate being especially low in France and Switzerland. Rates were broadly constant over time, with the exception of a noticeable decrease in Ireland 2, where HED rates are highest. 44. To some extent, rates of hazardous drinking have tended to converge over time, and differences across countries are diminishing. These changes are associated with changes in the types of alcohol consumed. In Europe, regional differences in dominant beverage types remain, with more wine drinkers in France and Switzerland, and more beer drinkers in central Europe, but Nordic countries are no longer predominantly spirit-drinking (Mäkelä et al., 2005). WHO reports indicate that geographical differences in the types of alcohol consumed are diminishing (WHO Global status report on alcohol and health, 2011). In addition, although HED was traditionally more common in northern Europe, drinking habits are becoming more homogeneous across European countries. The social stigma about HED in countries like France, Italy, and Spain seems to be eroding, at least in the younger generations (Burki, 2010) Trends in different age groups 45. While overall trends for adults appear rather stable over time, differences emerge when focusing on specific age groups. Rates of alcohol drinkers have increased in younger age groups in France and Germany, and in older groups in the United States (men only) and New Zealand. Among alcohol users, rates of hazardous drinking increased in younger age groups in England, Switzerland (men) and United States (men). In Ireland and Germany, hazardous drinking remained stable among young men and women, but declined in other age groups. Finally, an increasing number of young adults were found to engage in HED at least on a weekly basis. Rates of HED in young drinkers have been increasing in several countries, including France (men), Switzerland (men), New Zealand (women), Canada, Germany, Italy, and United States, although rates of HED remain relatively low in these countries. Trends in HED in France and Germany are shown in Figure 5. Figure 5. Trends in HED by age group, France and Germany Probabilities of HED in men in France Probabilities of HED in women in Germany 10% 10% 8% 8% 6% 6% 4% 4% 2% 2% 0% % Note: Probabilities are adjusted for age, ethnicity, marital status, smoking status, occupation status, education level, and socioeconomic status. Sources: France: Enquête Santé et Protection Sociale ; Germany: Epidemiological Survey on Substance Abuse , OECD analyses. 2 Findings for Ireland must be interpreted with caution since survey methods changed over the period covered in the analysis (from postal questionnaires in 2002 to face-to-face interviews in 2007) (Morgan et al, 2008). 18

19 46. OECD analyses of longitudinal Canadian data confirm the above finding, showing increasing cohort effects in hazardous drinking. Hazardous drinking was found to increase with age within each cohort (except in the elderly), with younger cohorts displaying higher rates of hazardous drinking, especially in women. BOX 1. Alternative sources of alcohol consumption data in the United States Some countries have different national survey data. In the US for example, in addition to the NHANES data, information on alcohol drinking is available from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) (adults aged 18 and over), and from the National Survey on Drugs Use and Health (NSDUH) (with a focus on younger ages). Although it would be more suitable to use the NESARC for the purpose of this study on adult drinking, this survey only includes two data points, which is not sufficient for a meaningful assessment of trends. A 2010 report examines changes in alcohol drinking between the two NESARC waves ( and ) (NIAAA, 2010). The report highlights that among current non-drinkers in , 76% remain non-drinkers in , 17% are current drinkers, and 7% are current harmful drinkers (i.e. drinking above the daily or weekly limits 3 ). These proportions vary by age, namely the share of new harmful drinkers is 22% in the age group 18-24, 10% in the age group 25-44, 4% in the age group 45-64, and 1% in the age group 65 or above. Similarly, among current non- harmful drinkers in , 17% become non-drinkers in , 61% remain current non- harmful drinkers, and 22% are harmful drinkers. The proportion of harmful drinkers varies by age: 47% in the age group 18-24, 27% in ages 25-44, 17% in ages 45-64, 8% in ages 65 or above. These findings suggest that younger people are more affected by harmful drinking behaviours compared to other age groups. A comparison is shown below between rates of hazardous and heavy episodic drinking cased on NSDUH and NHANES data for the period Figure A shows overall (population) rates in men and women; Figures B and C show trends in HED in men and women by age group. The NHANES data show a clear increase in HED in the age group 20-24, while NSDUH data show a significantly milder, or no increase. Figure A. Comparison of age-standardised rates from alternative data sources, age % 25% 20% Men Hazardous drinking (NSDUH) Hazardous drinking (NHANES) 30% 25% 20% Women Hazardous drinking (NSDUH) Hazardous drinking (NHANES) 15% 10% HED (NSDUH) 15% 10% HED (NSDUH) 5% 0% HED (NHANES) 5% Note: NHANES and NSDUH data, % HED (NHANES) 3 Daily limit is 4 drinks per day for women and 5 for men. Weekly limit is 14 drinks per week for women and 21 for men. 19

20 Figure B. Regression-based estimates from NHANES HED rates in men in the USA HED rates in women in the USA 25% 25% 20% 15% 10% 5% 0% Note: NHANES data Multivariate regression estimates (adjusted for ethnicity, marital status, smoking status, education, socioeconomic status, and a quadratic term for the survey year) 20% 15% 10% 5% 0% Figure C. Regression-based estimates from NSDUH 25% HED rates in men in the USA (NSDUH) 25% HED rates in women in the USA (NSDUH) 20% 15% 10% 5% 0% Note: NSDUH data Multivariate regression estimates (adjusted for ethnicity, marital status, smoking status, education, socioeconomic status, and a quadratic term for the survey year) How to explain changes in alcohol drinking? % 15% 10% 5% 0% Several reasons for these increases in alcohol drinking among young people are plausible, including: the low cost of alcoholic beverages, the wider availability and accessibility of alcohol, the use of alcohol products (e.g. alcopops ) designed for younger drinkers, and changing social norms. 48. How much young people can spend on alcohol is an important determinant of consumption. Evidence from Finland shows that the rise in drinking among adolescents can be explained with an increased alcohol availability and an increase in money available to young people for their leisure time activities, including alcohol consumption (e.g. Finland, as discussed by Rahkonen and Ahlström, 1989). In addition, alcohol products have become more affordable in many countries, and this is notably true for younger consumers (Rabinovich et al., 2009). 49. Burki (2010) discussed possible causes for the increases in average consumption and emphasised the correlation between consumption and affordability (and availability) of alcohol, citing a report by the British Medical Association showing an increased affordability of alcohol in the UK between 1980 and 2006, and a parallel increase in per capita alcohol consumption of 1.5 litres during the same period of time. 20

21 50. The trend towards increased consumption in young people seems to be consistent with increased commercial pressures and the formulation of alcohol products expressly designed to appeal to younger drinkers. Although the emergence of these new products may not be directly correlated to increased consumption, it may contribute to changing attitudes to alcohol drinking in young people and may lead to an earlier drinking initiation. 51. Heavy drinking in young adults is a way of socializing. Drinking alcohol is often seen as the standard way of fitting in with peers. For the younger generations surrounded by alcohol offers (e.g. multiple delivery points, price offers, new products), the pursuit of excessive alcohol consumption has become part of a normal experience of young adulthood (Seaman and Ikegwuonu, 2010). Besides, younger generations have different motivations for drinking and different patterns of drinking compared to older generations, such as abusing of spirits for the purpose of getting drunk rapidly in order to unwind and have fun (Pabst et al, 2010). 52. Another tentative explanation for increases in HED in young people may lie in the growing predrinking pattern that consists of consuming cheap alcohol before going out to party (Hughes and Bellis, 2012). Indeed, consumers can buy alcohol at a lower cost in supermarkets than in pubs, bars and nightclubs. This is in line with the observed shift from alcohol consumption from on- to off-premise in many countries (Hughes and Bellis, 2012). 3. Social disparities in drinking patterns 53. Patterns of drinking may differ by gender, age, ethnicity and socioeconomic status. The international literature on such variations is vast and findings are not always consistent, largely due to different measures of drinking outcomes being used in the analyses. 54. Numerous international studies find gender differences in drinking. Men are more often drinkers and consume more alcohol than women. Bloomfield et al. (2006) showed gender differences in 13 European countries, as well as in Brazil and Mexico. They found the smallest gender differences in drinking behaviour in northern, followed by western and central European countries, with the largest gender differences in countries with less developed economies. Their results also show that gender differences in drinking behaviour are smaller in countries with greater overall gender equality. Gender differences in drinking are intertwined with educational and socioeconomic gradients in problem drinking patterns (Bloomfield et al., 2006; Gmel et al., 2005; Huerta and Borgonovi, 2010; Grittner et al., 2012a). 55. However, the gender gap tends to narrow as women s drinking behaviours increasingly resemble men s and overall drinking patterns in men and women tend to develop similarly. This trend should be viewed in the context of changes in cultural and social norms, including changes in women s social position in societies as well as new market forces (e.g. market products directed towards women). The following sections focus on the social aspects of drinking, by examining social disparities and trends over time Disparities in alcohol consumption 56. Many studies have looked at the association between socioeconomic condition and patterns of alcohol drinking. However, this association varies with the choice of drinking outcome (e.g. any consumption vs. harmful consumption; frequency vs. quantity; etc.). For instance, there is evidence of a positive association between education and the frequency of consumption but also of a negative association between education and heavy drinking (Bloomfield et al.,2006; Caldwell et al., 2008; Casswell et al., 2003). 21

22 57. In a meta-analysis of 33 countries, part of the GENACIS project, women with more education were found to be more likely to consume alcohol than women with less education. The same was true for men in most countries (with the exception of India) (Grittner et al., 2012). 58. In a number of countries (Switzerland, Germany, France, United Kingdom, Sweden, Finland, Czech Republic and Hungary), both men and women with higher incomes were found to be more likely to consume alcohol (Gmel et al., 2005). Similar findings were reported based on data from Estonia, Latvia and Lithuania (Mc Kee et al. 2000), with women in the highest income group being more likely to consume alcohol compared to women in the middle or lowest income group. 59. OECD estimates show a similar distribution of drinking status: for both genders, adults with higher education and higher socioeconomic status (SES) are more likely to consume alcohol in the past 12 months than their lower-education and lower-ses counterparts. This is observed in all countries, although the educational gradient is not significant among Finnish adults and the socioeconomic gradient is not significant for Swiss men. Figure 6 illustrates differences in any alcohol drinking rates by education level. Figure 6. Adjusted probabilities for any alcohol drinking by education level 100% Adjusted probability for drinking status in men by education level 100% Adjusted probability for drinking status in women by education level 80% 80% 60% 60% 40% 40% 20% 20% Low Medium High Low Medium High Note: Rates correspond to adjusted probabilities of having consumed any alcohol in the past 12 months in people aged 25 and over. (*) people aged 16 and over in Czech Republic, Slovak Republic and Slovenia. Source: OECD estimates on national survey data, most recent years (see Annex 2) Hazardous and heavy episodic drinking by education level and socioeconomic status 60. The association between education level and high-risk alcohol drinking patterns varies by gender. Huerta and Borgonovi (2010) found a positive association between educational qualifications and problemuse of alcohol in England and highlighted important gender differences. They found a strong positive link between educational attainment and frequency of alcohol consumption as well as life-time drinking problems in women, while they found no such association in men. They also showed that poor school performance in childhood is linked with future alcohol abuse in both men and women, but to a lesser degree in men. 61. In many countries, less educated men are more likely to be hazardous drinkers (Bloomfield et al. 2006); Gmel et al. 2005). In a smaller number of countries, men with low and average education were more likely to be binge drinkers than more educated men (Bloomfield et al., 2006). For instance, men with higher levels of education in the United States reported fewer days with 5 or more drinks consumed (Kerr et al., 2008). In the Baltic countries (Latvia, Lithuania and Estonia) and Finland, education was found to be negatively associated with alcohol abuse in men, and positively associated with heavy, but not binge, 22

23 drinking in women (Helasoja et al., 2007). In the Czech Republic, men, single persons and people with low education were found to be at higher risk of hazardous alcohol drinking (Dzurova et al., 2010). 62. Existing studies show that women with more education were more likely to be hazardous drinkers in Austria, France, Germany, the Netherlands, Norway and Switzerland, (Bloomfield et al ; Gmel et al ), while a gradient in the opposite direction was found in the Czech Republic, Finland, Hungary, and Sweden (Gmel et al ). 63. According to a meta-analysis, women with more education were more likely to engage in HED in lower income countries, while the opposite was true in higher income countries (Grittner et al., 2012a). In the same study, education was negatively correlated with HED for men. Reasons for these patterns are entrenched in social norms. 64. OECD analyses confirm these findings. The degree of inequality in high-risk drinking behaviours can be assessed using the concentration index (CI). This index indicates to what extent high-risk consumption is concentrated among people at the top or bottom of the socioeconomic ladder. For men, the pattern of disparity in hazardous drinking is mixed (Figure 7). Among 14 countries studied, England, Finland, and Germany show significantly higher hazardous drinking rates in the more-educated, while the reverse is true in ten countries -with the largest degree of inequalities in Chile, Hungary, Japan, Korea, New Zealand, Spain, and the United States. For women, lower socioeconomic groups are less likely to engage in hazardous drinking in 11 of 14 countries. Larger disparities are found in Canada, Finland, France, Germany, Ireland, and Switzerland, where more-educated women are significantly more likely to be hazardous drinkers. 65. Although the educational gradient in HED is generally mild for both men and women, different patterns of social disparity in HED seem to emerge in certain countries (Figure 8). Less educated men are more likely to engage in HED in most of the countries studied, especially in Chile, Portugal and the US, whereas more-educated women are more at risk of HED in ten out of 17 countries. 66. The distribution of adjusted rates of hazardous and heavy episodic drinking across education levels may be for some purposes- more informative than concentration indexes. These findings are presented in Annex 3B. Figure 7. Concentration index for hazardous drinking by education level, men (left) and women Chile Korea USA Hungary Japan New Zealand Spain Australia Switzerland Ireland France Canada Finland England Germany Korea New Zealand Japan USA Spain Australia Hungary England Switzerland report heavy drink Ireland Canada Germany Finland France People with more education more likely to drink at risk People with less education more likely to drink at risk People with more education more likely to drink at risk People with less education more likely to drink at risk Source: OECD estimates on national survey data, most recent years (see Annex 2). 23

24 Figure 8. Concentration index for HED by education level, men (left) and women Chile USA Portugal Switzerland Slovenia* New Zealand Ireland Australia Hungary Canada Germany Slovak republic* Japan Italy France Czech republic* England Portugal Chile USA Hungary Germany France Switzerland Slovenia* Ireland Italy Japan New Zealand Canada England Slovak republic* Australia Czech republic* People with more education more likely to drink at risk People with less education more likely to drink at risk The better-off more likely to drink at risk The worse-off more likely to drink at risk Note: (*) people aged 16 and over in Czech Republic, Slovak Republic and Slovenia. Source: OECD estimates on national survey data, most recent years (see Annex 2). 67. In the United States, HED was shown to be more prevalent in higher-income groups, but people with lower incomes engaging in HED did so more often and consumed, on average, a higher number of drinks (Kanny et al., 2011). 68. Young people with higher incomes tend to drink more often, and this feature is persistent over time (Casswell et al., 2003). Some studies found that people with a less privileged socioeconomic background drink more (Kuntsche et al, 2004; Leigh, 1996; Mossakowski, 2008); others found the opposite (Grossman et al, 1995; Maggs et al., 2008; NHS, 2008; McKee et al., 2000). Some of the inconsistencies between findings may be explained by the wide variety of definitions of drinking used (e.g. weekly or monthly frequency of heavy drinking) and by the variety of definitions of socioeconomic background used (e.g. income, occupation, employment status). In addition, the relationship between socioeconomic status and drinking behaviours is complex and difficult to study, as it may be influenced by a large number of factors to do, among other things, with alcohol price levels and individual sensitivity to prices and price changes. 69. In line with findings on education-based disparities, OECD analyses show a mixed pattern of SES-related disparities in hazardous drinking for men, and a clearer pattern for women. Men with higher- SES are significantly more likely to be hazardous drinkers than those with lower-ses in 6 of 15 countries (Australia, Canada, England, Finland, Germany, and Ireland), while women with higher-ses are more likely to be hazardous drinkers than lower-ses women in 12 of 14 countries. 70. The SES-related gradient in HED is mild in both men and women, and no clear pattern emerges in the set of countries studied. In men, SES-related disparities are stronger in Chile, Hungary, Portugal, and United States, where men with lower SES are more likely to engage in HED. In women, a reverse gradient is observed: women in higher SES are more at risk of HED in 8 out of 13 countries, with larger inequalities in Australia, England, and France. 24

25 3.3. Time-trends in social inequalities 71. The OECD analysis of trends over time revealed no major changes in education and SES gradients in (any) alcohol drinking, but with exceptions for Finland and Switzerland. Finland displays an increasing rate of any alcohol drinkers in the less educated and the lower-ses women. Switzerland shows a decreasing rate among the less educated. 72. Trends in social gradients in hazardous drinking showed: narrowing social disparities in six countries with decreasing rates of hazardous drinkers among the more educated and/or better off (England, Finland Germany, Ireland, Switzerland, and New Zealand). In contrast, increasing social disparities are observed among women in England with rates of hazardous drinkers growing among the more educated. 73. No major trends were found in disparities in HED, but again with some exceptions. Ireland, which has the highest proportions of binge drinkers, displayed a remarkable trend of narrowing social disparities. In contrast, social disparities have been increasing in Germany, New Zealand, and the United States (men only), with the worse-off increasingly at risk over time compared to other SES groups (although education-related inequalities in HED narrowed in Germany at the same time). 4. Disparities by ethnicity 74. Some studies suggest evidence of disparities in problem drinking by ethnic group in Australia, the UK, and the US. In Australia, evidence suggests that rates of high-risk consumption of alcohol among Indigenous Australians are generally twice those in the non-indigenous population (Gray and Wilkes, 2010). In the UK, people from white backgrounds have lower rates of abstinence and higher drinking levels compared to most minority ethnic groups, although drinking patterns vary both between and within minority ethnic groups (Hurcombe et al, 2010). Likewise, in the United States, non-hispanic Whites display a higher prevalence of HED compared to other ethnic groups (Kanny et al., 2011). 75. An analysis of the relationship between ethnic status and alcohol drinking in England, New Zealand, and USA shows that minority groups drink less than the white population, except in the US where Mexican American men have a higher prevalence of any alcohol drinking than non-hispanic white men. Concerning problem drinking, results show that in England, white men and women are more likely to engage in hazardous drinking (cf. Figure 9), while findings in other countries are seldom statistically significant. Figure 9. Odds ratios for hazardous drinking by ethnic status, England White (Ref.) Men - hazardous drinking Indian & Pakistani & Bangladeshi Black Women - hazardous drinking Mixed & Others (including Chinese) 25

26 Source: OECD estimates on national survey data, most recent years (see Annex 2). 76. These differences may be associated with strong ethnic identity, strong family and local community ties, continuing links with the host country and maintaining religious values (Hurcombe et al, 2010). Some minority ethnic groups may have strong religious ties that forbid alcohol drinking. Acculturation is a complex process, broadly understood as the adoption of cultural traditions and values of the host society by immigrant groups. In the US, higher acculturation among Hispanic is associated with a greater risk of alcohol abuse as well as HED in women (Chartier and Caetano, 2010). 77. In addition, genetic factors may influence alcohol drinking patterns, and certain genes have protective effects on the risk of alcoholism (Edenberg, 2007). For example, in Asian population, some people tend to drink less since genetic predispositions create unpleasant effects like facial flushing and other aversive symptoms. 26

27 SECTION II THE IMPACTS OF ALCOHOL CONSUMPTION ON LABOUR MARKETS AND PRODUCTIVITY 2.1. Alcohol consumption and labour market outcomes Alcohol and employment 78. The impact of alcohol consumption on employment, as well as on other labour market outcomes discussed in the following sections, depends on the quantity consumed and pattern of consumption. In Sweden, long-term light drinkers were shown to have better employment opportunities than any other group, including former drinkers, former abstainers, long-term heavy drinkers and abstainers (Jarl and Gerdtham, 2012). On the other hand, heavy drinking was shown to reduce the probability of being in employment for both men and women (Booth and Feng, 2002; MacDonald and Shields, 2004; Johansson et al., 2007), although a number of studies found no significant relationship between alcohol abuse and employment (Feng et al, 2001; Asgeirsdottir and McGeary, 2009). 79. The effects of problem drinking on employment appear to vary over the life cycle. Some evidence suggests that alcohol-dependent people above age 60 are more likely to be in employment than their non-dependent counterpart while this relationship is not true in younger age groups (Mullahy and Sindelar, 1993; Johansson et al., 2007). 80. The relationship between problem alcohol drinking and employment is complex, as lack of employment in turn may be a cause of alcohol problems. An increase in unemployment was found to be associated with higher suicide rates for people below age 65, and with a higher alcohol-related mortality (Stuckler et al., 2009). Working conditions, such as long working hours and job insecurity, have been linked with an increased likelihood of high-risk alcohol consumption (Marchand et al., 2011). 81. Heavy alcohol consumption during early adulthood seems to impact on people s lives in the long run, affecting employment opportunities. A US study found that those who drank regularly during early adulthood had lower occupational prestige 15 years later, than occasional drinkers, the finding was particularly strong among African-Americans (Sloan et al., 2009). 82. Some types of jobs may be affected by problem alcohol drinking more than others. For instance, in England and Wales, higher alcohol-related mortality was found among workers of the broader alcohol industry, including bar managers and staff (Romeri et al., 2007) Alcohol and wages 83. There is some evidence that moderate drinking is positively associated with wages, with the underlying cause being that moderate drinkers have a better health and better job performance than heavy drinkers and abstainers (Hamilton and Hamilton, 1997; Barret, 2002; Peters, 2004; Lee, 2003; MacDonald and Shields, 2001). Moderate drinkers spend more time with their colleagues out of work and they tend to be in good health, which influences positively their wages. They have a higher degree of life satisfaction 27

28 than abstainers and have stronger social networks. Social and networking skills are important factors in the labour market and determine wages to a high degree. A positive impact of moderate drinking on wages was also found in Germany, increasing in size and significance with age, starting from age 35 (Ziebarth and Grabka, 2009). 84. However, Lye and Hirschberg (2010) reviewed evidence of the labour market outcomes of alcohol consumption and found older studies indicating that alcohol consumption boosted wages (e.g. above studies) may be biased by the omission of relevant variables and inadequate accounting for selection and reverse causality. 85. In contrast, studies found heavy drinking to be associated with lower earnings (Mullahy and Sindelar, 1991; 1993). Problem drinking has a negative impact on both earnings and productivity, and it impairs people differently depending on their gender and personal traits associated with the propensity to be heavy drinkers (Jones and Richmond, 2006). In a study by Renna (2008), alcoholism (defined on the basis of DSMMD criteria), but not alcohol abuse, was found to be linked with reduced earnings, and the mechanism identified was reduced working hours, rather than a lower wage rate. 86. Some research examined gender differences in the relationship between alcohol abuse and wages, and found this to be stronger in men and weaker, or absent, in women (Soydemir and Bastida, 2006; Van Ours, 2004). These findings were deemed to reflect differences in the propensity to risk taking as a common determinant of both drinking behaviour and professional success Alcohol and labour productivity Absenteeism 87. Alcohol abuse has acute and chronic health consequences such as injuries from accidents, psychiatric and somatic diseases, all of which are likely to increase short- and long-term sickness absences. In a number of countries, there is evidence for a link between high-risk alcohol consumption and sickness absence among men, though not in women (Roche et al., 2008; Norström and Moan, 2009; Norström, 2006; Johansson et al., 2008). 88. A curvilinear relation between alcohol intake and sickness absence was observed in Finland, where medically certified absences were 20% higher among lifetime abstainers, former and heavy drinkers compared with light drinkers (Vahtera et al., 2002). Additional evidence for the U-shaped relationship between alcohol consumption and sickness absence is found in the literature (Salonsalmi et al., 2009; Jarl and Gerdtham, 2012). 89. A more pronounced association between quantity of alcohol consumed and sickness absence was observed in men with low levels of education (Johansson et al., 2008), while another study suggested that perceived co-worker support may attenuate the relationship between alcohol abuse and absenteeism (Bacharach et al., 2010). Early retirement 90. In addition to increasing the risk of sickness absence, alcohol abuse may also be a cause of early retirement on disability grounds, with receipt of disability benefits, in middle-aged workers (Upmark et al., 1999). Problem drinking was found to increase the risk of being the beneficiary of a disability pension in Finland, Norway, and Sweden (Salonsalmi et al., 2012; Skogen et al., 2011; Sidorchuk et al., 2012) although there is also evidence for a link between abstinence and receiving a disability pension (Upmark et al., 1999; Skogen et al., 2011; Salonsalmi et al., 2012). 28

29 Presenteeism 91. Worker productivity is not only affected by time away from work or early retirement, but also by a reduced performance at the workplace, known as presenteeism. The latter is far more difficult to assess than absenteeism, and there is no agreement on how it should be measured. Estimates are generally based on the costs associated with reduced output, increased number of errors, and failure to meet production targets (Schultz et al., 2009). Alcohol abuse is a cause of presenteeism. The after-effects of heavy drinking include inability to concentrate at work, deterioration in job performance and relationships with colleagues, higher safety risks and reduced overall output. These in turn can lead to disputes, grievances, loss of working time and reduced productivity (ILO, 2012) The value of lost productivity 92. A review of 22 studies from different countries observed a substantial economic burden of alcohol on society (Thavorncharoensap et al., 2009). In the UK, nearly 11 million working days were lost by alcohol-dependent workers in 2001, and the total cost of absenteeism due to alcohol was estimated to be 1.2 billion (UK Cabinet Office, 2003). In the European Union alcohol accounted for an estimated 59 billion worth of lost productivity through absenteeism, unemployment and lost working years through premature death in 2002 (Anderson and Baumberg, 2006). Productivity losses were found to be an important part of alcohol-related costs in Scotland, France and Canada (Rehm et al., 2009). In Ireland, the value of output lost due to alcohol-related absences from work was of 330 million (9% of the total costs of absences from work) and the cost of alcohol related accidents at work of 197 million (5% of total work-related accident costs), in The total cost of alcohol ( 3.7 billion) represented 1.9% of GDP in Ireland in 2007 (Byrne, 2010). 93. Lost labour earnings were found to account for the largest part of the economic costs associated with alcohol abuse in the Unites States (Harwood, 2000). In 2006, lost productivity represented 72.2% of the total economic cost of excessive drinking. The bulk of the value of lost productivity was attributable to impaired productivity at work (46%) and premature mortality (40%), while absenteeism accounted for 2.6% of the total value (Bouchery et al., 2011). In Australia, based on data from the 2001 National Drug Strategy Household Survey, the cost of alcohol-related absenteeism was estimated to be $437 million, or 36% of the cost due to all absenteeism (Pidd et al., 2006). 29

30 SECTION III AN ECONOMIC CASE FOR GOVERNMENT INTERVENTION 3.1. Policy objectives and rationale for government intervention 94. Governments have been trying for a very long time to reduce the negative social impacts of alcohol use, mainly by making alcoholic beverages more expensive through taxation, by limiting their availability, and by controlling their supply through state monopolies. An increased emphasis on the public health consequences of alcohol use in the past few decades has raised the importance of alcohol in policy agendas worldwide and broadened the spectrum of actions countries are prepared to consider and finance to fight the problem. On the other hand, governments have seen both public support and opposition to actions that are seen by some as undue interference with individual choices. Alcohol-related harms are dreaded and stigmatised, but alcohol is highly valued by many consumers as a source of individual pleasure and social enjoyment, and its production and trade represent an important part of the economy in many countries. 95. While the goal of alcohol policies is typically to reduce alcohol-related harms, many policies pursue that goal by deterring all consumption, not just harmful consumption. This is certainly the case with taxation, for instance, although there are suggestions that alternative policies acting through the same price mechanism, i.e. minimum pricing policies, may be better suited for targeting consumers at higher risk. Policies to limit the availability of alcohol are also likely to hit all consumers. However, even these policies can be tailored in ways that will make them more effective at targeting harmful consumption (e.g. by reducing outlet opening hours at times of the day/week when harmful use is more common). The approach of targeting all consumption in alcohol policies is in line with the well established public health principle that shifting the entire distribution of a given risk factor (the population approach) tends to produce larger reductions in those who are most exposed than targeting the latter alone (the individual, or high-risk, approach). 96. When alcohol policies hit all types of consumption, a potentially large number of moderate drinkers (typically, but not exclusively, older men) might see their risk of certain cardiovascular conditions increased because of a reduction in alcohol use (although they will still benefit from that reduction in terms of a wide range of other disease outcomes, such as several types of cancer). OECD analyses of the impacts of alternative policies have made an attempt to quantify the potential loss in those consumers, as shown in Section IV. Governments may wish to consider the size of the tradeoff involved when implementing policies that are likely to affect all types of consumers, although the key measure of impact they are most likely to be looking at is the overall impact of a policy. The decision of where to strike a balance between population and individual approaches to alcohol policy must be left to individual governments, but the work undertaken by the OECD aims at providing detailed estimates of policy impacts at the population level, as well as in specific subgroups, which may contribute to an evidence-based approach to such decisions. 97. From an economic perspective, the justification for government action to reduce alcohol consumption by individuals depends crucially on whether such consumption is simply the outcome of efficient market dynamics, or the effect of market and rationality failures preventing individuals and society from achieving more desirable outcomes. Where failures exist, prevention policies may provide the 30

31 means to increase social welfare. As discussed in the OECD conceptual framework on the economics of prevention (Sassi & Hurst, 2008), several such failures would appear to exist in alcohol consumption. 98. Among the most important failures is the impact of drinking on individuals other than drinkers ( externalities ). Examples include the victims of accidents caused by drivers under the influence of alcohol, or the victims of domestic violence fuelled by alcohol, but also the victims of the social externalities of alcohol, i.e. those who may take up hazardous drinking behaviours following the example provided by their parents or peers. Another classical area of market failure that applies to alcohol is imperfect information by consumers on the effects of consumption. Beyond a generic perception that drinking large quantities of alcohol is bad for one s health, consumers have a very imprecise knowledge of the exact modalities in which alcohol poses a risk to humans, partly because there are still many uncertainties in the underlying science. 99. Economists have studied alcohol consumption as a rational addiction, due to the strong correlation ( adjacent complementarity ) between present, past and future consumption. Although rational addictions are not incompatible with utility maximisation by consumers, and therefore with operating in a market context, there are several reasons for viewing the addictive nature of alcohol consumption as contributing to the justification for government intervention. First, addiction to harmful (for oneself and for others) consumption is arguably a justification for external intervention. In addition, rational addictions are compatible with utility maximisation in the presence of perfect information, which is unlikely to be available to alcohol consumers (as discussed above). Finally, myopic (present-oriented) consumers are likely to display more strongly addictive consumption, as well as inconsistent time preferences (wanting to change their consumption, but constantly deferring) Types of policies 100. The harmful effects of alcohol consumption can be reduced though effective prevention strategies. In 1979, the World Health Assembly called upon WHO member states to develop and adopt appropriate legislation and measures to tackle alcohol misuse (WHO, 1979). Such efforts culminated with the endorsement, in 2010, of the global strategy on the harmful use of alcohol (WHO, 2011) that supports ten target areas for national actions including: health sector response, community actions, drink-driving policies, limitation of the availability of alcohol, action on the marketing and pricing policies, reducing the negative consequences of intoxication and reducing the public health effect of illegally and informally produced alcohol. Policy makers and researchers alike agree that tackling harmful drinking requires a range of policies from the different areas identified in the Global Strategy The OECD conceptual framework on the economics of prevention classifies prevention policies in four main groups, characterised increasing degrees of interference with individual lifestyle choices, but also with decreasing cost of implementation. The four groups include: (a) interventions aimed at expanding individual choice by making available new options or lowering the price of existing options; (b) interventions aimed at influencing individual behaviour through means other than prices; (c) interventions which restrict choice by increasing the price of certain options; and (d) interventions which restrict choice by banning or mandating certain options. A review of policies in place in OECD countries shows that no systematic actions are being taken by governments which could be classified in the first group (a). There are commercial initiatives aimed at reducing the alcohol content of certain beverages, and there is a widespread use of price promotions by alcohol manufacturers and retailers, but these are generally not aimed at the pursuit of public health goals. Government actions in other groups are described below. A summary of the main features of common policies adopted by OECD and key partner countries is available in a comprehensive table in Annex 1. 31

32 Interventions aimed at influencing individual behaviour through means other than prices 102. Information and education. Campaigns to raise awareness of the risks of excessive drinking are common in OECD countries as a way to encourage responsible alcohol consumption. Information and education may be delivered by different means, covering different population subgroups. School-based interventions target children while worksite interventions target working-age adults. Reviews of existing evidence suggest that such campaigns do increase knowledge about the risks of harmful alcohol consumption (Anderson et al., 2012), but the change in drinking behaviours tends to be limited and short in duration. However, information and education campaigns can contribute to a package of measures to tackle harmful alcohol generating possible synergies with other measures. The delivery of education messages by private sponsors was found to have no significant public health effects (Christie et al., 2001; Smith et al., 2006; Barry & Goodson, 2010) Community actions. These include a heterogeneous set of interventions characterized by a bottom-up approach. Most often, community actions share many features of education and information policies. However, the former also include changes in the wider environment, and often also aim at moving alcohol up in the political agenda. The term community does not always mean that interventions are implemented at a neighbourhood or local level. For instance, workplace programmes and media advocacy both fall into this category. Community actions may usefully support broader alcohol prevention programmes (Anderson, 2012) Health sector response. In most countries, treatments are available within the health sector for those with high levels of alcohol consumption or alcohol dependence problems (WHO, 2010). Two types of intervention are especially common. Brief advice delivered by a trained physician aims at informing the patient about the harms caused by excessive alcohol consumption and providing help in regaining control towards an acceptable use. For patients with more severe problems, such as dependence, interventions usually combine cognitive-behavioural therapies with pharmacological treatment. The main barrier to a greater effectiveness of this approach is represented by difficulties in reaching the target population, which limits overall coverage. A WHO-Europe (2009) review concludes that less than 10% of the population at risk for harmful alcohol consumption would be routinely identified, and less than half of those diagnosed would be offered advice Regulation of marketing. Countries use a range of polices to regulate the marketing of alcohol but, in general, these are largely based on self-regulation and co-regulation, which devolves responsibility for setting boundaries for marketing to the alcohol industry. However, existing evidence suggests that the effects of self-regulation of marketing are disappointing (WHO, 2007; Anderson et al., 2009). This is often due to poor participation and compliance, and to numerous breaches of self-imposed standards (Jones et al., 2008; KPMG LLP, 2008). When advertising bans are implemented, they are more likely to be partial, for example for certain products (e.g. only beverages with a higher alcohol content) or certain media (e.g. television) or during specific hours. Partial bans tend to have only limited effects on overall alcohol consumption, as advertising expenditure is simply shifted onto other media or targets. The increasing use of media that reach across national borders (e.g. internet) and social networks for marketing alcohol suggests the need to consider parallel cross-border regulation of advertising (WHO, 2011). Interventions which restrict choice by increasing the price of certain options 106. Pricing policies. Alcohol prices can be altered by using taxes or direct price controls, including minimum unit prices in order to change consumption. The most common approach is based on a combination of excise duties and value added taxes. Excise duties are applied to alcoholic beverages in two main ways. which in some cases may be combined. The charge may be specific to the product (e.g. bottle of beverage), or its alcoholic content (e.g. percentage of alcohol in the drink), or calculated as a proportion 32

33 of the value of the product (ad valorem excise). The effects of different taxes may vary substantially (e.g. Keen, 1998). A concise summary of VAT and excise tax rates across OECD countries as of January 2009 (in some cases approximations were required because of the complexity of certain taxation systems) is provided in Annex 1. Potential substitution effects must be accounted for in designing taxation policies. For instance, the introduction of a tax on alcopops in Germany simply shifted consumption from spiritbased to beer-based beverages (Anderson et al., 2012b). Minimum pricing policies have also been in the spotlight in recent years. Time-series analyses in Canada (British Columbia) showed that a 10% increase in minimum price for alcohol reduced consumption of spirits by 6.8%, wine by 8.9%, alcopops by 13.9% and beer by 1.5% (Stockwell et al., 2012). Interventions which restrict choice by banning or mandating certain options 107. Regulating alcohol availability. Virtually all countries set a minimum legal age for the purchase of alcoholic beverages. Most of the countries in table in Annex 1 use the same minimum age threshold for both on-premise (e.g. bars, restaurants) and off-premise (e.g. retail) sales while in few cases thresholds differ and, normally, are higher for on-premise sales. A second approach to limit alcohol availability is through controls on outlet density, either through State monopolies or commercial regulation. A further approach involves restrictions on licensing and outlet opening hours. Regulation of alcohol availability has the potential to produce significant effects on alcohol consumption (Stockwell & Chikritzhs, 2009; Livingston et al., 2007; Gruenewald, 2007) and health outcomes. However, a stringent policy on alcohol availability should be always coupled with effective enforcement, as informal market activities are likely to develop as a side effect Drink-driving policies. A high blood alcohol concentration is a major risk factor for traffic accidents, injuries and death. Most countries have policies in place to prevent the utilization of motor vehicles under the influence of alcohol. A number of countries have different, usually more restrictive, thresholds for professional drivers and young people as they are at higher risk of traffic accidents. Enforcement is usually through check-points and random breath testing. Consequences for drivers identified with levels of alcohol beyond the tolerated threshold vary from fines to suspension or revocation of the driving licence to compulsory treatments for drivers with alcohol dependence. A strict enforcement strategy is key to effective drink-driving policies The private sector response: how are markets adjusting to the new challenges? 109. As part of its previous work on the economics of prevention, the OECD invoked the adoption of a multi-stakeholder approach as the most sensible way forward in the prevention of chronic diseases linked to lifestyles. While it must be recognised that the interests at stake are often in conflict with each other, it must also be acknowledged that no party is in a position to produce meaningful changes in healthrelated behaviours and associated chronic diseases without co-operation from other stakeholders An open dialogue and cooperation with alcohol manufacturers, as well as major retailers, may be part of an effective policy approach in fighting the harms associated with alcohol consumption, although so far private sector actions have not been as supportive as they might be. Examples of practices which have gone in the opposite direction to the thrust of public policy include the widespread use of price promotions and the adoption of ever more sophisticated marketing techniques, and media. There have been instances of product reformulation aimed at reducing the alcohol content of certain beverages, although these were generally triggered by economic incentives (e.g. to prevent the application of higher rate duties, as it happened in 2012 for many beers sold in the United Kingdom Moylan, 2012). The effects of advertising self-regulation initiatives taken by the industry have also been disappointing (WHO, 2007; Anderson et al., 2009), mainly due to poor participation and compliance (Bergamini et al., 2013). 33

34 SECTION IV WHAT POLICY OPTIONS? ARE THEY COST-EFFECTIVE? 4.1. Assessing the impacts of policy options: the need for a population-wide modelling approach 111. A better understanding of which measures or strategies represent the best use of society's resources, and by how much they can reduce the harmful consequences of alcohol use, is key to an evidence-based approach to alcohol policy making. The scarcity of adequate decision support systems in this area is an important barrier to an efficient allocation of resources and often prevents national and international policy bodies from considering the full range of costs and consequences of different actions. Policy makers who face increasingly tight budget constraints, and need to prioritise their actions, must know what policy options are likely to generate the largest health and social returns. Potentially valuable investments in prevention are forgone because of uncertainty and ignorance about the likely returns on those investments and the time frame in which they will materialize An empirical assessment of the impacts of prevention strategies through appropriately designed experimental studies is virtually impossible in most instances. Causal inference based on observational data is also difficult, because of the heterogeneous information on possible interventions contained in such studies and the variety of concurrent factors that counfound the possible effects of policies. A data-driven approach that captures the mid- and long-term impacts of prevention strategies, quantifying ranges of uncertainty around them, and isolating these impacts from the effects of potential confounding factors, while at the same time accounting for the heterogeneity of actors, exposures and behaviors, are key steps to an accurate and thorough assessment of the value of prevention An individual-based, population-wide simulation modelling approach can provide a sound basis for evaluating the relative effectiveness and cost-effectiveness of a range of alcohol prevention and control strategies, overcoming the limitations of alternative approaches discussed in the previous paragraph. A comprehensive alcohol policy simulation platform is required to generate consistent estimates of resource inputs, costs and outcomes. Case-based microsimulation approaches, in particular, offer the best opportunities to model realistic individual life trajectories, taking into consideration the heterogeneity in populations and in individual behaviours that can influence the harm caused by drinking as well as the relative effectiveness of policies across population groups. Such models capture the complex set of interrelationships between the prior history and current use of, and demand for, alcohol and the resulting distribution of health, social and economic consequences The international dimension of the OECD alcohol project offers an unparalleled opportunity to develop a modelling platform capable of addressing a broader range of individual, social, environmental and policy factors than any national-level analysis could account for. As in the case of the analyses of policies to tackle obesity, unhealthy diets and physical inactivity, previously undertaken by the OECD (e.g. Sassi, 2010; Cecchini et al., 2010), a comparison of policy impacts across countries with different characteristics may provide unique insights on what dimensions are especially likely to determine what returns should be expected from the implementation of those policies. This is particularly important in the case of alcohol policies, because different historical, social and economic factors have led to considerable variability across countries in the treatment of alcohol and the policy approaches used to control alcoholrelated harm. 34

35 4.2. The CDP-Alcohol model and the rationale for the selection of policy options assessed 115. The OECD developed a computer-based simulation model called CDP-Alcohol, with the aim of undertaking a comprehensive assessment of the health and economic impacts of a range of alcohol policy options in selected countries, and providing governments with a tool that would enable them to undertake similar assessments at the national level. The CDP-Alcohol model builds on previous modelling efforts in addressing key technical challenges, but has some unique features compared to other alcohol policy models. In its design, special emphasis was placed on ensuring the relevance of the model to an international endeavour, such as its flexibility and adaptability to alternative demographic, epidemiologic and policy contexts and to different national patterns of alcohol use. On the other hand, the current version of the model is not designed to capture some of the impacts of alcohol policies (e.g. crime impacts), or to distinguish between types of alcoholic beverages or site of consumption (on trade, off trade), in line with advice from the OECD Expert Group on prevention and Health Committee The chief aim of the CDP-Alcohol model is to provide comparable measures of impact for a wide range of alcohol policies in given settings. The WHO Global Strategy to reduce the harmful use of alcohol provides a menu of policy options based on international consensus, which OECD used as a starting point in identifying a set of policies to be assessed using the CDP-Alcohol model. However, it must be noted that not all types of policies lend themselves to an economic analysis based on a quantitative modelling framework. In particular: a. Certain actions discussed in the Global Strategy (e.g. monitoring and surveillance) are important in the context of an overall strategy to fight alcohol-related harms, but have only indirect effects on alcohol consumption and harms, and these effects are very difficult to measure or estimate. b. Other actions have been shown to be effective in reducing alcohol-related harms, but their effectiveness was determined only in qualitative studies, often based on heterogeneous outcomes, hardly suitable for inclusion in a quantitative modelling framework. c. Of the remaining actions, some are supported by stronger evidence of effectiveness than others from existing studies, and a key condition for an economic analysis to be valid and meaningful is that the policies assessed are of proven effectiveness For the above reasons, and because of the need to prioritise analyses due to time and resource constraints, it was agreed that OECD analyses would focus on a subset of the policy options discussed in the Global Strategy and in the current policy debate, without implying that the inclusion of a policy in the set of those assessed corresponds to a form of endorsement of the policy, or otherwise for the policies excluded from the analyses. The OECD Secretariat spent a considerable amount of time reviewing policy options and discussing the scope of its economic analysis with Member Countries and relevant stakeholders, primarily within the context of the Expert Group on prevention and Health Committee. The aim of this work was not to devise a competing policy agenda to that endorsed by WHO, but simply to focus the subsequent analysis on policies for which a meaningful assessment could be made The conclusion reached by the Bureau of the Expert Group on prevention, after examination of a broad range of comments received from country experts and stakeholders (see also Annex 6) is that the policy options outlined below will be assessed in the economic analysis, divided into two subsets to reflect the breadth and strength of the existing evidence in support of their effectiveness (main and further analyses). Details of the evidence reviewed for each policy option and of the characteristics of the options to be assessed are available in Annex 5. 35

36 Main analysis 1. Regulation of advertising of alcoholic beverages Two alternative regulation scenarios leading, respectively, to a 10% and a 25% reduction in expenditure on commercial advertising, accounting for the likely smaller effects of regulation on the share of advertising channelled through the Internet. 2. Taxation of alcoholic beverages Modelled as an increase in existing taxes leading to price increases of 10% for all types of alcoholic beverages. 3. Measures to counter drink-driving Although the value of other approaches is recognised, the analysis will focus on increasing the intensity of enforcement (roadside checkpoints with random breath testing) of existing blood alcohol concentration limits. 4. Brief interventions in primary care Aimed at heavy drinkers who are not alcohol-dependent. Response and compliance rates modelled on the basis of evidence from existing studies. 5. Pharmacological and psychosocial treatments for alcohol dependence Modelled on the basis of the effects reported in existing Cochrane meta-analyses. 6. Measures to limit the availability of alcohol Further analysis Although the value of other approaches is recognised, the analysis will focus on the regulation of outlet opening hours along the lines of regulations tested in Norway and Finland, barring sales for a 24 hours period at the weekend, or involving equivalent changes during weekdays. 7. School-based interventions Modelled on the basis of the School Health and Alcohol Harm Reduction Project (Mc Bride et al., 2000; McBride et al., 2004). 8. Workplace programmes Modelled on the basis of a brief intervention at the workplace in a big Australian postal network (Richmond etal., 2000). 9. Minimum pricing Modelled on the basis of the only existing minimum pricing policies for which evidence of effectiveness is curently available (those in place in the Canadian provinces of British Columbia and Sasktchewan), assuming a further increase in minimum alcohol prices of 10% for Canada, 36

37 and an increase based on current market conditions for other countries where a similar policy is not in place. 10. Further scenarios for policies included in the main analysis, including: A tax increase scenario involving further reduced effects on heavy drinkers, based on findings by Ayyagari et al. (2011) (sensitivity analysis only). 4.3 Design of the OECD CDP-alcohol model 119. The basic structure of the CDP-Alcohol model is illustrated in Figure 10. The model was built to reflect a individual characteristics and outcomes, including demographic and socioeconomic characteristics, patterns and levels of alcohol consumption, disease incidence and disease-specific mortality. A detailed description of the modelling approach is provided in Annex 4. Figure 10. Simplified structure of the CDP alcohol model 120. The model was built to simulate the dynamics of a population starting in the year The population is projected forward for 40 years, until It should be noted that a time horizon of 40 years does not allow simulations to capture the full effects of interventions targeting young children. Appropriate mathematical algorithms were devised to allow the model to simulate individual trajectories of alcohol consumption over time The CDP-alcohol model explicitly accounts for 10 conditions whose incidence is affected by the volume and/or pattern of alcohol consumption. These include: alcohol use disorders; injuries; cirrhosis of the liver; liver cancer; epilepsy; ischaemic heart disease; cerebrovascular diseases; oesophagus cancer; 37

38 mouth and oropharynx cancers and breast cancer, as shown in Figure 11. The contribution of each condition to the alcohol-related burden of disease in developed economies is shown in Table 2. According to the most recent WHO estimates (WHO, 2011) this group of diseases accounts for about 98% of the morbidity (in DALYs) and 93% of the mortality due to alcohol consumption. Levels of disability associated with specific conditions are based on the disability weights originally used in the WHO (2008) global burden of disease study. Figure 11. Diseases and injuries included in the CDP-Alcohol model 38