drink-driving and the alcohol beverage industry: will reducing per capita consumption solve the...

14
Accid. Anal. & Prey. Vol. 19, No. 6, pp. 449-462, 1987 0001-4575/87 $3.00 + .00 Printed in Great Britain. © 1987 Pergamon Journals Ltd. DRINK-DRIVING AND THE ALCOHOL BEVERAGE INDUSTRY: WILL REDUCING PER CAPITA CONSUMPTION SOLVE THE PROBLEM IN THE UNITED KINGDOM? DAVID RILEY Research and Planning Unit, Home Office, 50 Queen Anne's Gate, London SWlH 9AT, England Abstract--The impact of drinking and driving is one focus of the mounting concern in the West over the widespread incidence of alcohol-relatedproblems. Conventionalwisdom, in the United Kingdom as well as in other countries, suggests that reducing average consumption levels will diminish the impact of the negative effects of alcohol includingdrinking and driving. But whether policies designed to achieve changes in per capita consumption by increasing alcohol taxes across the board constitute the most effective strategy to reduce drinking and driving is called into question. A number of competing interventions directed at the alcohol beverage industry are analysed and new directions for producers and policymakers are proposed. Any analysis of the role of alcohol in the incidence of drinking and driving should not neglect the evidence concerning the part played by variations in the consumption of ~,lcohol in the incidence of a wide range of alcohol-related problems. It seems likely that the impetus for change will reflect concern not only about road casualties and fatalities brought about by drinking drivers, but also concern about alcoholism, cirrhosis and so on. The impact of alcohol problems is significant: recently the resource costs of alcohol raisuse in the United Kingdom have been estimated at over £1500 million [McDonnell ~,nd Maynard, 1985]. Moreover, the development of policies for alcohol control has not c,nly a national dimension but is one of the major themes of the World Health Organ- i:;,ation [WHO, 1982]. Indeed, to the extent that attention is focussed exclusively on crinking and driving, the range of possible interventions may be unduly restricted to measures which will impact uniquely on this problem. Such initiatives as increasing law enforcement or the level of sanctions for convicted offenders clearly have a part to play, but need to be set against a range of alternatives which embrace a more comprehensive assessment of possible courses of action. There is strong evidence that the national incidence of alcohol-related problems is c !osely related to average levels of alcohol consumption [Popham, 1970; Brunn, Edwards, aad Lumio et al., 1975]. Kendell [1984], for example, examined the relationship between per capita consumption of alcohol in the U.K. and the incidence of alcohol-related problems since the beginning of the 1970s. Average consumption levels increased from 7 0 litres of ethanol per year in 1970 to 9.8 litres in 1979, then fell by more than 10% to 8.7 litres per year between 1979 and 1982. The decrease was accompanied by significant reductions in first admissions to hospitals for alcohol dependence and in cirrhosis mor- tality. Indeed, over the 1970-1982 period Kendell found significant correlations between per capita consumption and these alcohol-related events. Attention to the shorter-term sequelae of alcohol abuse--including drinking and driving--using aggregate level data is relatively more recent. In the aforementioned report, Kendell refers to the relationship between per capita consumption and the level o! convictions for drunkenness and for drink-driving offences. These showed a similar link with aggregate consumption levels. The reduction in drink-driving convictions, for example, at 7% was almost twice as great as the reduction in cirrhosis mortality (4%). Similar findings were reported by Wagenaar [1984] in a methodologically more sophisticated study of the link between alcohol consumption and motor vehicle crashes. Wagenaar was able to allow for seasonal effects and long-term trends in his data in a time-series analysis, based on a single jurisdiction in the U.S. (Michigan), and found a 449

Upload: david-riley

Post on 15-Jun-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Accid. Anal. & Prey. Vol. 19, No. 6, pp. 449-462, 1987 0001-4575/87 $3.00 + .00 Printed in Great Britain. © 1987 Pergamon Journals Ltd.

DRINK-DRIVING AND THE ALCOHOL B E V E R A G E INDUSTRY: WILL REDUCING PER CAPITA

CONSUMPTION SOLVE THE PROBLEM IN THE UNITED KINGDOM?

DAVID RILEY Research and Planning Unit, Home Office, 50 Queen Anne's Gate,

London SWlH 9AT, England

Abstract--The impact of drinking and driving is one focus of the mounting concern in the West over the widespread incidence of alcohol-related problems. Conventional wisdom, in the United Kingdom as well as in other countries, suggests that reducing average consumption levels will diminish the impact of the negative effects of alcohol including drinking and driving. But whether policies designed to achieve changes in per capita consumption by increasing alcohol taxes across the board constitute the most effective strategy to reduce drinking and driving is called into question. A number of competing interventions directed at the alcohol beverage industry are analysed and new directions for producers and policymakers are proposed.

Any analysis of the role of alcohol in the incidence of drinking and driving should not neglect the evidence concerning the part played by variations in the consumption of ~,lcohol in the incidence of a wide range of alcohol-related problems. It seems likely that the impetus for change will reflect concern not only about road casualties and fatalities brought about by drinking drivers, but also concern about alcoholism, cirrhosis and so on. The impact of alcohol problems is significant: recently the resource costs of alcohol raisuse in the United Kingdom have been estimated at over £1500 million [McDonnell ~,nd Maynard, 1985]. Moreover , the development of policies for alcohol control has not c,nly a national dimension but is one of the major themes of the World Health Organ- i:;,ation [WHO, 1982]. Indeed, to the extent that attention is focussed exclusively on crinking and driving, the range of possible interventions may be unduly restricted to measures which will impact uniquely on this problem. Such initiatives as increasing law enforcement or the level of sanctions for convicted offenders clearly have a part to play, but need to be set against a range of alternatives which embrace a more comprehensive assessment of possible courses of action.

There is strong evidence that the national incidence of alcohol-related problems is c !osely related to average levels of alcohol consumption [Popham, 1970; Brunn, Edwards, aad Lumio et al . , 1975]. Kendell [1984], for example, examined the relationship between per capita consumption of alcohol in the U.K. and the incidence of alcohol-related problems since the beginning of the 1970s. Average consumption levels increased from 7 0 litres of ethanol per year in 1970 to 9.8 litres in 1979, then fell by more than 10% to 8.7 litres per year between 1979 and 1982. The decrease was accompanied by significant reductions in first admissions to hospitals for alcohol dependence and in cirrhosis mor- tality. Indeed, over the 1970-1982 period Kendell found significant correlations between per capita consumption and these alcohol-related events.

Attention to the shorter-term sequelae of alcohol abuse--including drinking and driving--using aggregate level data is relatively more recent. In the aforementioned report, Kendell refers to the relationship between per capita consumption and the level o! convictions for drunkenness and for drink-driving offences. These showed a similar link with aggregate consumption levels. The reduction in drink-driving convictions, for example, at 7% was almost twice as great as the reduction in cirrhosis mortality (4%).

Similar findings were reported by Wagenaar [1984] in a methodologically more sophisticated study of the link between alcohol consumption and motor vehicle crashes. Wagenaar was able to allow for seasonal effects and long-term trends in his data in a time-series analysis, based on a single jurisdiction in the U.S. (Michigan), and found a

449

450 D. RILEY

close relationship between monthly sales of different forms of alcohol over the period 1972-1980 and motor vehicle accidents involving drinking-drivers. The pattern of lagged cross-correlations was consistent with a causal effect of aggregate drinking on crash involvement. Wagenaar concluded that, "a random change in the quantity of beer or wine distributed in a given month is positively related to the number of crash-involved drinking drivers in that same month and the following month" (p. 179). He also discussed the likely impact on short-term alcohol-related problems of measures designed to reduce aggregate consumption, although the distinction between sales and consumption is glossed over at this point. The use of sales or distribution data is clearly problematic. If more alcohol is sold in a given period, we could reasonably expect that more alcohol was consumed in or around that period. But what is less certain is the extent to which increased sales are attributable to an increase in the number of drinkers in the intervals studied, or in the number of drinking occasions, or in the amount of alcohol consumed on each occasion. The implications for policy, which stem from the way changes in consumption are brought about, will clearly differ according to the particular profile of changes involved.

It seems likely that the strong evidence of a close association between aggregate alcohol consumption levels and chronic alcohol-related problems will encourage further attention to the corresponding link between alcohol use and the incidence of a number of aspects of drinker behaviour, including drinking and driving. However, as Kendell [1984], among others, has pointed out, the crucial role assigned to per capita consumption in the production of alcohol-related problems is largely based on research which studied how problems increased as consumption levels rose. Until relatively recently there were few exceptions to the general upward trend in consumption levels. But crucial to any analysis of what to do to reduce the burden of alcohol on society is an assessment of what happens when consumption levels fall. The link between alcohol use and its ap- parent problems when consumption levels are rising need not imply anything about what would happen if consumption fell. Above all, we need information about how aggregate consumption levels reflect individual drinking patterns both cross-sectionally and over time, since decreases in consumption may result from factors quite different from those responsible for increases.

DISTRIBUTION THEORIES

Theories of the distribution of alcohol consumption have an important role to play in understanding how changes in mean consumption levels will be reflected in changes among the various subgroups which make up the drinking population at any one time. Speculation about why alcohol-related problems increase when aggregate consumption increases, and why these problems could be alleviated by a decrease in average drinking levels, often reflects the influential work of Sully Ledermann, who identified an invariant distribution of drinking frequencies among populations [Ledermann, 1956]. Using data from a number of countries, Ledermann proposed a consistent log-normal distribution to describe patterns of alcohol use, and suggested that the frequency of alcohol con- sumption in an homogenous population could be estimated with this distribution. This is usually taken to imply that as per capita consumption rises so too will the proportion of heavy drinkers--the population subgroup most likely to present alcohol-related prob- lems. Despite the subsequent criticisms made of the empirical basis and theoretical development of Ledermann's research [see Miller and Agnew, 1974, who point out that many of the populations on which Ledermann based his work were either small or unrepresentative], the Ledermann curve has become widely accepted as an adequate account of the link between per capita consumption and the incidence of alcohol-related problems. De Lint [1978], for example, argues that, "a change in the average con- sumption of alcohol is likely to be accompanied by a change in the same direction in the proportion of heavy drinkers and that, therefore, measures affecting total consump- tion are likely to affect the prevalence of alcohol problems associated with heavy drink-

Reducing per capita consumption in the U.K. 451

ing" (p. 268). Moreover, in the U.K., the Royal College of Psychiatrists in their report "Alcohol and Alcoholics" wrote:

"A mass of evidence, culled from international analyses of drinking data, supports the assertion that shifts in average per capita consumption will be accompanied by shifts irr indices of alcohol-related disabilities in the same direction. Moreover, the shift is not simply proportionate: a given change in average consumption will be accompanied by a disproportionately magnified change of casualty rates". [Royal College of Psychiatrists, 1979, p. 91].

The idea that changes in per capita alcohol consumption represent the most useful target of measures to reduce the burden of the problems brought about by excessive use of alcohol is a pervasive one [see Cook, 1982]. But the view that the best way to limit the incidence of alcohol problems is to decrease or prevent growth in the average amount of alcohol drunk by consumers has been criticised by, among others, Duffy and Cohen [1978] and by Tuck [1980]. Duffy and Cohen, point out that, while reductions in total alcohol consumption represent an appropriate target, if we see "any and all consumption af alcohol as a public health problem measures aimed at total alcohol consumption will Rave unpredictable effects on the proportion of heavy drinkers" (p. 263). Tuck [1980] :~imilarly argues that simple versions of reductionist theory carry no direct implications :!or policy. Much will depend on how changes in the numbers drinking are distributed among the various categories of consumption. If, as Skog [1985] points out, those who change their consumption pattern are light drinkers, then little reduction in alcohol- related problems would be evident.

Drawing a parallel with vehicle use and the incidence of road accidents, Tuck points out that if fewer people drove there would indeed be fewer accidents or fatalities on our highways, but that no effort is made to limit vehicle use for these reasons. Tuck ~Lrgues, "this is because it is believed that the advantages of motoring to the individual ~Lnd to society outweigh the disadvantages of road accidents" (p. 3). She also reminds tts that the great majority of the adult population who do drink claim to find both i adividual and social values in the consumption of alcohol. Policy on road accidents iacludes improving the driving environment and teaching safer driving skills. There is r:o logical difficulty over an alcohol policy which aims at "encouraging safer drinking t~abits, devising safer drinking environments, rather than simply discouraging drinking" (p. 3).

Secondly, Tuck shares the view that the link between per capita consumption and tile corresponding distribution of drinking patterns is not well established. In principle, the distribution of alcohol consumption within a population and its relationship to per capita consumption is open to study. Does changing per capita consumption necessarily lead to different numbers of heavy drinkers, as assumed by Ledermann or as suggested be those [such as De Lint, 1974] who have extended and refined his analyses? The e,,~sential links between data on aggregate consumption and alcohol-related problems are individual patterns of alcohol use.

Data on the distribution of these patterns is crucial to an understanding of how v~triations in per capita consumption may be reflected among population subgroups. T'ack presents data which question whether there is inevitably a change in the con- sumption of heavy drinkers associated with a change in national per capita consumption figures--a key issue for those who wish to extend Ledermann's analysis into the policy arena. An analysis of subgroup changes based on just three categories of drinkers (light, moderate, and heavy) would need to take account of 51 possible combinations of subgroup change which could underlie an increase in per capita consumption. Consumption the- ories based on the arguments put forward by Ledermann fail to consider the possibility of many of these movements. They hold, for example, that there is no way that numbers of average or moderate drinkers can increase without an accompanying increase in numbers of heavy drinkers. They similarly hold that an increase in light or occasional

452 D. R~LEY

drinking will inevitably result in a growth in moderate and in heavy drinking. As argued by Tuck [1980],

" . . . it is at least a logical possibility that more people could enter the bottom of the market (become light or occasional drinkers, perhaps by buying a bottle at Christmas time or having the occasional celebration on birthdays or anniversaries) without more people becoming heavy drinkers. Similarly more people could move from the occa- sional drinking category to becoming average drinkers (perhaps visiting a pub once a week) without more people becoming heavy drinkers. It is the essence of strong consumption theory that though these are logical possibilities, they are not possibilities in a real world and that any increase in the number of light or moderate consumers will be accompanied by an increase in the number of heavy consumers" (p. 5).

Similarly, it is known that a small proportion of drinkers account for a large pro- portion of total consumption. Dight [1976] found that just 3% of alcohol users accounted for 25% of consumption in the Scottish population surveyed. Changes in national per capita consumption could occur even if the raw numbers of heavy, moderate , and light drinkers remained unchanged. An increase in the amount consumed by heavy drinkers could result in large increases in per capita consumption without any change in the amounts consumed by light or moderate drinkers.

Much of the evidence concerning the relationship between per capita consumption and alcohol-related problems has difficulty in pointing the way for policy. Is the cor- relation between per capita consumption and, for example, cirrhosis deaths due to more of the population being at risk if increasing numbers of the population engage in drinking, or to increasing consumption by the high-risk heavy-drinking group? Neither of these two possibilities would imply that a restrictionist policy aimed at the drinking patterns of the mass of the normal population would be either necessary or useful in alleviating the burden of alcohol misuse.

The data reported by Tuck [1980] were drawn from the Target Group Index (TGI) of the British Market Research Bureau. The TGI is an annual survey based on 25,000 individuals aged 16 and over. Respondents are interviewed to provide classification data and then left a self-completion questionnaire which collects information about the pur- chase and use of a wide range of products, including alcoholic drinks. During the period examined (1974-1978), national per capita consumption of alcohol rose from 8.8 to 9.8 litres of ethanol per year. Changes in the frequency of drinking (known to be closely related to alcohol consumption at the individual level [see Wilson, 1980]) were examined for different age-sex groups in the population. These data were not consistent with expectations based on classic consumption theory, since some of the rise in per capita consumption could be attributed to an increase in the proportion of the population who ever drink. Similarly, the growth in the number of young women (aged 16-19) drinking was not accompanied by any increase in the number of heavy drinkers in this sub-group (the percentage drinking "2 or 3 times a week" declined and there was no increase in the numbers of "daily" drinkers). The TGI data also allowed an analysis which took account of the type of drink consumed. In the U.K. , beer (and similar drinks such as lager and cider) accounts for over two-thirds of alcohol consumed, and draught beer accounts for the majority (about 75%) of this consumption. Changes in per capita consumption of draught beer between 1974 and 1978 were associated with increases in the numbers who ever drank this type of drink, but not with increases in the proport ion of heavy drinkers. The increase in consumption shown by national per capita figures appears to have come only from an increase in the percentage drinking at the moderate "3-7 pints a week" level.

The distinction between average and individual consumption levels was also evident in further analyses of TGI data for other types of drink. Per capita consumption of whisky, for example, showed little change, rising by an average of just 0.5% a year over this period. Analysis of individual consumption patterns, however, showed a significant growth in the numbers ever drinking whisky and those drinking four or more units a week. Per capita consumption of imported wine (which accounts for over 80% of wine

Reducing per capita consumption in the U.K. 453

consumption in the U.K.) also changed little over the 1974-1978 period. TGI figures, in contrast, indicated that consumption was rising in every category of drinker.

More recent evidence for the U.K. is provided by an analysis of alcohol expenditure patterns derived from the Family Expenditure Survey (FES), based on a nationally representative sample of households (N = 7,000) and published annually. Looking at the changes over the period 1979 to 1984, when average consumption fell by about 9% (from 9.6 to 8.7 litres of ethanol per year), FES data showed that changes in consumption were related to household income levels. Adults from high- and middle-income house- holds decreased their estimated consumption of alcohol by 7% and 17% respectively, while those from low-income families showed a 16% increase in consumption. The link between income group and consumption patterns is even more striking in relation to estimated consumption of beer. Over the 1979 to 1984 period, average consumption fell by 12% (from 5.8 to 5.1 litres of ethanol equivalent per year). Adults in middle- and high-income groups decreased their consumption of beer by an average of over 20% while those from low-income households increased their consumption by more than 50% aver this period.

Such data as these cannot in isolation provide a reliable indication of the actual distribution of alcohol consumption. What they can do, however, is caution against the ~se of per capita consumption levels in association with untested assumptions about this distribution to draw conclusions about how changing per capita consumption will be reflected in alcohol use by different population subgroups.

It would be misleading to suggest, however, that changes in per capita consumption ~tre invariably associated with conflicting patterns of change at the individual level. Cartwright, Shaw, and Spratley [1978a] report a sample survey of a London suburb carried out in 1965 and again in 1974. Over the nine-year period, per capita consumption (measured in number of drinks) increased by almost 50%, but there was no change in the percentage of abstainers; nor did the population drink more frequently in 1974 than they had in 1965. Instead, Cartwright et al. found that the increase in per capita con- sumption resulted from an increase in the average amount of alcohol consumed on each c!rinking occasion. As the authors put it, "on a drinking day in 1974, the 'average drinker' ~rank 56% more alcohol than he would have done in 1965" (p. 246). These patterns of r~sults were also described as "the first of their kind to confirm the hypothesis that (a) an increase in mean consumption over time implies the redistribution of increased pro- portions of drinkers into higher consumption categories, and that (b) this in turn is likely to be associated with an increase in the prevalence of alcohol-related problems" ]Cart- wright, Shaw, and Spratley, 1978b, p. 256]. As we have seen, the conclusion about the inevitability of changes in per capita consumption and changes in consumption categories ~as premature.

Nevertheless, more recently Skog [1980, 1985] has proposed a "social interaction theory of drinking behaviour" which describes how, "as a main rule the population tends to move in concert up and down the scale of consumption" (1985, p. 83). But again, as with Ledermann, the data on which Skog builds his theory are based on the cross- sectional distribution of consumption patterns in either general populations or in specific sttbgroups, and as such may fail to address directly the question of what happens when overall levels of consumption change with time.

The relationship between per capita consumption and the underlying distributions ot alcohol consumption may be influenced by the direction of change in aggregate drinking levels. Knibbe, Drop, and Van Reek et al. [1985], in their analysis of alcohol cc,nsumption in Holland, used survey data from 1958, 1970, and 1981 to measure changes in average consumption and in the distribution of consumption of different subgroups in the population. Per capita consumption increased by over 300% between 1958 and 1981, but in most subgroups the percentage of abstainers did not show significant changes. Knibbe et al. summarise their data thus: "All subpopulations with an increase in mean consumption show a strong increase in the percentage of drinkers in the highest con- sumption category" (p. 411). This pattern of results is clearly consistent with the Led- ermann account of the distribution of alcohol consumption. But the increase in con-

454 D. RILEY

sumption between 1958 and 1981 was uneven, and the bulk of the increase in per capita consumption in Holland took place in the interval between the first two of the three surveys. For younger males (aged 21-40), the average number of standard glasses per person per week increased by 330% between 1958 and 1970, but actually fell between 1970 and 1981 by 6%. For older males (over 40 years), the rise in per capita consumption was more consistent: it rose up 200% between the first two surveys, and 160% between the second and third. An analysis of national per capita figures would, however, have seen a rise from an average of 2.5 standard glasses per person per week in 1958 to 6.2 in 1970 and 7.5 in 1982. This apparent discrepancy--a 20% rise in per capita consumption between 1970 and 1981 accompanied by an actual decrease in consumption levels by younger men (the group most likely to be involved in drinking and driving)--raises important questions about the impact of policies designed to influence per capita con- sumption on the incidence of alcohol-related problem behaviour.

D R I N K I N G B E H A V I O U R

We have seen that changes in per capita consumption often provide an unreliable indicator of changes at the individual or group level, whether categorised by drinking level or by sociodemographic characteristics. Where does this leave attempts to control drink-driving and other alcohol-related problems through changing average consumption levels? Approaching the drink-driving problem the other way, not from aggregate figures on per capita consumption but from the study of drinking behaviour itself, provides a different perspective on the range of interventions with more immediate potential impact on drink-driving problems.

Young drivers One way forward would be to consider how any intervention could most usefully

be targeted on the population subgroup most likely to be involved in drink-driving. The sociodemographic characteristics of drink-driving offenders have been examined in the context of both official statistics on convicted offenders, and through driver surveys which have collected self-report information on the prevalence and the frequency of drink-driving violations. This research consistently shows that offenders are dispropor- tionately represented among young men, particularly those in the 20-29 age group [see Jones and Joscelyn, 1978; Riley 1984, 1985].

Consumption levels Driver surveys [see Berger and Snortum, 1985, 1986] additionally show that the

incidence of offending is closely related to individual levels of alcohol consumption after allowing for age and sex effects. This pattern is also evident in data from the first British Crime Survey [Riley, 1985; for a general account of the survey see Hough and Mayhew, 1983]. An analysis based on over 3,000 drivers in the survey in 1981 examined the relationship between self-report drink-driving and a range of respondent characteristics. These included measures of total alcohol consumption, involvement in other motoring and nonmotoring offences, alcohol beverage preferences, beliefs about legal blood al- cohol concentrations (BACs) when driving and about the legal penalties for convicted offenders, and "lifestyle" (such as frequency of drinking away from home). A multi- variate analysis showed that the prevalence of drinking and driving was most strongly related to total alcohol consumption. Allowing for age and sex characteristics, those who drank more than the average of the sample as a whole were three times more likely to admit to drink-driving than those who drank less than average. Similarly, the analysis indicated that drivers who preferred beer to other alcoholic drinks were about twice as likely to drink with illegal blood-alcohol levels.

Beverage preference These findings concerning beverage preference parallel those reported by Berger

and Snortum [1985]. They proposed that, "beer drinkers differ from wine and spirit

Reducing per capita consumption in the U.K. 455

drinkers in important ways that affect traffic safety" (p. 232). Berger and Snortum found that regardless of such driver characteristics as age and sex, beer drinkers usually drank to higher levels of intoxication and were more likely to drive after drinking than were drivers who preferred wine or spirits. The results of Berger and Snortum's regression analysis (based on whether drivers were estimated to have driven with a BAC over 100mg/100ml) showed, as did the analysis of the British Crime Survey data, that beverage preference was related to offending even after controlling for other driver characteristics strongly related to drink-driving including age, sex, and total alcohol consumption.

The observed relationships between offending and beverage preference were, how- ever, different in the two surveys. In the U.S. study, men drivers across all age groups who preferred beer were more likely to have driven with BACs over 100mg/100ml. In the U.K. survey, drink-driving was highest for beer drinkers only in the youngest age group (under 25 years). For those in the next age category (25-40) drink-driving appeared largely unrelated to preference beverage. For drivers in the over-45 group, although least likely to be offenders, those who preferred spirits had the highest offending rates. Despite these age-related differences in the results of the two surveys, the research does suggest that what drivers drink may have an effect on drink-driving over and above how much they drink and provides additional support for measures directed at beer con- sumption in particular. Using aggregate state-based U.S. data, Colon and Cutter [1983] also report results which direct attention from alcohol consumption in general to beer zonsumption in particular. They report a significant positive relationship between the rate of fatal traffic accidents and per capita consumption of alcohol contained in beer. iXlo comparable effects were evident for wine or spirits.

~ars Further, we know from drinking surveys [e.g., Wilson, 1980] that heavy drinkers

(defined in that survey in terms of consumption of at least 50 standard units of alcohol i~ the previous week for men and over 35 units for women) are more likely to consume their alcohol in settings which are more likely to lead to drinking and driving. Wilson f~und that heavy drinkers were more likely to consume their alcohol in bars and less l~kely to drink at home. The proportion of drinking occasions for men in the heavy- drinker category taking place in bars was 80% compared with 50% for men classed as light drinkers (up to 10 units of alcohol in the previous week). Such surveys also tend to show that heavy drinkers consume proportionately more of their alcohol in the form of beer or lager (86% for heavy drinkers against 65% for light drinkers in the survey reported by Wilson). Light drinkers in this survey were more likely to drink their alcohol in the form of spirits (17% against 9% for heavy drinkers) and wine (15% against just 4 % for heavy drinkers). This survey also showed that heavier drinkers drank faster than fight drinkers (drinking speed was calculated by dividing weekly consumption by the total number of hours spent drinking). Although part of the explanation for this differ- ence is that heavier drinkers were more likely to drink in bars--a venue generally characterised by relatively high drinking speeds [see Babor etal . , 1980; Storm and Cutler, 1981J--the survey found that heavier drinkers also drank faster at home. Given the limited ability of the human body to metabolise alcohol (about 10-15 cl per hour), faster drinking speeds tend, as Wilson [1980] points out, to lead to higher concentrations of alzohol in the blood and to "more acute short-term impairment" (p. 16).

The importance of bars as the venue prior to driving is clear: most male drinkers drive to their next destination (usually home) [see Clayton, McCarthy, and Breen, 1984]. Ttle rates of vehicle use after drinking were found to be highest for those drinking in bars. Howe~'er, the role of bars as the main setting for driving after drinking does not seem simply to reflect the preference of heavy drinkers to drink in bars. Both the Wilson [1!)80] and Clayton et al. [1984] surveys found that rates of vehicle use were not highest in the heaviest drinking categories. Drinkers in this category were those most likely to walk to drinking venues away from home. However, when heavy drinkers did use their ve]hicles, they were more likely to be illegally intoxicated on the road. In the survey reported by Wilson [1980], within each of the three drinking categories (light, moderate,

456 D. RILEY

Table 1. Differences in patterns of alcohol consumption associated with drinking categories and drinking locations

Hours S p en t Drinking Average Drinking Usual Drinking D r i n k i n g Occas ions Consumption Category Location per week per week per Occasion

Light In bars 2.3 1.9 3.0 units Elsewhere 2.4 2.8 1.7 units

Moderate In bars 7.8 5.3 5.5 units Elsewhere 9.4 8.4 2.8 units

Heavy In bars 13.5 7.5 6.3 units Elsewhere 13.2 7.6 4.8 units

and heavy), the total amount of time spent drinking and the average number of drinking occasions per week were, in fact, lower (with one exception) for those whose usual drinking location was in bars. These data are set out in Table 1 above.

The dramatic difference associated with drinking in bars concerns the average con- sumption per drinking occasion. Wilson suggests that, "people who mainly drink in bars may be more vulnerable to developing a heavy drinking style mainly because one extra drinking occasion in the week would lead to a larger increase in their weekly consumption than would be the case for people who usually drank at home. In addition, because people drinking in bars have a more concentrated drinking style they are more likely to be affected by various problems linked to drunkenness" (p. 17). There is similar evidence that venue may have a direct influence on drinking styles. Henderson and Freedman [1979], in an Australian survey of drinking male drivers, found that more than half of the youngest driver group (aged 17-19) said they found it difficult to either limit or even count their drinks when drinking in bars.

A strategy The accumulated evidence may be sufficient to support the suggestion that the point

of intervention most relevant to the control of alcohol-related problems generally and drink-driving specifically centres on the retailing of beer in bars.

As this paper has argued, interventions aimed at changing per capita consumption may not change the drinking behaviour of those most likely to be drinking drivers. The available evidence indicates that the primary target group consists of drivers who drink beers in bars. This focus on the specific behaviour of specific individuals in specific contexts avoids some of the difficulties of targeting initiatives on such broad sociode- mographic groups as young male drivers per se. In the latter case, the age and sex characteristics used to define the group--a l though providing useful pointers for further research-- imply little about possible intervention strategies and may fail to capture the significance Of offending by those outside the defined group. (For example, data from the 1982 BCS shows that among drivers under 25 for whom beer is the preferred beverage, the prevalence of offending is the same for women (37%) as for men (38%). The question now becomes: "What strategies can we identify to reduce the alcohol consumption of (driving) beer drinkers in bars?"

Price controls Leaving aside the possibility of reducing driving itself, it is well known that aggregate

alcohol use is closely associated with the relative price of alcohol in a given community [Cook, 1982; Kendell, 1984]. Despite the objection [see Stacey, 1983] that little is known about the elasticity of demand among those in the heavy-drinking category, it has been proposed that alcohol consumption and the incidence of alcohol-related problems can be reduced by increasing alcohol taxes and, hence, the price of alcohol. While such a policy is uncontroversial as a potential means of reducing alcohol consumption generally, its capacity to decrease actual consumption in specific groups is questionable. Studies of other forms of consumer behaviour, for example, might lead us to expect that relative price changes are more likely to impact on light or infrequent product users rather than frequent or heavy users.

Reducing per capita consumption in the U.K. 457

Relatively few studies in the U.K. have looked in any detail at changes in individual consumption patterns over time and their relationship to changes in the relative price of alcohol. However, Kendell, De Roumanie, and Ritson [1983], in a panel survey of "regular drinkers" in Scotland, found that at a time when the price of alcohol was rising faster than the retail price index and average disposable incomes, heavy drinkers reduced their consumption--compared with three years earlier--as much as light or moderate drinkers. But, as Kendell et al. point out, respondents who were in full- or part-time employment at both phases of the survey showed little or no reduction in consumption. In contrast, men who became unemployed in the intervening period showed a dramatic reduction in alcohol consumption (down from an average of 52.5 units per week to 23.1 units). This fall was ten times greater than that for men who kept their jobs. Thus, it seems likely that the apparent effects attributed to relative price changes were limited to those whose incomes were severely reduced, that is, to those whose consumption levels were likely to have fallen anyway even if alcohol had not become more expensive. During periods of relative stability in employment, the effects of changes in the relative 9rice of alcohol on consumption patterns may prove less predictable. Davies and Walsh ]1983] cite the example of the growth in per capita alcohol consumption in Ireland between 1960 and 1970 at a time when the real price of alcoholic beverages was also rising. What may be more significant than the simple relationship between the relative price of alcohol and per capita consumption may be the link between consumption and disposable income. Clearly, disposable income can change independently of the price of alcohol.

PRODUCT DEVELOPMENT

There is a parallel with cigarette smoking which provides a useful pointer to other forms of intervention. Extensive evidence of the dangers of disease and premature death caused by cigarette smoking encouraged manufacturers--partly in an attempt to prevent loss of sales-,to modify their product. (A similar process is currently evident in the steps food manufacturers are beginning to take to eliminate unnecessary additives from their products). Clearly, changing a product is just as effective a response to criticisms of the risks involved in its use as is changing the incidence of its use. And given what we know ai3out behaviour change in other fields, modifying the product itself is no less likely to be successful in reducing alcohol problems than attempts at persuading users to modify ttteir drinking behaviour [see Riley and Mayhew 1980].

In what ways therefore could the alcohol beverage industry be encouraged to change it:; products to reduce the incidence of alcohol-related problems? Few would disagree that alcohol content may be the most relevant dimension to consider, and there appears tc be little or no obstacle to broadening the range of alcohol control policies to include ar~ attempt to reduce the alcohol content as well as the volume of drinks consumed. Reducing the alcohol content of beer by 10% may have more of an impact on the inzidence of alcohol-related problems because the impact of change is likely to be more evenly spread than reducing the overall consumption of beer by the same amount. But why should a switch to lower-alcohol drinks offer advantages over a reduction in per capita consumption as a target for policy and what effects could we expect such a development to have on the problems posed by drinking and driving?

REDUCING ALCOHOL CONSUMPTION

A ~ : o h o l taxes Given the uncertainty about the link between the price of alcoholic beverages and

inci'ividual consumption patterns, it seems appropriate to consider alternative means of limiting alcohol consumption through alcohol taxes. Attempts at decreasing con- sumption using price as a lever are aimed at the level of alcohol taxes. However, the structure of such taxes may also be an appropriate focus of intervention. Given the possible resistance, particularly by alcohol producers [see Casswell, 1985], to measures designed to control alcohol consumption, this distinction between the sale of drinks and

458 D. RILEY

the sale of alcohol may be crucial. Recognising this difference also reduces the risk that policy initiatives will be opposed by advertisers and others who derive income or other support from alcohol sales. If the amounts of alcoholic beverages sold are unchanged and profit margins are maintained, alcohol content may not be a matter of central concern to the industry. Moreover, in the West, alcohol taxes make an important contribution to government revenue. In the U.K., for example, alcohol taxes represent about seven percent of all revenue raised by government [Brewers' Society, 1986]. Restructuring alcohol taxes need not threaten--and could even augment--this fiscal base.

The impact of introducing progressive alcohol taxes on alcohol consumption seems likely to depend on how such an initiative influenced the choices available to the con- sumer. For example, most popular beers sold in bars in the U.K. vary little in alcohol content. If alcohol taxes were revised simply by taxing stronger drinks proportionately more heavily than weaker alternatives, the effects on alcohol consumption of extending the price differential between competing brands may be only marginal. A more effective system would rely on the operation of thresholds in the alcohol tax structure. Setting thresholds below the average alcohol content of popular beers (and other drinks) would provide producers with an incentive to reduce the alcohol content of drinks and would introduce or strengthen their price advantage. Moreover, the spacing and size of alcohol tax thresholds might provide an unusually precise means of influencing the alcohol content of drinks.

The available evidence suggests that the potential impact of changes in the alcohol content of drinks is likely to be greatest among those more prone to offending. The incidence of drinking-driving is higher among younger drivers; it is also among this age group that we find the highest proportion of offending drivers with BACs in the region closest to the legal limit (see Table 2).

About 30% of those aged under 30 convicted of drinking-driving offences have BACs in the 81-120 mg/100 ml range. This contrasts with about 20% of convicted drivers aged over 50. If we assume that the upper end of the 81-120 rag/100 ml range involves the average male driver consuming, say, 8-10 units of alcohol over a two-hour period before taking to the road, reducing the average alcohol content of beer (and other drinks) consumed by 25% would put the majority of offending drivers in this BAC group on the right side of the legal limit. The modest marketing success of zero-alcohol beers in the U.K., however, underlines the importance of modifying existing market leaders or at least selling such new "light" beers (and other drinks) on grounds other than their low alcohol content. The product image is likely to be all-important, given possible public resistance to any advertising messages implying that even moderate alcohol con- sumption is harmful to the average drinker. (Such messages may, however, have a part to play in the context of any campaign by the medical profession warning of the general health risks of alcohol use, rather than misuse).

Interventions which have influenced the strengths of alcoholic beverages are not unprecedented in this country. In the now-classic example of the possible benefits of close regulation of the alcohol industry, restrictions on the supply of beer during the First World War in the U.K. led to the average gravity of beer falling from 1,052 OG (original gravity) in 1912 to 1,030 OG in 1918, that is, from about 6% to about 4% alcohol by volume. In addition, alcohol taxes during this period meant that beers with

Table 2. Percentage distribution of blood-alcohol concentrations of convicted offenders (1983 data; N = 77,647)

Blood-Alcohol Concentrations Age 81-120 mg/100 ml 121-150 mg/100 ml 150+ rng/100 ml

Under 21 33.9 23.9 42.2 21-24 25.9 20.9 53.2 25-29 22.3 18.4 59,4 30-39 19.5 16.9 63,6 40-49 17.7 16.0 66.3 50+ 20.4 16.4 63.2

Reducing per capita consumption in the U.K. 459

OGs over 1,054 cost three times as much as beers with OGs under 1,020 [Wilson, 1940]. Some of the observed reductions in alcohol-related problems were probably due to the drastic restrictions in the supply of alcohol. There were, for example, likely to have been fewer drinking occasions--pubs opened later and closed earlier and hours of em- ployment increased [Smart, 1974]. Nevertheless, it seems likely that the BAC distribution of beer drinkers would have experienced a marked downwards shift. While the U.K. wartime restrictions on alcohol production and distribution were short-lived, the present alcohol tax policies in Norway and Sweden (although not developed with drink-driving specifically in mind) show strong similarities with the scheme discussed in relation to :he U.K. and have been linked to the relatively low prevalence of drink-driving in those ,:ountries [Snortum, 1984].

One criticism such a policy would attract concerns drinkers modifying their con- ~;umption patterns to offset changes in the alcohol content of drinks. Clearly, little or no reduction in alcohol-related problems could be expected if drinkers in any significant numbers switched to higher strength, though more expensive, brands or if more drinks were consumed. While these would remain possibilities, the available evidence [Schach- ter and Singer, 1962; MacAndrew and Edgerton, 1969, Connors and Maisto, 1979] suggests that alcohol users may alter their consumption behaviour relatively little. If the circumstances of alcohol use remain unchanged and, for example, beer drinkers expect tile effects of drinking lower-alcohol beers to be similar to those of their previously preferred beverages, then it is seems likely that their experienced effects will also stay c gnstant [Steele and Southwick, 1985; Hull and Bond, 1986]. Moreover, in respect to other offences, measures designed to reduce the opportunity for offending have proved successful in reducing the incidence of problem behaviour [Mayhew, Clarke, and Sturman e~ al . , 1976; Cornish and Clarke, 1986].

OTHER INITIATIVES

The situational analysis of drinking contexts does not inevitably lead to a single set ot possible interventions, and other developments such as automobile design and highway ertgineering also have an important role to play [Stacey, 1983, Thomson, 1985]. This p~per has focussed on the potential beneficial effects which may derive from a restruc- turing of alcohol taxes, but targeting measures on bars could take other forms. Denny [1979], for example, proposed that more could be done to provide facilities for drivers to test their own BACs before taking to the road. Reliable and accurate equipment properly used would have a role to play in providing feedback to drivers on their BACs and may be effective in reducing drinking and driving for those drivers who underestimate the; relationship between alcohol intake and BAC. There is some evidence that drink- drivers tend to overestimate the amount they can drink and stay on the right side of the legal limit [Riley, 1984; Clayton et al. , 1984]. While feedback seems ineffective for those drivers faced with the choice of drink-driving or abandoning their vehicles [Meier, Brigham, and Handel, 1984], it may have beneficial effects on subsequent occasions. Moreover, although it is illegal to continue to sell alcoholic drinks in bars to someone who is clearly intoxicated, the effect of this legal restriction may not have much effect on reducing drink-driving in contrast to other forms of server behaviour. Earlier inter- ve~tion--before the stage of discernible intoxication is reached--offers more scope for bringing driver BACs closer to the legal norm. However, the experience of those areas of ~r:he U.S. which have introduced legislation creating legal liability for suppliers of alcohol to drivers later involved in injury accidents has not been encouraging to date, at least in the commercial sector [Mosher, 1985].

While the consumption of beer in bars appears to contribute disproportionately to the incidence of drinking and driving, other contexts are also relevant. There are, for example, a great number of actual and potential drink-drivers who rarely, if ever, drink beer or visit bars. But measures designed to influence the alcohol content of drinks generally might be expected to mitigate the adverse effects of alcohol, whatever the particular types of drink consumed in whatever contexts.

460 D. RILEY

Similarly, there are ways of reducing alcohol consumption without reducing the alcohol concentration of drinks. For example, changing from imperial to metric units for the sale of beer in bars (from about 570 ml to 500 ml) or for bottled spirits (from about 630 ml to 500 ml) could reduce overall alcohol consumption without necessarily reducing profits correspondingly. But changes which posed the threat of reducing the number of drinks sold may be less than enthusiastically received by those involved in the manufacture and supply of alcoholic drinks. The potential conflict here is an im- portant one, especially in the U.K. where producers dominate the retail sector of the alcohol beverage industry.

Increased restrictions over the supply of alcohol through tighter controls on liquor licensing [Jeffs and Saunders, 1983; Tether, Robinson, and Wicks, 1984] and over the number and location of outlets [Mosher, 1985] have also been proposed as ways in which the industry might be encouraged to mitigate the impact of alcohol problems. In the U.K., for example, licensing laws prohibit the sale of alcohol in motorway service areas. Paradoxically, the apparent benefits of increased controls, such as a lower incidence of drink-related offences, appear to be matched, at least as judged by preliminary results, by the effects of more relaxed licensing laws which increase the opportunities for drinking [Adam Smith Institute, 1986]. This report, based on the 1976 extension of licensing hours in Scotland, points out that drink-driving convictions rose by just 1 percent in the first four years compared with 36% in England and Wales. Other offences also showed relative decreases: violent assaults went down 27%, under-age drinking went down 42%, and drunkenness went down 27%. These apparent changes were accompanied by reductions in the numbers of hospital admissions for alcohol-related problems [Eagles and Besson, 1985]. Assessing the impact of licensing changes themselves is, however, a complex task since alcohol consumption (indexed by family expenditure) in Scotland fell relative to England and Wales in the years following the changes in licensing laws. The effects of the economic recession, experienced throughout Europe from the late 1970s, respresent a competing explanation for the observed effects. It may be optimistic to look for clear evidence concerning the role of licensing controls in isolation from other factors influ- encing the incidence of alcohol-related problems.

Just as it has been argued that single solutions represent an unrealistic goal of any general alcohol policy, analyses of individual facets of alcohol misuse, including drink- driving, will need to recognise the diversity of situations involved. All too easily this could become a reason for inaction. But by recognising the potential role of--and interaction between--the different agencies responsible for the ways alcohol is used in our societies and the ways in which the adverse effects of alcohol manifest themselves, we may begin to believe that effective intervention is possible. The alcohol beverage industry, which already contributes to the funding of research on alcohol abuse in general and on drink-driving in particular, may have a significant further role to play.

OVERVIEW

It has been one of the major themes of this paper that attempts to control per capita alcohol consumption, through increasing the relative price of alcohol generally, represent a course of action with uncertain effects. Little is known about the demand elasticity of alcohol. Over the past decade in the U.K. there have been changes in per capita and individual consumption patterns inconsistent with expectations based simply on price- determined drinking habits. The present structure of alcohol taxes in the U.K. levies excise duty in more or less direct proportion to the alcohol content of drinks within each of the broad categories, but does not include any progressive or threshold factors which tax the relatively stronger beers (or wines or spirits) at a proportionately higher rate than their lower-strength counterparts. As a consequence, there is little or no incentive within each of the beverage types for the consumption of lower-strength drinks. Based on an analysis of drivers' drinking preferences concerning venues and types of beverages, an intervention strategy was proposed which focussed on the consumption of beer in bars. This strategy was intended to provide an economic incentive for drinkers to switch

Reducing per capita consumption in the U.K. 461

their consumption in the direction of lower-strength beers (and other drinks) by the introduction of progressive alcohol taxes. An integral aspect of such an objective involves the recognition that consumption patterns of drinks of different types (and thus alcohol consumption) may be shaped by increasing the price advantage of lower-strength drinks. Such a development would not necessarily aim to reduce the number of drinks consumed by drivers. Instead, the target would involve limiting drivers' alcohol consumption. This distinction between sales of drink and sales of alcohol is a crucial one for the alcohol beverage industry. Just as the industry has perhaps the most to lose, it too may have the most to offer attempts to control the problems posed by drinking and driving. Product development could benefit both the industry and the consumer, not only in the U.K. ~ut also in other countries where the production and supply of alcohol are closely regulated.

R E F E R E N C E S

Adam Smith Institute. Time To Call Time. Adam Smith Institute, London, 1986. ]~abor T., Mendelson J. H. and Uhly B. Drinking patterns in experimental and barroom settings. J. Stud.

Alcohol 41, 636-651, 1980. Berger D. E. and Snortum J. R., Alcohol beverage preferences of drink-driving violators. J. Stud. Alcohol

46, 232-239, 1985. Berger D. E. and Snortum J. R., A structural model of drinking and driving: Alcohol consumption, social

norms, and moral commitments. Criminology 24, 139-153, 1986. Brewers' Society, Statistical Handbook. Brewing Publications Ltd., London, 1986. Bruun K., Edwards G., Lumio M., Makela K., Pan L., Popham R. E., Room R., Schmidt W., Skog O.-J.,

Sulkunen P. and Osterburg E., Alcohol Control Policies in Public Health Perspective. The Finnish Foun- dation for Alcohol Studies, Helsinki, 1975.

Cartwright A. K. J., Shaw S. J. and Spratley T. A., The relationships between per capita consumption, drinking patterns, and alcohol related problems in a population sample, 1965-1974. Part I: Increased consumption and changes in drinking patterns. Brit. J. Addiction 73, 237-246, 1978a.

Cartwright A. K. J., Shaw S. J. and Spratley T. A., The relationships between per capita consumption, drinking patterns and alcohol related problems in a population sample, 1965-1974. Part II: Implications for alcohol control policy. Brit. J. Addiction 73, 247-258, 1978.

Casswell S., The organisational politics of alcohol control policy. Brit J. Addiction 80, 357-362, 1985. Clayton A. B., McCarthy E E. and Breen J., Drinking and driving habits, attitudes, and behaviour of male

motorists. Department of Transport, Transport and Road Research Laboratory Supplementary Report 826, Crowthorne, England, 1984.

Calon I. and Cutter H. S. G., The relationship of beer consumption and state alcohol and motor vehicle policies to fatal accidents. J. Sa f Res. 14, 83-89, 1983.

C~ok P., Alcohol taxes as a public health measure. Brit. J. Addiction 77, 245-250, 1982. C,~nnors G. J. and Maisto S. A., Effects of alcohol, expectancy, and drinking rate on mood. Psychopharmacol.

62, 261-266, 1979. Cornish D. and Clarke R. V. G., The Reasoning Criminal, Springer-Verlag, New York, 1986. Davies E and Walsh D., Alcohol Problems and Alcohol Control in Europe. Croom Helm, London, 1983. D .~ Lint J., The epidemiology of alcoholism, in Alcoholism: A Medical Profile N. Kessel (ed.), Edsall, London,

1974. D,. • Lint J., Total alcohol consumption and rates of excessive use: A reply to Duffy and Cohen. Brit. J.

Addiction 73, 265-269, 1978. Dcmny R. C., Drinking and Driving. Robert Hale, London, 1979. Dight S., Scottish drinking habits. Office of Population Censuses and Surveys, Social Survey Division, HMSO,

London, 1976. Duffy J. C. and Cohen G. R., Total alcohol consumption and excessive drinking. Brit. J. Addiction 78, 259-

264, 1978. Eagles J. and Besson J., Changes in the incidence of alcohol-related problems in Northeast Scotland 1974-

1982. Brit. J. Psych. 147, 39-43, 1985. Henderson J. M. and Freedman K., Attitudes to drinking and driving in Australia. Paper published in

Proceedings of the Seventh International Conference on Alcohol, Drugs and Traffic Safety, Melbourne, 23- 28 January, 1977. Australian Government Printing Service, Canberra, 1979.

Hc ugh M. and Maybew E, The British Crime Survey: First Report. Home Office Research Study No. 76. HMSO, London, 1983.

Hull J. G. and Bond C. E, Social and behavioural consequences of alcohol consumption and expectancy: A meta-analysis. Psych. Bull. 99, 347-360, 1986.

Je fs B. W. and Saunders W. M., Minimising alcohol related offences by enforcement of the existing licensing legislation. Brit. J. Addiction 78, 67-77, 1983.

Jones R. K. and Joscelyn K. B., Alcohol and highway safety 1978: A review of the state of knowledge. U.S. Department of Transportation, Washington, D.C., 1978.

Kendell R. E., The beneficial consequences of the United Kingdom's declining per capita consumption of alcohol in 1979-1982. Alcohol and Alcoholism 19, 271-276, 1984.

Kendell R. E., De Roumanie M. and Ritson E. B., Influence of an increase in excise duty on alcohol consumption and its adverse effects. Brit. Med. J. 287, 809-811, 1983.

AAP 19/6-D

462 D. RILEY

Knibbe R. A., Drop M. J., Van Reek J. and Saenger G., The development of alcohol consumption in the Netherlands 1958-1981. Brit, J. Addiction 80, 411-420, 1985.

Ledermann S., Alcool--Alcoolisms--Alcoolisation. Cahier No. 29. Institut National d'Etudes Demogra- phiques. Presses Universitaire de France, Paris, 1956.

MacAndrew C. and Edgerton R. B., Drunken Comportment: A Social Explanation. Aldine, Chicago, 1969. Mayhew P., Clarke R. V. G., Sturman A. and Hough J. M., Crime As Opportunity. HMSO, London, 1976. McDonnell R. and Maynard A., The costs of alcohol misuse. Brit. J. Addiction 80, 27-35, 1985. Meier S. E., Brigham T. A. and Handel G., Effects of feedback on legally intoxicated drivers. J. Stud. Alcohol

48, 528-533, 1984. Miller G. M. and Agnew N., The Ledermann model of alcohol consumption. Q. J. Stud. Alc. 35, 866-898,

1974. Mosher J. E , Alcohol policy and the presidential commission on drunk driving: The paths not taken. Accid.

Anal. and Prey. 17,239-250, 1985. Popham R. E., ed., Indirect methods of alcoholism prevalence estimation: A critical evaluation, in Alcohol

and Alcoholism. University of Toronto Press, Toronto, 1970. Riley D., Drivers' beliefs about alcohol and the law. Home Office Research Bulletin 17, 32-35, 1984. Riley D., Drinking drivers: The limits to deterrence? Howard J. Crim. Just. 24, 241-256, 1985. Riley D. and Mayhew P., Crime Prevention Publicity: An Assessment. Home Office Research Study No. 63,

HMSO, London, 1980. Royal College of Psychiatrists, Alcohol and Alcoholism. London, Tavistock, 1979. Schachter S. and Singer J. E., Cognitive, social, and physiological determinants of emotional states. Psych.

Rev. 69,379-399, 1962. Smart R. G., The effect of licensing restrictions during 1914-1918 on drunkenness and liver cirrhosis deaths

in Britain. Brit. J. Addiction 69, 109-122, 1974. Skog O.-J., Liver cirrhosis epidemiology: Some methodological problems. Brit. J, Addiction 75, 227-243,

1980. Skog O.-J., The collectivity of drinking cultures: A theory of the distribution of alcohol consumption. Brit.

J. Addiction 80, 83-99, 1985. Snortum J. R., Controlling the alcohol-impaired driver in Scandinavia and The United States: Simple deter-

rence and beyond. J. Crim. Justice 12, 131-148, 1984. Stacey B. G., Drinking and driving: Alcohol association with traffic accidents. Austr. J. Soc. Issues 18, 2-17,

1983. Steele C. M. and Southwick L., Alcohol and social behaviour I: The psychology of drunken excess. J. Pers.

Soc. Psy. 48, 18-34, 1985. Storm T. and Cutler R. E., Observations of drinking in natural settings: Vancouver beer parlours and cocktail

lounges. J. Stud. Alcohol 42,972-997, 1981. Tether P., Robinson D. and Wicks M., Liquor licensing: Its role in a preventive strategy. Alcohol and

Alcoholism 19, 277-279, 1984. Thomson G. A., Drinking-driving accident countermeasures: Why not try changing the environment? Accid.

Anal. and Prey. 17, 207-210, 1985. Tuck M., Alcoholism and social policy: Are we on the right lines? Home Office Research Study No. 65, HMSO,

London, 1980. Wagenaar A. C., Alcohol consumption and the incidence of acute alcohol-related problems. Brit. J. Addiction

79, 173-180, 1984. Wilson G. B., Alcohol and the Nation. Nicholson and Watson, London 1940. Wilson P., Drinking in England and Wales. Office of Population Censuses and Surveys, Social Survey Division,

HMSO, London, 1980. World Health Organisation, Community response to alcohol related problems: Phase 1: Final Report. WHO,

Geneva, 1982.