The views expressed in this paper are the author’s. They do not necessarily represent Treasury positions.
CONTENTS
- the distribution across drinkers of their marginal net externalities of alcohol consumption; and
- the weighting to be applied to each drinkers’ marginal net externality of alcohol consumption, determined by the price elasticity of demand of drinkers.
- we have almost no knowledge of what goods or services would be substituted for alcohol, if a higher excise on it were imposed: if the basket of substitutes collectively generate negative externalities, the net external benefit of reducing alcohol consumption would be diminished. However, my subjective judgement is that it is unlikely that, if other harmful goods were consumed as substitutes for alcohol, they would generate negative externalities on the scale that alcohol does;
- in equilibrium the price of a good to each consumer reflects the marginal costs to society associated with consuming it ("social costs");
- - consumers are well-informed about the costs and benefits of consuming the good; and
- consumers are rational.
describe the externalities, both negative and positive, associated with alcohol consumption;
quantify these externalities, to the extent this is feasible; and
describe the relationship between externalities and differing levels of alcohol consumption at an individual level, and then, by describing the distribution of alcohol consumption (and hence externalities), at a population level.
- emotional and monetary costs to family and friends from the misuse of alcohol, to the extent that drinkers do not internalise these costs;
- the costs to third parties of physical violence to their persons or property;
- costs to the community as a whole of damage repair or containment, such as public health costs, social welfare services, policing, prevention programmes, etc.;
- costs of lower productivity which are paid by the drinker’s colleagues (by lower wages) or employer (by lower profit), or by society as a whole (for example, the tax revenue and associated deadweight cost required to pay for sickness benefits for beneficiaries with alcohol-related illnesses);
- excess premiums paid by an insurance company’s other clients, if they pay for costs incurred by drinkers; and
- the cost to the community of industrial or recreational accidents arising from alcohol-induced physical or mental incapacity.
Several categories of these costs may result from one event - such as a car crash caused by a drunk driver.
The costs of lower productivity are usually 50-80 percent of total (external and internal) costs. The externality comprises the part of these costs which the drinker’s colleagues or employer, or society as a whole, bear. If the labour market were perfect, the drinker’s income would exactly reflect his productivity: there would be no externality. However, the productivity of individual workers is difficult to measure, potentially resulting in excess pay for the drinker. Estimates of the external or internal costs of alcohol abuse make different assumptions about the extent to which productivity losses are external, from all to none. My judgement is that drinkers bear most of the costs of lower productivity, but that it is likely that others bear some. I suggest a conservative working assumption that 5 percent of the costs of lower productivity are externalities.
The question arises whether damage varies among the beverages (beer, wine, spirits). There are two possible sources of variation:
- a given quantity of alcohol consumed in the form of different beverages, but in the same drinking circumstances, may cause a different level of damage. However, epidemiological evidence shows this is not so: alcohol causes damage, with no distinction among the beverages; and
- different patterns of drinking among the beverages may lead to different levels of damage. In particular, if alcohol is absorbed into the bloodstream slowly, as is more likely to happen, for example, if alcohol is drunk with food, it is less likely to cause intoxication. Claims have been made for each of the beverages that actual patterns of drinking lead to less harm in their case than in the case of the other beverages. However, there is no formal empirical study in New Zealand, as far as I am aware, of such claims. Until robust empirical evidence on this is available, it is reasonable to assume that the costs of drinking a given quantity of absolute alcohol is the same across beverages.
- the CHD benefit is mainly from alcohol rather than the other components of each drink; and
- although some studies show that wine is the most beneficial of the beverages, this is probably an artefact of methodology or due to differences in patterns of drinking or to differing associations with other risk-factors.
In a survey of the cardioprotective effects of alcohol, Edwards et al. state that the consensus among epidemiologists is that such effects are similar among the alcoholic beverages. They also warn against "the comfortable delusion that our own favoured beverage is not really ‘alcohol’, but an essential article of food or an emblem of national virtue (beer or alternatively wine, are often thus viewed in certain countries)".
Light to moderate drinking may reduce the risk of gallstone formation, diabetes, cardiovascular disease and ischæmic stroke, although these findings have not been established as firmly as in the case of the reduction of CHD. Most of the health benefit of alcohol consumption derives from the CHD benefit, not from these other benefits. The positive externality is the cost savings to the public and emotional benefit to third parties (to the extent the drinker does not internalise this benefit) associated with reduced risks of illness and with longer life.
Some commentators have suggested that drinking encourages a relaxed social atmosphere and social cohesion, and this constitutes a positive externality. This argument is highly doubtful:
- it is a subjective judgement that the effect of drinking on the tone of society is benign. Some, rather than seeing "a relaxed social atmosphere", would perceive a undesirable loosening of social discipline, in particular if the comparison in question were increased drinking above current levels, compared with current levels of drinking;
- any positive social externalities would for the most part be restricted to the immediate group of the drinker, which is likely to internalise the benefits. An example is the practice of "shouting" drinks, which may be seen as a means for a group to reinforce the group dynamic of drinking, and hence internalise benefits external to the individual drinker; and
- it is difficult to see why consuming substitutes for alcohol (such as, say, restaurant meals) would generate a significantly less "relaxed social atmosphere".
- chronic and acute personal physical and emotional suffering from alcohol-induced sickness;
- the drinker’s premature death;
- medical costs resulting from alcohol use which the drinker pays. Alcohol can damage nearly every tissue and body system;
- property damage or loss from anti-social behaviour when drunk, if paid for by the drinker;
- costs of lower productivity (through unemployment, poor work performance, loss of promotion opportunities, etc.) which the drinker himself pays;
- loss of personal income and liberty arising from offences committed under the influence of alcohol; and
- reduction in social welfare through impaired capacity to maintain social relationships and to make "rational" decisions.
Internal benefits include æsthetic, social and psychotropic benefits. By standard economic assumption, the marginal utility associated with these benefits reduces with increasing quantity consumed.
- causality is clearer in some costs (e.g., cirrhosis) than others (e.g., some types of social damage);
- some costs are easily expressed in monetary terms (e.g. property damage), while some are not (e.g. the emotional costs to family and friends of someone’s excessive drinking). Studies of the costs of alcohol abuse typically do not include emotional costs;
- the counterfactual is often difficult to specify and quantify. The main approaches are:
- a "prevalence costing", which describes the cost in a given year of alcohol consumption in that year and before that year. The counterfactual is no alcohol consumption before and in the year in question; and
- an "incidence costing", which describes the present value of the cost in a given year of alcohol consumption in that year and future years. The counterfactual is no alcohol consumption in the year in question and thereafter.
To illustrate the difference between the two approaches, consider the cost of a rapidly expanding epidemic: a prevalence costing will be far less than an incidence costing. When the aim of the costing is to cost the benefit from a proposed policy intervention, an incidence costing is appropriate. However, information on costs over more than one year is usually not available. In either case, the counterfactual assumes that goods without significant externalities would be consumed instead of alcohol;
- most monetary estimates of the cost of drinking do not separate internal from external costs;
- unlike, for example, research on many of the medical costs of drinking, research on the cost of externalities must be specific to New Zealand, since the distinction between some internal and external costs is specific to the institutions of each country. For example, some medical costs which are paid for privately in the United States, and hence would be internal, are paid for by taxpayers in New Zealand, and hence are external;
- data on publicly-funded heath costs has until recently been poor; and
- there is currently no quantification of the monetary externality of the reduction in CHD.
Causality is a particularly important issue. It is inevitable that causality cannot be exactly determined for some types of damage, particularly damaging social behaviours, associated with alcohol consumption. Indeed, culture (for example, whether drunk driving is socially acceptable), as well as direct pharmacological effects, influence the behaviour of drinkers. However,
- medical experts can estimate and have estimated the extent to which alcohol causes various physical illnesses; and
- a variety of methodological approaches, such as statistical studies of the incidence of the damage associated with alcohol, psychological studies of the social effect of alcohol and pharmacological studies, implies causality between alcohol and various social behaviours. At the least, it is highly plausible that alcohol is a contributor or catalyst to such behaviours.
As an example, suicide is statistically associated with alcohol. Heavy drinking leads to a deterioration of social ties and is associated with depression, a precursor to suicide. Intoxication may reduce an individual’s self-control and thereby trigger a suicidal inclination. Studies which aim to correct for sample selection bias (the problem that drinkers are those who are for other reasons more likely to commit suicide) still suggest an additional link between alcohol and suicide.
- it understates external costs by excluding non-monetary external costs, such as the emotional suffering of the victims of alcohol-related damage, and the friends and family both of drinkers and their victims, to the extent drinkers do not internalise these costs;
- it understates external costs by excluding the monetary costs of fœtal alcohol syndrome and of property damage associated with alcohol, and by excluding some of the monetary costs associated with the suffering of those abused by drinkers, such as the cost of half-way houses for victims of alcohol-related violence, and the cost to government of the care and protection of children who are victims of alcohol-related violence;
- it minimises external costs by valuing each third-party death at $183,000, a lower bound among valuations of a life. If a $2.0 million value of a life - the value used by the Land Transport Safety Authority - were used, the external cost of deaths caused through drunk-driving would be $258 million, about $234 million higher than if a $186,000 value were attributed to each life;
- it excludes all externalities other than the direct negative externalities associated with consuming alcohol. Examples are the public-health costs directly avoided by reducing CHD and the additional public-health and pension costs associated with longer-lived drinkers who have avoided CHD by their drinking. It is unclear whether this overstates or understates external costs; and
- the data used represent a prevalence rather than an incidence costing. It is unclear whether the prevalence costing overstates or understates external costs compared with an incidence costing. It overstates costs (for example, cirrhosis) which are incurred significantly later than the drinking with which they are associated. However, it would also overstate a monetary measure (for example, the reduction in CHD) of benefits which occur significantly later than the drinking which generate them.
These sources of overstatement and understatement of costs are largely unquantified. However, I judge that it is unlikely that the total effect of the above sources of inaccuracy (including the positive externality of reductions in public-health costs associated with reduced incidence of CHD) would be to reduce the estimate of net externalities below the calculated range. Therefore it is unlikely that the net external costs of consuming alcohol in 1991 were less than, and they may be more than, the range of $432 million to $713 million (in March quarter 1996 dollars). This range thus represents a conservative estimate of the annual net external cost associated with alcohol consumption in New Zealand.
Externalities as a Function of Alcohol Consumption: General Comments
- individuals react differently to the same quantity of alcohol drunk in the same manner, due to differing individual constitutions, past drinking patterns, physical state (for example, tired or not), physiological state (e.g. on medication or not) and emotional state;
- individuals react differently to alcohol depending on whether they consume it with or without food, or slowly or quickly;
- some costs are incurred from the first drink (for example, the risk of car crash), but others only after heavy drinking (for example, a hangover);
- the two sexes are not equally susceptible to some of the effects of alcohol;- some consequences of alcohol consumption are cumulative (for example, cirrhosis of the liver), others are not (for example, the acute effects of a hangover on the next day); and
- some damage results from binge-drinking, some from chronic drinking.
Despite these difficulties, epidemiological, statistical and other studies have related alcohol consumption at an individual level to net damage (i.e., health or other types of damage and benefits). This net damage generates net monetary costs, some of which will be internal and some external. Increasing net damage is likely to be associated with roughly-proportionally increasing net externalities (i.e., external monetary costs). For example, the greater the net damage:
- the greater the health costs paid for by others;
- the greater the emotional costs imposed on others; and
- the greater the loss of productivity paid for by employers and work-colleagues.
Therefore the relationship between alcohol and net damage is a plausible proxy for that between alcohol consumption and net externalities (i.e., the net external costs associated with net damage).
Surveys of drinking patterns, combined with information on the relationship between alcohol consumption and net damage at an individual level, allow inferences to be made about how net damage (as a proxy for net externalities) is related to alcohol consumption at a more aggregated level.
- damage and alcohol consumption; and
- benefits and alcohol consumption.
With regard to damage as a function of alcohol consumption:
- the risks associated with some of the chronic physiological damage associated with alcohol, such as cirrhosis, are an exponentially increasing function of average chronic alcohol-consumption. However, in the case of other chronic risks, such as cancers, the relationship is more linear, in some cases with upward curvature at higher drinking levels. For some pathologies, risks increase at light to moderate consumption levels, in a number of cases by between 10 and 30 percent at a reported average consumption level of about 1½ drinks per day; and
- some types of damage associated with intoxication tend to have "moderately curved [convex] risk functions" (i.e., such damage increases more than linearly with consumption). Examples are assault, suicide and falls. However, the risk of road crash in New Zealand is an exponential function of alcohol drunk, with some increased risk even at low levels of drinking. Even at minimal levels of alcohol consumption there is a risk of damage associated with intoxication.
- if a drinker is not suffering from or will not suffer from CHD, or, if he is at risk of CHD but drinks infrequently and to harmful levels, all drinking confers (expected) net external costs. The (expected) external costs at low levels of drinking are slight and comprise a low risk of a harmful event (such as a car crash);
- damage is a convex function of alcohol consumed;
- if a drinker is suffering or will suffer CHD, the CHD benefit increases for drinking up to one drink on average per day and is approximately constant for higher drinking levels. If such a drinker does not drink in binge sessions:
- drinking up to about two drinks per day (on average) is likely to confer net benefits. This comprises a positive marginal benefit up to one drink per day and a marginal cost above about one drink per day. At about two drinks per day the two effects net out; and
- drinking more than about two drinks per day confers net damage, with increasing net damage at higher drinking levels.
As discussed above (see page 14), the relationship between net damage and alcohol consumption can be used as a proxy for the relationship between net externalities and alcohol consumption. If we use this proxy, we can infer the approximate relationship between (a) internal and external costs and benefits, and (b) individual alcohol-consumption. The following diagram illustrates this for drinkers who would otherwise suffer from CHD who do not binge-drink.
Notes:
internal marginal (æsthetic, social and psychotropic) benefits diminish with quantity consumed;
total marginal benefit includes the external marginal benefit (the reduction in incidence of CHD), which is positive up to about one drink daily. If the drinker would not otherwise suffer from CHD, the internal benefit curve would be the same as the total benefit curve;
the internal marginal cost is the market price of each unit of alcohol, which is constant with consumption, and internal marginal costs which increase with consumption such as hangovers, self-borne medical costs, the risk of sanctions from the Police and justice system etc.;
the external marginal cost is the vertical distance between "internal marginal cost" and "total marginal cost";
Q* is the consumption level at which society’s welfare is maximised; and
Q1 is the consumption level if externalities are not included in the price of alcohol.
We may combine the marginal external cost and the marginal external benefit of alcohol consumption, to produce the marginal net externality function:
Diagram 2 is drawn on the assumption that if a drinker would otherwise suffer from CHD and does not binge-drink, the marginal positive externalities of light drinking exceeds the marginal negative externalities. Whether this assumption is correct for a particular individual depends on the person’s risk of the various causes of death associated with light alcohol consumption, and his risk of CHD. Further, if an individual drinks at a moderate average daily level, but in binges, it is likely that a marginal increase in drinking is harmful at all levels.
- the frequency distribution of per capita alcohol consumption;
- the distribution of total alcohol consumption among heavy and light drinkers;
- the distribution across drinkers of the number of drinks per typical drinking occasion; and
- the distribution of health damage and benefits across age groups.
Diagram 3 illustrates a lognormal frequency-distribution of alcohol consumption. This frequency distribution is found in the per capita consumption of many types of commodities. It is the pattern of alcohol consumption across societies and in particular societies over time, and even among different subgroups of drinkers, such as both those dependent and those not dependent on alcohol. The lognormal distribution also holds for different average levels of per capita average consumption: the reason is that the cultural and social context influences the drinking levels of both heavy and light drinkers.
The frequency distribution of New Zealanders’ per capita alcohol consumption fits this pattern. In Auckland in 1990-92 the distribution of individuals by average daily drinking levels, among drinkers aged 14 to 65 years (87 percent of the population was in that age range) was as follows:
| percentile of drinkers | average number of drinks per day |
| 96% - 100% | 10.4 |
| 91% - 95% | 4.6 |
| 76% - 90% | 2.5 |
| 51% - 75% | 1.2 |
| 1% - 50% | 0.2 |
Although the consumption levels of the top 5 percent and 10 percent of drinkers may seem implausibly large, they are consistent with other New Zealand and overseas surveys, and below even average population consumption levels in some periods of history. It is also consistent with reported typical drinking levels per drinking occasion: among male drinkers, who consume about 71 percent of alcohol, 32 percent have six or more drinks on a typical drinking occasion, and 16 percent have nine or more drinks on a typical drinking occasion.
We also have information on the distribution of per capita alcohol consumption in New Zealand by age. Drinking levels among the young increase rapidly with age, peaking in the 20-24 age group, and then gradually decline with advancing age. Harmful drinking patterns (such as binge-drinking) are more common among the young than the middle-aged or elderly.
These statistics build up a picture of the distribution of net damage, the proxy for net externalities, at a population level:
- a small proportion of drinkers drink a large share of total consumption. Since damage is a convex function of alcohol consumption, their share of total damage is more than their share of the total number of drinkers. Survey findings confirm this;
- conversely, a large proportion of drinkers drink a small share of total consumption. Their share of total damage is less than their share of the total number of drinkers. However, although a light and moderate drinker has less damage than a heavier drinker (the convex damage function), because of their large numbers such drinkers may nonetheless experience in aggregate a larger share of total damage from alcohol than heavy drinkers. In particular, the bulk of problems associated with intoxication occur among light and moderate drinkers;
- 60-65 percent of drinkers drink about a drink per day (on average) or less. Only about 35 percent of these 60-65 percent of drinkers (that is, 21-23 percent of all drinkers) would otherwise suffer from CHD. Thus up to 21-23 percent of drinkers potentially benefit from marginal drinking. However, those of this group who binge-drink probably have a negative marginal externality from drinking;
- most alcohol drunk is likely to confer net harm, but most drinkers are drinking at a level (under two glasses daily) which would confer net benefit if they would otherwise suffer from CHD and if they do not binge-drink;
- since the distribution of alcohol consumption is to some extent fixed, policies which reduce consumption among particular subsets of drinkers, such as heavy drinkers, are likely also to reduce consumption among light drinkers. Equivalently, it is likely to prove difficult to change the drinking levels of only one part of the drinking-distribution curve;
- it is a property of the lognormal distribution of a per capita alcohol consumption that, unless damage were a declining function of per capita drinking levels (which is not the case), a reduction in per capita drinking levels would have a more than proportional reduction of damage associated with alcohol consumption; and
- the young do not generate health benefits in their young years from drinking alcohol, since they do not suffer from CHD when young. However, they do generate damage, mainly from intoxication. On the other hand, although the elderly suffer chronic damage from chronic alcohol use, they benefit from the reduction in CHD from drinking both currently and in earlier years. The following table shows an estimate of the effect of consuming alcohol on the number of deaths in New Zealand in 1987:
| Age Group | |||||
| Effect of Alcohol | 0-14 | 15-34 | 35-64 | 65+ | All Ages |
| Deaths Caused (+) | 25 | 266 | 416 | 757 | 1464 |
| Deaths Prevented (-) | - | -5 | -446 | -1429 | -1880 |
| Net Deaths Caused | 1678 | 13,384 | -318 | -5219 | 9525 |
Although in total more deaths are saved than caused, the key number of the table is that in 1987 there were 9525 person-years lost due to alcohol. This arises because, in terms of deaths caused or prevented, the young are harmed and the elderly, with fewer years to live than the young (on average), benefit from drinking.
Externalities lead to an inefficient allocation of resources. Drinking alcohol is associated with diverse negative positive externalities. Both negative and, probably, positive externalities relate to consumption of absolute alcohol, irrespective of the beverage (beer, wine, spirits).
A conservative estimate of the annual net external cost associated with alcohol consumption in New Zealand is a range $432 million to $713 million (in March quarter 1996 dollars).
When considering policy instruments for reducing the inefficiencies associated with externalities, it is useful to know the functional relationship between alcohol consumption (both at an individual and more aggregated level) and net externalities. In particular, it is useful to know the distribution of marginal net externalities of alcohol consumption, since policy instruments (other than prohibition) affect drinking at the margin.
At an individual level, marginal negative externalities (measured in absolute dollars, as illustrated in diagram 1 by the vertical distance between marginal internal cost and total marginal cost) are an increasing convex function of chronic and non-chronic alcohol consumption. However, there is a marginal positive externality in the consumption range of up to about one drink daily (on average) for drinkers who would otherwise suffer from CHD and do not binge-drink. They can also reduce their risk of CHD by other means (for example, by exercise). Whether this marginal positive externality exceeds the marginal negative externality for these individuals depends on each drinker’s risk of the various causes of death associated with light alcohol consumption, and his risk of CHD. Except for those who generate a positive net externality in this drinking range, the marginal net externality associated with drinking is negative and is related to alcohol consumption by a concave negative decreasing function at all levels of drinking, as illustrated in diagram 2. Even for individuals who would otherwise suffer from CHD who do not binge-drink, the negative externality outweighs the positive externality at average drinking levels of about two drinks per day or more.
The frequency distribution of per capita alcohol consumption in New Zealand, as in other countries, is lognormal. This being so, although most drinkers are drinking at a level (under two glasses daily) which would generate positive externalities if they would otherwise suffer from CHD and if they do not binge-drink, most alcohol drunk is likely to generate negative externalities.
The few very heavy drinkers generate a disproportionately large share of total negative externalities. However, because of their numbers, light and moderate drinkers may nonetheless generate a larger share of total negative externalities than heavy drinkers.
Competitive markets produce an efficient allocation of resources only if consumers are perfectly informed. If consumers of alcohol are poorly informed about the internal damage from consuming alcohol, they might consume more than they would if they were well informed. If this were so, some form of government intervention to reduce consumption may be justified, in part because the public pays the external costs resulting from excessive drinking.
Although consumers in all markets lack perfect information, an approximately-efficient outcome results. This applies to complex purchases with long-term effects, such as cars, houses and pension schemes. I am aware of no empirical evidence that consumers of alcohol lack information to some unusual extent which would justify some policy intervention. A 1982 British survey found 90 percent of the surveyed population were aware that alcohol could damage health. Moreover, common intuition suggests that it is common knowledge that alcohol consumption in more than moderate amounts damages health.
I conclude that we do not have evidence that lack of information significantly impedes an efficient allocation of resources with regard to alcohol consumption.
Competitive markets allocate resources efficiently only if consumers are rational. If a large number of drinkers were significantly irrational, some form of government intervention may be justified, in part because the public pays the external costs resulting from excessive drinking.
There are two main types of irrationality associated with alcohol consumption. One is the long-term irrationality of the dependent drinker. Economists have developed "rational addiction" models which suggest that seemingly irrational behaviours may be consistent with stable preferences and utility maximisation over time. The second type of irrationality is a short-term irrationality from intoxication. Again, intoxication may be rational risk-taking: the drinker rationally decides that the benefit to him from intoxication exceeds its costs. However, unless "irrationality" is to lose all meaning, it cannot be denied that some proportion both of dependent drinking and intoxication is irrational.
An excise tax would be a poor instrument for addressing the problems of information and irrationality, to the extent these problems exist. Better-targeted policies are, in the case of lack of information, to provide information, and in the case of irrationality, to implement appropriate programmes directed at potentially irrational drinkers, or in extreme cases to force abstention or appoint an agent, probably through some form of medical or custodial care.
Whether an excise tax on alcohol improves welfare therefore depends on whether it reduces the inefficiency associated with externalities. Other instruments available for achieving this aim are:
- - assignment of property rights;
- - regulation;
- - persuasion; and
- - provision of information.
The following sections discuss the assignment of property rights, which has been proposed in recent years as a preferable alternative to the excise, and the alcohol excise. The discussion of the alcohol excise assumes that current regulation, persuasion and provision of information are broadly unchanged from the present situation.
Cameron and Dwyer advocate assigning and enforcing property rights as a first-best means to reduce the inefficiency associated with externalities. This involves assigning all the costs of consuming alcohol to the drinker. If this happened, there would be no externalities. Cameron and Dwyer propose the following policies, along with supporting policies, to achieve this:
"charging most users for the health care they consume (with a commensurate reduction in taxation and an increase in private insurance arrangements). There would be grounds for the government to provide a modest level of services on a targeted basis or to require compulsory insurance at a prescribed minimum level (if the free-rider problem would otherwise be excessive)";
"accident insurance should be subject to competition, the right to sue should be reintroduced, and the Accident Rehabilitation and Compensation Insurance Corporation should be set up on a competitively neutral basis as a state owned enterprise and subsequently sold". This aims to enable costs from accidents committed by alcohol consumers to be paid by the offending drinker or his insurance company; and
"compensation should be payable by offenders to victims [of crime] to recompense them".
These policies are unsatisfactory as means of reducing the inefficiency generated by the externalities associated with alcohol consumption, for the following reasons:
- each of the policies is a possible response to complex, wide-ranging and difficult public-policy issues in the sectors concerned. These issues would need to be considered in deciding policies. Equity concerns and complex market-failures are likely to make approaches which focus only on the allocation of property rights undesirable;
- since private insurance is collectively financed, it does not necessarily lessen the externalities associated with alcohol consumption. Since the cost of identifying and monitoring those at risk of incurring health and other costs from consuming alcohol would be prohibitive, adverse selection and moral hazard ensure that the alcohol-related costs of insured drinkers would to some extent continue to be paid by others. We may speculate that if insurers had the power to tax, their search for a means of charging a higher premium related to the risk of incurring costs from consuming alcohol might lead them to impose a tax based on the amount of alcohol consumed;
- unless government were prepared to see medical and other services not provided to drinkers, it would be obliged to compel private insurers to cover drinkers and drinkers to buy insurance. Such a regulatory régime is likely to be complex and to generate inefficiencies;
- the transactions costs of enforcing property rights would be large. An often-cited example is the transactions costs of suing for private injury, currently precluded under the ACC scheme; and
- no likely New Zealand government will accept them, at least in toto;
Taxation: The Optimal Excise (Theory)
The Optimal Excise: Representative Individual
A partial-equilibrium analysis demonstrates that, if a representative consumer reduces alcohol consumption when its price increases (the empirics of this are discussed below), an excise tax will change economic welfare by:
- reducing the consumer surplus of all alcohol consumers;
- reducing the positive marginal net externality associated with non-binge drinkers who drink on average up to about one drink daily and would otherwise suffer from CHD;
- reducing the negative externality associated with all other drinking; and
- raises revenue, which increases the welfare of society.
When each of the above welfare changes is taken into account, resources are efficiently allocated - welfare is maximised - if an excise is applied per unit of the good equal to the net marginal external cost of consuming the good at the consumption level at which total (internal and external) marginal costs and benefits are equal (the optimal consumption-level). This is intuitively plausible: it replicates the standard microeconomic price equilibrium at which welfare is maximised for goods without externalities, except that in this case a tax offsets the effect of externalities.
It in general ensures that drinkers face an incentive to adjust their consumption to the level at which total (internal and external) marginal costs equal total (internal and external) marginal benefits.
However, the most accurate information a policy-maker is likely to have about the level of externalities is about that generated by current, rather than optimal, consumption. Therefore, a move to the optimal excise must be based either on an inevitably imprecise one-off estimate of the net marginal externalities at the estimated optimal consumption-level, or on an iterative process.
The same external costs should not be used to justify more than one tax.
The net external marginal cost may - as in the case of alcohol - be the total of more than one type of cost, for example, health costs and productivity costs. However, in the case of costs which may be mitigated by the victim’s action, the calculation of marginal cost should not in general include both the costs of relieving suffering and the residual suffering after the relief. This is because the sufferer or society acting on his behalf would purchase relief to the level where the marginal cost of the relief equals the marginal cost of not reducing the suffering: this is the marginal benefit of which should be included in the net external marginal cost. The estimated costs in the section "An Estimate of the Monetary Cost of Externalities" (see pages 11-13) are calculated on this basis.
Diagram 4, a modified version of diagram 1, illustrates the optimal excise when the relevant range has marginal net externalities that are negative.
If no tax is imposed, the individual consumes Qno tax. The cost to society of marginal consumption (the vertical distance from the horizontal axis to the "total marginal cost" curve) exceeds the benefit to the individual (the vertical distance from the horizontal axis to the "internal marginal benefit" curve). If the optimal excise tax (t*) is imposed, the consumer consumes Q*. The total cost to society and the individual reduces by A+B+C. The total benefit to the individual reduces by B+C. Society’s welfare increases by A. A similar analysis shows that society’s welfare is reduced if:
- an excise greater than the marginal external cost is imposed (thigh). In this case consumption reduces to Qhigh tax, below Q*; and
- an excise lower than the external cost is imposed (tlow). In this case consumption is at Qlow tax, between Q* and Qno tax. The cost to society of the individual’s alcohol consumption exceeds the benefit to him, but the difference is less than if the individual drank at Qno tax. Diagram 5 illustrates.
Although marginal and total externalities at Q* are lower than at Qlow tax, because marginal negative externalities (the vertical distance between "internal marginal cost" and "total marginal cost") are an increasing function of alcohol consumption, t* is necessarily larger than tlow tax.
The Optimal Excise: Heterogeneous Individuals
Since drinkers differ in their drinking patterns and other characteristics, the marginal net externality associated with each drinker’s marginal consumption to differ. This requires the analysis of the preceding section to be extended.
If government had perfect information and were able to perfectly price discriminate, it would equate its tax or subsidy of each drinker to the net marginal externality of consumption of that drinker. However, this is not feasible.
As a second-best tax, the uniform subsidy or excise (that is, uniform across numbers of drinks and across drinkers) which will maximise aggregate welfare is equal to the weighted-average across drinkers of their marginal net externalities of alcohol consumption at drinkers’ socially-optimal levels of consumption (Q* in diagram 5), the weights being the price derivatives of demand of drinkers. That is, the more a price change affects an individual’s absolute consumption and the more a given change in his absolute consumption affects the net externalities he generates, then the more important is that individual in the calculation of the optimal excise.
Since the optimal uniform excise is based on marginal (not average) externalities, it would only be coincidental if it recovered the total net external costs of alcohol consumption, the latter being equal to the average (not marginal) net externality of consumption multiplied by the quantity of consumption.
To set the optimal uniform excise, information is needed on:
- the distribution across individuals of their marginal net externality of alcohol consumption; and
- the weighting to be applied to each drinker’s marginal net externality of alcohol consumption. The weights are the price derivatives of demand of the drinkers.
Exact information on these matters is unavailable, and we therefore cannot calculate the optimal excise.
However, we do have the information which will allow us to determine whether the weighted-average marginal net externality at drinkers’ socially-optimal levels of consumption (the optimal uniform excise) is likely to be more or less than their average net externality at present consumption levels (the current excise). The following section discusses this empirical information.