Chapter 2 from the International Guidelines for Estimating the Costs of Substance Abuse (2ed) written by Eric Single, David Collins, Brian Easton, Henrick Harwood, Helen Lapsely, Pierre Kopp and Ernesto Wilson. I did the first draft of this particular chapter, and then we collegially revised it. (WHO, 2001) see http://www.eastonbh.ac.nz/?p=108 and http://apps.who.int/bookorders/anglais/detart1.jsp?sesslan=1&codlan=1&codcol=15&codcch=561)
Like other professions, economists rely on a substantial body of common understanding, which they assume in their professional communications. As a consequence, non-economists can be confused by the writings and conversations of economists, even though these communications may be perfectly intelligible within the profession. This problem of communication with outsiders is especially unfortunate where the issue involves other professions, as is the case in the multidisciplinary field of substance abuse.
Thus it is appropriate to begin the discussion of economic cost studies by elaborating for the non-economist the implicit assumptions that the economists use. Simplification has been necessary, which means that some of the subtlety of the professional discourse may be lost. Obviously then, this discussion cannot cover all the points in the guidelines, nor does it superseded them. But hopefully it will enable those from other professions involved with substance abuse to have a better insight into the issues which trouble economists.
The discussion in this section is organized around three interrelated topics: what economic cost studies are, what different approaches may be taken to estimating economic costs and what economic cost studies are not.
2.2 Economic cost studies
In brief, the study of the economic costs of problems associated with the use of psychoactive substances is (1) a type of cost-of-illness study (2) in which the impact of substance abuse on the material welfare of a society is estimated by examining (3) the social costs of resources expended for treatment, prevention, research and law enforcement, plus (4) losses of production due to increased morbidity and mortality, plus (5) some measure for the quality of life years lost, relative to a counterfactual scenario in which there is no substance abuse. Each part of this statement bears elaboration.
2.2.1 Economic cost studies: a type of cost-of-illness (COI) study
The evaluation of the economic and social costs of substance abuse belongs to the genre of cost-of-illness studies (COI). Superficially a COI study involves combining an epidemiological database with financial information to generate an amount valued in monetary terms which purports to say something about the costs to society of a particular disease. Typically the magnitude is large, or large enough, to be used to draw attention to the condition as one to which policy makers, research funders, and researchers, ought to pay attention.
It will be clear from the intensity in which economists debate each calculation, that they have in mind some conceptual framework. The total is not just some gee-whiz figure designed to give a significant place to this or that illness in the public debate. So what are economists think they are doing? Moreover, why do they disagree?
At the heart of the economist’s approach is that all relevant costs are opportunity costs, that is one activity (such as an illness) prevents resources being used for some other purpose, and so an opportunity is forgone. Thus COI studies rest on the proposition that if the illness were not to exist, then the resources that a society uses for treatment and other related purposes could be deployed in some other way.
Sitting behind the opportunity cost is a counterfactual scenario, that is, a description of an alternative state of affairs, by which the opportunity cost would be assessed. Often the counterfactual proposition is not controversial. For instance we might assume in a COI study of some viral infection, that the alternative scenario was no viral infection.
In substance abuse the alternative can be more arguable. For instance, the counterfactual to a situation of alcohol abuse might be that the abusers switch their consumption to mineral water and other health enhancing commodities, or it might be that abusers switch their consumption to narcotics. The latter is an extreme example, and usually a COI study assumes a switch to non-damaging activities, but sometimes the specific counterfactual situation is unclear.
However, it is not the counterfactual proposition which causes the most dispute between economists. More often the disagreement arises over what should or should not be included as a cost, as well as how that cost is to be valued or measured. These differences do not arise because of different underlying fundamental frameworks, but because the common framework has been applied in different ways, so practical considerations have led to the treatment of the same issue in different ways.
This is true for COI studies. The fundamental framework is “value theory”, the role and interpretation of market price, which has been developed rigorously over the post-war period. Two practical applications of that theory are (1) in the System of National Accounts (SNA) and (2) cost-benefit analysis (CBA), which is a method of evaluation of alternative actions or treatments.
An integral assumption of value theory is that consumers value their own consumption, and that they rationally seek to maximize the value of their consumption as best they can, subject to various limitations such as their income and borrowing power. Thus, it is assumed that when a person buys a potato, a beer or an illicit drug, the cost of the purchase is offset by the benefits the consumer obtains from its use. Many who are knowledgeable about alcohol, tobacco and drug dependence would challenge the veracity of this assumption. Addictive behaviour seems to violate the assumption of rational consumer behaviour. How then is the economic analysis to deal with this situation?
One approach is to treat the psychoactive substances as conventional commodities, assuming that even dependent users are consuming rationally, according to their lights if not that of wiser counsel. In this case the perceived benefits of consumption exceed the outlay on the substance, and the transaction is treated as rational, just as purchases of other commodities, such as potatoes.
Collins and Lapsley offer another approach, which attempts to modify the assumption of rational consumer behaviour in value theory, without destroying the entire paradigm.  They estimate a proportion of drug consumption which is judged to be abusive. For that portion of consumption, they treat drug expenditure by users as zero under the counterfactual scenario in which there is no drug abuse, on the basis that dependent users receive no benefit from use. (Indeed, many users wish they had never taken up drug use.) This permits them to count expenditures on drugs by dependent users as a cost in the current actual scenario. This is a plausible alternative, which does not undermine value theory, although it leaves the difficult task of determining what proportion of substance use is dependent use.
The problem of how to resolve additive consumption with the assumption of rational consumer behaviour is not fully resolved. The key point for the non-economist is that a main objective of the economist’s approach in economic cost studies is to retain the well-established value theory paradigm, but to adapt it for the consumer behaviour which addiction implies.
2.2.2 Costs to whom?–cost studies in the System of National Accounts framework
Newspaper stories often cite “costs” of various social problems or negative events, such as the loss of a business convention, the impact of poor weather on tourism or the dire economic consequences of a professional sports strike. Frequently, such estimates lack credibility, as costs are magnified by following the flow of dollars from one party to the next, with little or no consideration of alternative uses for the money. For example, the money spent by a tourist or baseball fan at a restaurant is counted, then the money spent by the restaurant to food wholesalers, then the farmer’s income, as well as income and other taxes paid by workers at each stage, and so forth. The fact that the money spent by local residents could have been used for alternative purposes is not considered. It is not surprising that cost figures such as these are viewed with scepticism. A cost for one person is typically a benefit to another. To the business manager who no longer has pay for his or her staff to attend the cancelled conference, to the potential tourists who stayed at home and to the baseball fan whose game was cancelled by a strike, the “costs” of these cancelled events are really savings.
In short, to be credible, estimates of the costs of substance abuse must be clear with regard to what constitutes a cost, who bears these costs, and the boundaries which should be placed on the economic ramifications of negative impacts.
Cost-of-illness studies are quite precise in this regard: they estimate the impact of illness on a measure of material welfare in a society, closely related to the Gross Domestic Product. The GDP is generated in the System of National Accounts (SNA) by combining expenditure and production data with accounting information to produce an aggregate statistic valued in monetary terms. The Gross Domestic Product is not an arbitrary set of decisions about what is to be included and how each item is to be valued. Rather the aim is to encompass all market transactions valuing them at their marginal private value (or utility), which is usually equal to the market price (including indirect taxes). An increase of so many monetary units in the GDP can be interpreted as an increase in the sum of consumer utility of the same amount of monetary units.
In order to appreciate the significance of COI estimates in this framework, consider the counterfactual situation where people choose not to eat potatoes, but switch their expenditure to other products (say pasta). Clearly there will be a disruption among potato producers, but this will be offset by an expansion of pasta production. We take these two effects as (largely) balancing out, in which case there will be no change in GDP, even though there is a change in the composition of GDP.
What, we might ask, is the cost-of-potatoes to the economy? An answer might be that it is the cost of production and distribution. However, this is offset by the benefits to consumers of eating potatoes. More formally, we see that in the counterfactual situation, where there are no potatoes (but there is more pasta), there is no change to GDP. So we assess the social cost-of-potatoes to be zero. Note that we are not here assuming that the cost of potatoes are zero to consumers. They are a real cost to them but it is assumed to be offset by the benefits to them of the potatoes, and which is taken into account when they make the (private) decision to buy them.
2.2.3 What constitutes a cost?–social vs. private costs
Thus, where the costs of a commodity are largely limited to private costs, the economic impact is estimated at approximately zero. None of this should appear extraordinary. What is unusual is when we consider the same situation as it applies to a psychoactive substance prone to abuse, which carries social costs as well as private costs.
To simplify, we shall illustrate the argument with tobacco, because it is probably the simplest of all the drugs, and its economic impact is the most transparent.  This time our counterfactual scenario is that there is no tobacco consumption, and there has been none in the past, so that smokers switch their consumption to some standard commodities (such as potatoes). In effect we are assuming that tobacco was never introduced to the society under consideration, and that potential smokers did not choose another drug (such as cannabis).
At first it might appear that this story was no different from the one about potatoes. But tobacco generates ill health, which requires medical care. The lack of smoking would mean that a significant quantity of medical resources would no longer be need for the care of smoking induced sickness, and could be used for some other purposes. (It would also result in other savings, such as the cost of cleaning up litter and the costs of smoking-related fires.) The counterfactual is a little vague on what exactly is the alternative, but in the context it is not likely to matter. What is critical here is that we have a resource use consequent on the smoking, which is not being offset by some benefit to smokers when they decided to smoke.
The terminology which is being used here differs from that used in most of the Economics literature on the subject. What we here call “social costs” are usually called “external costs” or “externalities” in the literature. Costs which accrue only to the people engaged in the activity in question (for example, the consumption of alcohol or tobacco) are called private costs. Thus the normal economic terminology is that private costs plus external costs equal social costs. It can be argued, however, that to the layperson the use of the term “social cost” to describe the costs which are imposed upon the rest of society is more intelligible. Thus, in the rest of this publication the phrase “social cost” has this meaning (that is, what the economist calls “externalities”).
Why is so much attention paid to the distinction between private and social costs and benefits? As the Australian Productivity Commission report on gambling states (1999, p. 43), it is not because private costs are unimportant. In fact, often they are far more significant than the social benefits and costs of an activity. Rather, private costs generally do not justify government action on the basis that:
- individual actions based on adequately informed and rational decision-making will generally accord with the best interests of the individual concerned;
- if there are no impacts on other people resulting from these actions which are not accounted for, then what is in the individual’s best interests will also be best for society; and
- if this is the case, there is no way that governments could intervene in individuals’ decisions that would improve the welfare of either the individuals concerned or society more broadly (Australian Productivity Commission, 1999).
Thus the existence or private benefits and costs does not normally provide a justification for government intervention, unless the distribution of private benefits and costs is seen to be in conflict with society’s concept of fairness. It can be argued that if lack of fairness is a problem it would be more efficient for the government intervention to take the form of broad social security or tax measures rather than measures targeted specifically at the activity under review. It may, however, be that a particular public sector intervention designed to reduce external costs will influence the size and distribution of private benefits and costs. In these circumstances public policy should not ignore private benefits and costs.
The key distinction appears to be between the private costs which rational and fully-informed smokers incur by their own activities, and the social costs which are borne by others. However, the distinction is in reality more complex than this. Consider the case of a substance abuser, say a smoker. Conventional economic analysis of consumer behaviour assumes that rational consumers will undertake an activity only if the private benefits received at least equal the private costs of that activity, so that there is almost certainly a positive net benefit in the form of what is known as “consumer surplus” (the difference between what consumers would be willing to pay for a good or service and the market price that they are actually required to pay). Consumers are better off, in their own estimation, as a result of the consumption activity.
But this analysis refers quite specifically to the costs as perceived by the consumer. What if consumers (say smokers) are uninformed or misinformed as to the costs which the consumption imposes on them? For example, smokers may not be aware of the full health consequences of smoking (perhaps as a result of ignorance or of misinformation resulting from advertising) or they may not realise that the highly addictive nature of nicotine means that quitting will turn out to be much more difficult than they expected.
If the smokers’ actions are determined by perceived costs that are less than their actual costs, the difference between the two is a social cost even though it is borne by the smokers themselves. This is because the smokers have not adjusted their behaviour in response to these unperceived costs and so these costs are unaccounted for. The smokers are not necessarily behaving irrationally. They are simply adjusting their behaviour to the best available, relevant information. It must be emphasised that costs borne by the substance abusers themselves can represent social costs if these costs have not been knowingly incurred.
It is sometimes suggested that these types of study should also estimate the extra value to the abusers (for example, the benefits to smokers of smoking) over and above the costs to them of that activity, and that these net private benefits should be set off against the social costs. However, from the point of view of public policy, it is social costs that are relevant, not private costs. In determining the appropriate levels for society of any activity, government is interested in the costs that this activity imposes on the rest of the community. As an illustration, in determining the appropriate levels of activities such as pollution, environmental degradation or even violence, society does not take into account any private benefits that the perpetrators may enjoy. They are seen as being irrelevant to the interests of the community as a whole. In the same way, studies of the social costs of substance abuse should estimate only the net social costs.
In COI studies, only the social costs are considered. Some of the medical costs may fall upon the smoker if, e.g., there is a co-payment for public care or if smokers pay higher medical insurance premiums. Other institutional arrangements may require similar careful distinctions between the payments the smoker contributes to medical care (and other expenses) and the payments from other sources. These costs are not part of the costs-of-illness, but a part of the private costs borne by smokers, just as they pay the costs of their cigarettes.
Social costs may be incurred by other persons in the private sector (e.g., when private insurance premiums are increased due to payouts to smokers) as well as by public sector expenditure. Thus, in the context of COI studies, “social” is not a synonym for “public”, nor “private” for “private sector”.
Another issue which arises in relation to drug abuse, is whether costs imposed by abusers upon other members of their own family constitute private costs or social costs. On the one hand, it is argued (or asserted) that potential substance abusers will take into account the effects on other family members in deciding the extent of their substance abuse and that these costs are, therefore, internalised as private costs. On the other hand, how can we ignore the costs of substance abuse upon other people who have had no part in the initial decision and who may find the effects intolerable (for example, resulting in marriage breakups)? The size of abuse cost estimates will depend very significantly on whether family costs are treated as social costs. Practical questions will also arise about what constitutes a family member – de factos, same sex partners, in-laws etc. It is, in fact, difficult to believe that the effects of substance abuse on other family members should be considered to be solely private costs.
Measuring the social costs of substance abuse is no easy matter. There is strong evidence, for example, that the consumption of alcohol is related to a variety of health consequences. The probability and severity of adverse health effects of alcohol are strongly related to level of intake, often in a non-linear fashion and sometimes in a manner which is also situation dependent (as with regard to accidents). The dose‑response relationship is most evident with respect to cirrhosis of the liver, but adverse effects of high intake of alcohol have also been found for many other disorders including delirium tremens, impaired brain function, cancer of the esophagus and digestive tract, chronic calcifying pancreatitis and congenital defects in the fetus among pregnant women. High and even moderate alcohol use is also associated with increased risk of trauma, such as that caused by impaired driving accidents.
The proportion of each of these causes of morbidity and mortality which can be attributed to alcohol use must be estimated, ideally for different age and gender groups. Where large-scale population based epidemiological studies have established the relative risk of particular disorders at different levels of alcohol consumption, the attributed fractions of alcohol-related morbidity and mortality can be determined with a fair degree of confidence. In many situations, however, such studies are lacking, and one is forced to estimate the attributable fractions from less reliable sources, such as studies of the excess morbidity and mortality among clinical populations. In such cases, one is forced to make the dubious assumption that rates of morbidity and mortality in the general population of heavy alcohol users can be estimated from clinical populations.
For other adverse consequences of alcohol use, the issue of causality can be even more daunting. For example, consider a person who consumed alcohol prior to committing a crime. Even if this person had been intoxicated, it is not clear whether the crime can be attributed to alcohol consumption. The alcohol may have caused the person to become aggressive or less inhibited, or precipitated the crime in some other fashion. On the other hand, the person may simply have happened to have a few drinks before engaging in a crime which he or she would have committed anyway. Alternatively, the person may have already decided to commit the crime and used the alcohol “for courage”. Thus, even when drinking immediately precedes a criminal act, the attribution of alcohol as a causal factor in the crime is not at all clear.
Furthermore, the attributable fractions for each disorder vary between societies and within societies over time, so no one set of attributable fractions can be applied to all societies. Thus, the assignment of medical costs associated with adverse health consequences arising from the use of a particular substance such as alcohol is a very complicated and difficult task.
2.2.4 Further costs: productivity losses
The reduction in medical expenses in the counterfactual scenario is not the only important change as far as GDP is concerned. Total production may be increased because the former substance abusers are more productive at work, with lower morbidity and lower absenteeism. This additional production under the counterfactual scenario is the productivity loss from substance abuse in the actual situation. 
It is possible that to a certain extent, smokers may carry the burden of the lost production themselves, e.g. in lower remuneration. In practice, however, it seems unlikely that the entire burden of the productivity loss is carried by the smoker, and that at least some is carried by the employer in lower profits, by other employees in lower wages, and/or by the taxpayer in lower tax receipts. These losses are a part of the COI.
2.2.5 The ultimate cost: placing a value on life itself
There is one further major difference between the actual situation and that of the counterfactual which factors into economic cost studies. Substance use and abuse may cause death. Compared to the counterfactual scenario, the population is less. The resulting lower production should be included as part of the COI, by considering the loss of income due to premature mortality.
But what are we to do about the deaths of those who are not in the workforce, such as homemakers and the retired? It is insufficient to ignore this loss of life due to substance abuse. There must be an explicit recognition of the different life years experienced under the two scenarios. As discussed in Section 4, a dollar value can be assigned to the labour of persons outside of the workforce, and in the case of retired persons, some measure of the value of life years lost must be assigned. Better still life years should be adjusted for the quality of the living experience. Someone suffering from terminal cancer is not experiencing the same quality of life as their non-smoking equivalent who is leading a full life.
It is not easy to value these life years or quality life years (QALYs). Insofar as this is an appropriate thing to do, the difference between the actual situation and counterfactual scenario is a part of the COI.
The notion of placing a dollar value on human life is troublesome to many. Some cultures and religions could not contemplate doing so. What right have economists to place a dollar value on life?
Unfortunately, when it comes to policy advice, an economist cannot always avoid putting some value on life. Consider the question of whether to install traffic lights at a crossroad, one of which effect would be to reduce accidents which lead to deaths. If the evaluation ignored lives saved by the lights, that would be equivalent to treating the value of life as zero. As a result some life saving traffic systems would not be recommended. On the other hand if the value of life was set as infinity, every traffic system which reduced the probability of death, no matter how small that probability, would be installed, with the result that we could barely move given the density of life saving traffic lights.
So in practice we incorporate some value of lives saved, when we make policy decisions, even if a dollar value is not stated. All economists are doing explicitly is what others, policy advisers and policy makers do implicitly.
The issue of placing a value on human life cannot be avoided by ignoring the issue, for that would be equivalent to setting the value of life at zero. However, because economic cost studies do not strictly offer policy advice, they can avoid the issue by enumerating the number of years of life lost due to premature mortality without placing a dollar value on those years. For example, the result might be reported that the annual cost of a particular illness was $100 million plus 10,000 quality life years lost. That meaning would be that under the counterfactual scenario, there would be $100 million of extra resources for consumption, and 10,000 additional quality life years saved.
However the cost of another illness might be $50 million plus 20,000 quality life years. Some may wonder which illness is the more costly, a question that can only be answered by combining the dollars with the quality life years in some way. Whatever way would be equivalent to putting a value on life. 
Another difficulty is that it is not clear that all life should be treated equally. This is especially pertinent in the context of substance abuse. Does the life of a chronically unemployed drunk driver have the same value as that of a young victim killed by the drunk driver? Does the life of a junkie have the same value as that of a productive, law-abiding citizen? The question of valuing the life of the junkie, compared to a good citizen, may turn out to be trivial, providing the counterfactual is kept mind. Suppose the counterfactual is to eliminate the substance abuse. Then the counterfactual scenario has the junkie as a good citizen, and her or his death is just as great a loss to society. Alternatively one might want to say the loss of the junkie’s life is much less valuable than that of the good citizen, because the quality of life is lower. But, in addition, the counterfactual scenario is about the recovery of that low quality life to a standard one. Thus the total valuation, summing the two components, will be the same as the loss of the good citizen’s life. The COI study includes the value of the existing damage to the life of an addict, as well as mortality effects.
2.3 Demographic approach vs. the human capital approach
There are two different approaches to the estimation of the economic costs of substance abuse: the more widely adopted “human capital” approach and the more recent “demographic” approach. The key difference, discussed in detail in Section 4, concerns the manner in which the costs of premature mortality are treated. In the human capital approach, the lost value of a deceased worker’s production is estimated by present earnings plus a discounted rate of future earnings. In the demographic approach compares the actual population size and structure to that of an “otherwise healthy” population, i.e. an alternative population in which there were no drug-related deaths.
The key point is that these different approaches are complementary rather than contradictory. The demographic approach addresses the question: “Suppose there had never been any substance abuse or problems associated with the use of psychoactive substances?” The human capital approach addresses the question: “Suppose all substance abuse and problems associated with the use of psychoactive substances were to end today?” The human capital approach generates an estimate of the present and future costs due to drug-related mortality in the current year, while the demographic approach estimates the present costs of drug-related mortality in past and present years.
Because these two alternative approaches to the estimation of the economic costs of substance abuse address different questions, it should not be expected that they would arrive at the same answers in all circumstances. During a period of increasing or decreasing consumption, one would expect somewhat different results. It is only during a prolonged period of stable consumption with no major impact from treatment or prevention programming that one would expect to achieve equivalent results.
Thus, there is no need to reconcile the two approaches. The need is to be clear about their origins and significance. The choice depends on the counterfactual situation being addressed. The preferred procedure will often be to conduct economic cost studies which utilize both the demographic and human capital approach, and compare the results. 
2. 4 Prevalence vs. incidence based approaches
Estimates of the economic costs of substance abuse may be either prevalence-based or incidence-based. Prevalence-based studies estimate the number of cases of death and hospitalisations attributable to substance abuse in a given year and then estimate the costs that flow from those deaths or hospitalisations (as well as other costs, such as prevention, research and law enforcement costs). Incidence-based studies estimate the number of new cases of death or hospitalisation in a given year and apply a lifetime cost estimate to these new cases. Thus, prevalence-based estimates generally measure the costs of substance abuse in the present and the past in a given year, while incidence-based studies generally estimate the present and future costs of substance abuse in a given year. For ongoing health and social problems such as illicit drug use, the results of prevalence-based and incidence-based estimates are often similar. For health problems that are declining in magnitude (such as smoking in some countries), prevalence-based estimates will generally be lower than incidence-based estimates. For emerging health issues such as epidemics of HIV or Hepatitis infection, incidence-based estimates generally provide higher estimates than prevalence-based estimates, because many infected persons may still be in the latency phase of the diseases. The use of prevalence-based vs. incidence-based estimates is discussed in Section 3.11.
2.5 What economic cost studies are not
Part of the appeal and desire for economic cost estimates of alcohol, tobacco and other drugs may unfortunately be based on confusion with other types of economic analyses. While useful and relevant to policy decisions, COI studies are not studies of avoidable costs, they are not studies of budgetary impact nor are they cost-benefit analyses.
First, economic cost estimates do not indicate the amount of money and life years which could realistically be saved via effective government and social policy and programming. The counterfactual situation in economic cost studies is one in which there are no problems associated with the use of psychoactive substances. This counterfactual situation is hypothetical and generally not realizable under any circumstances. The estimated costs include both avoidable and unavoidable costs. Even if completely effective policies could be found with no appreciable costs for enforcement, treatment and prevention programming, implementation would not be instantaneous and there would still be lingering adverse consequences from past use of the psychoactive substances. The calculation of avoidable costs associated with the use of psychoactive substances is discussed in Section 4.
Second, economic cost studies are not studies of the budgetary impact of alcohol, tobacco and other drugs on governments. The costs included in COI studies are in reference to the whole of society and just to the government accounts. A study of the economic costs of substance abuse would be very useful in conducting an accounting of the budgetary impact of psychoactive substances, as it would provide estimates for many of the government outlays. However, government costs do not include all of the costs imposed on the community. Further, budget impact includes consideration of government revenues and other benefits, which are not part of COI studies. The relationship between economic cost studies and estimates of budgetary impact is discussed further in Section 4.
Finally, economic cost studies do not attempt to fully consider the economic benefits of alcohol, tobacco and other drugs, and they should not be confused with cost-benefit or cost effectiveness analyses. These are two in a range of a range of tools that economists and others use to evaluate policy proposals. In the medical area they are frequently used to evaluate the usefulness of costly treatments or policy proposals (such as prohibiting drinking in certain circumstances) by, for example, weighing the costs of interventions against their benefits.
Cost-benefit analysis is based on the same value theory as COI studies, and economic cost studies can be used to provide important cost components in a cost-benefit analysis. However there are slightly different assumptions which means that cost-benefit analysis may give different outcomes and estimates. The most important differences involve the counterfactual scenario, and the treatment of the non-market sector.
Typically the cost-benefit analysis asks what would happen if the costs associated with a particular behaviour – such as tobacco smoking – were to cease from today. This contrasts with the counterfactual scenario in a COI study, which is to ask what would happen if the smoking had never started. Even if all smoking were to stop instantly there would still be the consequences of past smoking on mortality, morbidity, and health care. For instance the public sector would still be required to provide assistance for those who smoked in the past and were in need of medical care. Such social costs are unavoidable, and so are not included as a cost in the typical cost-benefit analysis. 
The other major difference between cost-benefit analyses and economic cost studies utilizing the SNA framework is that cost-benefit analysis has been concerned with the impact of an event on non-market activities. For instance, if the a problem of substance abuse involves keeping patients in hospital beds, and a counterfactual scenario of returning them to the community, the cost-benefit analysis usually includes the extra unpaid work that might be involved in the second scenario (as when extra house and care work is imposed on family members).
The extension of economic cost studies to cover such unpaid and non-market activities does not represent a major difficulty in principle, but it has not been given priority in development. If economic cost studies were extended to include unpaid and non-market activities and if they were able to distinguish avoidable from unavoidable costs, it would be enormously helpful for those who wish to carry out cost-benefit analyses of alcohol and other drug policies and programmes. For an economic cost estimate is almost the benefit side of a cost-benefit analysis, and if done properly it could be readily adapted into the full benefit side.  The reason why the COI study is close to the benefit side of a cost-benefit analysis is that the avoidable costs associated with the use of psychoactive substances represent the benefits (i.e. negative cost) in a cost-benefit analysis contrasting the current situation with a counterfactual situation in which a policy or programme is introduced. Thus there are potentially strong practical advantages to integrating economic cost studies with cost-benefit analyses.
Thus economic cost studies, while based on the same value theory, involve differences with cost-benefit analysis. Some of those differences can be eliminated with development. There is no ultimate reason why the economic cost studies in the SNA framework should ignore non-market activities. Experience derived from cost-benefit analyses in dealing with non-market activities and estimating avoidable costs will hopefully inform the further development of economic cost estimates in the SNA framework.
Finally, it should also be noted that COI studies, like most forms of economic analyses, only concern economic costs to the legitimate market economy. There may therefore be significant economic costs in drug-producing countries arising from substance misuse that are not measured in the COI framework. For example, the costs of corruption are not generally included in COI studies. Nor do COI studies generally attempt to measure costs arising from the economic disruption to legitimate business enterprises caused by large-scale illicit drug production and distribution. To do so would require a more extensive and demanding economic framework such as general equilibrium modelling.
2.6 Interpretation of substance abuse cost estimates
Estimates of the aggregate costs of substance abuse tend to attract a great deal of political and public attention. However, while the meaning of individual components of the aggregate costs (for example, the costs of health care or crime) is relatively straightforward, the interpretation of the aggregate estimates requires great care and precision. To understand this point we need to return to the distinction between the human capital and demographic approaches to estimation.
Both approaches relate to the valuation of the loss of production arising from the abuse-related deaths of otherwise productive members of society. Both approaches compare production and abuse costs in the actual situation with those in a hypothetical alternative situation which would have existed had there been no past or present substance abuse. The difference between the two approaches relates to the way in which the production costs of premature mortality are treated.
The essential difference between the two approaches is summarized earlier in the following way. The human capital approach calculates the present and future production costs of abuse-induced deaths which occur in the present year. The demographic approach calculates the present production costs of abuse-induced deaths which have occurred in past and present years.
When looking at the human capital approach we are estimating the present value of the future time stream of lost productivity resulting from abuse-induced deaths. Although we talk of the “costs of abuse in year X”, in reality a high proportion of these production costs will be borne in years subsequent to year X. In relation to the demographic approach, we are looking at the costs actually borne in year X but resulting from deaths not only in year X but also in many years prior to year X.
Thus, interpretation of aggregate estimates is difficult, but unfortunately there is no way round this problem. If we calculated the costs borne only in year X as a result of deaths only in year X, the resulting costs would be very substantial underestimates of the overall costs borne by society since we would not account for the fact that deaths can impose costs over many years, not just in the years in which they occur.
One implication of the way in which abuse costs are estimated is that the aggregate figures are not likely to change significantly over short periods of time. This is because rates of abuse and disease prevalence, the primary determinants of abuse costs, tend to change slowly. Thus it may well be a waste of research resources to undertake these estimates at intervals of less than three to five years.
 Theoretical Issues in Abuse Cost Estimation, paper prepared after the symposium.
 Even so we ignore the effects of passive smoking, in order to keep the story as simple as possible.
 Again the concept needs to be treated with care. Implicit in the counterfactual scenario is that the potential production from the greater productivity of the substance abuse is realised or, insofar that it is not, the potential is taken up in voluntary leisure with the same value as the additional production (and not involuntary unemployment).
 Pressed between the economic and policy logic, and cultural and religious or spiritual sensitivities, participants at the International Symposium on Estimating the Social and Economic Costs of Substance Abuse could not resolve the question of the treatment of the valuation of life. Instead it was suggested to use a deliberately clumsy term of social gains from additional (quality) life years. The term “social” is not meant to connote a gain in a religious sense, but to indicate that society may (or may not) value any improvements in the quality of life as a result of reduction is substance abuse. In making this suggestion the proposers were aware they were putting the matter into a temporary limbo, rather than ultimately resolving the philosophical issue. That will depend upon a wider range of professions than even those at the symposium. After that resolution economists can turn to the question of the best valuation method, if any. In the interim a number of methods are advocated, their choice depending on the resolving the deeper philosophical issue.
 There is another variant of COI studies which might be considered a third approach. Manning and his colleagues (1991), in contrast to the studies by Dorothy Rice (Rice, 1966; Rice et al., 1985; Rice et al., 1986; Rice et al., 1990; Rice et al., 1991; Rice, 1993), strictly limits cost estimates to external costs (paid by others). Furthermore, Manning’s external cost approach is incidence-based and utilizes a lifetime model of use. Thus the cost-benefit totals represent the current value of present and future substance use.
 Assuming that it had done so in the past.
 Over time the unavoidable costs diminish, so that the CBA counterfactual scenario has a growing stream of cost savings from the smoking cessation. Typically these are discounted to give a present value of the avoided costs. Discounting amounts to summing together all those costs, but the further a cost is in the future, the less weight it is given, because it is generally taken that income and spending in the future is less valuable than the same activity in the present. Typically the weighting involves a discount rate, whose magnitude is a matter of contention among economists, although there is widespread agreement that the concept is correct in principle. It follows that not only should the costs of medical care be estimated through time, discounting them to a present value, but so should other costs such as productivity and mortality losses.
 A common difference is that the treatment the evaluator is looking at, may reduce but not eliminate the illness, so only a proportion of the COI items will be relevant. Sometimes the proportion will differ from item to item.