Further Developments in Estimating the Social Costs Of Substance Abuse


The views in this report on the Avoidable Costs of Substance Abuse Workshop (Ottawa June 21-22) are my own and do not reflect those of the others involved. It focusses on issues particularly pertinent to New Zealand. The paper was presented to a seminar of officials on 7 August, 2005. Comments Welcome [1]

Keywords: Health;


This report is on The Avoidable Costs Workshop held in Ottawa, Canada June 21-22, under the sponsorship of the Office of Research and Surveillance, Health Canada (Bureu de la rechercher and de la surveillance, Sante Canada).

The purpose of the workshop was to review and develop the preliminary draft of Developing International Guidelines for Estimating the Avoidable Costs of Substance Abuse by David Collins and Helen Lapsley, with supplementary papers on epidemiological and criminological issues by Jurgen Rehm and Serge Brochu respectively. Other papers were also reviewed.

Around 19 researchers attended from 10 countries. The meeting was chaired by Louise Déry, of the sponsoring office.

The workshop is a continuation of a series which began in 1994, and which has led to the major publication International Guidelines for Estimating the Costs of Substance Abuse 2ed (WHO 2003) [2] (all of whom attended this workshop). Subsequently, International Guidelines have been used to calculate the social costs of substance abuse in a number of countries including New Zealand. [3]

The 2005 workshop was concerned with a number of new issues – particularly avoidable costs and the social costs of crime – which can be expected to be included in any 3rd edition of The International Guidelines. [4]

Theoretical Background

An early decision was made to anchor the work in the ‘theory of value’ the economists’ account of costs and prices. The most important reason for doing so is that we can draw upon the entirety of economics when confronted with a novel situation. Or a silly argument, because there is considerable rhetoric in the area of drug abuse where advocates grab invalid arguments which seem to support their case. [5]

While we do not seem to be able to prevent neophytes from misusing economics, any more than we can stop teenagers smoking and misusing alcohol, one can but hope that mature debate will reflect the coherent framework that economics offers, just as we hope teenagers will mature into sensible adults.

While the term ‘costs’ indicates that the research program has a considerable input from economists, as will soon be evident, other disciplines are also involved, especially epidemiologists and – increasingly in the future – their equivalents in criminology.

Opportunity Costs

At the core of the economist’s notion of the ‘cost’ of a resource is its opportunity costs that is the value of the next-highest-valued alternative use of that resource. It is what the community forgoes when the resource is used for its current purpose.

The notion is a subtle one, but at its simplest if someone consumes a muffin for which they paid $1.00 they have given up the opportunity to consume something else which cost $1.00: $1.00 is its opportunity cost. Note the consumer may value the muffin in excess of the $1.00 – perhaps they might have still purchased it were it $1.50, so they have the benefit of additional value over the resource cost to society. (We know, incidentally, that the consumer must normally value it at a minimum of $1.00, since had they not, they would not have purchased the muffin, but held onto the dollar to purchase something more worthwhile.)

The analysis becomes more complex for products where not all the resources used are included in the commercial cost – suppose the muffin maker was polluting the local stream. Drug abuse is riddled with these ‘externalities’. Indeed an economist might argue that there is no ‘abuse’ unless there is an externality – that is a cost to someone else that the consumer does not pay – or the probability of one. Thus a glass of wine, say, does not generate an externality, but a couple of bottles does if the imbiber goes off in a car and crashes or beats his wife. The first glass is a ‘use’, by the fifth, say, is it ‘abuse’. Where there are externalities there will be both public costs borne by others in the community, as well as private costs borne by the consumer.

The total private and public costs that the drug abuse generates are usually called ‘social costs’. Because the social costs of drug abuse are high relative to private costs that there is much public intervention. The economic ideal would be a price for the drug which reflected full social costs, say by the addition of a tax. In practice this is not always possible because some drugs are illicit, some consumption behaviour is irrational, and because it is not practical to impose the optimum tax (since for instance the external cost differs between the first drink and subsequent ones).

The Counterfactual Scenario

If costs are about alternative uses, it is necessary to define the next best alternative, especially where there is as complicated phenomenon as drug abuse. This alternative is the called ‘counterfactual scenario’.

However, it is important to appreciate there is no uniquely right counterfactual scenario. This is illustrated in the case of alcohol consumption. Some researchers’ scenarios have been based upon the assumption that there is no unsafe drinking, while others have investigated the implication of there being no drinking of alcohol at all – safe or unsafe. Perhaps the distinction here is the former refers to the social costs of alcohol abuse, the latter to the social costs of alcohol. Whatever, it is important to be clear about the scenario, since it is usually one of many.

Another complication that there may be a displacement from one drug use to another. For instance some countries’ ban on alcohol appears to have increased its consumption of cocaine. Usually (and often implicitly) the scenario posits that there is no displacement, but this may not always be a reasonable assumption. At least displacement should be discussed when describing the details of the counterfactual scenario.

In principle then the researcher need to identify precisely the counterfactual scenario, although those who use the research are often more casual.

In practice the choice of counterfactual is often determined by the data availability and other technical issues. When I did the New Zealand estimates, I used the assumption that there had never been tobacco or alcohol in New Zealand, and there was no displacement. (There was insufficient data to do illicit drugs or misuse of pharmaceuticals).

An important alternative counterfactual scenario is where everyone stops using (or abusing) the drug, but that there is a carryover from past use. We shall see that this leads to the ‘avoidable cost’ estimates which the workshop was about.

Constructing a Counterfactual Scenario

In the researcher’s mind eye, if not actually in the published document, there is a table with the actual scenario in one column and the counterfactual in the other. Line by line, the table highlights the differences between the scenarios. For instance an analysis of smoking will have deaths from lung cancer in the left hand column of the actuality (a lot of them), and deaths from lung cancer in a counterfactual scenario where, say, nobody has smoked a (few of them). The table has an infinite number of rows in principle, but in practice identical rows are omitted. Consolidation (as in the causes of death) leads to between one and two dozen rows.

Constructing a Counterfactual Scenario

Actual Counterfactual
Deaths from Lung Cancer = Deaths from Lung Cancer =
…. ….
…. ….

The table entries need to be quantitative estimates. They are not largely the work of economists, but are derived by other disciplines, most notably epidemiologists, but also social statisticians , sociologists and others. The entries should include those by criminologists, but until recently they were omitted because there was no useful data. I return to the crime dimension in a later section.

In principle the collection of many of the data entries in the table is a bit mysterious to an economist – it is in practice too. Data sources can be quite ad hoc. (In New Zealand there may be only one, and one prays that the work is reasonable quality, it being difficult to independently check).

Epidemiological Fractions

‘Epidemiological fractions’ are at the core of the health impacts of drug abuse, and increasingly important in the criminological estimated.

I am not an expert on epidemiological fractions. I doubt any economist is. I have sat through intense and focussed – even heated – discussions between epidemiologists, and not understood the details of the points at issue. (I have watched those same epidemiologists looking equally mystified as economists have rowed over some subtleties in the theory of value.) Here is my short explanation, using lung cancer to illustrate the general principle.

People suffer and (usually) die from lung cancer. The scientific evidence is that the main cause of lung cancer is tobacco smoking (including passive smoking). However there is a very low incidence of lung cancer among people who have never smoked, and appear not to have been passive smokers. What is needed for two scenarios is the number of people with lung cancer (on the actual side) and the number who would have it were there no smoking (assuming that is the counterfactual). Miraculously, epidemiologists calculate these ‘epidemiological fractions’.

As I understand it they compare the incidence of the disease between those who smoke and those who do not smoke – after adjusting for the effects of passive smoking. From this they can infer the proportion (epidemiological fraction) of those who die (or suffer) from long disease as a result of smoking. Since the incidence and intensity of smoking varies from country to country (as well as by cohort, gender, social class …) the fractions vary from country to country too. So a country aspiring to do a social cost study needs its own epidemiological fractions.

If they dont exist, the social costs cant be done, except by borrowing the fractions from other countries, which is not particularly robust. To foreshadow a later section, how do we get the equivalents fractions for crime?

Valuing the Scenarios

Once we have got the tabulation comparing the two scenarios, (that is when the hard grind has been done by the data gatherers) then the economics come into play (and have the fun?).

The process might be described as follows. To the tabulation of actual and counterfactual scenarios add three more columns. In the first new one (the third) put the difference between the two columns. In the second new one (the fourth) put the price relevant to the activity shown in the row. Multiply the two, putting the product in the fifth and final column. Sum the final column and voila!, one has the social cost difference between the two scenarios – that is the social cost of the drug abuse.

Calculating the Social Cost

Actual Counterfactual Difference Price Value
…. …. …. …. ….
…. …. …. …. ….
…. …. …. …. ….

Column Sum of Final Column = Social Cost Difference between Scenarios

Choosing the correct prices is complicated. Basically the price should reflect the resource cost of the activity – what it is worth to society – or, where a resource is not directly involved, of how it is valued by society in resource costs terms. That means the sum total represents the difference in resources used in the two scenarios, valued in social prices. So if the figure is say, 3.0 percent of GDP, then we may say that material standards of living would be 3 percent higher under the counterfactual scenario.

The Value of Life

One complication is how to treat deaths and poorer quality of life as the result of drug abuse. The scenarios will have different mortality and morbidity histories. Without going into the technical details, the differences can be summarised in a measure called ‘quality adjusted life years’, or QALYs. Substance abuse results in significant loss of QALYs, both from early death and also form an inferior quality of living while alive. .

We can calculate the reductions in QALYs. That is tricky, and on the technical frontier, but it is not impossible. More contentious is the valuing of a QALY. The issue is agonisingly set out in the International Guidelines. What must be said is that the valuation of life can not be zero, for that would mean we never made sacrifices to save life; it can not be infinite because that would mean we would never risk life no matter how small that risk; and the somewhere in between is not easily decided.

However it would seem that any realistic choice for a value of life results in a social cost which is very large, say a large proportion of GDP. That is because GDP only covers the value of material goods, whereas the social cost includes the value of life which is not in GDP. My view is that while life quality must be valued, and the aggregate can be quoted, the better comparison is to state that the social costs are equivalent to a reduction in the material standard of living (the tangible costs) by X percent of GDP, and a reduction in the aggregate quality of life (intangible costs) by Y percent of the total.

Interestingly, X is often sort-of close to Y as the following tabulation illustrates, although there is a differing relative importance between material consumption and life between the two:

Social Costs of Drug Abuse [6]

  Units Tobacco
Tangible Costs (X) % of GDP 1.7 4.0
Intangible Costs (Y) % of Total Life Quality 3.2 2.0

The Use of Social Costs Calculations

The precise meaning of the social costs of substance abuse estimate depend upon the precise counterfactual scenario. In practice any estimates prove to be large, evidenced by the much tinier estimates for other diseases (although they do not always use the advice, methods, or rigour of the International Guidelines). The estimates are usually used as ‘Gee whiz’ figures to draw attention to the seriousness of the public policy issue.

As a general rule there is still not enough consistency to make robust international comparisons (use instead the direct epidemiological figures), and comparisons through time in a country are often invalidated by improvements in data sources and methods. The estimates have also been used by dividing the aggregate value by some physical consumption figure (say kilos of tobacco or an illicit drug, or litres of absolute alcohol) to get a unit social cost of a drug.

The International Guidelines also recommend that the aggregate cost be disaggregated to where the costs fall directly on the user, households, government and business., which can be useful for policy purposes.

A not always appreciated role of the aggregate figure is that it reveals deficiencies in the data. I found in my 1995 study that the poorer quality of lives as a result of alcohol misuse seem high (of a similar order of magnitude to the mortality costs), but that there is virtually no data on them. Alas there is still none. In the case of road accident costs outside the health system, there were none and I had to borrow the Australian data.

Sometimes the estimates indicate that some effects are very small. I spent a tedious day chasing up the costs of fires caused by smoking to discover that they are trivial compared to, say, health costs, despite the wont of the anti-tobacco lobby to emphasise them.

However the biggest New Zealand data lacuna is that there is hardly anything useful on illicit drugs (including crime and justice effects). Studies from other countries suggest that they are probably a small proportion of total costs of drug abuse, because their consumption is much less widespread than consumption of the licits. However they fall very heavily upon some groups and services in the community, and so have a very high unit cost.

In any case, size is an indicator of public importance, it is not an indicator of whether or what public initiatives should be taken. That is where avoidable costs come in.

Avoidable Costs

Consider the following counterfactual scenarios:
Scenario 1: There has never been any tobacco consumption in New Zealand;
Scenario 2: All tobacco consumption ceases from a particular point in time.

The second scenario will have a number of consequences not evident in the first. For instance there will still be a overhang from those who smoked in the past and suffer from health disabilities (and mortality) after the everybody-ceases-smoking day. The differences between the two counterfactual scenarios, valued in the standard way, is the ‘avoidance cost’.

There is an important difference between the two scenarios. Scenario 1 (there has never been any smoking) has broadly the same values for every year. However in Scenario 2 the values will vary over time, perhaps (on some items anyway) at first rising as the disease consequent on the abuse rise, and then falling off, as the diseased die off. Thus avoidance costs vary over time.

This leads to the need to consolidate the stream of costs. There is a straightforward economic procedure: ‘discounting’, a kind of weighted averaging of each year, giving greater weight to years closer to the cessation decision. We need not go into it here, except to say there is a well established economic procedure and that care needs to be taken between comparing the capitalised value of the stream of costs and an income stream such as GDP. [7]

More contentious is the discount (weighting) rate. [8]

The principle of avoidable costs was well known to the Working Party, and is mentioned in International Guidelines. However it is only recently that the epidemiological data base has become available to pursue them effectively. What is needed, in the case of tobacco, is quantitative estimates of mortality and morbidity if the subject stops smoking. This is far harder to calculate than the traditional epidemiological fractions, since it usually requires following the health course of a group of smokers, ceased smokers and non-smokers.

While there is increasingly good information on the time profile of smoking cessation, there is far poorer data on those for alcohol and illicit drugs. The workshop spent a considerable amount of time discussing alternative strategies (such as identifying some long term level and assuming a path to it). At this stage, those outside the intense discussion of the workshop (future ones will be conference calls) may wish to observe that there is considerable uncertainty, which wont be really addressed until some actual estimation is done. In this area, the new guidelines are likely to be permissive, rather than directive.

Policy Evaluation and Avoidable Costs

Thus far the discussion has been in terms of total cessation, and so may be thought of as a ‘macroeconomic’ evaluation. However, because the approach of the International Guidelines has been rigorously based on economic theory, the method and underlying theory is almost identical with that which is used for standard evaluations such as cost-benefit analysis. So the macro-statistics may in whole or part be useful for evaluating policy interventions.

At the simplest level consider a policy which results in ten percent abandoning smoking (or even one which applied to a particular group results in ten percent abandoning smoking). Ten percent of the macro-avoidable cost estimate (on a per capita level where a group was concerned) may be the return on this intervention, and so with very little extra work the cost to benefit analysis (CBA) can be completed, and the policy’s economic effectiveness evaluated (that is, whether the policy outlay can be justified, in standard CBA terms).

I have said ‘may’ here because there may be complicating factors. For instance, suppose 10 percent of the group ceased smoking and the remainder of the group reduced their consumption by 20 percent. The analysis would need to take the reduction into account (or take 10 percent of the macro figure to provide only a minimum estimate of the intervention return).

In my view the most powerful use of the avoidable costs scenario will be the basis for evaluation of policy interventions, which involves a useful extension of the social costs approach from that set out in the second edition of the International Guidelines. Far too many policy interventions seem based upon ‘it seems like a good idea’ rather than any careful analysis of their effectiveness. Rarely is there a post-implementation assessment, which reflects this casualness. However the application of CBA does depend on the availability of the micro-data for each policy intervention.


To begin with a caveat emptor. Over the year, economists have had little to say useful about crime. The economic studies I have read are either pathetically trivial or arrogantly ignorant. This is not because the issues are unimportant, but that economists have not really been able to offer useful insights in the area (in contrast to that what they have done for health). Thus the following remarks are made by an economist working in very unfamiliar territory. [9]

The working party which produced the International Guidelines was well aware of the significance of drugs in crimes, but at the time the report was prepared it, could not see any coherent way of obtaining the required data. To illustrate the problem with an example of the costs of prisons.

The epidemiological fractions depend upon identifying those in hospital (or who have died) who smoke (or whatever) against those who dont by each disease category. However in the case of prison the exact question is much more complicated, because causation is more complicated. Consider the following statements by prisoners :

“I take drugs, but they are not the reason I am in prison.”
“I am in prison, because I committed a crime because I take drugs (say robbery) to pay for my addiction.”
“I was under the influence of drugs when I committed the crime.”
“I took drugs in order to commit the crime.”
“I am in prison because I committed the crime of possessing drugs, but for no other crimes.”
And so on.

Would one even believe the prisoners’ account of causality? A paper presented to the workshop by Serge Brochu, professor of criminology at the University of Montreal, Drug-related Crime: Definitions and Avenues For Reduction, indicated how difficult the exercise is. [10]

Brochu has designed a questionnaire which he says gives the proportions in prisons (or whatever) who are there because of drugs (analogous to those who are in hospital because off smoking). If the approach is successful. the epidemiological fractions can be calculated and the same techniques used for health can be applied to criminally associated expenditures.

I do not have the skills to evaluate the success of the procedure. My impression is this is front line research, and not yet widely adopted within the criminology profession, in contrast to the widespread acceptance of the epidemiologists’ approach. However, the method has been applied in Australia, with Collins and Lapsley including the results in their estimates of the social cost of drug abuse in Australia. What we know from their work is that the crime induced social costs from drugs are significant contributors to total costs of substance abuse.

Brochu also included in his paper reports of meta-analyses of studies of various programs aimed at reducing or abandoning the consumption of drugs. They give the impression that criminologists have done better than health professionals in evaluating their interventions. Some popular interventions may not be very effective, and may not cross the approval threshold of a cost-benefit analysis, once dollar values are included. But first we need the macro-avoidance of cost studies to provide the framework.

One theme in the workshop was the needs of poorer countries. A particularly troubling class of countries are those where drugs are a central part off society, politics and the economy. (For example Colombia.) Fortunately their problems are not directly relevant to New Zealand. [11]


It cannot be over-emphasised that good quality social cost estimation involves good quality data bases. Especially for illicit drugs, and for the licits to a lesser extent, some of the data is hard to collect. An interesting contribution to the workshop was from Eric Single on the problem of under-recording of alcohol consumption.

Conclusions and Directions

The primary purpose of the workshop was to progress the development of international guidelines for estimating the avoidable costs of substance abuse. The draft paper is to be revised and circulated, after comments by the workshop. I shall continue to be involved.

However this report back enables one to reflect on what steps New Zealand might next take.

1. Do nothing.


2. Have we sufficient new data to make it worth updating my 1995 study?

3. Have we sufficient data to include illicit drugs in an updated study? Should we systematically collect some of it? (Especially in the area of crime, given that public spending is rising.)

4. It was suggested that there should be country pilot studies of avoidable costs for particular drugs. It was proposed that perhaps New Zealand should do the tobacco use study. Could we then go on and do alcohol misuse, perhaps coordinated with David Collins and Helen Lapsley, who are based in Sydney?

5. What institutional framework would best contribute to any active answers to the above questions? Who should be involved?

Go to top

[1] Attendance at the conference was made possible by Health Canada’s Office of Research and Surveillance (Bureu de la rechercher and de la surveillance, Sante Canada). Previous attendance at workshops was made possible by Alcohol Advisory Council of New Zealand (ALAC)and the Canadian Centre for Substance Abuse (CCSA). This paper has been commissioned by the Office of the Commissioner of the New Zealand Police which is not responsible for any opinions or error in it. I would also like to express gratitude to Louise Déry for chairing the workshop, and her team at the Office for administrative support, to my colleagues at the workshop for sharing in a challenging and rewarding meeting, and to the paper presenters, especially David Collins and Helen Lapsley, for interesting their contributions.
[2] E. Single, D. Collins, B. Easton, H. Harwood, H. Lapsley, P. Kopp & E. Wilson (2003) International Guidelines for Estimating the Costs of Substance Abuse 2ed (Geneva. WHO)
[3] B.H. Easton (1995) The Social Costs of Tobacco Use and Alcohol Misuse (Wellington, Department of Public Health, Wellington School of Medicine)
[4] A couple of terminological issues: the expressions ‘substance’ and ‘drug’ are used interchangeably in the literature, partly reflecting national practices. Sometimes the term ‘abuse’ is interchanged with ‘use’ or even ‘misuse’. This partly reflects the impacts of different drugs, nicely captured in the title of my report ‘The Social Costs of Tobacco Use and Alcohol Misuse’. The area is divided into licit drugs (alcohol and tobacco) and illicit drugs (narcotics). There is also others areas, hardly studied by economists, such as the misuse of pharmaceuticals, household substances such a glue, and party pills. Some countries pay less attention to tobacco.
[5] For instance ‘What about the contribution of all the tobacco farmers?’ Aside from that contributing something which may have no positive value is hardly of public worth. the standard economic model assumes that the farmers switch to an almost as lucrative crop, so there is little impact on the value of farming. Where the switch is not easy – as for tobacco growing in Mali – there is a straightforward method to deal with the situation.
[6] B.H. Easton (1995) op. cit. p.26.
[7] It is better to convert the capitalised value back to an annuity.
[8] Which is one of the advantages of the first (never any abuse) scenario, since it avoids the need, because every year is the same. The annuity approach, however, reduces the overall effect of the choice of discount rate.
[9] The workshop got into a complicated discussion of the case for penalties for possession of illicit drugs: resolving it is important because it reflects upon the treatment of some of the valuations. In its course there popped up the novel idea, to me, that making drugs illegal was a means of reducing pressure on the health system, thereby transferring a problem from heath to crime. The thought that here was an interesting tradeoff analysis, which economists could contribute to, occurred to the economists around the table, although it was outside the scope of the workshop.
[10] S. Brochu (2005) (Paper for Avoidable Costs of Substance Abuse Workshop (Ottawa June 21-22).
[11] A weakness of the workshop was that Brochu was the only criminologist there, in contrast to a number of economists and epidemiologists from various countries. This is not to demean Brochu. But I have learned much about the robustness of epidemiology from the debates between epidemiologists. I do not have the same knowledge of criminology. The under-weighting of criminological representation reflects, I think, that we are still coming to grips with the general issue.

Go to top