Conceptual problems with studies of the social cost of alcohol and

Research
report
HANS OLAV MELBERG
Conceptual problems with
studies of the social cost
of alcohol and drug use
ABSTRACT
H. O. Melberg: Conceptual problems with
studies of the social cost of alcohol and
drug use
■ Existing studies of the social cost
of substance abuse give very different
answers. Part of the reason for the
differences is that the contributions
Introduction
differ in their definition of costs and
The literature on the social cost of substance
the types of costs that are included
use dates back to the debates about prohibi�
or excluded from the study. Closer
tion in the early twentieth century (Fisher
attention to the concept of cost
1927; Österberg 1983). It is a controversial
employed in these studies also reveals
field with some arguing that the methodology
some weaknesses with many of the
is now “well accepted within the scientific
existing studies. The policy relevant
community” (Kopp & Fenoglio 2001) while
concept of cost necessarily implies a
others argue that the calculations represent
comparison between all realistically
“an exercise in hubris” (Reuter 1999). This
expected consequences (both positive
paper is located on the sceptical side of the
and negative) of two policy alternatives
controversy and it makes three contributions
and many cost studies do not use the
compared to previous research.
concept of cost in this way. Finally, it is
First of all it looks at how developments in
argued that cost estimates in this area
the economic theories of substance abuse af�
cannot be neutral or scientific because
fect the debate. For instance, Ainslie’s work
the results depend on our views on
on hyperbolic discounting and addiction is
when a choice can be characterized
important to the estimation of costs because
as voluntary and rational, how we
it undermines the assumption of stable pref�
deal with the problem of inconsistent
erences ��������������������������������������
preferences, and which preferences we
cause costs are related to the satisfaction – and
consider to be legitimate. Since there is
lack of satisfaction – of preferences. If prefer�
no agreed consensus on these issues,
ences are instable, then the estimation of cost
the cost estimate necessarily becomes
will differ depending on which of the prefer�
subjective.
ences the researchers pick in the calculation.
■ KEYWORDS
Social costs, alcohol, illegal drugs,
definition of cost, cost of illness,
submitted 15.01.2010;
initial review completed 22.01.2010
final version accepted 09.03.2010
conceptual problems, revealed
preferences, weakness of will
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7 . 2 0 1 0 . 4
287
Conceptual problems with studies of
the social cost of alcohol and drug use
Second, the essay connects the litera�
However this is only one of several prob�
ture on the cost of substance abuse to the
lems and although using the correct cost
theoretical literature on the concept of
concept would be an improvement, it does
cost. The weak link between applied stud�
not prevent other problems with cost stud�
ies and theory has been noted in the ex�
ies which are discussed below.
isting literature, for instance by Atkinson
Third, and most importantly, this pa�
and Meade (1974), Markandya and Pearce
per draws more heavily than is usual on
(1989)������ E
�����������������������
.�� ����
H
� �
the philosophical literature on the nature
ever, the mentioned articles focus mainly
of self, free choice and legitimate prefer�
on the distinction between social and pri�
ences ������������������������������������
(E
�
vate costs as opposed to the definition of
cause cost estimates will differ depending
cost itself. In classical economics the con�
on whether one views the consumption of
cept of cost was very much debated (Bu�
����
drugs as a disease or a free and informed
chanan 1969). For some it was a measure
choice and whether we exclude some pref�
of pain associated with an alternative, for
erences from the calculation of costs and
others it measured what was sacrificed
benefits. It is also important because some
when a choice was made. These debates
economists working on the topic tend to
are interesting in themselves, and the two
assume away or ignore these problems.
conceptions of cost seem to clash also in
For instance, in an otherwise excellent
modern studies of the cost of substance
critique of cost studies in the related field
use, but the point of bringing this into the
of gambling Walker and Barnett �����������
discussion is not to re-run the old debates.
gue that it is important to understand the
Instead, the point is to take the theory of
economics of costs since this “removes the
cost that became dominant – the “Austri�
subjectivity in the classification of patho�
an” opportunity cost concept – and exam�
logical gambling costs.” I question this
ine the implications of this for the study
conclusion by investigating some of the
of the cost of substance use. As I show in
underlying philosophical problems such
the second part of this paper, such a theo�
as free will, rational choice, weakness of
retical investigation makes it easier to un�
will and legitimate preferences.
derstand the different answers found in
Taken together these arguments imply
the empirical studies. Some, for instance,
that many existing cost studies are of little
only try to estimate external costs, while
use. They do not agree with each other, they
others try to estimate social costs. Closer
use a policy irrelevant concept of cost and
attention to the concept of cost, however,
in any case the conclusion reflect more an
also reveals some weaknesses with many
opinion than a neutral estimate of a fact.
of the existing studies. I shall argue that
The conclusion however is not all negative.
the policy relevant concept of cost neces�
While it may be useless to aggregate all the
sarily implies a comparison between all
costs to a single estimate of social costs,
realistically expected consequences (both
some of the less aggregated information is
positive and negative) of two policy alter�
useful for policy purposes. For instance, for
natives and that several cost studies did
the purpose of making well informed policy
not use the concept of cost in this way.
decisions it is useful to know more about
288
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
the causal affects of drugs and alcohol on
rational deliberation. If the latter is the
the tendency to commit crimes or to incur
case, no amount of empirical sophistica�
various illnesses. It is less obvious that it is
tion and advance will help in producing
useful to convert all of these consequences
a scientifically valid estimate of the cost
to monetary units and aggregate them.
of substance abuse. Hence, the main ques�
Different studies
– different cost categories
In this section I want to give some exam�
tion below is what to include in an esti�
mate of the social cost of substance use,
not the problems involved in the actual
measurement.
ples of cost-studies and how they differ.
At a general level the cost categories of
The aim is mainly descriptive. In other
many studies in this field can be catego�
words, I want to point out a few areas of
rized under four headings: 1. Lost earn�
disagreement and the issues involved
ings; 2. Crime/Law enforcement; 3. Health
without necessarily taking a position on
(Treatment – direct and indirect, preven�
who is right or wrong. The aim is not to
tion, research); and 4. Other effects (ac�
be comprehensive and the focus is mainly
cidents, administration). However, as the
on the cost of illegal drugs. For a broader
discussion below shows, there is consider�
overview of different cost studies and vari�
able and significant disagreement both on
ous approaches, the reader is referred to,
the general categories and on which costs
for instance, Collins and Lapsley (1991),
should be included under the headings.
Culyer et al. (2002), Robson and Single
As an example, consider Harwood et
(1995), and Single et al. �������������������
al.’s estimate of the costs of alcohol and
ested reader may also consult, for instance,
drug abuse in the US in 1992 (Harwood
Xie et al. (1999), Harwood et al. (1999) and
1999). According to this study “In 1992
Manning et al. (1989) for an introduction
the economic cost to society from alcohol
to the various approaches and a discussion
and drug abuse was an estimated $246
of some of the issues I have left out (inci�
billion.” The largest cost category in the
dence vs. prevalence vs. demographic ap�
study is “Lost Earnings” which represents
proach; human capital vs. willingness to
72% of all costs. Clearly, if one disagrees
pay method; and discount rates).
with the inclusion of this category, the cost
The exclusion of some topics does not
estimate will change significantly. For in�
imply that they are insignificant. There
stance, Healey et al. (1998) argue that lost
is, however, an important difference be�
earnings should not be included in the es�
tween empirical and conceptual prob�
timate of social cost.
lems. Empirical problems are less damag�
The study by Healey et al. (1998) can also
ing in the sense that they can be reduced
be used to illustrate a second point of disa�
by improved data collection and statisti�
greement. It was based on a sample of 1075
cal methods. Conceptual disagreements
drug users and the conclusion was that
might also be reduced through discussion
each user cost society about £17 000. Crime
(e.g. by pointing out logical inconsisten�
represented 78% of these costs and the rest
cies), but they might also reflect deeper
was mainly health care costs. One of the
differences that are unresolved even after
reasons crime turned out to be so costly in
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
289
Conceptual problems with studies of
the social cost of alcohol and drug use
Table 1. Some controversial cost categories
Cost category
Importance *
Included in …
Not included in …
Lost earnings
72%
Harwood et al. (1999)
Henley et al. (1998)
Value of stolen goods
61%
Healey et al. (1998)
Single et al. (1998)
Cost of drugs itself
35%
Collins and Lapsley (1991)
Harwood et al. (1999)
Human suffering
45%
Kleiman (1999)
Single et al. (1996)
* Change in estimate if the category is excluded.
that study, was that Healey et al. – unlike
ing to pay $1000 each. Assume, moreover,
Harwood et al. – included the value of the
that about 10% of the population have a
goods stolen as a cost of crime (61% of the
problematic relationship with alcohol or
total costs). Hence, whether we should in�
drugs. Then human suffering amounts to
clude the value of stolen goods represents a
about $200 billion i.e. an increase of 81%
second point of disagreement.
on Harwood’s estimate.
A third example of a contested cost-catego�
In short, different studies and different
ry is the money spent on drugs and alcohol.
authors use very different cost categories
In the context of drugs, Collins and Lapsley
when estimating costs and these differenc�
(1991)�����������������������������������������
es produce very different estimates (Table
mate since in the words of Robson and Sin�
1). Given this confusion one is forced to
gle (1995): “the consumer does not receive
ask: Who is right and who is wrong?
a benefit equal to the cost of the product.”
This is no small issue since the cost of drugs
The concept of cost
represented more than one third of the total
In this section I shall argue that a closer
costs in the mentioned study. Many other
look at the concept of cost itself can re�
studies – for instance Single et al. (1998) –
duce some of the confusion from the first
do not include this cost.
section. I shall also argue that many of the
As a final example of categories that
studies are weakened by the failure to use
cause problems, consider Kleiman’s (1999)
a realistic alternative when they compare
suggestion that we should include human
a “world with drugs” to a “world of no
suffering as a cost category. The argument
drugs.”
against doing so is usually that it is impos�
In everyday language cost often means
sible to measure human suffering in dol�
monetary outlays, but in economics cost is
lars. Kleiman is clearly aware of this but
defined as the value of the next best alter�
he seems to believe that it is at least possi�
native. For instance, in Figure 1, the cost
ble to derive some kind of lower bound on
of policy A is the (expected) value of the
the cost of human suffering. Assume that
(best) alternative we have to give up in or�
suffering can be measured by willingness
der to do A. This opportunity concept of
to pay (to avoid the suffering) and that eve�
cost avoids the problem of counting only
ry addict is willing to pay $10 000 every
direct monetary outlays since it includes
year to stop the habit or that he/she has
all the consequences that are relevant –
ten people (friends/family) who are will�
monetary and non-monetary.
290
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
One problem with the concept of op�
portunity costs is that it often involves a
Figure 1. Cost as the sum of positive
and negative effects of an alternative
comparison between hypothetical alter�
natives. We do not know exactly what we
have to sacrifice in order to do A. Tech�
nically speaking we have to compare a
Positive
Negative
Total
20
15
A
B
policy (A) against a hypothetical counter�
factual: What would have happened if not
10
A? To answer this we must use a model
that allows us to work out the expected net
sum of positive and negative consequenc�
es measured in the same unit. Only then
can we work out what we sacrifice by not
choosing A.
Does the concept of cost used in costof-illness studies conform to the ideal?
Although they often pay lip service to the
idea of opportunity cost, they fail to use
the correct alternative for comparison. The
basic structure of most studies of the social
costs of substance use is to measure the
monetary value of a situation with drugs
5
0
–5
–10
Note: Each policy alternative (A and B) has some
negative and positive consequences. The cost of
choosing one alternative is represented by the height
of the “total” of the other alternative, not the size of
the negative effects. In this example, project A is
clearly more costly, but it is still the one we would
choose because­the cost is more than balanced by the
positive­effects. The cost of choosing A would then be
the height of “total” in B.
to a world without drugs. To estimate how
a world of “no drugs” or “no alcohol”
would look like they use the current world
as a starting point. They then add the in�
the illustration in Figure 2. The horizontal
come we could have had if there were no
axis measures the number of users of a sub�
addiction since in this world (they argue)
stance. The government can try to reduce
there would be less crime, fewer people in
the number of users and this will reduce
prison, more people working and reduced
the external cost as a result of abuse (crime
health expenses. Many only add those var�
and health costs), but it requires higher
iables which will make the total income in
control costs (police, prison, treatment).
the imagined world higher although this is
When we try to estimate the opportunity
clearly wrong since the concept of cost –
cost we should – by definition – compare
as explained above – requires comparison
realistic and possible worlds. Cost of ill�
of the total sum of positive and negative
ness studies assume a world of no users
changes (see Figure 1).
(U = 0 in the figure) and elimination of all
In order to be useful the concept of cost
control costs (zero control costs at U = 0).
not only has to compare the net effect of
This is clearly unrealistic. To achieve U =
two alternatives, but the comparison also
0 we are required to spend resources on
has to be between realistic alternatives. To
control efforts and if we do not include
formalize the problem slightly, consider
these costs the resulting gap between a
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
291
Conceptual problems with studies of
the social cost of alcohol and drug use
Figure 2. Control costs, external costs and the optimal level of control
Costs
Optimal level of control
1.2
External costs
Control costs
1
Total costs
0.8
U*: That
0.6
level of abuse
which mini-
0.4
mizes total
costs
0.2
0
0
0.2
0.4
U*
0.6
0.8
1
Level of abuse (Number of abusers)
Note: If we use the common framework, then the optimal level is always zero use, but when we include control
costs outlays (as we should according to the definition of cost) we find that the optimal level of use is not zero,
but U* (if the aim is to minimize cost).
“world with drugs” and a “world without
which cost categories we should include
drugs” cannot be labelled as a cost.
in our estimation of social costs. Once
Except for Godfrey et al. �����������������
again it can be interpreted in at least two
ies in the literature distinguish between
different ways. First, one could say that
control costs and external costs. One might
cost to society imply that we should in�
ignore this if the costs were small, but
clude all the individuals in the society in
there is strong indication that the control
our study. If we do this, the theft of a car
costs are substantial. For instance, in the
does not represent a direct cost. As long as
US it has been argued that the control costs
we include everybody, the fact that a dif�
are four times larger than the cost of the
ferent person now controls the car is only
drug itself �������������������������������
������
a transfer, not a loss. On the other hand
though these estimates are controversial,
if we interpret cost to society as the cost
there is little doubt that the sums are sub�
that non-users of drugs and alcohol suffer
stantial enough to justify conceptual sepa�
as a result of others’ use of drugs/alcohol
ration and empirical estimation.
(external cost), then theft is a direct loss.
I have so far examined the concept of
Hence, depending on the differences in
cost as employed by economists and com�
their interpretation of the phrase cost to
pared this to the cost concept used in cost
society, different studies tend to include
of illness studies. However, I have said lit�
different cost categories.
tle about the second part of the title, i.e.
The ambiguity in the phrase “to society”
what the phrase “to society” implies for
is compounded by the failure of the litera�
292
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
ture to employ standard economic defini�
One might argue that there is nothing in�
tions that could differentiate between dif�
herently wrong with using different defi�
ferent meanings of the phrase “to society.”
nitions. It is perfectly possible to make one
The standard definition is that social cost
cost estimate in which one tries to meas�
is the sum of private costs and external
ure the cost of drug use in general and
costs (Stiglitz & Walsh 2002; Varian 1992).
another in which one tries to measure the
However, commonly in cost of illness
cost of drugs to one sub-group of the popu�
studies, and by their own admission, the
lation. However, the terms social costs and
“definition of social costs correspond to
externalities have normative implications:
what economic theory usually calls ‘exter�
High social costs and negative externali�
nalities’” �������������������������������
����������
ties are used as arguments for interven�
stance, after claiming that “The social cost
tion. Because of this it is important to be
of smoking tobacco is a subject … which is
aware that not everything that goes into the
littered with conceptual confusion, inap�
definition of externalities on the wide in�
propriate measurement, and invalid infer�
terpretation is equally uncontroversial in
ence”, Merkandya and Pearce (1989) goes
terms of justifying intervention. The least
on to define social costs as follows:
controversial are negative technological
externalities imposed on others which are
“To the extent that the costs are know�
often accepted as grounds for intervention.
ingly and freely borne by the consumer
However, even with respect to technologi�
or producer himself, they are referred
cal externalities, there is some controversy
to as PRIVATE COSTS but to the ex�
as to whether there is enough information
tent that they are not so borne but fall
to make an optimal intervention and what
on the rest of society, they are called
type of intervention that is justified. Some
SOCIAL COSTS. Hence, the total cost
argue in favour of taxes or regulation,
of any activity is the sum of the private
while others – like Coase (1960) – argue
and the social costs.” (italics added)
that the solution is to assign and enforce
property rights. Most agree, however, that
Evidently, social costs is only a part of
some form of intervention is desirable in
the total costs in Merkandya and Peace,
the case of large technological externali�
while social cost in Kopp and Fenoglio
ties. A much weaker justification for in�
is the same as total costs. Clearly in this
tervention is the type of externality that is
literature people mean different things by
sometimes called internalities i.e. conse�
the same terms. Although confusing, this
quences imposed on yourself which you
realization is also clarifying because all
fail to consider, perhaps due to misleading
the different cost estimates are more un�
information. Morally there is an important
derstandable once it is realized that they
difference between costs imposed on oth�
are measuring different costs. Some only
ers and costs born by the actor himself. It
include the negative consequences, some
is easier to justify interventions to prevent
try to include all consequences, some fo�
harm from one person on another, while
cus on the consequences for the whole of
paternalistic policies aimed to prevent a
society, some only on the external costs.
person from doing harm to herself are of�
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
293
Conceptual problems with studies of
the social cost of alcohol and drug use
ten more difficult. Hence, the policy im�
1998). Hence, one might argue that pure
plication of internalities is not on the same
academic curiosity may justify spending
footing as traditional externalities. Adding
time on the question of the social cost of
the two together may produce a large sum,
drugs. If this is the aim, however, one can�
but it would be misleading to present this
not justify the common tendency to focus
sum as a traditional externality and derive
only on the negative consequences of the
policy implications based on this.
substance or the unrealistic counterfactual.
Closely related to the issue of “What do
Indeed, it seems difficult to justify this fo�
we mean by social cost?” is another possi�
cus unless the aim is simply to find a very
ble response to the question of social cost:
large estimate for political purposes. Policy
“Why do we want to know the social costs
makers and lobbyists use cost estimates to
of drugs?” This is an important question,
justify intervention and to fight for higher
because different answers correspond to
budgets internally. For this purpose a high
different investigations. For instance, if
sum is often deemed politically most inter�
the aim is to find the tax rate that would
esting which makes the focus on negative
make the price of the substance reflect its
consequences more understandable. From
social cost, then only an estimate of the
an academic point of view, it is more prob�
external costs are relevant. Cost estimates
lematic since cost estimates say nothing
are sometimes used in this way. For in�
about whether the costs can be reduced.
stance, in Journal of Economic Perspec-
Moreover, when only the negative con�
tives Grossman et al. (1993) report that
sequences are included in the estimate it
cost estimates from the already mentioned
does not even tell us whether it is desirable
Harwood et al., imply that the price of a
to reduce the use of the substance. Hence,
alcohol should be increased from 35 USD
the most commonly used method may pro�
per gallon pure (average) to somewhere
duce numbers that are politically useful,
between 70 and 80 USD per gallon.
but not politically relevant from an eco�
Tax considerations do not justify cost
studies of illicit drugs, and in this case one
nomic point of view.
not to be immediately useful but that the
Cost, free choice, multiple self
and acceptable preferences
question is of academic interest in the same
There are at least three problems which
sense that curiosity makes us ask what the
make the estimation of social costs de�
US GDP would have been if the railway had
pendent on normative responses to diffi�
not been invented or if there had been no
cult philosophical questions. In this sec�
slave trade. (Fogel 1964; Fogel & Engerman
tion these problems will be discussed un�
1974) Even the topic of the deadweight loss
der the following headings: Uninformed
of Christmas has been honoured with eight
and involuntary actions, multiple self, and
publications in the American Economic
legitimate preferences.
might retreat to the position that the aim is
Review, although nobody proposes to can�
cel Christmas (List & Shogren 1998; Ruffle
& Tykocinski 2000; Solnick & Hemenway
1996; 1998; 2000; Waldfogel 1993; 1996;
294
NORDIC STUDIES ON ALCOHOL AND DRUGS
■ Uninformed and involuntary actions
Consider the inclusion of the cost of the
drug itself in the estimate of cost to society.
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
In standard economic theory it is left to the
Simply labelling something a disease,
consumer to evaluate whether something
addictive, or ill-informed, however, is not
is worth the cost (consumer sovereignty)
a good justification for deviating from the
and the fact that they pay for it by defi�
standard approach. For instance, Husak
nition demonstrates that they believe it is
(1989) makes the following argument:
worth the cost (revealed preference theo�
ry). If we then apply standard economic
"… it may be contended that consum�
reasoning, the money spent on drugs is
ers do not want, enjoy, or derive sat�
always balanced by the utility the person
isfaction from recreational drug use.
gets from using drugs. Based on this line of
This strategy will be dismissed by
reasoning, one should not include the cost
economists who take wants as given
of drugs as a social cost.
and make policy recommendations
Against the standard account based on
designed to maximize the satisfaction
consumer sovereignty and revealed prefer�
of existing consumer preferences. And
ences some argue that addiction is a dis�
it is almost certain to be rejected by
ease or the act of taking drugs is involun�
philosophers familiar with the sleight-
tary or ill-informed. Because of this, they
of-hand that results from positing an
argue, the consumers do not gain utility
alleged 'rational will' and then sup�
from their consumption. This is an impor�
posing that a person's 'true wants' can
tant assumption since it makes a signifi�
be identified apart from his expressed
cant difference in the cost estimate. For in�
desires."
stance, Collins and Lapsley (1991) include
30% of the expenditure on alcohol in the
Husak is right to warn against the politi�
estimation of social costs, arguing that this
cal dangers of the conceptual distinction
is the percentage of alcohol consumption
between false and real selves, but the fact
that is consumed by addicts. Without this,
that a conceptual distinction may have
their estimate of the social cost of alcohol
undesirable political consequences, does
would be 36% lower. Pogue and Sgontz’
not imply that it is a false distinction.
(1989) discussion of the optimal tax rate
Conceptually a preference is not the same
on alcohol in an economy with both light
as a choice. For instance: A person is of�
and heavy users is another example of the
fered the choice between an apple and an
importance of the debate on disease, free
orange. He selects the apple because he
will and addiction. In their paper the op�
believes that the orange is poisoned. If this
timal tax rate to cover the external costs of
is a true description, then the inference
alcohol is estimated to be six times higher
from "he selected the apple" to "he prefers
if alcohol is viewed as a disease compared
apples to oranges" is misleading. Thus,
to conceiving it as an informed and vol�
there is not necessarily always a "sleight of
untary choice. The reason for the large
hand" involved when preferences are said
difference is that viewing alcoholism as a
to exist separately from actions.
disease is assumed by the authors to im�
One might object that adherents of re�
ply that alcoholics do not gain utility from
vealed preference theory do not make this
their consumption.
obvious mistake since all they would ad�
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
295
Conceptual problems with studies of
the social cost of alcohol and drug use
mit is that the action shows that “the in�
several other less artificial examples of in�
dividual prefers the joint alternative of ‘an
voluntary actions. Following Elster (1999)
apple and being alive’ to the alternative of
one might use the decision to sleep as an
‘getting an orange and being poisoned.’”
illustration. Usually it is clearly voluntary,
This argument is advanced, for instance,
but at some point you simply cannot help
by Dowding ��������������������������
yourself and you fall asleep whether you
ever, is that all we observe is the action.
want it or not (i.e. involuntary). There
We do not observe the beliefs and often we
might be grey cases in between and this
do not know the full context of the action.
could lead us to conclude that instead
Hence, even on this approach one is forced
of using the dichotomous distinction be�
to admit that the inference from action to
tween a free choice and an involuntary ac�
preferences is not obvious. Dowding is
tion, we should use a continuum.
clearly aware of this, but he argues that
As shown by Sen’s attack on the use of
these practical problems do not invalidate
revealed preferences in economics theory,
the theory of revealed preferences. This is
there are many more cases in which it is
an interesting issue in itself, but the ad�
wrong to infer welfare relevant preferenc�
mission of practical problems is enough
es from actions ���������������������������
to get the argument about possibly mis�
guments briefly mentioned by Sen is that
taken inference from observed actions to
the inference from choice to preferences
preferences off the ground. One need not
is invalid when the choice is not based
agree or conclude that revealed preference
upon rational deliberation. To simplify
theory is wrong in order to discuss practi�
slightly we may use the following illustra�
cal problems related to the inference from
tion: If I hit a person out of anger, then it
actions to preferences. What is needed,
is wrong to treat that action as if it were
however, is a good justification for when
selected to express my preferences using
and how actions are misleading indicators
the process of rational and cold weighting
of preferences.
of costs and benefits. More problematic,
Mistaken beliefs are one of the factors
there are situations, such as the prison�
that could make actions deviate from the
ers’ dilemma in game theory, in which it
underlying preferences we are interested
would be wrong to generalize from actions
in. Pure accident is another, although this
to preferences even after observing many
is perhaps less interesting than a third
actions over time. Picking up on the same
problem: involuntary actions. At first sight
theme Hausman (2000) has argued that the
this may seem like an oxymoron. To con�
"notion of 'revealed preferences' is unclear
vince the reader of the conceptual possi�
and should be abandoned." Whether one
bility one might use the following exam�
agrees or not, the statement at least shows
ple: If the person selecting the apple did so
that not all economists simply equate ex�
because a third person threatened to shoot
pressed choice and preferences and work
him if he went for the orange, then the ob�
from there.
served action cannot be translated into “he
Some of the disagreement between
prefers apples over oranges.” This may
economists working on the topic seems to
seem like a silly example, but there are
boil down to different conceptions of pref�
296
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
erences. Some argue that preferences are
claim that the exceptions have more to
defined by choice, not just revealed. For
do with incomplete description of prefer�
instance, the approach in formal choice
ences (and the context) than people acting
theory is to define preference relations in
against their own preferences. Thus, both
terms of choices (bets) made by agents. On
approaches can at least in theory handle
this approach no action can go against our
the situations described above in which
preferences. However, a preference can
the naïve inference from action to prefer�
also be viewed as one of several inputs
ence is mistaken.
(such as beliefs and emotions) that gener�
In sum, it is possible that observed ac�
ate an action. On this approach the prefer�
tions do not directly demonstrate a pref�
ence is not always defined by the action.
erence in the sense that is relevant for
The distinction is important to welfare
welfare evaluations. However, the mere
economics because on the latter approach
existence of possible exceptions does not
preferences have some independent exist�
imply that drug use falls into either of the
ence. This implies that before evaluating
categories above. The justifications for
a policy proposal we need to decode the
deviating from the general presumption
actions to find the pure preference. It was
that actions tend to indicate preferences,
precisely this claimed independent ex�
were based on examples involving poor
istence of pure preferences that inspired
information and degrees of involuntari�
Samuelson’s (1938) article on the concept
ness. While it may be possible to claim
of revealed preferences. He was dismiss�
that some youths take drugs based on poor
ive of the metaphysical concept of pref�
information, it seems wrong to argue that
erences and wanted to base the theory of
all recreational drug users are unaware of
consumption only on observable entities.
the potential harmful side-effects of drugs.
Once again, however, the practical dif�
It is more difficult to determine whether
ferences of the disagreements may be
regular and heavy drug use is an act of
smaller than the theoretical heat indicates.
free will. It may not be analogous to falling
Modern defenders of revealed preference
asleep, but it may be argued that the habit
theory – like Dowding – argue that theory
of drug use creates a desire so strong, or a
does not apply to single acts or individu�
situation so restricted, that the scope for
als, but that it is necessary for the analy�
choice is greatly reduced. However, once
sis of aggregates. On this point even those
again it seems difficult to argue that all
who hold a more metaphysical conception
drug use falls into this category. At least
of preferences would agree that in general
what Husak calls recreational drug use
it is useful to assume that actions reveal
does not have the compulsive nature that
the welfare-relevant preferences. They
would justify such an argument. Empiri�
would, however, claim that there might be
cally one also observes many recreational
exceptions to the general rule even on an
users who simply quit drugs which sug�
aggregate level. Modern defenders of re�
gests that there is room for choice.
vealed preference theory are not so naïve
The problem of free will does not only
that they do not understand these excep�
apply to the cost of drugs itself. It also ap�
tions, the difference just being that they
plies to the classification of costs into ex�
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
297
Conceptual problems with studies of
the social cost of alcohol and drug use
ternal or private. Negative externalities are
of freedom makes the analysis of external
sometimes defined as costs imposed on
and social costs very much dependent on
others i.e. costs that are not voluntarily in�
the researchers’ own views on what it is to
curred by third parties. Based on this defi�
be “free” to do something.
nition one might try to determine whether
ity. Clearly, some would argue, the costs
■ Multiple self
Economist trained in the theories of ra�
tional choice may not be convinced by
related to crime are costs imposed on the
the disease analogy or philosophical ar�
non-users by the drug-users. Others might
guments about free will. However, even if
argue that it is equally obvious that it is
we discard the disease theory, it does not
the non-users who impose the laws that
automatically follow that we must include
make the use of drugs costly (both by mak�
all the utility from consuming drugs as a
ing drugs illegal and deciding the punish�
benefit. One might, for instance, argue that
ment). In that sense, a large part of the cost
people have inconsistent preferences as
of crime is not an unavoidable externality
a result of a split between a long-run self
suffered by the non-users.
and a short-run self. In this case the an�
the cost of enforcing laws on drugs and al�
cohol should be interpreted as an external�
The same reasoning applies to the cost
swer to whether the cost of drugs should
of treatment. On the one hand, it is true
be included or not, depends on whether
that non-users pay for treatment of drug
you believe it is possible to talk about a
users – so it sounds like an externality. On
person being split between a false and true
the other hand, it is often the non-users
self, and whether we should only count
who decide to provide and pay for treat�
what we perceive as the “true” self when
ment and because of this it seems wrong to
calculating costs.
label it an external cost. Society could de�
To economists with rational choice in�
cide not to pay for treatment. Once again
clinations, any talk of a split self may
the answer depends greatly on the defini�
initially sound too much like mysterious
tion of free will. If one adopts a definition
Freudian theories of id, ego and superego.
with a continuum, one might argue that
There is, however, one way of looking at
we do not have a fully free choice when
it that may be convincing even to econo�
faced with a dying person who needs help.
mists. All that is needed to find examples
If the same kind of reasoning can be ap�
of preference inconsistencies over time
plied to treatment, one might argue that
in standard economic choice theory, is
a portion of the treatment cost should be
to replace the assumption of exponential
considered an externality (correspond�
discounting with hyperbolic discounting.
ing to the degree to which you believe
When making a choice individuals are as�
the non-users do not have a choice but to
sumed to put less weight on payoffs that
provide treatment). One could of course
are far away. Using exponential discount�
argue about the extent to which drug us�
ing one might, for instance, decrease the
ers are analogous to “people dying from a
value of a future payoff by 3% every year
deadly disease”, but this only highlights
you have to wait. Hyperbolic discounting
the problem that the possibility of degrees
is different in the sense that the individual
298
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
is not assumed to discount future payoffs
tion of consumer sovereignty so that it
at a constant rate. This, in turn, leads to
could work even when individuals have
situations in which we on Monday, Tues�
inconsistent preferences. Instead of using
day and Wednesday are firmly committed
meta-rankings of preferences or relying on
to do something, e.g. go to the dentist, but
theories of a split self to solve the problem,
that we change our minds on Thursday
Sugden argues that consumer sovereignty
once the alternative is close in time. Next
should be reframed to focus on opportu�
week the same pattern may repeat itself. It
nity as opposed to preference satisfaction.
has been argued that this kind of discount�
This proposal, in turn, may be criticised
ing may explain why we yield to tempta�
for placing too much emphasis on the val�
tions – such as taking drugs – that look
ue of having many opportunities and not
good in the short run, but with damaging
enough on their satisfaction. People may
long run consequences (Ainslie 1992).
not prefer to maximize only one of these,
Hyperbolic discounting creates a prob�
preferring instead some kind of trade-off.
lem for the evaluation of costs since the
In short, the existence of inconsistent
costs will differ depending on whether
preferences creates a problem for the es�
one uses the preferences expressed by the
timation of costs. There is no universally
short-run self or the long-run self. One
accepted solution to this, and any solution
possible answer would be to treat the pref�
will necessarily involve normative judge�
erence expressed most of the time as an
ments about true selves, and what is valu�
expression of the true self, but for believ�
able in life, such as the relative value of
ers in consumer sovereignty it is not clear
having many options versus more satisfac�
why priority should be given to the long-
tion within fewer options.
term self. Instead some economists have
suggested to split the person altogether
and simply let both selves count as if they
were different persons (Cowen 1993). To
my knowledge, however, no one has made
the logical extension of weighting the pref�
erences proportional to their intensity or
the length of time each preference was
held.
The proposal of treating preferences
as persons has been criticised by Sugden
(2004) for leaving us with “no preferencebased concept of welfare that applies
to the person as a continuing entity”. It
is slightly unclear what this means and
why it is a problem, but Sugden goes on
to present a radical solution to the prob�
lem of unstable preferences and welfare
evaluation. The aim is to redefine the no�
■ Legitimate preferences
There is a third line of argument for the
subjectivity of cost estimates that does not
rely on judgements about free will or a the�
ory of a split person. The argument is sim�
ply that according to a set of criteria some
preferences are legitimate and others are
not in a calculation of costs. Consider the
following example: Many people dislike
seeing drug addicts in the street. Should
we count this as a cost to society? If we
insist on counting all kinds of preferences,
it seems impossible to avoid including this
as a cost. On the other hand, if the argu�
ment is generalized – say to skin-colour –
most people would strongly object if we
argue that an overview of the costs of im�
migration should include the dislike some
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
299
Conceptual problems with studies of
the social cost of alcohol and drug use
people have for individuals with a differ�
drugs. A very ethically conscious econo�
ent skin-colour. This shows that it is pos�
mist comes along and tells you that your
sible to find examples of preferences that
current preference for drugs should not be
most people agree should not be counted
counted since it can be traced back to poor
in an analysis of costs and benefits, and
information. You then object that any cal�
that it is not obviously wrong to discrimi�
culation should take into account the per�
nate between preferences. The mere pos�
son you are right now, not the kind of ideal
sibility, however, does of course not imply
preferences you would have had if things
that the same logic can be applied to the
had been different. More generally, as ar�
use of drugs. We must ask what it is that
gued by Cowen (1993): “Fully informed
makes us exclude some preferences and
preferences do not offer an Archimedean
whether this also applies to the prefer�
point for value theory in a world of imper�
ence for drugs and alcohol. Since there is
fect information.”
no neutrally accepted theory of acceptable
In sum, there is disagreement on the rel�
preferences we are once again in the realm
evance of incomplete information for the
of the subjective.
evaluation of preferences. The disagree�
Being in the realm of the subjective does
ment is inherent to the topic because it is
not imply that it is impossible to have an
impossible to prove scientifically that one
informed discussion or that we are free to
position is correct. There may be scope for
choose whatever position we want. The
rational reflection, but it is not certain that
conclusions must still be based on correct
it is possible to reach a reflective equilib�
information and valid logical reasoning.
rium that we all agree to. In the end we
Given the constraint of “informed reason”
may simply disagree on which preferences
we can ask whether it is possible to find a
to include or exclude. Since the estima�
set of criteria that excludes the preference
tion of costs depends on the set of prefer�
for drugs as illegitimate. For instance,
ences we include, the cost estimates will
Goodin ��������������������������������������
also differ depending on our views on the
tions for censoring utility functions.” The
issue of legitimate preferences. This is an
list includes, for instance, the argument
unavoidable problem and to argue that
that there is no moral obligation to respect
“all preferences should be counted” is not
preferences based on incomplete informa�
a solution. First of all, it is easy to find ex�
tion. Many argue that addictive consump�
amples of preferences that are deemed un�
tion of drugs, alcohol and tobacco fall into
acceptable to almost everyone, in which
this category.
case the principle breaks down. Second,
While it may seem initially plausible
the principle is itself normative. When
that preferences based on poor informa�
two people argue about which preferences
tion should not be counted, the issue is
to include in a calculation of social costs,
not obvious on second thoughts. For in�
the person arguing in favour of including
stance, assume you experiment with drugs
all preferences is advancing a normative
without knowing that you might become
position in the same way that a person
addicted. After some time you discov�
arguing that some preferences should be
er that you find it difficult to stop using
excluded.
300
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
To conclude this section: The calcula�
this distinction is not enough for meaning�
tion of the social costs of substance abuse
ful estimation since the concept of cost
depends on the answer to philosophical
necessarily implies a comparison between
questions about free will and addiction,
all the realistically expected consequences
the possibility of split selves and the mor�
(both positive and negative) of two policy
al standing of different preferences. These
alternatives and many studies did not use
are issues with no clear answers, and it is
the concept of cost in this way. Finally, cost
not plausible to expect neutral answers
estimates cannot be neutral or scientific
to be forthcoming since the answers ul�
because they depend strongly on our
timately depend on normative positions.
views on when a choice can be character�
For this reason the calculation will un­
ized as voluntary, whether we admit to the
avoidably be a subjective exercise where
possibility of a split self, and which prefer�
different people arrive at different results
ences we consider to be legitimate. Since
depending on their values.
there is no universally agreed answer on
Conclusion
these issues, the cost estimate necessarily
becomes subjective.
There is disagreement on the elements that
should be included in a study of cost to
society. Some of this disagreement can be
solved by distinguishing between “cost
to society” and “external cost”. However,
Hans Olav Melberg, Researcher
Department of Health Economics
and Health Management,
University of Oslo and SIRUS, Norway
E-mail: [email protected]
REFERENCES
Ainslie, G. (1992): Picoeconomics. Cambridge:
Cambridge University Press
Atkinson, A. B. & Meade, T.W. (1974): Meth�
ods and preliminary findings in assessing
the economic and health consequences of
smoking with particular reference to lung
cancer. Journal of the Royal Statistical Soci�
ety (Series A) 137: 297–312
Buchanan, J. M. (1969): Cost and Choice. An
Inquiry in Economic Theory. Chicago:
Markham Publishing Company
Coase, R. H. (1960): The problem of social
cost. Journal of Law and Economics 3: 1– 44
Collins, D. J. & Lapsley, H. M. (1991): Estimat�
ing the Economic costs of drug abuse in
Australia, Canberra: National Campaign
againt Drug abuse
Cowen, T. (1993): The Scope and limits of
preference sovereignty. Economics and
Philosophy 9: 253–269
Culyer, A. & Eaton, G. & Godfrey, C. & Kout�
solioutsos, H. & McDougall, C. (2002):
Economics & Social cost of substance
misuse in the United Kingdom: Review of
the methodological and empirical stud�
ies of the economic and social costs of
illicit drugs, University of York, Centre for
Criminal Justice, Economics, and Psychol�
ogy, York
Dowding, K. (2002): Revealed preference and
external reference. Rationality and Soci�
ety14 (3): 259–284
Ellemann-Jensen, P. (1991): The Sosial cost of
smoking revisited. British Journal of Addic�
tion 86: 957–966
Elster, J. (1986): The Multiple self, Cambridge:
Cambridge University Press
Elster, J. (1999): Strong feelings: emotion, ad�
diction, and human behavior, Cambridge,
Mass.: MIT Press
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
301
Conceptual problems with studies of
the social cost of alcohol and drug use
Fisher, I. (1927): Prohibition at its worst, New
York: Alcohol Information Committee
Fogel, R. W. (1964): Railroads and American
economic growth: essays in econometric
history, Baltimore, Johns Hopkins Press
Fogel, R. W. & Engerman, S. L. (1974): Time
on the cross; the economics of American
Negro slavery. 1st ed., Boston, Little Brown
Godfrey, C. & Eaton, G. & McDougall, C. &
Culyer, A. (2002): The Economic and social
costs of Class A drug use in England and
Wales. Home Office Research, Development
and Statistics Directorate
Goodin, R. E. (1986): Laundering Preferences.
In: Elster, J. & Hylland, A. (eds.): Founda�
tions of Social Choice Theory. Cambridge,
Cambridge University Press
Grossman, M. & Sindelar, J. & Mullahy, J. &
Anderson R. (1993): Policy Watch: Alcohol
and Cigarette Taxes. The Journal of Eco�
nomic Perspectives 7 (4): 211–222
Harwood, H. (1999): Cost estimates for alcohol
and drug abuse. Addiction 94 (5): 631–647
Hausmann, D. (2000): Revealed Preference,
Belief, and Game Theory. Economics and
Philosophy 16: 99–115
Healey, A. & Knapp, M. & Astin, J. & Gossop,
M. & Marsden, J. & Stewart, D. & Lehmann,
P. & Godfrey, C. (1998): Economic burden
of drug dependency. British Journal of
Psychiatry 173: 160–165
Husak, D. (1989): Recreational drugs and pa�
ternalism. Law and Philosophy 8: 353–381
Kleinman, M. A. R. (1999): Economic cost’
measurements, damage minimization and
drug abuse control policy. Addiction 94 (5):
638–641
Kopp, P. & Fenoglio, P. (2001): Calculating
the social cost of illicit drugs: Council of
Europe Publishing
List, J. A. & Shogren, J. F. (1998): The dead�
weight loss of christmas: Comment. Ameri�
can Economic Review 88 (5): 1350–1355
Manning, W. G. & Keeler, E. B. & Newhouse,
J. P. & Sloss, E. M. & Wasserman, J. (1989):
The taxes of sin. Do smokers and drinker
pay their way? JAMA 261 (11): 1604–1609
Markandya, A. & Pearce, D. W. (1989): The
Social Costs of Tobacco Smoking. British
Journal of Addiction 84: 1139–1150
Pogue, T. F. & Sgontz, L. G. (1989): Taxing to
302
NORDIC STUDIES ON ALCOHOL AND DRUGS
control Social Costs: the Case of Alcohol.
The American Economic Review 79 (1):
235–243
Reuter, P. (1999): Are calculations of the eco�
nomic costs of drug abuse either possible or
useful? Addiction 94 (5): 635–638
Robinson, M. B. &. Scherlen, R. G (2007): Lies,
damned lies, and drug war statistics : a crit�
ical analysis of claims made by the office of
National Drug Control Policy, Albany: State
University of New York Press
Robson, L. & Single, E. (1995): Literature Re�
view Of Studies On The Economic Costs Of
Substance Abuse, A Report of The Cana�
dian Centre On Substance Abuse, Canadian
Centre on Substance Abuse, Ottawa, ON
Ruffle, B. J. & Tykocinski, O. (2000): The dead�
weight loss of Christmas: Comment. Ameri�
can Economic Review 90 (1): 319–324
Samuelson, P. A. (1938): A Note on the Pure
Theory of Consumer's Behaviour. Econom�
ica 5 : 61–71
Sen, A. (1986): Behaviour and the concept
of preferences, In: Elster, J. (ed.): Rational
Choice. Cambridge, Cambridge University
Press
Single, E. & Robson, L. & Xie, X. & Rehm, J.
(1998): The Economic costs of alcohol,
tobacco, and illicit drugs in Canada, 1992.
Addiction 93 (7): 991–1006
Single, E. & Collins, D. & Easton, B. & Har�
wood, H. & Lapsley, H. & Kopp, P. &
Wilson, E. (2003): International guidelines
for estimating the costs of substance abuse,
World Health Organization, Geneva
Solnick, S. J. & Hemenway, D. (1996): The
deadweight loss of Christmas: Com�
ment. American Economic Review 86 (5):
1299–1305
Solnick, S. J. & Hemenway, D. (1998): The
deadweight loss of christmas: Reply. Amer�
ican Economic Review 88 (5): 1356–1357
Solnick, S. J. & Hemenway, D. (2000): The
deadweight loss of Christmas: Reply. Amer�
ican Economic Review 90 (1): 325–325
Stiglitz, J. E. &. Walsh, C. E (2002): Economics
3rd ed., New York: Norton
Sugden, R. (2004): The opportunity criterion:
Consumer sovereignty without the assump�
tion of coherent preferences. American
Economic Review 94 (4): 1014–1033
V O L . 2 7. 2 0 1 0 . 4
Conceptual problems with studies of
the social cost of alcohol and drug use
Varian, H. R. (1992): Microeconomic analysis
3rd ed., New York: Norton
Waldfogel, J. (1993): The Deadweight Loss of
Christmas. American Economic Review 83
(5): 1328–1336
Waldfogel, J. (1996): The deadweight loss of
Christmas: Reply. American Economic
Review 86 (5): 1306–1308
Waldfogel, J. (1998): The deadweight loss
of christmas: Reply. American Economic
Review 88 (5): 1358–1359
Walker, D. M. & Barnett, A. H. (1999): The
Social Costs of Gambling: An Economic
Perspective. Journal of Gambling Studies
15 (3): 181–212
Xie, X. & Rehm, J. & Single, E. & Robson, L.
(1999): Methodological issues and selected
economic consequences of substanse abuse
in Ontario. Addiction Research 7 (5):
417–431
Österberg, E. (1983): Calculating the costs of
Alcohol: The Scandinavian experience, In:
Grant, M. & Plant, M. (eds.): Economics and
alcohol. London, Croom Helm.
NORDIC STUDIES ON ALCOHOL AND DRUGS
V O L . 27. 2010 . 4
303