DP 2666
DP 2666
DP 2666
Panu Poutvaara
Lars-H. R. Siemers
March 2007
Forschungsinstitut
zur Zukunft der Arbeit
Institute for the Study
of Labor
Smoking and Social Interaction
Panu Poutvaara
University of Helsinki
and IZA
Lars-H. R. Siemers
RWI Essen
IZA
Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail: iza@iza.org
Any opinions expressed here are those of the author(s) and not those of the institute. Research
disseminated by IZA may include views on policy, but the institute itself takes no institutional policy
positions.
The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
company supported by Deutsche Post World Net. The center is associated with the University of Bonn
and offers a stimulating research environment through its research networks, research support, and
visitors and doctoral programs. IZA engages in (i) original and internationally competitive research in
all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research
results and concepts to the interested public.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 2666
March 2007
ABSTRACT
Keywords: smoking policy, social norms, guilt aversion, deviant behavior, social interaction
Corresponding author:
Panu Poutvaara
Department of Economics
University of Helsinki
P.O. Box 17 (Arkadiankatu 7)
FIN-00014 Helsinki
Finland
E-mail: panu.poutvaara@helsinki.fi
*
We thank Johannes Binswanger, Essi Eerola, Wolfgang Gick, Volker Hahn, Rainer Kambeck, Topi
Miettinen, Felix Mühe, Joel Stiebale, Harald Tauchmann, Nicolaj Verdelin, and participants at the
annual meetings of the European Public Choice Society (EPCS) in Turku, of the Verein für
Socialpolitik in Bayreuth, of the Public Choice Society (PCS) in New Orleans, and of the Scottish
Economic Society (SES) in Perth, and at an RWI Essen-seminar for very helpful comments on an
earlier draft of the paper.
1 Introduction
Active smoking causes health problems and often advanced death.1 Second-hand smoking
is less but still significantly dangerous.2 Hence, active smokers exert a dangerous, negative
externality on non-smokers whenever they smoke and socially interact with non-smokers.
Correspondingly, tobacco and smoking policies have become stricter in recent years in
several countries in Europe and in parts of the U.S.A.3 The major aim of these regulations
is to prevent involuntary passive smoking.
Passive smoking typically occurs in situations in which smokers and non-smokers socially
interact, for instance at work, in pubs or in restaurants. We often observe the behavior
that non-smokers hesitate to complain and agonize smoking, although they—including the
potential utility gain from being together—would prefer not being trapped in interaction
with smokers who smoke in their presence. This may appear paradoxical. However, it
is fully in line with the revealed preferences once taking into account social norms. We
analyze the social interaction between smokers and non-smokers as a sequential game.
Beside the utility of smoking and the disutility of second-hand smoke, individuals care
about behaving in line with social norms. Our model explains the observed behavior
of hesitating to complain by a weak level of strategic bargaining power, determined by
social norms. Social norms cause that non-smokers and smokers are often trapped in social
interaction so that smoking is unduly often accepted—inefficiency arises. Therefore, in
our scenario, a social norm is harmful to welfare. We show that the introduction of non-
smoking areas is not sufficient to cope with this specific inefficiency. Strict smoking bans
are only a second-best policy, but appear to be required in areas where smokers and non-
smokers socially interact. Therefore, our findings provide support for the strict smoking
1
Smoking is a documented risk factor, e.g. for cancer. The risk is increased when alcohol is consumed
additionally (Chowdhury and Rayford, 2001; DKFZ, 2005a; Li, 2001; Partanen et al., 1997; Schuller et
al., 2002; Silverman et al., 1995).
2
DKFZ (2002), IARC (2004), or Gruber (2001: 203-204). Smoking-related illness was identified to be
the leading preventable cause of death in the United States (McGinnis and Foege, 1993: 14).
3
The World Health Organization (WHO) even follows a policy on non-recruitment of smokers
(http://www.who.int/employment/recruitment/en/index.html).
2
policies followed by several countries in recent years.
Our paper is also related to the research on social conformity and customs (Akerlof, 1980;
Bernheim, 1994). Similar to these models, we extend the standard model, in which utility
is derived directly from consumption, to indirect social determinants. In contrast to
Bernheim, in whose model individuals additionally care about status, we include the will
to behave in line with social norms and customs, which is related to Akerlof’s idea of a code
of behavior. Moreover, our paper is related to the article of Charness and Dufwenberg
(2006) on “guilt aversion,” in which individuals suffer utility losses when they do not meet
the expectations of other individuals. The guilt-aversion mechanism is closely related to
Hammar and Carlsson (2005), above. Related to Charness and Dufwenberg, Miettinen
(2006) suggests an approach where agents feel bad about breaching social norms. While
Charness and Dufwenberg (2006) and Miettinen (2006) identify cases in which social
norms allow reaching a superior outcome, we highlight that social norms of politeness
may also result in inefficiency and call for policy intervention in form of smoking bans.
An interesting discussion of social norms is provided by Elster (1989).
3
2 Insights from social psychology
In social psychology, a major research field investigates the impact of social norms on
the behavior of individuals (e.g. Aronson et al., 2004; Zimbardo and Gerrig, 1999).
Social influence arises from confrontation with opinions or evaluations of the majority of
society or of the own group, which constitute social norms (van Avermaet, 2002: 412).
Social psychologists distinguish two channels via which social impacts may work: (a)
normative aspects that cause adopting norms of other people for being respected, and
(b) informative aspects that are followed to behave “correctly” (Zimbardo and Gerrig,
1999: 412). Erb and Bohner (2002) emphasize that many different experiments have
proved that individuals as members of a social group often show a behavior of conformity.
Levine (1989) found that individuals expect negative evaluations when they do not behave
conformable to social norms, and that deviance is sanctioned. According to Turner (1991),
missing consensus generates uncertainty, and thus causes subjective costs. Hence, there
exists a subjective strategy to reduce interpersonal conflict (Moscovici, 1985) and the will
not to behave deviantly in order to maintain social stability. But the wish to be accepted
by majority may often only cause public conformity with social norms (“compliance”),
without a change of private attitude (“conversion”). Therefore, we observe the change in
behavior only in social interaction (Erb and Bohner, 2002). An instance is the individual
decision to wait in line. If there is already a line where other people wait, most people
will probably also line up, even though they might not do so if there would be only two
or three other persons that are waiting.
The idea that departures from social norms impair the individual reputation or status or
entail other forms of social punishment is also a central building block in Akerlof (1980),
Bernheim (1994), Charness and Dufwenberg (2006), and Miettinen (2006). We believe
that such social and psychological aspects are also crucial for the behavior of smokers and
non-smokers when they socially interact. We argue that individual behavior is influenced
by whether an action is in line with social norms and conventions or not. This extension
generates an adjustment in the distribution of bargaining power, so that people may
actually behave in their own best interest when hesitating in asking for stopping smoking,
4
though seemingly suffering a net utility loss.
To our knowledge, no author so far has addressed and explained the behavior of non-
smokers in the social interaction with smokers. In the literature on smoking and in the
smoking debate, the role of social interaction and the consequences of social norms have
been completely neglected. We shed some light on this behavioral and health issue, and
deduce corresponding policy implications.
3 Model
For simplicity, we consider the social interaction of one smoker, player S, and one non-
smoker, player N ; the players are indexed by i = {S, N }. The smoker obtains utility of
B > 0 by smoking; potential utility losses in the case of addiction when she/he does not
smoke represent saved opportunity costs and increase B. That is, variable B is the net
benefit from smoking. The smoker’s utility when she/he is alone and does not smoke is
normalized to zero. The non-smoker, in turn, suffers a utility loss of size E > 0 by second-
hand smoking; utility loss E (external effect) also involves the subjective perception of
the danger of second-hand smoking. Moreover, both players might enjoy being together
and receive utility of size Ti > 0, i = {S, N }, from this social interaction.
So far, our model is standard. We now additionally assume that within society, or in
the narrow environment of social interaction, there exists a social norm or standard
behavior that determines whether or not smoking is generally accepted: We hypothesize
that if accommodating smoking is the social norm, then social interaction happens at
a location where smoking is accepted in general, and the non-smoker has to ask the
smoker not to smoke. Hence, our model has to take into account the findings of the
theories on social impacts and guilt aversion: If accommodating smoking is the norm, we
assume that the non-smoker will suffer a utility loss of size AN > 0 from asking for not
smoking; because asking for not smoking is uncommon, it potentially starts a conflict, and
it may be considered as a deviant behavior that may cause a feeling of guilt. Similarly,
if accommodating smoking is not standard, the smoker will have to ask for permission to
5
smoke, which costs her/him utility of size AS > 0.
Suppose the two players sit in a pub or restaurant, or the like. If accommodating smoking
is the norm the smoker will not ask for permission to smoke and smokes whenever she/he
wants to.4 Therefore, the two play the following sequential game:
Game 1:
Stage 1 The smoker decides to stay and smoke, to stay and not to smoke, or to
leave. If the smoker does not smoke both stay in the room together, if the
smoker goes she/he smokes alone. In both cases the game ends.
Stage 2 If the smoker chooses to stay and smoke, the non-smoker decides whether
she/he goes away, asks the smoker to stop smoking, or accepts smoking.
If she/he accepts smoking, the game ends and both stay together; if the
non-smoker directly goes away, the game ends as well.
Stage 3 If the non-smoker asks for stopping smoking, the smoker decides to stay
and stop smoking, to stay and continue smoking, or to leave. If the smoker
stops smoking, both stay together. If the smoker goes away she/he smokes
alone. In both cases the game ends.
Stage 4 If the smoker continues smoking, the non-smoker decides whether to ac-
cept smoking or to go. If the non-smoker accepts they will stay together,
otherwise she/he has to leave. The game ends.
The game is illustrated by the game tree in Figure 1; one can also find the payoffs Pj ,
j = {1, 2, . . . , 8}, there. Notice that the possibility of going away at every stage represents
an exit option. We assume that there also exists a social norm that determines that going
away is considered as a rude step. Breaking it generates a (strong) feeling of guilt and
therefore involves, for the one leaving, a loss of utility of Li . In the following we assume
4
The same holds if it is a social norm of politeness that non-smokers do not object if a smoker asks
for a permission to smoke.
6
Li > Ai , for i = {S, N }. Whenever the smoker chooses not to smoke, the non-smoker
chooses to accept smoking, or one of the players goes, the game ends. We assume that
both players exactly observe the actions of the other player (perfect information). We
also assume that both players know each other’s type and payoff function (complete in-
formation), for simplicity.5 The social interaction between the players during the game
does not take so much time that we would have to discount the payoffs correspondingly.
Moreover, for simplicity, we introduce the following tie-breaking rule: if a player is indif-
ferent between two actions, the player chooses that action that results in being together,
that is, for instance, the smoker is then willing not to smoke. We solve the game by
backwards-induction and obtain:6
(i) B > TS , the unique subgame-perfect Nash equilibrium is described by the se-
quence of actions (smoke, go) and payoff P3 = (B, −LN ).
7
Proof: See appendix.
In case (a), the non-smoker’s preferences are such that she/he prefers being together
suffering second-hand smoke to leaving and being without the smoker: if the non-smoker
values the smoker’s company plus the costs of being “impolite” by leaving higher than the
danger of second-hand smoking, E, the outcome will always be that the smoker smokes
and the non-smoker accepts this. The smoker knows that the non-smoker’s threat of
leaving would not be credible, and hence the smoker will not stop smoking if asked for,
which the non-smoker knows, in turn: there is no point in asking the smoker to stop
smoking. Thus TN + LN ≥ E is a necessary and sufficient condition for smoking being
accepted.
However, in the contrary case, the non-smoker’s threat of leaving is credible. Now the
smoker must consider whether she/he prefers being together with the non-smoker re-
nouncing smoking (case (b)(ii)) or smoking alone (case (b)(i)). If the smoker prefers the
latter alternative (B > TS ) she/he will smoke at stage 1 knowing that the non-smoker will
immediately go. The smoker will not go at stage 1, because then she/he would behave
impolitely, and the non-smoker will not ask the smoker to stop smoking because she/he
knows that she/he will have to leave, anyway. If the smoker prefers being together with the
non-smoker, it is clear that knowing that the non-smoker might leave, the smoker decides
not to smoke. However, because the smoker knows that the non-smoker prefers asking
her/him to stop smoking—compared to directly going away—(case (b)(ii)), the smoker
might also prefer that the non-smoker first asks for stopping smoking before she/he stops
smoking.
Notice that the smoker will never go away, since the norm “allows” to smoke, while non-
smokers might have to leave the pub (case (b)(i)) or suffer smoking. Part (a) of the
proposition tells us that a non-smoker will accept smoking even though her/his subjective
perception of the danger of second-hand smoking, expressed by E, is higher than her/his
utility from being together, i.e. TN < E. This seemingly paradoxical behavior occurs
because smoking is not considered as impolite and asking for stopping smoking involves
social costs due to the social norm. The social norm thus reduces the non-smoker’s
8
bargaining power in our game, and the non-smoker hesitates to ask.9 Normally we would
argue that the non-smoker should just leave if TN < E. But going away is a step that is
considered as impolite, whereby the non-smoker hesitates to leave. The smoker, in turn,
has no reason to regard her/his behavior as impolite, because she/he acts in line with the
social norm. Without the social norm that accommodating smoking is standard, asking for
stopping smoking would not involve any costs (AN = 0): condition TN − AN ≥ −LN was
more often fulfilled, so that the smoker would more often decide not to smoke. Moreover,
without the social norm that going away is impolite—that is, when LN = 0—smoking
would be less often accepted, since condition TN ≥ E − LN was less often fulfilled;
especially the behavior that non-smokers accept smoking though their perception is TN <
E would not occur.
If accommodating smoking is not the social norm, then it is the smoker who has to ask
whether she/he may smoke. The two players then play the following three-stage game:
Game 2:
Stage 1 The smoker decides whether to ask for the non-smoker’s approval to
smoke. If the smoker goes away or does not ask permission to smoke
the game will end. In the first case, the smoker smokes without the non-
smoker, in the latter they stay together.
Stage 2 If the smoker asks for permission to smoke, the non-smoker decides
whether or not to allow smoking. If the non-smoker allows the smoker
to smoke, the game ends and both stay together with the smoker smok-
ing. Moreover, the non-smoker has the option to go away, in which case
the game ends.
9
Our concept of bargaining power is similar to the Best Alternative To a Negotiated Agreement
(BATNA) (cf. e.g. Korobkin, 2004; Spangler, 2003; Breslin and Rubin, 1991; Fisher et al., 1992):
the smoker can stay and smoke, the non-smoker can only go away.
9
Stage 3 If the non-smoker does not want the smoker to smoke, the smoker decides
whether she/he stays or goes. The game ends.
The game is illustrated by the game tree in Figure 2; one can also find the payoff vectors
Pj , j = {1, 2, . . . , 6}, there. We obtain:
(i) TN < E, the unique subgame-perfect Nash equilibrium is described by the se-
quence of actions (go) and payoff vector P1 = (B − LS , 0);
10
knows that the non-smoker will not reject her/his request, so that she/he will ask whether
she/he may smoke and receive permission to do so (case (b)(ii)).
We obtain the reversed image of the case where accommodating smoking is the norm:
the non-smoker will never go away, since the social norm of not accommodating smoking
strengthens the non-smoker’s bargaining power. With LS = 0 inequality B − LS ≤ TS
was fulfilled less often, so that the smoker would smoke more often. Again, since people
want to avoid behaving in a way that by the majority of people is considered as impolite
(i.e. is against the social norm), now it can happen that the smoker does not smoke even
if B > TS , that is, when she/he rather would prefer to smoke instead of not smoking in
companion with the non-smoker. One might say that this case is comparable to that where
accommodating smoking is the norm. However, smokers produce a dangerous externality,
non-smokers do not. Therefore, the two cases differ qualitatively.
4 Policy implications
Proposition 3. If both players cooperate and maximize the sum of their payoffs, the
optimal payoffs are given by:
(TS , TN ) if E > B and TS + TN > B;
(1) P ∗ = (B + TS , TN − E) if E < B and E < TS + TN ;
(B, 0) otherwise.
11
from the externality of second-hand smoke. If the group benefits from being together more
than from smoking and the damage from smoking is higher than its benefits, it is optimal
to stay together without smoking. Finally, if for the group as a whole being together
neither compensates for the damage of smoking nor bears more utility than smoking, it is
optimal to go separate ways, so that the smoker can smoke without aggrieving the non-
smoker. Comparing these conditions of social optimum with these of the private game, it
becomes clear that the private outcome cannot guarantee socially optimal outcomes.
Obviously, only good friends will do so and achieve the efficient cooperative solution.
With (sufficiently) selfish individuals a game is played and the corresponding private
arrangements may produce inefficient outcomes. Thus government intervention may be
justified. In most countries accommodating smoking at least has been the norm, and in
a lot of countries or situations it still is. As a consequence, social interaction would often
be accompanied by smoking unduly often. To contain the problem of excessive smoking,
politicians have introduced smoking bans at many places all over the world.10 However,
a theoretical scrutiny of alternative instruments is missing.
In the context of smoking policy the introduction of separated smoking and non-smoking
areas is often discussed. It is argued that the establishment of smoking and no-smoking
areas is sufficient to overcome the problem of second-hand smoke. However, this is ul-
timately not the case. To demonstrate this, we simply reinterpret the games already
analyzed in Section 3.
Imagine there are indeed separated smoking and non-smoking areas. The smoker and
non-smoker are together and have to decide where to go. If accommodating smoking is
the norm, at stage 1 of the game, the smoker can go away, opt for the non-smoking area
10
Australia, Cuba, England, Estonia, Finland, France, Ireland, Italy, Latvia, the Netherlands, Macedo-
nia, Malta, New Zealand, Norway, Scotland, South Africa, Spain, Sweden, Switzerland (Kanton Tessin),
Tanzania, Thailand, several states and cities in the United States (e.g. California, New York, Montana
and Washington), Wales.
12
or for the smoking area. If the smoker goes away or opts for the non-smoking area the
game ends. However, if the smoker opts for the smoking area, the non-smoker can go
away, accept going to the smoking area, or ask for going to the non-smoking area. If the
non-smoker accepts or goes away the game ends, but if the non-smoker asks for going to
the non-smoking area, the smoker at stage 3 can go away, accept going to the non-smoking
area, or insist on going to the smoking area. If the smoker really insists on going to the
smoking area, the non-smoker could accept going to the smoking area, or can go away.
Therefore, both players play the same game as analyzed in Section 3, the only difference
is that the actions are relabeled (see Figure 3). Analogously, one can reinterpret Game
2. It directly follows that the establishment of smoking and non-smoking areas is not
sufficient to overcome the identified problem of social norms in the social interaction of
the smoker and non-smoker. The introduction of the two areas simply doesn’t change
the fact that accommodating smoking is or is not the social norm. If demanding going to
the non-smoking area is not in line with the norm, it represents a deviant behavior and
causes a feeling of guilt which involves costs AN . The power of the social norm is likely
to be weakened by official anti-smoking policy, so that the cost AN and LN are lower,
but the bias in the distribution of bargaining power remains. Overall, the bargaining
power of non-smokers would be strengthened by the introduction of non-smoking areas,
but inefficiency may persist as long as the social norm favors smokers.
In the Appendix, we show, too, that one yields qualitatively the same result if the non-
smoker moves first (see Lemma 1). Therefore, the establishment of non-smoking and
smoking areas is not a tool that solves our problem.11
In many countries it has been decided to enact more or less strict smoking bans in the
last couple of years (cf. footnote 10). Smoking bans, however, involve the drawback
that they even do not allow smoking when it is efficient. All the efficient outcomes in our
11
However, designated smoking areas might solve the problem of passive smoking of non-smokers, who
are not interacting with smokers.
13
model that involve smoking would be excluded. Hence, bans may only represent a second-
best solution. The private outcomes, however, rely on the subjectively perceived negative
effect of second-hand smoke, E, and many individuals presumably still underestimate
the hazard of second-hand smoke.12 The objective, real damage of smoking is probably
significantly higher, so that it is unclear how often a socially optimal outcome arises from
private action. Be that as it may, a presumably second-best smoking ban should represent
only a last resort.
The alternative classical tools are Pigouvian taxes or subsidies, the creation of markets
and establishing property rights so that a Coasean solution takes place (Cropper and
Oates 1992). The Pigouvian subsidy involves well-known problems, since they generate
bad incentives. Taxes, in turn, are widespread and are found to reduce smoking (e.g.
Chaloupka, 1991; Hammar and Carlsson, 2005). However, they do not influence the
decision of the remaining smokers whether they start smoking in interaction with a non-
smoker.13 Hence they do not solve our problem; in fact, accepting smoking would have
to be taxed, or starting smoking in companion of non-smokers, which is not feasible. The
Coasean solution, in turn, may also fail to obtain because of the psychological transaction
costs associated with asking an individual to refrain from smoking, addressed in this paper.
Therefore, external effects in the area of consumption, like the instance of second-hand
smoke, may have to be solved by bans:14 limited smoking bans in closed spaces where
12
There is no doubt that smoking is significantly dangerous for passive smokers (Chowdhury and
Rayford, 2001; DKFZ, 2005a, 2002; Gruber, 2001; IARC, 2004; Li, 2001; Partanen et al., 1997; Schuller
et al., 2002; Silverman et al., 1995). Passive smoking is linked to higher rates of cancer and heart disease
in non-smokers (Evans et al., 1999: 728). It can also cause other health problems like asthma. The
number of early deaths caused by second-hand smoke in Germany, for instance, is estimated to amount
to 3300 per year (DKFZ, 2005b).
13
It is open whether excise taxes have a significant effect on the decision to start or quit smoking
(Hammar and Carlsson, 2005).
14
Sohmen (1976, p. 270) emphasizes that solving external effects in consumption is much more difficult
than in production, since there prevail traditional patterns of behavior that prevent solutions based on
compensations in the Coasean sense. Schelling (1980, pp. 32-33) also emphasizes that an important
limitation of solving conflicts via bargaining is absence of a custom of bidding to pay for a particular
right. Beyond that, if compensations would actually be paid, everybody would feign externalities involving
14
people socially interact ought to be introduced.15 An open issue in reference to smoking
bans, however, is whether they will be accepted and enforced in practice.
5 Conclusion
The paper is twofold. In the first part, we incorporate insights of social psychology and
experimental economics on guilt aversion into a game-theoretic model. We highlight the
crucial role of social norms in determining the behavior of smokers and non-smokers in
social interaction. Asking for something that is not in line with social norms represents
deviant behavior and may cause a feeling of guilt, as experiments in social psychology
and economics have proved. If accommodating smoking is the norm, non-smokers will
hesitate to ask smokers to stop smoking, since asking is not customary and thus involves
utility losses. Additionally, going away is considered as rude and causes a feeling of guilt.
Extending the standard model correspondingly, we explain why non-smokers may, in social
interaction with smokers, accept smoking even though they would, overall, actually prefer
not to be trapped in social interaction with a smoker who smokes. Contrarily, if tolerating
smoking is not the social norm, the smoker hesitates to ask whether she/he may smoke.
Thus social norms and the will to behave politely determine and distort the distribution
of bargaining power among smokers and non-smokers when they socially interact. This
generates, in both cases, a social inefficiency. Since inefficiently much smoking involves
inefficiently high risk of health damage and of death, while inefficiently low incidence of
smoking solely involves decreased pleasure from smoking, the former case appears more
severe. A typical instance for the problem is the case of teenagers in schools.16 A child
would rarely ask smoking classmates to stop smoking because the social costs of doing
15
so—namely being considered as very “uncool”—are significantly high.
In the second part, we embed our results in a smoking policy debate. We show that the
introduction of smoking and non-smoking areas does not suffice to overcome the distortion
of bargaining power generated by social norms. Without a well-founded welfare analysis,
enriched by empirical facts, we cannot provide definite policy implications from within
our model. However, we argue that social norms produce transaction cost and render
Coasean bargaining inefficient, so that all methods but smoking bans turn out to be
inadequate.17 Accordingly, we suggest smoking bans in all closed spaces where smokers
and non-smokers socially interact, for instance in restaurants, pubs, bars and cafés—
though bans represent only a second-best instrument. Our model especially suggests
introducing smoking bans at places where the identified social transaction costs caused
by social norms are substantially high, for instance, at schools where the social pressure
among teenagers is massive. Hence our analysis supports corresponding policies already
implemented all over the world. In addition, models of limited self-control and weak
will (O’Donoghue and Rabin, 1999; Suranovic et al., 1999) suggest that smoking bans
support many smokers who want to give up smoking anyway. This conclusion is especially
in line with the results of Gruber and Mullainathan (2002) who find that taxation of
cigarettes—i.e. restricted access to tobacco—makes smokers happier as the tax provides
a valuable self-control device. Finally, in the model of Bernheim and Rangel (2004: 1580),
a restriction of public consumption of goods like tobacco—for instance a smoking ban—
reduces people’s exposure to cues that can cause addictive behavior by mistaken decisions.
Our paper opens many avenues to future research. An interesting question to investigate
is which effect the extension of the model to more than one smoker and one non-smoker
would have? The bargaining power of a group may increase in its number of members,
but starting a conflict by asking smokers not to smoke becomes more costly when a non-
smoker has to ask more than one smoker. Additionally, the non-smoker asking smokers
to stop smoking might disturb other non-smokers by starting a conflict when the group of
17
Coase (1960: 15-19) himself stated that his theorem fails if transaction cost is too high. See also
Schweizer (1988).
16
non-smokers consists of heterogeneous members. Another extension is to analyze repeated
games where players could play dynamic strategies like trigger strategies. Smokers, e.g.,
could initially follow the strategy to continue with smoking more often, to strengthen
their bargaining power.
Moreover, future research could elaborate under which environment which social norm
prevails and how strong such norms are. Related, an interesting aspect to investigate is
why smoking in the public in some time period has been generally accepted and in others
not. Future research could also address the question whether there exists a potential trade-
off between public and private smoking. On the one hand, it is possible that smokers who
are not allowed to smoke publicly so often anymore will smoke more often at home, so
that their children will suffer more second-hand smoke than before. On the other hand,
O’Donoghue and Rabin (1999), Suranovic et al. (1999), and Bernheim and Rangel (2004)
suggest that there will be less smokers and less smoking of smokers in general, once a ban
has been established.
17
Appendix
A Proofs
Proof of Proposition 1. Consider the game tree depicted in Figure 1 and the payoffs
Pj , j = {1, 2, . . . , 8}, there. Beginning at stage four, the non-smoker’s optimum choice
is “go”, if −LN > TN − E. If TN ≥ E − LN , the non-smoker accepts the smoking of the
smoker.
18
compares her/his payoffs in P1 , P2 and P3 . Since B > TS , “do not smoke” is no option,
and “go” is also no option. Thus the smoker smokes and we end in the terminal node
with payoff P3 . In contrast, if we consider scenario TN < E − LN combined with the
constellation B ≤ TS , the smoker effectively compares payoffs P1 , P2 and P6 . Outcome P1
is strictly dominated by P2 and P6 . Between payoff vector P2 and P6 , in turn, the smoker
is indifferent, and we obtain two subgame-perfect equilibria, (do not smoke) and (smoke,
ask, stop smoking).
At stage two, in turn, the non-smoker will play “allow” if B − LS > TS and additionally
TN ≥ E holds. If B − LS > TS holds together with TN < E, in contrast, the non-smoker
chooses “do not allow”. If B − LS ≤ TS holds, the non-smoker definitely decides to select
“do not allow”.
Eventually at stage one, the smoker compares P1 , P2 , and P4 , if the parameter con-
stellation is such that B − LS > TS and TN ≥ E. Because of LS > AS , we know that
TS + B − AS > B − LS , so that we can drop option P1 . If it now holds that B − AS ≥ 0,
the smoker plays “ask”, and we arrive at the end node with payoff P4 .18 If B − LS > TS ,
TN ≥ E but B − AS < 0 the smoker will choose “do not smoke” and payoffs are given by
P2 . If we now turn to the constellation B − LS > TS and TN < E, the smoker considers
P1 , P2 , and P5 . We can directly exclude P5 and P2 , so that the smoker will play “go”
right at the beginning of the game. Finally, if B − LS ≤ TS , the smoker must compare
P1 , P2 , and P6 . We immediately see that the smoker will decide to play “do not smoke”,
and the outcome is described by P2 .
2
18
We assume that the smoker prefers being together and smoking to being together without smoking.
Therefore, the smoker chooses “ask” also when B − AS = 0.
19
Proof of Proposition 3. We have to select the maximizing option from the following
three sum of payoffs:
B (both separated)
The optimum alternative is found by comparing the three options with each other, given
the parameter constellation. It is easy to prove that the ranking of the alternatives
depends on E R B, B R TS + TN , and E R TS + TN . As there are three alternatives that
have to be compared pairwise, there are 23 = 8 permutations. If E > B and TS + TN > B
the optimum is that both come together without smoking, irrespective of whether E R
TS + TN , which covers two permutations. If B > E and TS + TN > E the optimum is
that both come together and the smoker smokes, irrespective of whether B R TS + TN ,
which covers further two permutations. Moreover, if B > TS + TN and E > TS + TN
the optimum is that both stay separated, so that the smoker can smoke alone, which
again covers further two permutations. Two permutations remain to analyze, namely, (i)
E > B, B > TS + TN , and TS + TN > E, and (ii) B > E, TS + TN < E, and B < TS + TN .
Both constellations are inconsistent and cannot exist; from (i) E > B > TS + TN , for
!
instance, it directly follows that TS + TN < E.
20
B Smoking and non-smoking areas: when the non-
smoker moves first
When there are separated smoking and no smoking areas and the non-smoker moves first,
the following game is played (see Figure 4):
Game 3:
Stage 1 The non-smoker decides whether to propose going to the smoking area
or to the non-smoking area, or to go directly away. If she/he proposes
going to the smoking area both stay there together and the game ends. If
the non-smoker decides to leave instead, the smoker smokes alone and the
game ends.
Stage 2 If the non-smoker has chosen to propose to go to the non-smoking area,
the smoker decides whether she/he goes away, accepts going to the non-
smoking area, or to veto the non-smoker’s proposal. If she/he accepts
going to the non-smoking area, the game ends and both stay together; if
the smoker goes away, she/he smokes alone and the game ends as well.
Stage 3 If the smoker vetoes going to the non-smoking area, the non-smoker has
to decide whether to join the smoker and to go to the smoking area or not.
If he/she joins, both stay together at the smoking area and the smoker
smokes. If the non-smoker decides not to join she/he has to leave and
the smoker goes to the smoking area and smokes alone. In both cases the
game ends.
We obtain:
21
Lemma 1. Depending on parameter constellation, Game 3 possesses the following sub-
game perfect Nash equilibria:
(i) B > TS , the unique subgame-perfect Nash equilibrium is described by the se-
quence of actions (go) and payoff P1 = (B, −LN ).
Proof of Lemma 1. Beginning at stage 3, the non-smoker will accept going to the
smoking area if −LN ≤ TN − E and leaves when −LN > TN − E. At stage 2 the smoker,
in turn, will clearly play “veto” if TN − E ≥ −LN . However, if TN − E < −LN she/he
compares B − LS , B and TS . Therefore, the smoker will accept going to the non-smoking
location if B ≤ TS , but will play “veto” otherwise. Arriving at stage 1, the non-smoker
has to compare TN − AN − E, −LN , and TN − E, if TN − E ≥ −LN . Hence, she/he
plays “going to the smoking area”. In contrast, if TN − E < −LN holds, her/his decision
depends on inequality B R TS . If B > TS , she/he plays “go”, and “go to the smoking
area” else.
22
References
[1] Akerlof, George A.. “A Theory of Social Custom, of Which Unemployment May Be
One Consequence,” The Quarterly Journal of Economics, 1980, 94(4), pp. 749-775.
[2] Aronson, Elliot and Timothy D. Wilson and Robin M. Akert. Sozialpsy-
chologie. Munich: Pearson Education, 2004. English version: Social Psychology, 5th
ed., 2004, Prentice Hall (Pearson Education).
[6] Bernheim, Douglas and Antonio Rangel. “Addiction and Cue-Triggered Deci-
sion Processes,” The American Economic Review, 2004, 94(5), pp. 1558-1590.
[7] Breslin, J. William and Jeffrey Rubin, eds. Negotiation Theory and Practice.
Cambridge, MA: Program on Negotiation Books, 1991.
[8] Chaloupka, Frank. “Rational Addictive Behavior and Cigarette Smoking,” The
Journal of Political Economy, 1991, 99(4), pp. 722-742.
[9] Charness, Gary and Martin Dufwenberg. “Promises and Partnership,” Econo-
metrica, 2006, 74(6), pp. 1579-1601.
[10] Chowdhury, P. and P.L. Rayford. “Smoking and Pancreatic Disorders,” Euro-
pean Journal of Gastroenterology & Hepatology, 2001, 12, pp. 869-877.
[11] Coase, Ronald H. “The Problem of Social Cost,” Journal of Law and Economics,
October 1960, 3, pp. 1-44.
23
[12] Cropper, Maureen L. and Wallace E. Oates. “Environmental Economics: A
Survey,” Journal of Economic Literature, June 1992, 30(2), pp. 675-740.
[13] DKFZ. “Increased Health Hazards due to Additives of Tobacco Products – Conse-
quences for Product Regulation,” Deutsches Krebsforschungszentrum (German Can-
cer Research Center), Heidelberg, 2005a. http://www.dkfz.de/en/index.html
[16] Elster, Jon. “Social Norms and Economic Theory,” The Journal of Economic Per-
spectives, 1989, 3(4), pp. 99-117.
[17] Erb, Hans-Peter and Gerd Bohner. “Theorien zum sozialen Einfluss durch
Mehrheiten und Minderheiten (Theories on Social Impact due to Majorities and Mi-
norities),” in D. Frey and M. Irle, eds., Theorien der Sozialpsychologie (Theories of
Social Psychology). Gruppen-, Interaktions- und Lerntheorien (Theories of Groups,
Interaction and Learning). Vol. 2, 2nd ed. Bern: Verlag Hans Huber, 2002, pp. 47-61.
[18] Evans, William N., Matthew C. Farrelly, and Edward Montgomery. “Do
Workplace Smoking Bans Reduce Smoking?,” The American Economic Review, 1999,
89(4), pp. 728-747.
[19] Fisher, Roger, William L. Ury, and Bruce Patton. “Original Explanation of
BATNA,” in same authors, eds., Getting to Yes: Negotiating Agreement Without
Giving In, 2nd ed. Boston: Houghton Mifflin, April 1992, pp. 101-111.
[20] Gibbons, Robert. A Primer in Game Theory. New York: Harvester and Wheat-
sheaf, 1992.
24
[21] Gruber, Jonathan and Sendhil Mullainathan. “Do Cigarette Taxes Make
Smokers Happier?,” NBER Working Paper 8872, National Bureau of Economic Re-
search, Cambridge, MA, 2002. http://www.nber.org/papers//w8872.
[22] — . “Tobacco at the Crossroads: The Past and Future of Smoking Regulation in
the United States,” The Journal of Economic Perspectives, 2001, 15(2), pp. 193-212.
[23] — and Botond Köszegi. “Is Addiction Rational? Theory and Evidence,” The
Quarterly Journal of Economics, 2001, 116(4), pp. 1261-1303.
[24] Hammar, Henrik and Fredrik Carlsson. “Smokers’ Expectations to Quit Smok-
ing,” Health Economics, 2005, 14, pp. 257-267.
[25] IARC. “Tobacco Smoke and Involuntary Smoking,” IARC Monographs on the Eval-
uation of the Carcinogenic Risks to Humans, 2004, 83, International Agency for
Research on Cancer, Lyon.
[27] Laux, Fritz L.. “Addiction as a Market Failure: Using Rational Addiction Rules to
Justify Tobacco Regulation,” Journal of Health Economics, 2000, 19(4), pp. 421-437.
[28] Levine, J. M.. “Reaction to Opinion Deviance in Small Groups,” in P. Paulus, ed.,
Psychology of Group Influence. 2nd. ed. Hillsdale, NJ: Erlbaum, 1989, pp. 187-231.
[29] Li, D.. “Molecular Epidemiology of Pancreatic Cancer,” Cancer Journal, 2001, 7,
pp. 259-265.
[31] Miettinen, Topi. “Promises and Conventions,” Discussion Papers on Strategic In-
teraction 2006-29, Max Planck Institute of Economics, Jena.
25
[32] Moscovici, Serge. “Social Influence and Conformity,” in G. Lindzey and E. Aron-
son, eds., Handbook of Social Psychology. 3rd. ed. Reading, MA: Addison-Wesley,
1985, pp. 347-412.
[33] O’Donoghue, Ted and Matthew Rabin. “Doing It Now or Later,” The American
Economic Review, 1999, 89(1), pp. 103-124.
[34] Orphanides, Athanasios and David Zervos. “Myopia and Addictive Behavior,”
The Economic Journal, 1998, 108(446), pp. 75-91.
[35] — and — . “Rational Addiction with Learning and Regret,” The Journal of
Political Economy, 1995, 103(4), pp. 739-758.
[36] Partanen, T.J., H.U. Vainio, I.A. Ojajarvi, and T.P. Kauppinen. “Pancreas
Cancer, Tobacco Smoking, and Consumption of Alcoholic Beverages: a Case Control
Study” Cancer Letters, 1997, 116, pp. 27-32.
[37] Schelling, Thomas C.. The Strategy of Conflict, 1980 ed. Cambridge, MA; London:
Harvard University, 1980.
[39] Schweizer, Urs. “Externalities and the Coase Theorem: Hypothesis or Result?,”
Journal of Institutional and Theoretical Economics, 1988, 144(2), pp. 245-266.
[40] Silverman, D.T., L.M. Brown, R.N. Hoover, M. Schiffman, K.D. Lillimoe,
J.B. Schoenberg, G.M. Swanson, R.B. Hayes, R.S. Greenberg, J. Benicou,
A.G. Schwarz, J.M. Liff, and L.M. Pottern. “Alcohol and Pancreatic Cancer in
Blacks and Whites in the United States,” Cancer Research, 1995, 55, pp. 4899-4905.
26
[42] Spangler, Brad. “Best Alternative to a Negotiated Agreement
(BATNA).” in Guy and Heidi Burgess, eds, Beyond Intractability.
Conflict Research Consortium, University of Colorado, June 2003.
http://www.beyondintractability.org/essay/batna/
[43] Spinnewyn, Frans. “Rational Habit Formation,” European Economic Review, 1981,
15(1), pp. 91-109.
[44] Suranovic, Steven M., Robert S. Goldfarb, and Thomas C. Leonhard. “An
Economic Theory of Cigarette Addiction,” Journal of Health Economics, 1999, 18(1),
pp. 1-29.
[45] Turner, J. C.. Social Influence. Pacific Grove, CA: Brooks/Cole, 1991.
[46] Van Avermaet, Eddy. Soziale Einflüsse in Kleingruppen (Social Impacts in Small
Groups). Heidelberg: Springer Verlag, 2002.
[47] Zimbardo, Philip G. and Richard J. Gerrig. Psychology and Life, 7th ed. Allyn
& Bacon / Longman Publishers, 1999.
27
go S
P1
go N
P
3
P2
ask accept
go S
P
5
P6
go accept
P P
7 8
Figure 1: Game tree of Game 1 where accommodating smoking is the social norm and
payoffs are given by P1 = (B − LS , 0), P2 = (TS , TN ), P3 = (B, −LN ), P4 = (B +
TS , TN − E), P5 = (B − LS , −AN ), P6 = (TS , TN − AN ), P7 = (B, −AN − LN ), and
P8 = (B + TS , TN − AN − E).
28
go S
P
1
go N P
2
P
3
S P4
go stay
P P
5 6
Figure 2: Game tree of Game 2 where accommodating smoking is not the social norm
and payoffs are given by P1 = (B − LS , 0), P2 = (TS , TN ), P3 = (B − AS , −LN ), P4 =
(B + TS − AS , TN − E), P5 = (B − AS − LS , 0), and P6 = (TS − AS , TN ).
29
go S
P1
go N
P
3
P2
ask accept
go S
P
5
non−smoking area
P4
insist
P6
go accept
P P
7 8
Figure 3: Game tree of Game 1 with relabeled actions when there exist smoking and non-
smoking areas; payoffs are again given by P1 = (B−LS , 0), P2 = (TS , TN ), P3 = (B, −LN ),
P4 = (B + TS , TN − E), P5 = (B − LS , −AN ), P6 = (TS , TN − AN ), P7 = (B, −AN − LN ),
and P8 = (B + TS , TN − AN − E).
30
go N
P
1
smoking area
non−smoking area
go S P
P 2
3
accept
veto
N P4
go accept
P P
5 6
Figure 4: Game tree of game where there exist smoking and non-smoking areas and the
non-smoker moves first; payoffs are given by P1 = (B, −LN ), P2 = (B + TS , TN − E),
P3 = (B−LS , −AN ), P4 = (TS , TN ), P5 = (B, −AN −LN ), and P6 = (B+TS , TN −AN −E).
31