On the Eﬀectiveness of Smoking Discouragement Policies

(1)

On the Effectiveness of Smoking Discouragement

Policies

Master Thesis Research Master in Economics and Bussiness

Patrick Hullegie

December 18, 2007

(2)

2

Acknowledgements I would like to thank Peter Kooreman for helpful suggestions,

com-ments, and advice during the project. Frank Chaloupka kindly provided data on U.S. tobacco prices and taxes. I am grateful to David Vonka, who provided excellent help at

(3)

1. INTRODUCTION

The first Surgeon General’s report on smoking and health, published in 1964, initiated an ongoing research on the effects of smoking on health. As a consequence, at present

it is well known that smoking is associated with a significant increase of cardiovascular diseases, lung and other types of cancer, and chronic lung diseases. At the same time

it is well known that smoking also imposes negative externalities on non-smokers. An epidemiological study by Peto, Lopez, and Boqi (1999) predicts that tobacco consumption

will become the leading cause of death in developing countries during the first third of the 21st century. Not surprisingly, governments have come under pressure by anti-smoking

lobbying groups and the general public to limit the exposure of non-smokers in public places and generally to discourage smoking. The following two newspaper reports shed

some light on the current attitude toward smoking:

On November 5th, 2007, the New York Times reported: “This year, two

Cali-fornia cities passed laws restricting smoking inside multiunit residential build-ings. In the last 14 months, two large residential real estate companies with

apartment complexes in several states banned smoking inside units.”

NRC Handelsblad reported on December 6th, 2007, that a member of the Dutch parliament asked the Minister of Health, Welfare and Sport to propose

legisla-tion that prohibits smoking in the presence of children within a car.

On the contrary, however, tobacco growing, processing, trade and consumption may bring economic and social benefits. A government has the difficult task to appropriately

(6)

1. Introduction 6

In the last three decades governments in Europe and the United States have taken

several measures to discourage smoking; cigarette excise taxes have dramatically increased for most US states, laws prohibiting smoking in a variety of public places and workplaces

have passed, and advertising possibilities have been restricted. A few examples will illus-trate this. In the United States many states raised its excise tax on a pack of cigarette by

significant amounts, e.g., on June 1, 2005 Kentucky raised its excise from $0,03 to $0,30; in North-Carolina the excise tax was raised from $0,05 to $0,30 on September 1, 2005; Maine

increased the excise tax from $1 to $2 on September 19, 2005. The first city in the world that banned smoking in all public building is San Luis Obispo (CA., U.S.), where the ban

became effective on August 2, 1990.1 At present, many countries (or states/cities) lim-ited or banned smoking in workplaces and public places and imposed cigarette advertising

bans as well. Already in 1970 the Public Health Cigarette Smoking Act banned radio and television cigarette advertising in the United States.

There have been many studies on the economics of smoking, the most recent review of this literature is the chapter in the Handbook of Health Economics by Chaloupka and

Warner (2000). The early literature focused on the effect of prices on quantity consumed, and reported an price elasticity of demand of, on average, -0.4. Chaloupka and Warner

(2000) argue that “because excise tax comprises an important component of price, the resultant literature has played a prominent role in legislative debates about using

taxa-tion as a principal tool to discourage smoking in individual states, in the U.S. as a whole, and in numerous other countries as well.” The results of this literature are limited,

how-ever, in that they do not consider smokers’ behavioural response. Smokers who want to maintain their current level of tar or nicotine intake may switch to longer cigarettes, to

cigarette higher in tar and nicotine or change their smoking habits. I will refer to such behavioural responses as compensating behaviour. The first studies that explicitly

anal-ysed the compensating behaviour of smokers had to rely on self-reported data, which has some disadvantages. The results of these studies suggest that increasing taxes decreases

(7)

1. Introduction 7

consumption, but smokers will switch to longer cigarettes or cigarettes higher in tar and

nicotine. Recently, the use of biomarker information was introduced in economic research which allows researcher to study, in the context of smoking, changes in smoking habits.

Adda and Cornaglia (2006b) exploit this source of information, which is more reliable com-pared with self-reported data, and conclude that smokers also smoke more intensively in

the presence of higher taxes.

Economists have also analysed the effectiveness of advertising bans. These studies

indicate that banning advertising decreases smoking, particularly when a large number of media is banned since then substitution possibilities to non-banned media are small.

There are also a few articles which analyse the effect of smoking bans in public places. Banning smoking in public places might also result in more exposure to smoking by

non-smokers, especially by young children. As might be clear to reader, developing an coherent smoking discouragement policy will certainly not be easy. Chapter 2 provides a more

detailed discussion of the literature.

The literature on compensating behaviour so far focused on smokers. As far as I am

aware there are no studies that explicitly analyse compensating behaviour by tobacco firms. However, there are a few studies that model the supply side, but focused on the

measurement of monopoly power (Sumner (1981), Sullivan (1985)), or on the question whether tobacco firms price-discriminate (Keeler, Hu, Barnett, Manning, and Sung (1996)).

The aim of this thesis is to analyse compensating behaviour of tobacco firms, with an emphasis on the effect of excise taxes on the producer price of a pack of cigarettes. In

theory it follows from a simple monopoly market model that cigarette producers respond to higher excise taxes and advertising bans by cutting prices. This can also be the case

for a duopoly model, as shown in Schoonbeek and Kooreman (forthcoming). However, the results of Benham (1972) suggest that the price actually increases as advertising is banned.

Studying the effect of an increase in excises is interesting, since it is a priori unclear what the effect on the consumer price will be. That is, since producers respond to an excise tax

(8)

1. Introduction 8

For governments it is not only relevant to know how smokers respond to a tax increase,

but also to understand how tobacco firms respond. In particular, will firms partly offset the tax increase, or will they strengthen the increase. In this thesis, I will empirically estimate

the effect of an increase in excise taxes on the price of a pack of cigarettes. This is discussed in chapter 4. The results of the empirical analysis suggest that the producer price of a pack

of cigarettes actually rises in response to a tax increase. Thus, the retail price increases by more than the amount of the tax increase. As a consequence, cigarette consumption

will decrease more than governments intended. However, though addicts smoke less they will compensate for the tax increase switching to longer cigarettes, cigarettes higher in tar

and/or nicotine, or change their smoking habits.

Finally, chapter 5 concludes this thesis and present several suggestions for future

re-search. The appendices to this thesis provide additional information. Appendix A provides more information on the data and data sources. Appendix B provides the

(9)

2. LITERATURE REVIEW

In the past governments have taken several measures to discourage smoking. At present, there are limitations by age groups, limitations on points of sale, bans on smoking in public

places, health warning on packs and adds, excise taxes and advertising bans. This chapter will mainly review studies that have analysed the effectiveness of excise taxes and

adver-tising bans. First, I will discuss articles which studied the effectiveness of taxes. Cnossen (2005) provides three objectives of excise taxation. First, excise taxes on tobacco “are good

potential sources of revenue, because the products are easy to identify, the volume of sales is high, and the fact that there are a few producers simplifies collection.” For addictive goods,

like tobacco, close substitutes are usually absent and consequently demand is predicted to be inelastic. There are many empirical studies in economics, surveyed in Chaloupka and

Warner (2000), that analyse the effect of prices on quantities smoked. The price elasticity estimates for overall cigarette demand fall in the range -0.3 to -0.5 for most studies,

in-dicating that cigarette demand is indeed inelastic. Economists are often concerned about the distortive nature of taxes, however, as is explained in Cnossen, for addictive goods

“the potential for distortion of economic decisions by the imposition of excise taxes is rel-atively small.” As the collections of excises are a substantial part of total tax revenues, an

inelastic demand implies that it will be a stable source of revenue as well. Obviously, this is attractive for governments.1 _{It is likely, that revenue considerations have been the main}

reason why excise taxes have been enacted. The second objective of excise taxation is to

1_{It is illustrative to compare ratios of taxes on specific goods and services to total tax revenues for}

(10)

2. Literature review 10

reflect external costs. Cnossen (2005) provides a clear description: “Charging consumers

or producers for external costs is known as the Pigouvian prescription (Pigou,1918), which holds that efficient consumption or production can be achieved through the tax system

by imposing an excise on the activity equal to the marginal cost of the damage caused to other people.” Some research made an effort to measure the external costs of smoking and

studied whether existing taxes, both in the United States and Europe, correspond to the Pigouvian prescription. Cnossen and Smart (2005) provide a survey of this literature which

found that the external costs associated with smoking are quite small. Moreover, a purely Pigouvian tax on tobacco would be much lower than existing excise taxes. A possible

explanation for this observation is, according to Cnossen and Smart, that existing taxes “seem to reflect paternalistic attitudes to smoking, rather than externality considerations

appropriate to a government that respects consumer sovereignty.”2 The third objective of excise taxation mentioned by Cnossen (2005) is to discourage consumption. The excistence

of information failures justify government intervention, “even in the absence of explicit ex-ternal costs.” Hines (2007) confirms this by arguing that “irrational consumers may smoke

tobacco products without fully appreciating the regret they will experience years later, and experience what have been called ‘internalities’. In such settings, excise taxation might

help consumers by making the overconsumption of such goods more expensive, and there-fore less likely”. In particular this argument might apply to the young, who are likely to

underestimate the risks of smoking. Moreover, research indicates that the price elasticity of demand amoung the young is, on average, -0.8. Since the elasticity of the young is twice

as large as that of adults, the young are much more responsive to tax changes than adults. Furthermore, it is also interesting to study whether smokers suffer from information

fail-ures. Research by Khwaja, Sloan, and Chung (2007), which compares subjective beliefs about mortality with objectively estimated probabilities, finds that current smokers are

relatively optimistic (i.e. they underestimate the risks). On the contrary, an article by Khwaja, Silverman, and Sloan (2007), finds that smokers have quite accurate beliefs about

2_{In NRC Handelsblad (December 6, 2007) the Dutch Minister of Health, Welfare, and Sport said that}

(11)

mortality. Furthermore, they do not find evidence that smokers are misinformed.

The results of the literature reviewed in this thesis indicate that excise taxes are a useful instrument to raise revenues, and to reduce the incidence of smoking among the young.

Additionally, although demand is inelastic increasing of excise taxes reduces smoking. Unfortunately, the effectiveness of excise taxes is not as unambiguous as the literature

cited above might suggest. In contrast to the studies referred to in Chaloupka and Warner (2000), a few recent studies do not assume that cigarettes are homogeneous, since they vary

in tar and nicotine levels, as well as in size. This enables them to analyse compensating behaviour of smokers. Harris (1980) is one of the earliest contributions. Harris’ analysis

suggests that in case cigarette taxation is independent of tar and nicotine content higher excise taxes may change smoking behaviour. Under the assumption that the quantity

smoked and the nicotine content of cigarettes are substitutes, the results of Harris (1980) indicate that, cigarette consumption will decline and average tar and nicotine levels will

rise as excise taxes are raised.

The study of Evans and Farrelly (1998) is an empirical analysis of compensating

be-haviour of smokers, thereby relying partly on self-reported data. They analysed whether smokers exhibit compensating behaviour using various measures, such as cigarettes per

day, tar and nicotine per cigarette, cigarette length, total millimeters smoked and tar and nicotine smoked per day. Evans and Farrelly find that “for most age groups, the

compen-sating behaviour of smokers is so large that average daily tar intake is unaffected by taxes.” In accordance with the studies surveyed in Chaloupka and Warner (2000), Evans and

Far-relly find that young adults (aged 18-24) are more responsive to tax changes than older smokers are. However, they additionally conclude that young smokers are such responsive

that higher excise taxes actually increase average daily tar and nicotine consumption. The work by Farrelly, Nimsch, Hyland, and Cummings (2004) uses a different dataset than

Evans and Farrelly (1998) but finds very similar results.

Evans and Farrelly (1998) make an important remark: “Our analysis may in fact

(12)

medicine, as is noted by Adda and Cornaglia (2006b), “have shown that smokers can

reg-ulate the amount of nicotine they extract from a given cigarette.” Adda and Cornaglia explain that this can be done “by varying the number of puffs, by varying the degree

and length of inhalation, or by blocking the ventilation holes on the filter”, as is noted in Adda and Cornaglia (2006b). Self-reported data does not allow researcher to study

this type of adjustment, but biomarker information does. This type of data was recently introduced in economic research, and in the context of smoking exploited by Adda and

Cornaglia (2006b). The data source exploited by Adda and Cornaglia contains informa-tion on the number of cigarettes smoked and informainforma-tion on cotinine concentrainforma-tion (a

marker for nicotine concentration) in saliva, for a large number of smokers in the United States over time.3 _{Cotinine concentrations provides helpful information because they “are}

influenced by size and type of the cigarette smoked, but also by the intensity the cigarette is smoked, and therefore it is a good proxy for the intake of health-threatening substances

in cigarettes” (Adda and Cornaglia (2006b)). Adda and Cornaglia empirically estimate the effect of excise taxes on the number of cigarettes smoked, as well as on the amount

of nicotine consumed. Note that they use taxes instead of prices, because the latter are potentially endogenous. The authors conclude that “increases in taxes during the 1990s

led to an increase in the intensity of smoking.” Since it is known that smoking a cigarette more intensively, up to the filter, leads the smoker to inhale more dangerous chemicals and

causes cancer deeper into the lung, the results question the usefulness of excise taxes as an effective smoking discouragement policy.

It is worthwhile mentioning that excise taxes are independent of tar and nicotine con-tent. Since tar is acknowledged to be the main cancer producing substance of cigarettes

and nicotine causes addiction taxes should be higher as the tar and/or nicotine content is greater. However, as is noted by Cnossen (2005), “ some research [. . . ] shows that addicts

smoke low-tar and low-nicotine cigarettes differently, inhaling more to increase the amount of nicotine they ingest. So Pigouvian taxes might not be proportional to tar content, but

(13)

some differentiation is likely to be appropriate still.” At the same time, a tax based on tar

and nicotine levels “would also give the companies an incentive to develop safer cigarettes” (Bulow and Klemperer (1998)). From the articles discussed above one can conclude that

the effectiveness of excise taxes as an instrument to discourage smoking is at least question-able. A more optimal tax scheme, based on tar and nicotine contents, should be balanced

against higher administrative costs and still might generate perverse effects.

I will proceed by discussing articles which analysed the effectiveness of restrictions on

advertising. Several studies have made an effort to estimate the effect of advertising bans, among others, Hamilton (1972), Baltagi and Levin (1986), Baltagi and Levin (1992), and

Saffer and Chaloupka (2000).

Hamilton (1972) studies to what extent United States cigarette consumption has been

affected by cigarette advertising and by cigarette related health-scare messages.In par-ticular, Hamilton evaluates the Congressional ban of broadcast advertising of cigarettes

effective in January 1971. An important consequence of this ban is, as Hamilton explains, that it “ [. . . ] ended the free broadcast time subsidy given to antismoking advertisement.”

Thus, the ban eliminated health-scare messages. If these health-scare messages were more effective than cigarette advertising, the net effect of the ban may be that cigarette

con-sumption increases. This is exactly what Hamilton’s analysis concludes: “[. . . ] eliminating both cigarette advertising and the antismoking messages surely would give per capita

con-sumption a net stimulus.”

The study of Baltagi and Levin (1986) also tries to asses the net effect of this advertising

ban. Baltagi and Levin (1986) estimate a dynamic demand equation for cigarettes using panel data of 46 states over the period 1963-1980. Their results indicate that there is

mild support for the effect of subsidised anti-smoking messages in reducing consumption of cigarettes and no support for the hypothesis that the advertising ban has the effect of

increasing consumption rather than reducing it. The results of Baltagi and Levin (1992) confirm these conclusions.

(14)

dummy variables: weak, limited, and comprehensive ban, depending on the number of

media banned. The results of Saffer and Chaloupka suggest that a comprehensive adver-tising ban reduces tobacco consumption, whereas a more limited set of restrictions have

little or no effect. Intuitively, this can be understood by noting that a limited advertising ban will not reduce the total level of advertising expenditures but will most likely result

in substitution to the remaining non-banned media.

A third instrument governments have at their disposal to discourage consumption is

limiting or banning smoking in public places. A recent study by Adda and Cornaglia (2006a) studies the effect of smoking bans in public places on the exposure to tobacco

smoke of passive smokers in contrast with the effect of excise taxes. Adda and Cornaglia find that excise taxes are an effective instrument to reduce exposure to tobacco. On the

contrary, the conclusion with respect to the effectiveness of bans is more differentiated: “While bans in public transportation, shopping malls, and schools lead to the desired

decrease in exposure of non smokers, we find that bans in recreational public places can perversely increase tobacco exposure of non smokers by displacing smokers to private places

where they contaminate non smokers. Children seem to be particularly affected by this displacement.”

This chapter so far focused on the behavioural response of smokers to government pol-icy. I proceed this review by discussing the behavioural response by tobacco producers,

which will be studies in more detail in Chapters 3 and 4 of this thesis. From a simple monopoly market model it follows that cigarette producers respond to higher excise taxes

and advertising bans by lowering before tax prices, at least theoretically. However, the analysis of Benham (1972) suggests that an advertising ban raise prices. Benham studied

the effect of advertising on the price of eyeglasses and finds evidence for the hypothesis that “restrictions on advertising reduce competition and raise prices.” Thus, if Benham’s results

also hold for the tobacco market, banning advertising is an effective instrument for dis-couraging smoking. Since prices rise cigarette consumption decreases, though those addicts

(15)

3. ECONOMIC THEORY

In the introduction of this thesis and in the previous chapter I claimed that, in theory, a monopolist responds to an increase in excises or an advertising ban by lowering the producer

price. Before showing that this claim is valid, I will discuss why the United States tobacco industry is modeled as a monopoly. Finally this chapter studies the behavioural response

of a monopolist to an increase of excise taxes for a general demand function.

3.1 The monopoly assumption

For the empirical analysis of this thesis we use data for the United States, thus it is

relevant to discuss the monopoly assumption for the US tobacco industry. The US tobacco industry is dominated by a few number of firms for decades. Nowadays the industry is

dominated by four large firms: Phillip Morris (49%), R.J. Reynold (25%), Brown and Williamson (16%), and Lorillard (9%), with market shares between parentheses.1 _These

four firms produce over 98 percent of the cigarettes smoked in the United States. As Gruber (2001) notes: “As one might expect based on this high level of concentration,

the tobacco industry is also very profitable. [. . . ] Given these facts, the lack of entry in this industry likely reflects fear of lawsuits and strong existing brand loyalty, as well as

the informational barriers posed by advertising restrictions (such as the ban on television advertisements for cigarettes).” In their discussion of the Master Settlements Acts, Bulow

and Klemperer (1998) are concerned about the collusive price increases agreed to settle lawsuits: “By agreeing with the attorneys general that each of the companies would pay

a per-pack tax, the companies would push the price of cigarettes closer to the monopoly

(16)

3. Economic Theory 16

level [. . . ].” Another, more pragmatic, reason is that the data only allow me to model a

monopoly. Considering these arguments, in my opinion, modeling the US tobacco industry as a monopoly might be a reasonably well approximation to reality.

Alternatively one might model the U.S. tobacco industry as an oligopoly in which each firm has some local monopoly power (due to brand loyalty). As Phillip Morris has a market

share of 50%, a Stackelberg type of model might be reasonable. Since there is little price difference, within each class of cigarettes, between high-tar and low-tar cigarettes this may

indicate price competition or market power of firms. Note that empirical estimation of an oligopoly model, which is beyond the scope of this thesis, requires data on all firms.

3.2 The effect of an advertising ban on before tax prices

In this section I will show that it follows from a simple monopoly market model that producers respond to an advertising ban by lowering prices. Consider the following market

demand function

q = γ0+ αp?+ γ1A

1 2,

where p? _{= p + τ . Demand, the consumer price, and advertising level are, respectively,}

represented by q, p?, and A; the producer price is denoted by p, and τ represents excise taxes. Furthermore, I assume that the monopolist’s technology is represented by constant

marginal costs c, and c? = c + τ . Note that I assume that the excise tax is a fixed amount per pack, as this represents how excise taxes are levied in the United States.2 _{Note that}

the demand specification assumes that demand in linear in price, which is a common assumption in the literature. Also, the marginal effect of advertising is positive but at a

diminishing rate. The monopolist’s profits are given by

Π (p, A) = (p − c) q − A,

2_{This structure is in the literature known as a specific rate as opposed to the ad valorem rate, which}

(17)

and by noting that (p − c) = (p?_{− c}?_{) the monopolist’s profits can be rewritten as}

Π (p?, A) = (p?− c?) q − A.

The problem of the monopolist is to maximize profits, thus to

max

p?_,A Π(p

?_{, A).}

The first-order conditions with respect to the consumer price and advertising level are given by ∂Π ∂p? = γ0+ αp ?_{+ γ} 1A1/2 + (p?− c?) α = 0, (3.1) ∂Π ∂A = 1 2(p ?_{− c}?_{) γ} 1A−1/2− 1 = 0.

The second-order derivatives are given by

∂2_Π ∂p?2 = 2α, ∂2_Π ∂p?_∂A = 1 2γ1A −1 2, ∂2_Π ∂A∂p? = 1 2γ1A −1 2, ∂ 2_Π ∂A2 = − 1 4(p ?_{− c}?_{) γ} 1A−1 1 2.

Using equation (3.2) it is seen that −1₄(p?− c?_{) γ} 1A−1

1 2 = −1

2A −1_.

Consequently, the Hessian matrix (matrix of second-order partial derivatives) is given by

H =   2α 1₂γ1A− 1 2 1 2γ1A −1 2 −1 2A −1  .

For a maximum it is required that the Hessian is negative definite. An 2 × 2 matrix is

negative definite if and only if the elements on the principal diagonal are negative and its determinant is positive. This implies:

α < 0, α +1 4γ

2

(18)

The optimal demand, consumer price, producer price, and advertising level are given by

˜ qA = 2α (γ0+ αc?) 4α + γ2 1 , ˜ p?_A = (2α + γ 2 1) c?− 2γ0 4α + γ2 1 , ˜ pA = (2α + γ2 1) c − 2 (γ0+ ατ ) 4α + γ2 1 , ˜ A = γ₁2 γ2 0 + 2γ0c? + c? 16α2_{+ 8αγ}2 1 + γ14 .

In order to analyse the claim that prices are lower in case the government imposed an

advertising ban, the optimal demand and consumer price for the model without advertising nee to be estimated. The maximization problem that needs to be solved is as the one solved

above, except there is no advertising. The optimal demand, consumer price and producer price are given by

˜ q = 1 2(γ0+ αc ? ) , ˜ p? = 1 2 c?− γ0 α , ˜ p = 1 2 c − τ − γ0 α

Proving that the claim is correct amounts to showing that the optimal producer price with advertising is higher than without advertising, i.e. it should be true that

˜ pA > p˜ (2α + γ₁2) c − 2 (γ0+ ατ ) 4α + γ2 1 > 1 2 c − τ − γ0 α .

It can be shown that this holds if and only if c < − (τ + γ0/α). This condition holds

(19)

3.3 The effect of an increase in excise taxes on before tax prices

This section will discuss the effect of an increase in excises on before tax price. Take the

derivative of the optimal producer price (with advertising) with respect to τ

∂ ˜pA ∂τ = − 2α 4α + γ2 1 ,

since α < 0 and α +1₄γ₁2 < 0, see equation (3.2) it follows that the derivative

∂ ˜pA

∂τ < 0.

In the model in which I abstract from advertising the derivative equals exactly minus one-half. This results indicates that, in this model, the monopolist responds to an increase in

excise taxes by decreasing the producer price. That is, the monopolist partly offset the effect of an increase of taxes on the consumer price.

3.4 The effect of an increase in excise taxes on the monopoly price

The demand function assumed in the previous section might be too restrictive. Therefore,

in this section I assume a more general demand function, and only requiring that it is concave (i.e Q0 < 0, and Q00 < 0). I still assume that the monopolist maximizes profits,

and has constant marginal costs. In this section I abstract from advertising.

Suppose the monopolist faces a demand curve given by Q (p?_{), which is downward}

sloping, i.e. Q0(p?) < 0. The monopolists problem is to maximize profits, which are given by Π (p?) = (p?− c?_{) Q (p}?_{) ,} thus to max p? Π (p ?_{) .}

Taking the first-order condition with respect to p? yields: ∂Π

∂p? = Q + (p

?_{− c}?₎ ∂Q

(20)

Solving this first-order condition gives the following expression for the consumer price

p? = εc

?

ε − 1,

or the following expression for the producer price

p = εc + τ

ε − 1 , (3.3)

where ε, the price elasticity of demand, is defined as

ε = −∂Q (p

?₎

∂p

p?

Q (p?₎. (3.4)

This definition of the price elasticity of demand is perhaps not customary, but is used in

the industrial organization literature. For example, see Tirole (1988). From equation (3.3) it can be seen that a monopolist will always set his price such that ε > 1. In order to

analyse the effect of an increase in excises on the producer price it is useful to rewrite equation (3.3) 1 − 1 ε p = c + τ ε,

and then to take the total differential of this equation. This is given by

1 − 1 ε dp + p ε 0 ε2 (dp + dτ ) = εdτ − τ ε 0_{(dp + dτ )} ε2 .

Collecting terms and rearranging yields

dp dτ =

1 − τ ε_ε0 −pε_ε0

ε − 1 +pε_ε0 +τ ε_ε0, (3.5)

where ε0 ≡ ∂ε/∂p?_{. In order to analyse the sign of equation (3.5) it needs to be known}

whether the components of the numerator and denominator are either positive or negative. I already showed that ε > 1, and it is assumed that the producer price p, and the tax τ

are positive. So, only the sign of ε0 remains to be established. For this purpose take the derivative of ε (defined in equation (3.4)) with respect to p?_,

(21)

It is assumed that Q0 < 0, and using the sufficient condition for concavity of the profit

function (Q00 < 0) it can be seen that ε0 > 0.3 Hence it is known that ε > 1, and ε0 > 0, and using this it can seen that the denominator of equation (3.5) is positive. Thus it depends

on the sign of the numerator whether the sign of the total derivative is positive or negative. It can be seen that the sign of the numerator is not always positive or negative. First, I

will study in which case the sign of the numerator is positive, and consequently the total derivative is larger than zero. That is

dp dτ > 0

only if the numerator

1 −pε 0 ε − τ ε0 ε = 1 − p?_ε0 ε > 0.

Or, rewritten, only if

1 > p

?

ε ∂ε

∂p?, (3.6)

where the right-hand side of the above expression will be referred to as the elasticity of the price elasticity of demand. The interpretation of the result is as follows: if the elasticity

of the price elasticity of demand is smaller than 1, then the price set by the monopolist will actually increase in response to an increase of excise taxes. On the contrary, when

this elasticity of the price elasticity of demand in larger than 1, then the monopolist will decrease his price in response to an increase of excise taxes. In addition note that

dp

dτ > −1.

This means that a monopolist will not decrease his price by such an amount that this will

completely offset the tax increase. That is, in this model the wholesale price will not be lower as a result of a tax increase.

3_{The profit function is concave if Π}00 _{< 0, i.e. Q}0 _{+ (p}?_{− c}?_{) Q}00_{+ Q}0 _{< 0. It can be shown that a}

(22)

3.5 Summary of results

This section will briefly summarize the main findings of this chapter. In section 3.2 it

is shown that the claim that a monopolist responds to an advertising ban is by a lower producer price is valid. Section 3.3 shows that a monopolist responds to an increase in taxes

by a lower before tax price. The demand function assumed in these two sections, however, might be too restrictive. Therefore, in section 3.4 I analyse the effect of an increase in taxes

on the producer price for a more general demand function. The analysis indicates that the monopolist responds by either increasing or decreasing the producer price. Moreover, if

the monopolist sets a lower producer price, it will never decrease by more than the amount the tax increases. That is, the monopolist will not completely offset the effect of the tax

(23)

4. EMPIRICAL ANALYSIS

This chapter deals with the empirical analysis for which I will use data for the United States at the state level for the period 1976-2003. This chapters starts with a descriptive

analysis of the data. For more information on the data and the data sources see Appendix A. In section 4.2 the empirical model is formulated. The econometric methods employed

are discussed in section 4.3, and the estimation results are discussed in section 4.4.

4.1 Descriptive analysis

Figure 4.1 (a) depicts the number of cigarettes smoked per capita in the United States

for the period 1940-2006. Until the early 1950s cigarette consumption grew steadily, and perhaps as a consequence of successful health scare messages more slowly until the 1980s.

Per capita cigarette consumption declined afterward. The per capita number of cigarettes smoked in 2006 is approximately the same as the number of cigarettes smoked in 1940.

Figure 4.1 (b) shows the real price of a pack of cigarettes for the period 1970-2006. From the 1970s until the early 1980s real prices decreased, increased substantially in 1982-83

and grew steadily until 1992. The substantial increases in cigarette prices in 1982-83 and 1991 are due to implementation of two major federal excise tax increases. In 1983 the

federal excise tax rose from 8 to 16 cents per pack, and in 1991 from 16 to 20 cents per pack. A significant price reduction occurred in 1993-1994 as a result of a price war, which

was caused by a growing market share for the discount brands in the early 1990s. On April 2, 1993 - known as ”Marlboro Friday” - Phillip Morris announced a 40 cent per pack

(24)

4. Empirical analysis 24

taxes as a percentage of the retail price is shown in Figure 4.1 (c). The tax share has

fallen considerably from the 1970s until 1983, then rose due to the federal tax increase. In 1998 the share dramatically decreased as a consequence of the settlement-related price

increase. Figure 4.1 (d) shows the development of real total taxes. From the 1970s until the early 1980s real excise taxes declined steadily, real taxes the jumped and remained

approximately constant for a decade. Real taxes increased as a consequence of federal tax increases in 1991 and 1993 and due to state actions.

Since January 1, 1993 the federal government imposes a excise tax of $0.39 per pack of cigarettes. However, state taxes vary widely, from $2.40 in Rhode Island to $0.07 in

South Carolina in 2006. See also Figure 4.1, which shows the minimum, average, and maximum nominal state tax for the period 1970-2006. Another interesting observation is

that northern states tend to levy higher excise taxes than tobacco-growing southern states (Kentucky, Virginia, North Carolina, South Carolina, Georgia, and Tennessee). Table 4.1

provides some descriptive statistics.

4.2 Empirical model

In this chapter the effect of excise taxes on the price set by a monopolist is empirically

estimated. Notice that I abstract from advertising since, as far as I am aware, there is no advertising data available at the U.S. state level. In this section the empirical model will

be formulated. The empirical model is based on the model introduced in sections 3.1 and 3.2. There I derived the following equation from the monopolist’s maximization problem

p? = 1 2 c?− γ0 α ,

which is a reduced-form equation as all right-hand side variables are exogenous. The

assumed demand equation

q = γ0+ αp?,

(25)

Fig. 4.1: Some descriptive figures for the tobacco industry.

(a) (b)

(26)

Tab. 4.1: Summary statistics

Variable Mean Std. Dev.

Cigarette Sales (in millions of packs) 499.882 494.898

Per capita number sales of packs 109.052 32.531

Real price of a pack 192.713 64.928

Real federal tax rate 25.006 5.165

Real state tax rate 33.433 20.33

Real total tax rate 58.439 22.362

Real per capita disposable income 21195.548 3942.505

Unemployment rate 0.06 0.02 Female 15-24 Yrs 0.079 0.012 25-39 Yrs 0.116 0.012 40-64 Yrs 0.135 0.016 > 65 Yrs 0.068 0.016 Male 15-24 Yrs 0.081 0.011 25-39 Yrs 0.116 0.012 40-64 Yrs 0.131 0.015 > 65 Yrs 0.052 0.01 Population density 130.339 477.14

(27)

Fig. 4.2: Minimum, mean, and maximum nominal state tax

The parameter γ0 will be taken to be a function of real disposable income (INC ), the

unemployment rate (UNEMP ), and demographic variables (MALE and FEMALE ):

γ0 = β0+ β1INC + β2UNEMP + 4 X j=1 β2+jMALEj+ 4 X l=1 β6+lFEMALEl.

As in chapter 2, the technology of the monopolist is represented by a constant marginal cost function C (q) = c·q, and is assumed to consist of production costs, which are invariant

across states, and transportation costs which vary across states. In the empirical model I choose the constant c to be a function of the producer price of tobacco leave (TOB ) and

the population density (POPDEN ):

c = η1TOB + η2POPDEN.

The producer price of tobacco leave is invariant across states and represents production costs; population density is a proxy for the transportation costs. Since cigarette production

is located in a few southern states an additional variable that would proxy transportation cost is the distance between the capital city of each state and the capital cities of one of the

tobacco producing states. For example, Richmond VA., where the headquarters of Philip Morris USA is located. However, I will estimate fixed effects models and since the distance

(28)

It is important to note that the Tax Burden on Tobacco report sales and not

con-sumption. Therefore cross-border shopping and smuggling may be important to take into account. In fact, the reported data on sales may understate or overstate actual

consump-tion, depending on how important smuggling is for the particular state. A number of studies indicate the importance of smuggling, for example Baltagi and Levin (1986) who

note that “At any point in time, cigarette prices vary across states because cigarette tax rates vary. Smokers in high tax state are tempted to buy from neighbouring states with

lower tax rates [. . . ]. Or as Baltagi and Levin (1992) state: “The Advisory Commission on Intergovernmental Relations (ACIR, 1977) found that disparities in tax rates among

states cost the high-tax states $ 391 million in 1975 with 17 states losing more than 9% of their cigarette revenue due to tax evasion or tax exemption.” Following Baltagi and

Levin (1986) I model the problem of bootlegging by introducing Pn, which denotes the

minimum real price of a pack of cigarettes in any neighbouring state, and acts as a

sub-stitute price. As already realised by Baltagi and Levin (1986) this way of modeling the bootlegging problem is not free of problems, e.g. “it does not take into account that in

large states, cross-border shopping may occur in different neighbouring states and not just the minimum-price neighbouring state” (Baltagi and Levin (1992)).1 2

The market equilibrium is determined by both demand and supply, so the number of packs of cigarettes sold as well as its price are endogenous variables. In fact, it is a system

of two equations that can be jointly estimated. The system of equations consists of a structural-form demand equation and a reduced-form equation that determines the price.

Notice that it is assumed that both equation hold for each state and each year. The system

1_{Remark that Alaska and Hawaii have no other US states as their neighbours. Therefore I take their}

own price as the minimum price.

2_{Following Baltagi and Levin (1992), I also estimate the model with the maximum real price of cigarettes}

(29)

of simultaneous equations is given by:

qkt = αp?kt+ β0 + β1INC_kt+ β₂UNEMP_kt+ 4 X j=1 β2+jMALE_j,kt+ 4 X l=1 β6+lFEMALE_l,kt+ β₁₁P_n kt +β12TREND_kt+ β₁₃TREND2_kt+ u_1,kt

p?_kt = δ1TOB_kt+ δ₂POPDEN_kt+ δ₃+ δ₄INC_kt+ δ₅UNEMP_kt+

4 X j=1 δ5+jMALE_j,kt+ 4 X l=1 δ9+lFEMALE_l,kt +δ14TREND+ δ₁₅TREND2+ u_2,kt,

where there are the following restrictions on the parameters

δ1 = 1 2η1; δ2 = 1 2η2 δi = − βi−3 2α for i = 3, 4, . . . , 13.

Furthermore, I assume that the composite error terms u1kt = µ1k+ 1kt, and u2kt = µ2k+

2kt. The unobserved random variables µ1k, and µ2k are modeled as parameters to be

estimated and capture the unobserved heterogeneity over states. Since the producer price of tobacco leaves is invariant across states, including time-dummies will result in perfect

multicollinearity. Instead I include a variable named trend, and additionally its square (trend2 ).3

Besides the system of equation, I will also estimate the reduced-form of the demand equation which is also given by

qkt = φ0+ φ1INC_kt+ φ₂UNEMP_kt+ 4 X j=1 φ2+jMALE_j,kt+ 4 X l=1 φ6+lFEMALE_l,kt+ φ₁₁P_n kt

+φ12TOB+ φ₁₃POPDEN+ φ₁₄TREND_kt+ φ₁₅TREND2_kt+ u_kt,

where there are the following restrictions on the parameters

φi = 1 2βi, for i = 0, 1, . . . , 10; φ11= 1 2αη1; φ12= 1 2αη2.

It is explicitly taken into account that prices and quantities are endogenous. However, also the tax rate might be endogenous as politicians might respond to observed consumption

3_{Trend is defined as: trend}

kt = t − 1975, and trend2kt = (t − 1975) 2

(30)

of cigarettes. A simple regression of the tax rate on its one-period lagged valued and the

one-period lag of per capita consumption does not indicate that the tax rate is endogenous. However, still I use a variable denoted political as an instrument for the excise tax. This

variable is a binary variable and equals 1 when the governour of the state is a democrat, and 0 otherwise.

Before discussing the estimation methods, I will motivate the stochastic disturbance. In structural models there are three motives for the stochastic disturbance. First,

het-erogeneity in preferences is partly dependent on observed regressors but also depends on variables that are not observed (captured by the µ1,kt and µ2kt in out model). Second,

there are disturbances due to errors in the optimization. I assume that the monopolists maximizes profits and determines its price according to some formula. However, the

mo-nopolist might not strictly apply this derived pricing-rule. Third, measurement errors in the observed data motivate the stochastic disturbance.

4.3 On the econometric methods

The estimation of a simultaneous-equation model (SEM) with either cross-sectional data or time series data is well known and treated in most econometric textbooks. However, as

is noted by Krishnakumar (1996), “when panel data or pooled data are available to the researcher [. . .], the classical simultaneous model is no longer adequate as there may be

specific effects associated with the different cross-sectional units and/or the time periods.” See Baltagi (1995) and Krishnakumar (1996) for a treatment of several estimators in case

of a SEM using panel data. Both authors consider random effects estimators, Baltagi (1995) assumes a one-way error component structure and Krishnakumar (1996) a two-way

structure. Instead, I assume fixed effects which is an appropriate assumption if the focus is on a specific set of N American states (see Baltagi (1995)). The reduced-form demand

equation will be estimated by the usual fixed effects (within) estimamtor. The system of equations will be estimated with maximum likelihood assuming that the remainder error

(31)

4. Empirical analysis 31   1kt 2kt   ∼ N     0 0  ,   σ2 1 ρσ1σ2 ρσ1σ2 σ22    , where σ2

1, and σ22 are the variances of 1,kt, and ε2,kt respectively, and ρ is the correlation

coefficient. Appendix B provides the Stata likelihood-evaluation programme used for the

maximum likelihood estimation.

4.4 Estimation Results

In this section I will discuss the estimation results, which are reported in Tables 4.2 and 4.3.

Concerning the demand equation, prior to estimation, I expect a particular sign for some regressors. It is found that unemployed smoke more than people with a job (see e.g. Saffer

and Chaloupka (2000)), so states in which the unemployment rate is larger are expected to have higher the per capita sales of cigarettes. Because cigarettes are an inferior good

(see Bulow and Klemperer (1998)), the sign of real personal disposable income is expected to be negative. If bootlegging is indeed important the expected sign is positive. Obviously

the retail price should have a negative effect on demand. For the price setting equation it is expected that the producer price of tobacco leaves and population density have a positive

sign, and the sign of real total tax could be either positive or negative.

Consider the estimation results in Table 4.2. Column 1 provides the estimates for the

reduced-form demand equation. The coefficient of the unemployment rate is negative and significantly different from zero, which is at odds with prior expectations. The coefficient of

real income is also negative and significantly different from zero, confirming that cigarettes are an inferior good. Bootlegging seems not important, since the coefficient does not

significantly differ from zero at the 5% level (though it is close to signficance at 10%). The sign of the coefficient of real total tax is negative and also significantly differs from zero,

(32)

Most important, the coefficient of real price is negative and significantly different from zero.

Concerning column 3, which provides estimates of the reduced-form price setting equation, I am most interest in the coefficient of real total tax, the estimated value is 1.134. A

t-test indicates that the estimate statistically differs from one-half. Recall that the value of one-half was predicted by the monopoly model of chapter 3 without advertising. More

important, the estimate also differs significantly from one, as revealed by another t-test. From this I infer that the monopolist increases its producer price in response to higher taxes.

This result has also been found by others, see for example Evans and Farrelly (1998) who find a value of 1.17. I also performed a overidentifying restrictions test, which rejects the

nullhypothesis. So there is doubt about the validity of the instruments. However, this is not unexpected due to the restrictive model assumptions.

Table 4.3 report the results of the estimations for which I used the variable political as an instrument for real total tax. The results do not differ much from the ones just

discussed, therefore I will focus on the most interesting results. From the second column, which provides the estimates of the structural-from demand equation, I infer that the

coefficient of the unemployment rate confirms expectations. The coefficients of political are not significantly different from zero, probably due to weakness of the instrument.

4.5 Nonlinear model

The monopoly market model of chapter 3 predicted that as taxes increase the retail price of cigarettes would increase by only half the amount of the tax increase. However, as the

results of the empirical analysis indicate this prediction is not supported by the data. For this reason, one would like to have a demand function that allows for more flexibility of

the effect of taxes on the monopoly price. Such a more general demand function is

(33)

Tab. 4.2: Estimation results I

Fixed Effects Maximum Likelihood Demand Demand Price setting eq. Variable Coefficient Coefficient Coefficient

(Std. Err.) Std. Err. Std. Err. Unemployment rate -104.825† -46.475† 264.957† (19.694) (20.601) (39.631) Real income -0.0009† -0.0004 0.004† (0.0003) (0.0002) (0.0005) Pn 0.008 0.075† -(0.010) 0.011 -Female 15-24 Yrs 570.237† 407.739† 62.770 (129.889) (127.686) (263.513) 25-39 Yrs 345.979† 431.049† 130.538 (163.068) (161.655) (327.917) 40-64 Yrs 333.065† _43.015 _-598.802† (138.497) (137.361) (280.581) > 65 Yrs 338.105† 545.202† 743.639† (77.245) 78.923 (156.608) Male 15-24 Yrs 261.207 700.595† 427.070 (136.533) (131.432) (276.152) 25-39 Yrs 287.313 317.770† _524.746 (160.859) (157.747) (325.766) 40-64 Yrs 20.578 213.316 809.125† (112.910) (113.825) (226.694) > 65 Yrs 306.393† _167.278† _102.863 (82.151) (83.102) (163.312) Real tobacco 2.926† - -6.027† 1.122 - 2.110 Population density 0.035† - 0.023 0.007 - (0.014) Real price - -0.302† -- (0.017)

(34)

Tab. 4.3: Estimation results II

Fixed Effects Maximum Likelihood Demand Demand Price setting eq. Variable Coefficient Coefficient Coefficient

(35)

Solving the monopolist’s profit maximization problem yields as a reduced-form equation

for p

p = ζαck− γ0− ατk α (1 + ζ) .

The effect of an excise tax on the monopolist’s price is given by

− α

α (1 + ζ) = − 1 1 + ζ.

As one can see, this is a more general specification than with a linear demand function.

Joint estimation of the demand function and the reduced-form equation that determines the monopolist’s price is more difficult than in the linear case. There are at least two reasons

for this. First, for computational reasons it is relatively difficult to estimate a nonlinear function. Second, in case of panel data, it is most logical to assume that the ‘cross-section

specific effect’ is in the nonlinear part of the demand equation. For an estimator to be (asymptotically) consistent, it is now required that both the cross-sectional number of

observations as well as the time-series go to infinity. This is clearly not the case for our application, and the estimates will therefore be inconsistent. Then a simulation study is

(36)

5. CONCLUSION

Since a few years governments are under pressure not only to discourage smoking, but also to limit or even ban it via legislation. Governments have several instruments at their

disposal to discourage smoking, one of these instruments being excise taxes. There are three main objectives of excises taxation identified in the literature. First, raising revenue.

Second, reflecting external costs. Third, discouraging smoking. Excise taxes on tobacco products are a good source of revenue. The external costs of smoking are quite small,

and existing taxes are much larger than a Pigouvian prescription would require. As a instrument to discourage smoking the effectiveness of excises is questionable according to

results of recent studies. In response to higher taxes smokers change their smoking habits in such a way that it is worse in terms of health consequences.

Tobacco firms respond to higher taxes by increasing their producer price. That is what the results of chapter 4 suggest. For governments this may seem positive news since this

strengthens the effect of higher taxes and as a consequence smoking is discouraged more than intended. However, more addicts will switch to longer cigarettes or to cigarettes higher

in tar and nicotine or change their smoking habits. On the other hand, the incidence of smoking among the young is reduced more than intended, as well as exposure of smoke to

nonsmokers. A tax scheme based on tar and nicotine content may discourage compensating behaviour of smokers in response to higher excises and may give tobacco firms an incentive

to develop safer cigarettes.

Banning advertising is only useful if a comprehensive ban (in terms of Saffer and

Chaloupka (2000)) is enforced, and public bans of smoking are only effective if they limit smoking in public transportation, shopping malls, and schools, not in recreational places.

(37)

5. Conclusion 37

these issues and suggest bold and radical reforms. Their point of view is laudable, but too

simplistic in the sense that they do not take into account how public policies can generate perverse incentives and effects. The most important policy recommendation is therefore

that policymakers and politicians should carefully evaluate their proposed measures. Future research on compensating behaviour of tobacco firms should aim at estimating

models based on monopolistic competition. Furthermore, availability of advertising expen-diture data would enable researcher to estimate the effect of banning advertising. Research

(38)

(39)

A. DATA DESCRIPTION

The purpose of this Appendix is to describe the data search process as well as describe the data finally obtained.

Researcher who want to study the economics of smoking might encounter difficulties

finding useful data. Some surveys report data on the number of cigarettes smoked and numerous background variables of respondents. However, data on prices and excise taxes

on cigarettes are more difficult to find. For the United States the only source seems to be the Tax Burden on Tobacco, of which the most recent version is published by Orzechowski

and Walker (2006). There is no data at the firm level is available, which a reseacher is ultimately willing to have acces to. So, in case of the US cigarette industry, data for the

firms Phillip Morris, R.J. Reynold, Brown and Williamson, and Lorillard is required. Data on advertising expenditures are only available at the federal level, and is aggregated as

well. For OECD countries there is also a database available from Health New Zealand, but this comes at a serious expense. Moreover the data is not at the firm level, and no data

on excise taxes and advertising is available.

As I use data for the United States, the following describes the sources. The Tax Burden on Tobacco reports annual per capita consumption of cigarettes by state. This is measured

in packs of cigarettes per head. It also provides the average retail price per pack, as well as the tax rate per pack. Both are measured in US dollar cents. Per capita disposable

personal income data are obtained from the Bureau of Economic Analysis, price deflators and the unemployment rate are obtained from the Bureau of Labor Statistics. Population

(40)

A. Data description 40

Tab. A.1: Description of variables

Variables Description

Per capita packs Tax-paid per capita sales in number of packs

Price Average retail price of a package of 20 cigarettes in US dollar cents

State tax State cigarette tax rate in US dollar cents

Total tax Total cigarette tax rate in US dollar cents (stata tax + federal tax)

Unemployment rate Unemployed workers as a fraction of the labour force, not seasonally adjusted

Income Per capita disposable personal income Male (Females):

(1) 15-24 Yrs Males (females) in age group 15-24 as a fraction of total population

(4) >65 Yrs Males (females) that are 65 and over as a fraction of total population

Tobacco Producer price of tobacco leaves, varies only over time and not by state

Population density Population per square kilometre

(41)

B. STATA COMMANDS

Estimation of the several models is done with the use Stata. In order to be able to compute maximum likelihood estimates I wrote a likelihood-evaluation programme. To estimate the

system of 2 equation, of which is assumed that the errors are bivariate normally distributed, I use the following programme.

capture program drop mymle program define mymle

args lnf mm1 mm2 ss1 ss2 rrho #delimit ; quietly replace ‘lnf’ = ln( (1.0/(2.0*_pi*‘ss1’*‘ss2’*sqrt(1.0-‘rrho’^2)))* exp((-1/(2*(1.0-‘rrho’^2))) * (((($ML_y1-‘mm1’)/‘ss1’)^2) - (2 *‘rrho’*(($ML_y1-‘mm1’)/‘ss1’)*(($ML_y2-‘mm2’)/‘ss2’)) + ((($ML_y2-‘mm2’)/‘ss2’)^2) ))) ; #delimit cr end

ml model lf mymle (eq1: dependent variable 1 = independent variables)

(eq2: dependent variable 2 = independent variables) /sigma1 /sigma2 /rho ml init /sigma1=100000 /sigma2=100000

(42)

REFERENCES

Adda, J., & Cornaglia, F. (2006a). The effect of bans and taxes on passive smoking. Cemmap working paper.

Adda, J., & Cornaglia, F. (2006b). Taxes, cigarette consumption, and smoking intensity. American Economic Review, 96, 1013-1028.

Baltagi, B. (1995). Econometric analysis of panel data. John Wiley and Sons.

Baltagi, B., & Levin, D. (1986). Estimating dynamic demand for cigarettes using panel

data: the effects of bootlegging, taxation and advertising reconsidered. Review of Economics and Statistics, 68, 148-155.

Baltagi, B., & Levin, D. (1992). Cigarette taxation: raising revenues and reducing con-sumption. Structural Change and Economic Dynamics, 3, 321-335.

Benham, L. (1972). The effect of advertising on the price of eyeglasses. Journal of Law and Economics, 15, 337-352.

Bulow, J., & Klemperer, P. (1998). The tobacco deal. Brookings Papers on Economic Activity: Microeconomics, 323-394.

Chaloupka, F., & Warner, K. (2000). The economics of smoking. In A. Culyer & J. New-house (Eds.), Handbook of health economics (p. 1539-1627). Elsevier Science.

Cnossen, S. (2005). Economics and politics of excise taxation. In S. Cnossen (Ed.), Theory and practice of excise taxation (p. 1-19). Oxford University Press.

Cnossen, S., & Smart, M. (2005). Taxation of tobacco. In S. Cnossen (Ed.), Theory and practice of excise taxation (p. 20-55). Oxford University Press.

Evans, W., & Farrelly, M. (1998). The compensating behavior of smokers: taxes, tar, and nicotine. RAND Journal of Economics, 29, 578-595.

(43)

References 43

cigarette prices on tar and nicotine consumption in a cohort of adult smokers. Health

Economics, 13, 49-58.

Gruber, J. (2001). Tobacco at the crossroads: the past and future of smoking regulation

in the united states. Journal of Economic Perspectives, 15, 193-212.

Hamilton, J. (1972). The demand for cigarettes: advertising, the health scare, and the

cigarette advertising ban. Review of Economics and Statistics, 54, 401-411. Harris, J. (1980). Taxing tar and nicotine. American Economic Review, 70, 300-311.

Hines, J. (2007). Taxing consumption and other sins. Journal of Economic Perspectives, 21, 49-68.

Keeler, T., Hu, T., Barnett, P., Manning, W., & Sung, H. (1996). Do cigarette produc-ers price-discriminate by state? an empirical analysis of local cigarette pricing and

taxation. Journal of Health Economics, 15, 499-512.

Khwaja, A., Silverman, D., & Sloan, F. (2007). Are smokers misinformed? Working paper.

Khwaja, A., Sloan, F., & Chung, S. (2007). The relationship between individual expecta-tions and behaviors: mortality expectaexpecta-tions and smoking decisions. Journal of Risk

Uncertainty, 35, 179-201.

Krishnakumar, J. (1996). Simultaneous equations. In L. Matyas & P. Sevestre (Eds.), The

econometrics of panel data. Kluwer Academic Publisher.

OECD. (2006). Consumption tax trends, 2006 edition. Organisation for Economic

Coop-eration and Development.

Orzechowski, & Walker. (2006). The tax burden on tobacco: historical compilation (Vol. 41).

Orzechowski, and Walker.

Peto, R., Lopez, A., & Boqi, L. (1999). Global tobacco mortality: Monitoring the growing

epidemic. In R. Lu, J. Mackay, S. Niu, & R. Peto (Eds.), The growing epidemic. Springer-Verlag.

Saffer, H., & Chaloupka, F. (2000). The effect of tobacco advertising bans on tobacco consumption. Journal of Health Economics, 19, 1117-1137.

(44)

References 44

Sullivan, D. (1985). Testing hypotheses about firm behavior in the cigarette industry.

Journal of Political Economy, 93, 586-598.

Sumner, D. (1981). Measurement of monopoly behavior: an application to the cigarette

industry. Journal of Political Economy, 89, 1010-1019.

On the Eﬀectiveness of Smoking Discouragement Policies

On the Effectiveness of Smoking Discouragement

Policies

Master Thesis Research Master in Economics and Bussiness

December 18, 2007

CONTENTS

1. INTRODUCTION

2. LITERATURE REVIEW

3. ECONOMIC THEORY

3.1

The monopoly assumption

3.2

The effect of an advertising ban on before tax prices

3.3

The effect of an increase in excise taxes on before tax prices

3.4

The effect of an increase in excise taxes on the monopoly price

3.5

Summary of results

4. EMPIRICAL ANALYSIS

4.1

Descriptive analysis

4.2

Empirical model

4.3

On the econometric methods

4.4

Estimation Results

4.5

Nonlinear model

5. CONCLUSION

A. DATA DESCRIPTION

B. STATA COMMANDS

REFERENCES