Economics Essays - Economics Smoking Health

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Economics Smoking Health

The Economics of Smoking

Economics of Smoking - p. 1

1. Introduction and overview

Historically one of the oldest and most important crops in the United

States, tobacco has become embroiled in the second half of the twentieth

century in a struggle pitting American economic against public health

interests. While the tobacco industry ranks among the most substantial and

successful economic enterprises in the U.S., tobacco products are associated

Services, 1989, 1998). Tobacco products, and particularly cigarettes, whichwith more deaths than any other product (U.S. Department of Health and Human

account for 95% of U.S. tobacco product sales, are credited with approximately

one-fifth of the nation's annual death toll. Cigarettes cause fully a third

of deaths during middle age. The leading cause of lung cancer and chronic

obstructive pulmonary disease mortality, as well as a major cause of

cardiovascular death, cigarette smoking leads all other causes of death in

virtually all industrialized nations. According to an epidemiological

analysis sponsored by the World Health Organization, tobacco will become the

leading cause of death in developing countries during the first third of the

21st century. By 2030, tobacco will be responsible for 10 million deaths

annually worldwide (Peto et al., forthcoming), a toll that will exceed by far

that associated with any other cause of disease (Murray and Lopez, 1996).

Formal economic analysis of tobacco dates back at least half a century

(Tennant, 1950). At that time, most tobacco industry economic analysis was

motivated by the factors that prompted market analysis of any other product or

service, such as a desire on the part of an industry to understand the degree

of price elasticity of demand for its product, or the interest of government

and academic economists in the causes and implications of market

concentration. Beginning in the late 1960s, however, following publication of

seminal British and American reports on smoking and health (Royal College of

Physicians, 1962; U.S. Department of Health, Education, and Welfare, 1964),

the focus of economic research shifted from a general industrial organization

orientation toward analysis self-consciously relevant to the public health

Economics of Smoking - p. 2

damage wrought by tobacco. The subsequent economic analysis has been

motivated by a desire to determine how economic forces influence tobacco

consumption, with continuing emphasis on refining the scientific rigor of the

work; but the objective of much of the research is now to determine how to

harness economic forces and logic, how to use economic tools, to decrease

smoking, with the ultimate goal being to reduce the toll of tobacco. 1

Certainly the most important example of this phenomenon has been the

rapidly expanding and increasingly sophisticated body of research on the

effects of price increases on cigarette consumption. Because excise tax

comprises an important component of price, the resultant literature has played

a prominent role in legislative debates about using taxation as a principal

tool to discourage smoking in individual states, in the U.S. as a whole, and

in numerous other countries as well. In the United States in the late 1990s,

the findings of this literature have been showcased in the intense

congressional debate over whether to adopt comprehensive tobacco control

legislation, with a major price increase lying at the heart of all proposals

(Chaloupka, 1998).

1.1 Coverage

This chapter examines in detail economic analysis of the relationships

among taxation, price, consumption, and disease outcomes, as well as

considering how analysis has enlightened other debates about the economics of

tobacco. The relationship between price and cigarette consumption has been

the focal point of economic research on smoking, and the locus of increasingly

sophisticated and interesting development of theory and methodology. For

these reasons, and because the resultant literature constitutes the most

1 Not all of the research is motivated by a desire to decrease smoking.

Some authors express the opinion that more respect should be accorded consumer

sovereignty, despite the issues of addiction and youthful initiation of

smoking that have led many economists to perceive the market for cigarettes as

suffering from important market imperfections (Warner et al., 1995). See, for

example, Viscusi (1992) and Tollison and Wagner (1992).

Economics of Smoking - p. 3

important contribution of economics to understanding tobacco policy, this

chapter's principal emphasis is on this body of research. The chapter devotes

special consideration to recent attempts to model nicotine addiction in the

context of rational economic behavior. Borne of a generic interest in the

role of addiction in economic behavior, new theoretical models have received

their best empirical testing through the use of data on cigarette smoking. 2

The chapter also considers a variety of equity and efficiency concerns

that invariably accompany debates about cigarette taxation, including the

validity of the externality or social cost argument frequently invoked by the

public health community in calls for higher taxes; whether there are other

legitimate grounds in economic theory to support increased excise taxation;

whether cigarette tax increases are regressive, and if so how much; and

whether large tax hikes produce substantial cross-border smuggling of

cigarettes. Recent economic studies have explored subtle impacts of cigarette

taxation that receive attention here as well; for example, cross-price

elasticity issues include how cigarette taxation may shift demand toward other

tobacco products, such as smokeless tobacco, or, within the cigarette family,

from lower to higher tar and nicotine cigarettes.

Coverage in this chapter also includes attention to economic analysis of

the role of advertising in the demand for cigarettes, as well as the role of

restrictions or bans on advertising. The effects of advertising and of

advertising and other marketing restrictions are of special interest due to

their prominence in debates about tobacco control at all levels of government,

from municipalities, which have restricted cigarette advertising on public

transit and on billboards, to international bodies, which have called for

2 As is discussed below, numerous variables other than price influence the

demand for cigarettes, including consumers' knowledge of the hazards of

smoking, parental and sibling smoking behavior, smoking by peers, role

modeling, income, and education (U.S. Department of Health and Human Services,

1994). The near-exclusive focus of this chapter on price and taxation

reflects the facts that tax is the most policy-tractable variable influencing

the demand for cigarettes and that the economics literature has focused on

price, taxation, and addiction.

Economics of Smoking - p. 4

complete bans (Roemer, 1993). Although its value is constrained by obvious

limitations, econometric analysis offers insights into the role of

advertising, and of advertising restrictions, on the demand for cigarettes.

The chapter also examines what is known about the influence of other

tobacco control policies on the demand for cigarettes, including the

development and dissemination of information on the health consequences of

smoking; media advocacy by means of "counter-advertising;" the adoption and

implementation of laws or policies that limit smoking in public places; and

legal restrictions on youth access to tobacco products.

Given tobacco's role in employment, tax revenues, and, in selected

countries, trade balances, governments have a legitimate interest in the

"health" of their tobacco industries. Tobacco companies tout the industry's

economic contribution in attempts to combat tobacco control policy measures.

In recent years, independent economists have countered the industry's economic

argument by carrying out macroeconomic analyses that examine the net

contributions of tobacco to economies, rather than the gross contributions

featured by the industry. This literature, and its role in the debate over

tobacco control policy, are examined toward the end of the chapter. Also

considered briefly is the influence of tobacco agriculture support policies so

prominent in the agricultural policies of the United States and the European


Despite its wide scope of coverage, this chapter does not examine all of

the economic contributions relating to smoking and health. To illustrate with

two examples, the chapter does not consider the growing literature on the

cost-effectiveness of smoking cessation interventions ( Cromwell et al., 1997;

Warner, 1997) and it omits the newly emerging and potentially quite important

analysis of the economics of the market for nicotine replacement products

(Oster et al., 1996; Hu et al., 1998). Another limitation on coverage is that

we consider almost exclusively English language publications, believed to

comprise a very sizable majority of the peer-reviewed literature. Further,

Economics of Smoking - p. 5

reflecting the authors' knowledge of this field and familiarity with data,

examples draw heavily, although not exclusively, on the U.S. experience. In

particular, there is little coverage of the economics of smoking in developing

countries, the result primarily of the dearth of studies on the subject.

Although specific empirical conclusions from a given country may not apply

precisely to other nations, the general phenomena described and findings

presented should apply qualitatively to all countries, unless otherwise


Before turning to the economics literature, the remainder of this

introductory section presents a brief "primer" on the health consequences of

tobacco use. We deem this important background for understanding the nature

and social significance of the economic issues.

1.2 Health consequences of tobacco consumption

The health implications of tobacco have been contemplated for at least

the past millenium. During the first half of that period, the predominant

view held that tobacco afforded users a wide variety of health benefits. The

Amerindians employed tobacco as an analgesic and as a treatment for such

diverse ailments as intestinal problems, asthma, rheumatism, headaches,

toothaches, boils, worms, fevers, and the pains of childbirth (Goodman, 1993).

Serious medical and scientific attention to the health consequences of

smoking is a phenomenon of the present century, primarily of its second half. 3

This is a reflection of the development of the science of epidemiology during

3 Concern about the health consequences of smoking predates the "modern

era" by nearly four centuries. In 1604, for example, King James I of England

lambasted smoking as "a custome lothsome to the eye, hatefull to the Nose,

harmefull to the braine, dangerous to the Lungs, and in the blacke stinking

fume thereof, neerest resembling the horrible Stigian smoke of the pit that is

bottomlesse" (as quoted in Sullum, 1998, p. 18). King James subsequently

raised the tax on tobacco by 1000%, deriving significant revenues for his

coffers. This illustrates the profound dilemma that has confronted policy

decision makers ever since: whatever its health consequences, tobacco has

long been truly a "golden leaf" for farmers and politicians alike. Its role

in the very earliest commerce between England and the American colonies is

legendary, as is its role in contemporary politics (Taylor, 1984; Fritschler

and Hoefler, 1996).

Economics of Smoking - p. 6

this period and of the relatively modest number of victims claimed by tobacco

prior to the 20th century. Before this century, relatively few people reached

the ages at which tobacco takes its greatest toll (average life expectancy in

the U.S. was 47 in 1900; currently it is 75). More importantly, widespread

intensive use of the most dangerous form of tobacco consumption, cigarette

smoking, began only in the very late 1800s. Lung cancer, today the source of

30% of all cancer deaths in the U.S. (U.S. Department of Health and Human

Services, 1989), was a rarity until earlier cigarette smoking spawned the

epidemic first widely observed during the 1930s.

Although a few scientific studies associated smoking with disease prior

to mid-century (Broders, 1920; Lombard and Doering, 1928; Pearl, 1938), the

first evidence that strongly implicated smoking in disease (specifically, lung

cancer) was published in the 1950s (Wynder and Graham, 1950; Doll and Hill,

1954, 1956; Hammond and Horn, 1958a, 1958b). Since then, some 70,000

scientific articles have implicated smoking in a wide variety of ailments,

constituting the largest and best documented literature linking any behavior

to disease in humans (U.S. Department of Health and Human Services, 1994).

Today, cigarette smoking is esta blished as the leading cause of lung

cancer (responsible for approximately 90% of lung cancer deaths in the U.S.),

the leading cause of chronic bronchitis and emphysema (responsible for over

80% of chronic obstructive pulmonary disease deaths), and a major cause of

heart disease and stroke. Smoking also causes aneurysms, atherosclerotic

peripheral vascular disease, oral cavity and laryngeal cancer, intrauterine

growth retardation and neonatal death, including SIDS (Sudden Infant Death

Syndrome). It is associated with additional cancers (bladder, pancreatic,

renal, gastric, and cervical) (U.S. Department of Health and Human Services,

1989), as well as a host of other conditions affecting a wide variety of organ

systems and disease processes, including, for example, vision and hearing

problems, slowed healing from injuries, and increased susceptibility to

certain infections (Napier, 1996). Chronic inhalation of environmental

Economics of Smoking - p. 7

tobacco smoke (ETS) causes lung cancer in nonsmokers and an assortment of

diseases and functional limitations in the children of smokers (Environmental

Protection Agency, 1994). ETS may be responsible for tens of thousands of

heart disease deaths annually ( Glantz and Parmley, 1995).

All told, smoking is far and away the leading cause of premature death

and of avoidable morbidity and disability in the United States and in most

industrialized nations. As indicated above, the intensification of smoking in

the world's less affluent nations will soon bring the same distinction to

smoking in the developing countries. Barring substantial and unexpected

decreases in tobacco use worldwide, a few decades hence the global death toll

from tobacco will dwarf all other causes, with the majority of deaths

occurring in the developing nations. The World Health Organization estimates

that fully 500 million of the 5 billion people alive at the beginning of this

decade will die as a result of consumption of tobacco products ( Peto et al.,


The mortality toll of tobacco reflects not only the le thality of tobacco

products but also the prevalence of their consumption. In the United States,

approximately 45 million adults, almost a quarter of the adult population,

smoke cigarettes (down from a high of 42% in 1965 (U.S. Department of Health

and Human Services, 1989)). Worldwide, tobacco products are used by

approximately one billion people. The large numbers of tobacco consumers,

combined with their frequent use of tobacco products, account not only for the

disease toll of tobacco, but also for the substantial size of the tobacco

industry. Important features of the structure and economic importance of the

industry are reviewed in section 6 below.

2. The impact of price on the demand for tobacco products

Many researchers once viewed cigarette smo king and other addictive

behaviors as irrational and therefore not suitable for conventional economic

analysis (Elster, 1979; Winston, 1980; Schelling, 1984b). They believed that

Economics of Smoking - p. 8

the demand for cigarettes (and other addictive substances) did not follow the

basic laws of economics, including perhaps the most fundamental law, that

embodied in the downward-sloping demand curve. As the now-substantial body of

economic research demonstrates, however, the demand for cigarettes clearly

responds to changes in prices and other factors, as found in applications of

both traditional models of demand and more recent studies that explicitly

account for the addictive nature of smoking.

Conceptually, economists use a relatively broad definition of price that

includes not only the monetary price of purchasing a product, but also the

time and other costs associated with using the product. Restrictions on

smoking in public places and private work sites, for example, impose

additional costs on smokers by forcing them outdoors to smoke, raising the

time and discomfort associated with smoking, or by imposing fines for smoking

in restricted areas. Similarly, limits on youth access to tobacco may raise

the time and potential legal costs associated with smoking by minors, while

new information on the health consequences of tobacco use can raise the

perceived long-term costs of smoking. This section focuses on the effects of

monetary price on demand, while section 5 below considers the effects of other

aspects of full price.

In addition to price, a variety of other factors can affect the demands

for cigarettes and other tobacco products, including income, advertising and

other promotional activities, and tastes. In the industrialized nations, the

relationship between income and cigarette consumption has reversed. Early

demand studies (for example, Ippolito, et al., 1979; Fujii, 1980) concluded

that cigarette smoking was a normal good, with cigarette consumption rising as

income rose. More recent studies, however, have found that cigarettes have

become an inferior good, in that the likelihood of smoking declines as income

rises (Wasserman, et al., 1991; Townsend et al., 1994). The effects of

advertising and promotion on the demand for cigarettes have been the subject

of numerous studies; these are reviewed in detail in section 4 below.

Economics of Smoking - p. 9

Finally, nearly all econometric studies of cigarette demand use a variety of

factors to control for tastes, including gender, race, education, marital

status, employment status, and religiosity. Given the focus of this book on

economics, the impact of these socio-demographic determinants of demand will

not be reviewed.4

This section begins with a review of conventional studies of the impact

of money price on cigarette demand. This is followed by a discussion of

economic models of addiction and their applications to cigarette demand.

Implications for the effects of price on cigarette demand from the relatively

new field of behavioral economics are then reviewed. The section closes with

a short consideration of the relatively limited research on the effects of

price on the demand for other tobacco products.

2.1 Conventional studies of cigarette demand

Numerous investigators have estimated the effects of price on cigarette

demand using conventional models of demand that do not account for the

addictive nature of cigarette smoking. Their studies have used diverse

econometric and other statistical methods on data from numerous countries.

Many used aggregate time-series data for a single geographical unit, while

others employed pooled cross-sectional time series data; still others used

individual level data taken from surveys. The price elasticity estimates for

overall cigarette demand from recent studies fall within the relatively wide

range from -0.14 to -1.23, but most fall in the narrower range from -0.3 to -


4 The importance of these variables should not be downplayed, however. In

many instances, these and other variables, such as parental and peer smoking

behavior and societal norms, are as important or more important than the

variables which economists have studied (U.S. Department of Health and Human

Services, 1994). Variations in these and other variables help to explain why

large variations in prices across countries are often not associated with

comparably large variations in smoking prevalence. Economists' interests

focus on the marginal impact of price, advertising, and other economic

variables on the demand for cigarettes.

Economics of Smoking - p. 10

2.1.1 Analysis of aggregate data

Many recent studies use aggregate data and appropriate econometric

methods to examine the effects of price on cigarette demand, controlling for

income, tobacco control policies, and a variety of socioeconomic and

demographic factors. The exceptions ( Baltagi and Goel, 1987; Peterson et al.,

1992) compared changes in cigarette consumption in states that had raised

cigarette taxes to consumption in states where taxes had not changed. The

estimated price elasticities from these quasi-experimental studies, in the

range from -0.17 to -0.56, are consistent with those obtained from the

econometric studies.

Although there are numerous studies of the price-de mand relationship in

industrialized nations, until recently there were almost no estimates for

developing countries. Warner (1990) argued that price responsiveness in less

developed countries is likely to be greater than in more affluent countries,

given the relatively low incomes and relatively low levels of cigarette

consumption by smokers in the poorer countries. Findings from studies using

data from Papua New Guinea (Chapman and Richardson, 1990), China (Mao, 1996;

Xu, Hu and Keeler, 1998) South Africa (van der Merwe, 1998a), Zimbabwe

(Maranvanyika, 1998), and Taiwan (Hsieh and Hu, 1997) are consistent with this


Several difficulties are encountered in studies using time-series data.

Particularly troubling are the high correlations among many of the key

independent variables and price. Consequently, estimates of the impact of

price and other factors on demand can be sensitive to the inclusion and

exclusion of other variables. Including highly correlated variables can

result in multicollinearity and unstable estimates for the parameters of

interest. Excluding potentially important variables, however, can produce

biased estimates of the impact of price on demand. Recent studies using

state-of-the-art econometric methods have addressed many of these difficulties

(Seldon and Boyd, 1991; Simonich, 1991; Flewelling et al., 1992; Sung et al.,

Economics of Smoking - p. 11

1994; Barnett et al, 1995; Keeler et al., 1996). Nearly all of the estimates

from these studies have produced estimates for the price elasticity of demand

in a relatively narrow range, centered on -0.4.

Other problems are encountered when using pooled cross-sectional timeseries

data. The measure of cigarette smoking employed in these studies is

typically annual state-level tax-paid cigarette sales. Interstate differences

in cigarette prices, resulting from wide variation in state cigarette taxes

(Tobacco Institute, 1998), can lead to casual and organized smuggling of

cigarettes from low-tax to high-tax states, however (Advisory Commission on

Intergovernmental Relations (ACIR), 1977, 1985). As such, tax-paid sales data

are likely to overstate cigarette consumption in states with low cigarette

taxes and underestimate it in high tax states. 5 Failing to account for this

will produce upward-biased estimates of the impact of price on cigarette

demand. Many of the more recent studies employing pooled time-series crosssectional

state data have controlled for the potential for smuggling (ACIR,

1977, 1985; Baltagi and Levin, 1986; Chaloupka and Saffer, 1992; Keeler et

al., 1996). These studies have also produced estimates of the price

elasticity of cigarette demand generally falling in a relatively narrow range

centered on -0.4.

The fact that cigarette prices, sales, and consumption are

simultaneously determined creates an additional complication in the analysis

of cigarette demand and supply. Failing to account for this simultaneity

would lead to biased estimates of the price elasticity of demand. Again, many

of the recent studies employing aggregate time-series data for a single

country or other geographical unit, as well as many of those using pooled

cross-sectional time-series data, have avoided this problem by theoretically

and empirically modeling cigarette demand and supply (Bishop and Yoo, 1985;

5 The same problem exists in time-series studies using aggregate countrylevel

data for countries with relatively high taxes and prices compared to

neighboring countries. See Joossens (1998) for a discussion of factors other

than price that influence smuggling across country borders.

Economics of Smoking - p. 12

Porter, 1986; Showalter, 1991; Sung et al., 1994; Barnett et al., 1995;

Tremblay and Tremblay, 1995; and Keeler et al., 1996). Other studies have

taken advantage of natural experiments, most notably 25-cent increases in the

California and Massachusetts cigarette excise taxes, to look at the impact of

price on demand (Keeler et al., 1993; Hu et al., 1994, 1995b; Sung et al.,

1994; Harris et al., 1996). After accounting for the potential simultaneity

or taking advantage of natural experiments, most of these studies produce

estimates of the price elasticity of demand that fall into the same narrow

range found in other studies.

Finally, studies employing aggregate data are generally limited to

examining the impact of cigarette prices and other factors on aggregate or per

capita measures of cigarette consumption. Consequently, these studies are

typically unable to evaluate the differential impact of prices on smoking by

various population subgroups of particular interest, especially youth and

young adults. Nor can they differentiate between the impact of price on

smoking prevalence and quantity, or smoking initiation and cessation.

A few recent analyses have attempted to address these limitations. For

example, Harris (1994) used annual time-series data on U.S. smoking prevalence

taken from the National Health Interview Surveys, coupled with aggregate

measures of cigarette consumption, to estimate the effects of price on smoking

prevalence and average cigarette consumption by smokers for the period from

1964 through 1993. His estimate of the unconditional price elasticity of

demand fell into the same narrow range generally found in other studies. He

estimated that approximately half of the impact of price was on smoking

prevalence, with the price elasticity of smoking participation being -0.238,

while the unconditional price elasticity of demand was -0.47. Townsend et al.

(1994) looked at the differential effects of price on cigarette smoking for

various population subgroups defined by age, gender, and socioeconomic status,

using data aggregated from the 1972 through 1990 British General Household

Surveys. They concluded that women were more responsive to price than men,

Economics of Smoking - p. 13

that both men and women in lower socioeconomic groups were more sensitive to

price than those that were better off, and that youth (16-19 years) and young

adults (20-24 years) were less responsive to price than adults. 6

2.1.2 Analysis of individual level data

A relatively small but growing number of cigarette demand studies have

used data on individuals taken from large-scale surveys. In general, their

estimated price elasticities of demand are comparable to those estimated using

aggregate data. The use of individual-level data helps avoid some of the

problems inherent in using aggregate data. For example, because an

individual's smoking decisions are too small to affect the market price of

cigarettes, potential simultaneity biases are less likely. Similarly,

individual-level income data and measures of socio-demographic determinants of

demand are less correlated with price and policy variables than comparable

aggregate measures.

Other problems persist but can be addressed somewhat more easily using

individual-level data. For example, failing to account for interstate

differences in cigarette prices will again produce a biased estimate of the

price elasticity of demand (biased towards 0 in this case). Thus, given

information on where an individual resides, studies using individual-level

data have employed a variety of approaches to control for potential crossborder

shopping in response to interstate price differentials. Some have

limited their samples to individuals who do not live near lower-price

localities (Lewit and Coate, 1982; Wasserman et al., 1991; Chaloupka and

Grossman, 1996; Chaloupka and Wechsler, 1997). Others have included a measure

of the price differential (Lewit et al., 1981; Chaloupka and Pacula, 1998a,

1998b). Still others have used a weighted average price based on the price in

6 As we discuss below, other studies have derived the opposite conclusions

concerning the relative price responsiveness by gender (e.g., Lewit and Coate,

1982; Mullahy, 1985; Chaloupka, 1990) and different age groups (e.g. Lewit, et

al., 1981; Chaloupka and Grossman, 1996).

Economics of Smoking - p. 14

the own-locality and other nearby localities (Chaloupka, 1991).

As with the state tax-paid sales data, self-reported data on cigarette

smoking yield inaccurate measures of true consumption, given potential

reporting biases. Based on a comparison of self-reported consumption with

aggregate sales data, Warner (1978) demonstrated that survey-based selfreported

consumption significantly and substantially understated actual sales.

Studies using individual-level survey data have implicitly treated

underreporting as proportional to true consumption across groups of interest

(e.g., age, gender, or socioeconomic groups). If the assumption is true,

estimates of the price elasticity of demand will not be systematically biased.

The assumption has yet to be demonstrated, however.

Finally, as Wasserman et al. (1991) observed, studies using individuallevel

data may be subject to a substantial ecological bias in that omitted

variables affecting tobacco use may be correlated with the included

determinants of demand. Failing to account for this can produce biased

estimates for the included variables. For example, unobserved sentiment

against smoking may affect both cigarette sales and the strength of tobacco

control policies (including taxes and, consequently, prices). Ohsfeldt et al.

(1998) considered this possibility in their analysis of cigarette smoking and

other tobacco use that employed data from the 1992/93 Current Population

Survey Tobacco Use Supplements. Surprisingly, after modeling cigarette taxes

and other tobacco control policies as a function of cigarette smoking, various

other indicators of sentiment against smoking, and other factors, they found

that taxes have a larger impact on demand.

Using individual-level data allows researchers to examine issues that

generally cannot be addressed with aggregate data. For example, most studies

using individual-level data separately consider the effects of price on the

probability of smoking and on average cigarette consumption by smokers. In

addition, several consider the differential effects of price on demand for

various population subgroups (defined by age or gender, for example).

Economics of Smoking - p. 15

Finally, some have taken advantage of retrospective or longitudinal data to

examine the effects of prices and other factors on smoking initiation and

cessation decisions.

The earliest of the cigarette demand studies employing individual-level

data were conducted by Lewit and his colleagues (Lewit et al., 1981; Lewit and

Coate, 1982). Lewit and Coate used data from the 1976 National Health

Interview Survey to examine the effects of price on cigarette smoking,

estimating an overall price elasticity of demand of -0.42 and an elasticity of

smoking participation of -0.26. In addition, they found an inverse

relationship between (the absolute value of) price elasticity and age,

estimating a total price elasticity of demand for 20 through 25 year-olds more

than double that of persons 26 and older. The researchers found that most of

the effect of price for young adults was on the decision to smoke

(participation elasticity of -0.74 and conditional demand elasticity of -

0.20), but was about evenly split for those over 35 years of age

(participation and conditional demand elasticities of -0.15). Finally, they

also looked at differences in price responsiveness by gender, concluding that

men, particularly young men, were very responsive to price, while women were

generally insensitive to price.

Lewit et al. (1981) and Grossman et al. (1983) confirmed the Lewit and

Coate (1982) conclusion concerning the inverse relationship between price

elasticity of cigarette demand and age. Using data from Cycle III of the

Health Examination Survey, Lewit et al. estimated that the price elasticity of

smoking participation for 12-17 year-olds was -1.20, while the conditional

demand elasticity was -0.25. Their estimated total price elasticity of youth

cigarette demand of -1.44 was more than three times Lewit and Coate's (1982)

estimate for adults. These conclusions were generally supported by Grossman

et al.'s (1983) analysis of data from the National Household Surveys on Drug

Use conducted during the 1970s.

Lewit et al. (1981) offered two reasons why youth should be more price

Economics of Smoking - p. 16

sensitive than adults, at least in the short run. First, given the addictive

nature of smoking, long-term adult smokers are likely to adjust less quickly

to changes in price than youth who have been smoking for a relatively short

time, if at all. In addition, peer behavior is likely to be much more

influential for youth, multiplying the effects of price on youth smoking.

That is, an increase in cigarette price directly reduces youth smoking and

then again indirectly reduces it through its impact on peer smoking. Grossman

and Chaloupka (1997) offered two additional reasons. First, the fraction of

disposable income a young smoker spends on cigarettes is likely to exceed that

spent by an adult smoker. Second, compared to adults, youth are more likely

to be present-oriented. In the context of an economic model of addictive

behavior (discussed below), Becker et al. (1991) predicted that changes in

money price will have a greater impact on individuals with higher discount

rates since they give less weight to the future consequences of addictive


The conclusion that youth cigarette demand is more price elastic than

adult demand was widely accepted until an influential 1991 Rand study by

Wasserman and colleagues (1991). These researchers evaluated adults'

cigarette demand using data from several of the National Health Interview

Surveys from the 1970s and 1980s and youth demand with data from the Second

National Health and Nutrition Examination Survey of the late-1970s. Using a

generalized linear model, the authors concluded that adult demand in the

earlier years of their data was relatively unresponsive to price, but that

demand had become more price elastic over time. Based on the trends in price

elasticity, they predicted an overall price elasticity of adult cigarette

demand of -0.283 for 1988. Estimates from a two-part model of adult cigarette

demand implied that the effects of price on the decision to smoke were almost

double the impact of price on conditional demand. However, the authors did

not find a statistically significant impact of price on youth smoking. They

attributed their relatively low estimates of price elasticity, particularly

Economics of Smoking - p. 17

those for youth, to the inclusion in their models of an index of restrictions

on smoking. These restrictions, which they note are positively correlated

with price, had not been included in most previous studies of cigarette

demand. Indeed, they obtained very similar estimates to Lewit and Coate

(1982) when leaving the restriction index out of models estimated using the

1976 survey data.

Several more recent studies of youth and young adult smoking have

supported the earlier conclusions reached by Lewit and his colleagues (Lewit

et al., 1981; Lewit and Coate, 1982; Grossman et al., 1983) that the price

sensitivity of cigarette demand is inversely related to age. Chaloupka and

Grossman (1996) examined the impact of price, numerous tobacco control

policies (including smoking restrictions and limits on youth access to

tobacco), and a variety of other socioeconomic and demographic factors on

youth smoking, using data from the 1992, 1993, and 1994 Monitoring the Future

Surveys of eighth, tenth, and twelfth grade students. They estimated a total

price elasticity of youth cigarette demand of -1.31, strikingly similar to the

estimates obtained by Lewit et al. (1981) 15 years earlier. In contrast to

Lewit and his colleagues, however, Chaloupka and Grossman found that the

effects of price on smoking participation and conditional demand were similar

(-0.68 for smoking participation and -0.64 for conditional demand). Chaloupka

and Pacula (1998b) used the same data to look at the differential response by

gender and race, concluding that young men and young blacks are more

responsive to price than young women and young whites.

Chaloupka and Wechsler (1997) reached similar conclusions using data on

young adult smoking taken from the 1993 College Alcohol Survey. Also

controlling for numerous other determinants of cigarette demand, including a

variety of restrictions on smoking, they estimated a price elasticity of

smoking participation of -0.53 and an unconditional price elasticity of demand

of -1.11 for college students. Noting that their sample was not a random

sample of all young adults, Chaloupka and Wechsler suggested that the price

Economics of Smoking - p. 18

elasticity of cigarette demand by young adults may be even higher, given the

evidence that cigarette demand is relatively less elastic for more educated or

higher-income individuals (Townsend, 1987; Chaloupka, 1991; Townsend et al.,

1994; Farrelly et al., 1998).

Farrelly and his colleagues (1998) found similar evidence for young

adults and adults, based on 13 waves of the National Health Interview Survey

conducted between 1976 and 1992. They estimated that demand was more than

twice as elastic for their sample of young adults, ages 18 to 24 years (total

elasticity of -0.58), as for their full sample (total elasticity of -0.25).

Similarly, they estimated that blacks were about twice as responsive as whites

to cigarette prices, while Hispanics were even more price sensitive. In

addition, they found that men were more price sensitive than women. Finally,

they estimated that individuals with family incomes below the sample median

were about 70 percent more responsive to price than those with higher family


Additional support for the inverse relationship between price

sensitivity and age is provided by recent studies by Lewit and his colleagues

(1997), Evans and Huang (1998), and Tauras and Chaloupka (1998). Lewit and

his colleagues used data for ninth grade students in 1990 and 1992 collected

in the 22 North American communities involved in the National Cancer

Institute's Community Intervention Trial for Smoking Cessation (COMMIT). They

found that both youth smoking participation and intentions to smoke among

young non-smokers were inversely related to price, with estimated price

elasticities of -0.87 and -0.95, respectively. Evans and Huang used state

level aggregated data on smoking prevalence constructed from the 1977 through

1992 Monitoring the Future surveys to estimate a price elasticity of youth

smoking participation of -0.20. Unlike other studies on youth smoking that

largely rely on the cross-sectional variation in state cigarette taxes and

prices, Evans and Huang took advantage of the long time period covered by

their data and used the time series variation in state cigarette taxes to

Economics of Smoking - p. 19

identify the impact of cigarette taxes on smoking participation. While their

estimated elasticity for the 1977 through 1992 period is relatively low, Evans

and Huang concluded that youth smoking has become more price sensitive over

time, estimating an elasticity of -0.50 for youth smoking participation in the

period from 1985 through 1992. Most recently, Tauras and Chaloupka (1998)

used the longitudinal data from the Monitoring the Future surveys of high

school seniors conducted from 1976 through 1993 to estimate the price

elasticity of smoking for young adults; respondents in their sample ranged in

age from 17 to 35 years. In models controlling for unobserved state and

individual factors affecting demand, they estimated an overall price

elasticity of demand centered on -0.79.

In general, researchers examining the effects of price on smoking

participation using individual-level data from cross-sectional surveys have

assumed that much of the price effect estimated for youth reflects the impact

of price on smoking initiation, while the estimate for adults is largely

capturing the effects of price on smoking cessation. A few recent studies

have attempted to directly examine the impact of cigarette prices on smoking

initiation. With retrospective data from the smoking supplements to the 1978

and 1979 National Health Interview Surveys, Douglas and Hariharan (1994)

studied the ages at which survey respondents reported that they began smoking.

Based on current state of residence, they matched data on cigarette prices to

the survey data to estimate the impact of price on smoking initiation. They

estimated a hazard model in which "failure" was defined as a never smoker

taking up smoking and used a relatively general variation on standard duration

methods: the split population duration model developed by Schmidt and Witte

(1989). This model allows for a large part of their sample to never begin

smoking. Finally, Douglas and Hariharan's theoretical and empirical framework

was based on the Becker and Murphy (1988) rational addiction model (described

below). As anticipated, Douglas and Hariharan found that a number of

socioeconomic and demographic factors had a significant effect on smoking

Economics of Smoking - p. 20

initiation. However, their estimates for cigarette prices were insignificant.

Given the errors-in-variables problem associated with both the retrospective

data on smoking initiation and the cigarette price data, they noted that price

effects will be biased towards zero. Nevertheless, they found no evidence

that higher cigarette prices reduced smoking initiation.

Douglas (1998) extended this work by estimating a time-varying covariate

model that allows the hazard of smoking initiation to respond dynamically to

changes in prices and other factors. In addition to initiation, Douglas also

estimated the hazard of smoking cessation in a similar empirical framework, as

well as estimating the impact of smoking regulations and information on

initiation and cessation (these findings are discussed later). Using data

from the cancer risk factor supplement to the 1987 National Health Interview

Survey, Douglas again concluded that cigarette price has little impact on

smoking initiation. As with the earlier analysis, however, there are likely

to be errors-in-variables problems that could account for this finding.

DeCicca et al. (1998a) employed data from the National Education

Longitudinal Survey of 1988 to examine the impact of price on initiation of

daily smoking. This data set contains data on youth smoking at several points

in time (eighth, tenth, and twelfth grades). Treating the three waves as

independent cross-sections, they obtained estimates of the price elasticity of

youth smoking participation comparable to other recent estimates. In an

effort to examine the impact of price on smoking initiation, they attempted to

exploit the longitudinal aspect of their data by looking at the probability of

smoking in twelfth grade for a sample that excluded those who were smokers in

eighth grade. Their estimates for the effect of cigarette taxes on the

probability of starting to smoke between the eighth and twelfth grade are not

statistically significantly different from zero, supporting the findings of

Douglas and Hariharan (1994) and Douglas (1998) that raised doubts about the

hypothesis that higher cigarette prices lead to significant reductions in

youth smoking. DeCicca et al. attributed the inconsistency in their two sets

Economics of Smoking - p. 21

of results to the possibility that cigarette tax rates are a proxy for

unobserved sentiment against cigarette smoking. If true, then estimates based

on cross-sectional studies are likely to significantly overstate the impact of

price on smoking.

Dee and Evans (1998) reexamined the longitudinal data used by DeCicca et

al., arguing that their finding that price has no impact on smoking initiation

was largely the result of the way in which their sample was constructed. In

particular, rather than following DeCicca et al. in deleting the large number

of observations with missing values for key independent variables (including

income, parental education, and number of siblings), Dee and Evans included

these along with dummy variables indicating observations for which the data

are missing. In addition, they included a variety of binary indicators for

categorically collected data, rather than constructing "continuous" measures

from these data as did DeCicca et al. (e.g. parental and family attributes).

After making these changes but otherwise following the same basic approach,

Dee and Evans estimated a negative and significant impact of cigarette taxes

on smoking initiation. Their estimated price elasticity of smoking onset is -

0.63, consistent with several other recent studies of youth smoking employing

cross-sectional data.

In response to Dee and Evans (1998), DeCicca and his colleagues (1998b)

conducted a reanalysis of the NELS data that used an alternative approach for

dealing with the missing data problem. Where possible, they used information

from the longitudinal sample to fill in missing values; when this could not be

done, they used a conditional mean imputation approach. Their reanalysis

produced somewhat more significant estimates for the effect of cigarette taxes

on the onset of daily smoking, with implied price elasticities from

alternative specifications ranging from -0.025 to -0.505; somewhat smaller,

less significant estimates were obtained from models using price rather than

tax. In addition, their estimates for samples based on race/ethnicity implied

that higher cigarette taxes significantly reduced smoking onset among

Economics of Smoking - p. 22

Hispanics, but had little impact on whites and blacks.

Clearly, the use of longitudinal data to examine the impact of cigarette

tax and price changes on smoking initiation and cessation is an important

advance. The findings from studies using relatively longer panels that

control for unobserved state and/or individual factors affecting demand (i.e.

Evans and Huang, 1998; Tauras and Chaloupka, 1998) are consistent with the

findings that price sensitivity is inversely related to age, as found in

several earlier studies based on cross-sectional data. The inconsistent

findings from a few recent studies ( DeCicca et al., 1998a, 1998b; Dee and

Evans, 1998) directly addressing the effects of price on smoking initiation

with a relatively short panel should be viewed with caution.

Hu et al. (1995a) introduced an innovation in cigarette demand

estimation, using data from California's Behavioral Risk Factor Surveys for

1985 through 1991 to examine the possible effects on adult smoking of the

interdependence of cigarette smoking with other risk factors, including

alcohol use and obesity. Estimates of the smoking participation elasticity

from models that included other behavioral risk factors were significantly

lower than when these factors were ignored, while conditional demand

elasticities were generally unaffected. Using two-part methods, Hu et al.

estimated an overall price elasticity of -0.46 from the models that included

other risks, with the effects of price about equally divided between smoking

participation and conditional demand. The authors noted, however, that their

estimate of the price elasticity might be relatively high given that they did

not control for other tobacco control efforts.

Evans and Farrelly (forthcoming) recently examined a phenomenon not

previously studied by economists. Using data from the 1979 Smoking Supplement

and the 1987 Cancer Control Supplement to the National Health Interview

Surveys, the authors investigated the compensating behavior by smokers in

response to tax and price changes. The supplements contain unique information

on smokers' choices of types of cigarettes, which Evans and Farrelly combined

Economics of Smoking - p. 23

with data from the Federal Trade Commission on the tar and nicotine content of

cigarette brands to construct a variety of measures of daily smoking intensity

(including cigarette consumption, total length of cigarettes consumed, tar

intake, and nicotine intake). They also constructed comparable aggregate

measures for 1964-1993 from the data used by Harris (1994) on aggregate

smoking prevalence and cigarette consumption. They found consistent evidence

that, although smokers reduced daily cigarette consumption in response to

higher taxes, they also compensated in several ways. In particular, smokers

in high-tax states consumed longer cigarettes and those that are higher in tar

and nicotine, with young adults smokers also most likely to engage in this

compensating behavior. As a result, they argued that the perceived health

benefits associated with higher cigarette taxes are likely to be somewhat

overstated. Given this compensating behavior, Evans and Farrelly suggest that

if cigarette taxes are to be used to reduce the health consequences of

smoking, then taxes based on tar and nicotine content would be appropriate, an

idea first suggested by Harris (1980).

2.2 Addiction models and cigarette demand

The first discussion by an economist of the effects of addiction on

demand can be found in Marshall's (1920) Principles of Economics, where he

observed that

Whether a commodity conforms to the law of diminishing or increasing

return, the increase in consumption arising from a fall in price is

gradual; and, further, habits which have once grown up around the use of

a commodity while its price is low are not so quickly abandoned when its

price rises again. (Appendix H, section 3, page 807)

As Phlips (1983) noted, Marshall's statement clearly introduced the three

basic dimensions of addiction (U.S. Department of Health and Human Services,

1988) of gradual adaptation (tolerance), irreversibility (withdrawal), and

positive effects of habits (reinforcement) that are used in many of the more

recent formal models of addictive behavior. Until recently, however,

economists have either ignored the addictive nature of goods such as

Economics of Smoking - p. 24

cigarettes when estimating demand or have assumed that behaviors such as

smoking were irrational and could not be analyzed in the rational, constrained

utility maximizing framework of economics.

Many of the most recent studies of cigarette demand explicitly address

the addictive nature of cigarette smoking. Economic models of addiction can

be divided into three basic groups: imperfectly rational models of addictive

behavior, models of myopic addictive behavior, and models of rational

addictive behavior.

2.2.1 Imperfectly rational addiction models

Elster (1979), McKenzie (1979), Winston (1980), and Schelling (1978,

1980, 1984a, 1984b) best exemplify the economic models of imperfectly rational

addictive behavior. These models generally assume stable but inconsistent

short-run and long-run preferences. This is seen, for example, in Schelling's

(1978) description of a smoker trying to "kick the habit":

Everybody behaves like two people, one who wants clean lungs and long

life and another who adores tobacco.... The two are in a continual

contest for control; the "straight" one often in command most of the

time, but the wayward one needing only to get occasional control to

spoil the other's best laid plan. (p. 290)

Thus, the farsighted personality may enroll in a smoking cessation program,

only to be undone by the shortsighted personality's relapse in a weak moment.

Winston (1980) formally modeled this behavior and described how this contest

between personalities leads to the evolution of what he called "anti-markets,"

which he defined as firms or institutions that individuals will pay to help

them stop consuming.

Strotz (1956) was the first to develop a formal model of such behavior,

describing the constrained utility maximization process as one in which an

individual chooses a future consumption path that maximizes current utility,

but later in life changes this plan "even though his original expectations of

future desires and means of consumption are verified" (p. 165). This

inconsistency between current and future preferences only arises when a nonEconomics

of Smoking - p. 25

exponential discount function is used. 7 Strotz went on to suggest that

rational persons will recognize this inconsistency and plan accordingly, by

pre-committing their future behavior or by modifying consumption plans to be

consistent with future preferences when unable to pre-commit. Pollak (1968)

went one step further, arguing that an individual may behave naively even when

using an exponential discount function. Thaler and Shefrin (1981) described

the problem similarly, referring to an individual at any point in time as both

a "farsighted planner and a myopic doer" (p. 392), with the two in continual

conflict. While these models present interesting discussions of some aspects

of addictive behavior, they have not been applied empirically to cigarette

smoking or other addictions.

2.2.2 Myopic addiction models

The naive behavior described in some of the imperfectly rational models

of addiction is the basis for many of the myopic models of addictive behavior.

As Pollak (1975) observed, behavior is naive in the sense that an individual

recognizes the dependence of current addictive consumption decisions on past

consumption, but then ignores the impact of current and past choices on future

consumption decisions when making current choices. Many of these models

treat preferences as endogenous, allowing tastes to change over time in

response to past consumption (Gorman, 1967; Pollak, 1970, 1976, 1978; von

Weizsacker, 1971; Hammond, 1976a, 1976b; El- Safty, 1976a, 1976b).

These models are similar in spirit to those in which tastes change in

response to factors other than past consumption, including advertising ( Dixit

and Norman, 1978; Galbraith, 1958, 1972) and prices ( Pollak, 1977). Others

allow past consumption to affect current consumption through an accumulated

stock of past consumption (e.g., Houthakker and Taylor, 1966, 1970). These

models are comparable to those of the demand for durable consumer goods that

7 Vuchinich and Simpson (1998) provided an interesting application of this

idea to the demand for alcoholic beverages, comparing behavior under

hyperbolic versus exponential discounting.

Economics of Smoking - p. 26

use a stock adjustment process ( e.g, Chow's (1960) model of the demand for

automobiles, and Garcia dos Santos' (1972) analysis of the demands for

household durables). As Phlips (1983) noted, however, the distinction between

models with endogenous tastes and those with stable preferences within a

household production framework is purely semantic, since the underlying

mathematics of the two are the same.

The earliest theoretical models of demand in the context of myopic

addiction can be traced to the irreversible demand models ( Haavelmo, 1944;

Duesenberry, 1949; Modigliani, 1949; Farrell, 1952). Farrell, for example,

described an irreversible demand function as one in which current demand

depends on all past price and income combinations. As a result, price and

income elasticities are constant, but may differ for increases and decreases

in price and income. Farrell tested this model empirically, using U.K. data

on the demands for tobacco and beer from 1870 through 1938, in a model that

included not only current price and income, but also price, income, and

consumption in the prior year. In general, his estimates were inconclusive,

although he did find limited evidence of habit formation for tobacco use.

The notion of asymmetric responses to price and income reappeared in

Scitovsky (1976) and was applied to cigarette demand by Young (1983) and

Pekurinen (1989), using data from the U.S. and Finland, respectively. Both

found that smoking was almost twice as responsive to price reductions as it

was to price increases, which they interpreted as evidence of addiction.

Most empirical applications of myopic models of addiction are based on

the pioneering work by Houthakker and Taylor (1966, 1970) that formally

introduced the dependence of current consumption on past consumption by

modeling current demand as a function of a "stock of habits" representing the

depreciated sum of all past consumption. Houthakker and Taylor estimated