THE EFFECTS OF TAXES AND BANS ON

PASSIVE SMOKING

Jérôme Adda
Francesca Cornaglia

THE INSTITUTE FOR FISCAL STUDIES

DEPARTMENT OF ECONOMICS, UCL
cemmap working paper CWP20/05
Passive Smoking 1

The Effect of Taxes and Bans on Passive Smoking

Jérôme Adda and Francesca Cornaglia∗

University College London and Institute for Fiscal Studies.

November 2005

Abstract

This paper evaluates the effect of excise taxes and bans on smoking in public places on the
exposure to tobacco smoke of non-smokers. We use a novel way of quantifying passive smoking:
we use data on cotinine concentration- a metabolite of nicotine- measured in a large
population of non-smokers over time. Exploiting state and time variation across US states, we
reach two important conclusions. First, excise taxes have a significant effect on passive smoking.
Second, smoking bans have on average no effects on non smokers. While bans in public
transportation or in schools decrease the exposure of non smokers, bans in recreational public
places can in fact perversely increase their exposure by displacing smokers to private places
where they contaminate non smokers, and in particular young children. Bans affect
socioeconomic groups differently: we find that smoking bans increase the exposure of poorer
individuals, while it decreases the exposure of richer individuals, leading to widening health
disparities.

∗
[email protected] and [email protected]. We are grateful to a number of seminar participants and to David
Card, Ken Chay, Christian Dustmann, Steve Machin, Costas Meghir, and Imran Rasul for helpful discussions and
comments and to William Evans for supplying us with data on US tobacco prices and taxes. Funding through the
ESRC is gratefully acknowledged.
Passive Smoking 2

1 Introduction
In the US, 15% of the population smokes regularly. Yet, detectable levels of nicotine in body
fluids can be found in 84% of non smokers of all ages.1 A large medical and epidemiological
literature has stressed the dangers of exposure to environmental tobacco smoke. 2 Passive
smoking has been linked to a number of serious illnesses such as lung cancer or heart disease in
the adult population. It causes about 35000 deaths per year from heart diseases and about 3000
lung cancer deaths (Environmental Protection Agency, 1992, American Cancer Society, 2003,
IARC, 2004). Passive smoking affects particularly the health of young children and babies,
causing asthma, bronchitis or sudden infant death syndrome. Exposure to smoke causes about
200,000 lower respiratory tract infections in young children each year, resulting in 10,000
hospitalizations (Environmental Protection Agency, 1992). Medical studies consistently find that
smokers impose a negative externality on non-smokers. As a result, governments have come
under pressure by the general public and by anti-tobacco groups to limit the exposure of non-
smokers and generally to discourage smoking. Since the mid eighties, support for smoking bans
in public places has steadily risen. The proportion of individuals supporting a total ban in
restaurants has increased from 20% in 1985 to 54% in 2005. 3 The government uses two
instruments to discourage smoking: directly by limiting or banning smoking in public places, and
indirectly by raising taxes on cigarettes.

The economic literature has focused on the effect of prices or taxes on smokers. Following the
work of Becker and Murphy (1988), most papers estimate price elasticities both in the short and
the long run. 4 The evidence in these papers suggests that prices have an effect on cigarette
consumption. However, some recent papers dispute the effect of prices. DeCicca et al (2002)
show that cigarette prices do not affect initiation at young ages. Adda and Cornaglia (2005) show
that although taxes affect the number of cigarette smoked, smokers compensate by smoking more
intensively a given cigarette. Few papers analyze the effect of bans on smoking. Among these
Evans et al. (1999) show that workplace bans decrease the prevalence of smoking in those who
work.

1
See descriptive evidence in section 3.1
2
See for instance Law et al (1997), Hackshaw (1997), He et al (1999), Otsuka (2001), Whincup et al (2004),
for adults and Strachan and Cook (1997), Gergen et al (1998), Kriz et al (2000), Lam et al (2001), Mannino et
al (2001) for children who all find that exposure to passive smoke is harmful for non-smokers health.
3
Source: Gallup poll (http://poll.gallup.com/).
4
See for instance the paper by Becker et al(1994), Chaloupka (1991), and references in Chaloupka and Warner
(2000).
Passive Smoking 3

While the literature on regulations, either through taxes or bans, on smokers is quite large, there
is hardly any evidence of the effectiveness of these measures on the population of non-smokers.5
Yet, the debate in public circles and in the media on the effectiveness of different measures has
recently intensified, and policies to ban smoking are often justified by the protection of non
smokers rather than smokers 6. There is to our knowledge no study evaluating the response of
passive smoking to changes in excise taxes, or on the growing set of regulation and clean air acts
passed in the last decade7. A main reason why there is hardly any work in the economic literature
on the exposure of non-smokers to environmental smoke is the apparent difficulty of measuring
passive smoking directly.

This paper fills this gap. We propose a way of measuring passive smoking directly in non-
smokers. We use a unique data set, which reports a direct measure of exposure to passive
smoking: cotinine concentration in body fluids. Cotinine is a by-product of nicotine, and is a
good marker of exposure to second hand smoke, which has been used routinely in the medical
and epidemiological literature.8 Using cotinine measures for analysing changes in exposure to
passive smoking has several advantages. First, one can detect even small effects in exposure to
environmental smoke; second, cotinine measures are sensitive to changes in exposure; third, it is
a more reliable and objective measure than self-reported exposure which has been used as a
measure of passive smoking. An alternative measure would be to use changes in smoking related
diseases. However, most of these diseases are not specific to smoking and they usually take
several years to develop. This makes it difficult to correctly identify the effect of state
interventions. Cotinine is therefore a straightforward and precise measure of passive smoking
especially when evaluating public policies.

Our analysis proceeds in several steps. First, using cotinine levels for a large and representative
sample of non-smokers over time, including very young children, we document the extent of

5
One exception is the effect of maternal smoking on birth weight, see for instance Rosenzweig and Schultz (1983)
and Evans and Ringel (1999).
6
See for instance ASH (2005) for a summary of the case for smoke free public places.
7
A search in EconLit for the key words “passive smoking” generates only 4 hits that are unrelated to the issue
discussed here.
8
The epidemiological literature has examined the issue of passive smoking, mostly from its health consequences.
This literature has produced a measure of passive smoking by analyzing the concentration of cotinine, a metabolite of
nicotine, in blood, saliva or urine samples. The amount of cotinine is a good marker of the exposure to environmental
smoke (Jarvis et al 1984). The epidemiological literature has also tried to characterize the socio-economic groups that
are more prone to exposure to environmental smoke (Pirkle et al, 1996; Howard et al, 1998; Siegel, 1993; Jarvis et al,
2001; Whitlock et al, 1998; Jarvis et al, 2000; Strachan and Cook, 1997).
Passive Smoking 4

passive smoking in the US. We evaluate the effect of tobacco tax increases that took place in the
US over the last decade on exposure to environmental tobacco smoke (ETS). Our analysis
exploits changes over time in regulations on smoking in public places across different states. We
find that changes in tobacco taxes have a significant effect on the exposure to environmental
smoke. A 10% increase in the state excise tax reduces the cotinine concentration in non smokers
by about 3%. The effect is particularly sizable for children who are exposed to their parents’
smoke.

Second, we analyse the impact of smoking bans. Bans on smoking in public places have on
average no effects on non smokers. However, we show that bans have different effects when
imposed in different public places. While bans in public transports, shopping malls or schools
decrease the exposure of non-smokers, bans in bars, restaurants or recreational facilities appear to
increase their exposure. The reason is that such bans displace the smoking to places where non-
smokers are more exposed, especially young children. Moreover, bans have contrasting effects
on different social and demographic groups. We find evidence that smoking regulations increases
the exposure of poorer individuals, while it decreases the exposure of individuals in higher socio-
economic position. This suggests that smoking regulations may increase health inequalities
between socio-economic groups.

The remainder of the paper is structured as follows. Section 2 presents the theoretical framework
used for analyzing the effect of passive smoke exposure, and outlines the estimation strategy.
Section 3 contains a description of our data set. In Section 4, we investigate the effect of different
state interventions on passive smoking, measured by the cotinine concentration present in non-
smokers. Finally, Section 5 concludes and discusses the implications of our results.

2 Conceptual Framework and Methodology

This section discusses our framework for analyzing the effect of tax changes and smoking
regulations on passive smoking. In particular, we define our measure of passive smoking and
describe our identification strategy.

2.1 Effects of Tax Changes and Smoking Bans

Public interventions can have both a direct and an indirect effect on non-smokers. Smoking bans
have a direct effect on non-smokers as they guarantee a smoke-free environment. Changes in
excise taxes operate indirectly as they can only have an effect through the behaviour of smokers.
Passive Smoking 5

To some extent, this indirect effect is also present in smoking bans given that they may induce
changes in the way smokers smoke. To analyse fully the effect of bans on non-smokers, it is
necessary to understand the effect on smokers and the extent to which smokers and non smokers
cohabit and interact.

Smokers
The literature has shown the negative effect of taxes (and prices) on the demand for cigarettes.
Moreover it has been pointed out that taxes operate both on the intensive and the extensive
margin. 9 However, this does not mean that regular smokers reduce smoking in a uniform way:
during the day, some cigarettes may in fact be easier to cut down. If smoking is a social activity,
a smoker may reduce the number of cigarettes consumed when alone, and not those consumed in
company. In this case, non smokers may not benefit at all from a rise in excise taxes.

Regarding smoking regulations, there is evidence that bans in the workplace decrease the
prevalence of smoking (Evans et al, 1999). However, bans may also lead to a displacement of
smoking: smokers may shift their consumption within the day, or decide to avoid spending time
in places where bans are effective. For instance, a ban in bars and restaurants may induce
smokers to spend more time at home, and therefore contaminate more other members of the
household, especially children.

Non Smokers
Contamination from ETS is often termed passive smoking. Passive may, however, be in this
context a misleading terminology. Non-smokers may in fact choose whom to socialize with and
where, and the introduction of a smoking ban may change their choice of workplace,
transportation mode or social venues. In particular, adults who live with a smoking partner
represent a selected group of individuals who may care less about exposure to smoke. They may
therefore spend relatively more time in public places where smoking is permitted. On the one
hand, they may benefit more from tighter regulations, but on the other hand, they may be more
likely to switch to other places where smoking is permitted, accompanying their smoking partner.
The net effect depends on the relative importance of these two factors.

For some non-smokers, avoiding smokers may not be possible. Young children may have little
choice but to stay with their parents or carer.

9
See section 1 for references.
Passive Smoking 6

The effect of state interventions depends on the interaction between smokers and non-smokers. It
is therefore not straightforward to infer the effect of government interventions on non-smokers by
looking at the effect of these interventions on smokers (i.e. measuring the change in prevalence,
or the change in the number of cigarettes smoked). Passive smoking should be measured directly
in non-smokers. Next section describes the measure of smoking we use.

2.2 Cotinine as a Proxy for Smoking Intake

In order to analyze the effect of state interventions on non-smoker we need a measure of the
amount of tobacco smoke inhaled by non smokers. We use as a proxy the cotinine concentration
in body fluids. Cotinine is a metabolite of nicotine. While nicotine is unstable and is degraded
within a few hours of absorption, cotinine has a half-life in the body of about 20 hours and is,
therefore, a biological marker often used as an indicator of passive smoking.10 It can be measured
in, among other things, saliva or serum.
The use of cotinine has several advantages. First, cotinine is related to the exposure to cigarette
smoke. Figure 1 plots the relationship between the total number of cigarettes smoked in the
household and the cotinine level observed in the body fluids of non smokers sharing the house
with smokers.

[Figure 1]

The relationship between the number of cigarettes smoked in the household and the cotinine level
in non smokers living with smokers is upward sloping (Figure 1). Second, cotinine – and nicotine
from which it is derived- is a good proxy for the intake of health threatening substances in
cigarettes. The nicotine yield of a cigarette is, in fact, highly correlated with the level of tar and
carbon monoxide, which causes cancer and asphyxiation. 11 , 12 Cotinine is, therefore, a good
indicator of health hazards due to exposure to passive smoking. Third, cotinine levels reveal
rapidly variations in exposure due to changes in policy, which is not the case with other markers

10
The elimination of cotinine is slow enough to allow comparing measurements done in the morning or in the
afternoon.
11
Based on our data set (the National Health and Nutrition Examination Survey), which report for some years the
nicotine, tar and carbon monoxide yield of each cigarette, the correlations between nicotine and both tar and carbon
monoxide are high, 0.96 and 0.85.
12
The main health impacts of exposure to environmental tobacco smoke (ETS) are lung cancer (more than 50
epidemiological studies have examined the relationship between passive smoking and lung cancer; for a review see
NHS Scotland, 2005), coronary heart diseases, respiratory disorders, and ETS in pregnancy can lead to low birth
weight and poor gestational growth.
Passive Smoking 7

such as tobacco related diseases which take time to develop. Finally, there is minimal
measurement error, compared with self-declared exposure to cigarettes.

The novelty of our analysis is to use cotinine concentration in non smokers to evaluate the effect
of public intervention aimed at reducing tobacco exposure.

2.3 Methodology
In this section we outline the empirical methodology used for the analysis. We consider the
following econometric model of exposure to environmental smoke for a non smoker indexed by i,
in state s and in period t:

Cot ist = α 0 + α1 X it + α 2 log taxst + α 3 Rst + δ s + λt + uist (1)

where Cot ist is the cotinine concentration (expressed in ng/ml), X it is a vector of individual

characteristics that affect exposure such as age, sex, occupation or race; tax st is the state excise

tax on tobacco (adjusted for inflation) in a given state and period; Rst is a measure of restrictions

on smoking in the state at the period of interest; δ s is a set of state of residence dummies, while

λt is a set of year dummies. The identification of the effect of taxes and regulation comes from
variation across states and time, and not from cross-sectional differences in the level of state
regulations or taxes, which are taken account for by state dummies. Our identification relies on
the exogeneity of the timing of changes in taxes and regulation within states. The timing depends
in part on the electoral cycle which can be thought of exogenous to exposure to passive smoking.
The coefficients of interest are the effect of taxes and the effect of restrictions on cotinine
measures. We relate exposure to excise taxes as this is the relevant policy variable from a public
health point of view.

As discussed above, cotinine is constantly eliminated by the body, although at a slow rate. Some
of the variation in cotinine levels depends on the timing of the examination during the day. To
the extent that the timing of the examination is uncorrelated with changes in taxes and level of
regulation in the state, we do not expect a bias in the coefficient of interest. The same argument
can be made for biological diversity in the speed at which cotinine is cleared from the body.
The model is estimated by OLS, and standard errors are adjusted for heteroskedasticity and
clustered at state and year levels. This correction accounts for the presence of a common random
Passive Smoking 8

effect at the year-state level. We also check the robustness of our results to serial correlation in
the error term.

Taxes and regulation may affect non-smokers differently, according to their age and occupation.
For instance, children are probably unaffected by smoking bans in the workplace because they do
not spend any time there. The same may be true for individuals out of the labor force. We
therefore investigate the specific effect of state intervention on different socio-economic groups.
In addition, we also split the analysis according to the smoking status of other members of the
household. 13

After having considered a global measure of restrictions on smoking ( Rst ), we separate cigarette

smoking regulations according to the place where these regulations are enforced. Among the
different places where smoking bans may be enforced we consider: public transport, shopping
malls, workplaces, schools, and recreational places. We consider the following econometric
model for a non smoker indexed by i, in state s and in period t:

Cot ist = α 0 + α1 X it + α 2 log taxst + α 3GOst + α 4 PTst + α 5 SM st + α 6WPst + α 7 S st + δ s + λt + uist (2)

where X it is a vector of individual characteristics that affect exposure such as age, sex,

occupation or race; tax st is the tax of tobacco (adjusted for inflation) in a given state and period;

GOst is a measure of restrictions on smoking in recreational places (“going out”);

PTst restrictions in public transport; SM st restrictions in shopping malls; WPst restrictions in the

work place; S st restriction at school; δ s is a set of state of residence dummies, while λt is a set of

year dummies. As in (1) the identification of the effect of taxes and regulation comes from
variation across states and time, and the coefficients of interest are the effect of taxes and the
effect of restrictions in the different public places on cotinine measures ( Cotist ).

13
To the extent that the smoking status of other members of the household depends on taxes and regulations, we
expect the OLS estimates of the effect of these variables to be biased. If higher taxes and tougher regulation
encourage proportionally more light smokers to quit, the sample of non smokers in smoking household will shift
towards a population more exposed to passive smoking. This would bias upward the effect of taxes or regulations. As
a robustness check, we have also done the analysis by re-weighting the sample so that each year becomes comparable,
in terms of observables, to the first year of our sample. This methodology is developed in DiNardo et al (1996) to
study changes in wage inequality and relies on a change in composition which can be corrected by matching on
observables. In this way, we are comparing groups of individuals who are similar in a number of observable
characteristics. This will be further discussed in section 4.2.
Passive Smoking 9

3 The Data and Descriptive Statistics

3.1 Exposure to Passive Smoking

We use data from the National Health and Nutrition Examination Survey (NHANES III and
NHANES 1999-2002). NHANES is a nationwide representative sample of the US civilian
population. It provides information, from 1988 to 1994 and from 1999 to 2002, for 51835
individuals, aged zero and above. The data set reports information on the age, sex, race, health,
education and occupation of the individual, as well as information at the household level such as
family composition, income or geographical location. In addition, the cotinine concentration in
both smokers and non smokers (aged four and above), and the number of cigarettes smoked in
the household are reported. This last information allows distinguishing between non smokers that
are exposed to passive smoke at home, from non smokers that live in smoke-free households.

From the available sample we select non-smoking individuals. We drop all individuals who
report them-selves as smoker or report consuming cigarettes, cigars, pipe, snuff or chewing
tobacco. We also drop all individuals who have a cotinine level in excess of 10 ng/ml. This rule
is often use in epidemiological studies to distinguish smokers from non smokers.14 It represents
about 5% of the declared non smokers. In total, we observe 29667 non-smokers with a valid
measure of cotinine concentration (Table 1).15

[Table 1]

Table 1 provides a summary statistic of the data set. Column 1 refers to the whole sample,
columns 2 and 3 provide descriptive statistics for non-smokers living in household where the
other members either smoke or not. The average cotinine concentration is equal to 0.44ng/ml.
84% of the sample has a cotinine concentration higher than the detectable threshold of
0.035ng/ml, while 14% have a value higher than 1ng/ml. The amount of cotinine in non smokers
living in a non smoking household is more than five times lower than the amount of cotinine
present in individuals living with smokers (0.26 n/ml in non-smokers living in non-smoking
households compared to a level of 1.47 n/ml in individuals living with smokers). Individuals
living in households with smokers have almost all detectable levels of cotinine, and are much

14
See Jarvis et al, 1987. This threshold also constitutes the upper level of exposure of younger children (aged 6 or
less) for whom we can presumably assume that they are genuinely non smokers.
15
All valid cotinine measures below the detection threshold (0.035 ng/ml), were set to the threshold value.
Passive Smoking 10

more likely than non smokers living in non smoking households to have a concentration of
cotinine above 1ng/ml.

3.2 Health Effects of Passive Smoking

We briefly present some evidence of the effect of passive smoking on health for children and
adults. The purpose of this analysis is not to reproduce results established in the medical
literature, but to provide some light on the risks of passive smoking, evaluated within our sample,
with measures of exposure that are similar to the one we use to evaluate the effect of state
interventions.

We exploit the information on health outcomes contained in the NHANES III. We consider four
categories of passive exposure based on cotinine concentrations: 0 to 0.1 ng/ml, 0.1 to 0.2 ng/ml,
0.2 to 1.0 ng/ml and 1 to 10 ng/ml. We select different health outcomes for children and adults.
For adults we consider strokes and chronic bronchitis. As the incidence of strokes in children is
very low, we consider symptoms of respiratory diseases such as asthma and chest wheezing. We
run logistic regressions and control also for age, sex and race. The results are presented in Table
3 in the form of odds-ratios. The omitted category is the lowest exposure group.
[Table 3]

The first two columns displays the results for children. Exposures resulting in cotinine
concentrations lower than 1 ng/ml do not lead to higher risks of developing asthma or chest
wheezing. For the highest exposure group, the odds of having asthma or chest wheezing are
respectively 1.62 and 1.41, both statistically significantly different from one.
For adults, we find a similar pattern, low exposures up to 0.2 ng/ml do not lead to higher risks of
developing strokes or chronic bronchitis. For the highest exposure group, the odds are 1.70 for
strokes and 1.56 for chronic bronchitis.

The effect of passive smoking appears to be non-linear, with no or limited effect for light
exposure, and a significant effect when the exposure results in a cotinine concentration higher
than 1ng/ml. We shall return to these results below when we discuss the effect of state
interventions.

3.3 Excise Taxes and Smoking Restrictions to Tobacco Exposure

We merge information on state level excise taxes and smoking regulations to the NHANES
datasets. The data on excise taxes are from the Tax Burden on Tobacco, published by The
Passive Smoking 11

Tobacco Institute until 1998 and updated by Orzechowski and Walker (2001). It reports taxes by
state and year. We deflate taxes using the consumer price index. Most of the variation is cross-
sectional, where taxes can vary by about 80%. There are however differential variations over
time across states that we exploit to identify the effect of taxes. Figure 2 plots the excise taxes
over time (1988-2002) in each of the US states. Taxes have on average increased by 2 cents per
year.

[Figure 2]

We also merge information on smoke free law in the different US states to the NHANES datasets.
Regulations on smoking bans in the US are obtained from the ImpacTeen web site, based on state
clean air acts.16 This data set reports the regulation in place by year and by state in different
locations. The data set provides information on the severity of the restrictions and on the place
where the restriction is enforced (e.g. government worksites, private work sites, public transits,
schools, and restaurants). We refer the reader to Appendix A for a more detailed description of
the dataset. We recode the severity of the restriction into four categories: zero if no restrictions;
one if smoking is restricted to designated areas; two if smoking is restricted to separate areas;
three if there is a total ban on smoking.17 Figure 3 displays the average level of restrictions on
smoking, by US states over the period 1991-2000. The restriction index is the average, by year
and state, of all levels of restriction in all places.

[Figure 3]

Over the nineties, regulations have become more stringent. Moreover, the proportion of states
with no restriction in any places falls from 50% in 1991 to 36% in 2001. Similarly, in 1991 only
27% of the states had at least a total ban on smoking in one public space, whereas the figure is
51% in 2001.
[Table 2]

Our identification strategy relies on within state variation in excise taxes and smoking regulation.
Table 2 presents descriptive statistics for taxes and regulations. In particular, the last column
presents the within-state standard deviation, which is important to interpret the magnitude of the
effects presented in the next section. While the range of our regulation variable is between zero
16
http://www.impacteen.org
17
See appendix A for a detailed description of the regulations variables.
Passive Smoking 12

(no ban) and three (full ban), a one standard deviation within a state correspond to a change of
about 0.2 to 0.3 for most of the regulations we consider. In other words, we never observe in the
data a state going from no to a complete restriction on smoking.

3.4 Trends in Passive Smoking

In the US the cotinine concentration in non-smokers has halved over the nineties, from about 0.8
ng/ml in 1988 to 0.4 ng/ml in 2002 (Figure 4). This remarkable trend may indicate that policies
regulating smoking have been successful. This decrease in passive smoking can also be observed
in non smokers at the upper end of the distribution of exposure (Figure 5). These are non smokers
more at risk of developing smoking related diseases. Over this period, the proportion of
individuals with a cotinine level in excess of 1 ng/ml has decreased from 21% to 11%.18

[Figure 4 and Figure 5]

Next, we separate non smokers who share their household with smokers, from non smokers who
live in “smoke free” households. Figure 6 plots the cotinine concentration in non-smokers living
in non smoking households from 1988 to 2000. Figure 7 shows, for the same time period, the
cotinine concentration of non smokers sharing the house with smokers.

[Figure 6 and Figure 7]

The level of cotinine has been halved in non smokers living with non smokers over the period of
analysis (1988-2000), from about 0.4 ng/ml to 0.2 ng/ml (Figure 6). However, policies have been
less successful in reducing exposure of those who live with smokers. In the period considered
(1988-2002) the concentration of cotinine in non-smokers living with smokers does not show a
similar trend (Figure 7). Despite the increasing level of severity in regulations and higher excise
taxes, this evidence suggests that tobacco exposure of non smokers living in smoking households
did not decrease. 19

18
We arbitrarily look at the cotinine level of 1 ng/ml, which corresponds to the 15% upper percentile.
19
An alternative interpretation is that of a change in composition in the pool of smokers due to the fact that light
smokers are more prone to quit (see section 2.3). We reweighted the sample by matching on a number of observable
characteristics (sex, race, age group and income group). We found no substantially different results compared to the
analysis presented above.
Passive Smoking 13

4 Empirical Results: Passive Smoking and State

Intervention
In this section, we estimate the empirical model outlined in section 2.3. We start with the whole
sample and then break down the results by age groups, family smoking status and by type of
smoking ban.

4.1 Whole Sample of non smokers

We first analyse the impact of taxes and bans on passive smoking in the whole sample of non
smokers. The results are presented in Table 4.20 We consider first the effect of taxes on cotinine
concentration in non smokers (columns 1 and 2), then of bans (columns 3 and 4), and finally of
both taxes and bans together (column 5).

[Table 4]

Column (1) of Table 4 displays the effect of (log) excise taxes on passive smoking without
controlling neither for state of residence nor for year of survey. The effect is identified through
variations through time and state differences. Raising taxes by 100% reduces cotinine levels by
about 0.1ng/ml. Note that the average concentration of cotinine is equal to 0.44ng/ml. The
doubling of the excise tax would therefore lead to a reduction in exposure to smoke of about 25%.
From Table 2, a one standard deviation change in state taxes would lead to a reduction in
exposure of 0.27*0.097=0.026ng/ml.

Column (2) controls for year of survey and state of residence. This eliminates state level
characteristics and aggregate changes in passive smoking. The effect of taxes is stronger, with a
reduction of the cotinine levels of about 0.14ng/ml, corresponding to about a 30% reduction in
exposure to smoke. Column (3) displays the effect of regulations on smoking. As described in
section 2.3 we have scaled the regulation variable from 0 (no regulations) to 3 (smoking
prohibited in all public places). The result reported in column (3) implies that going from no
regulations to a total ban would reduce cotinine by 0.48ng/ml. This means that, given that the
average concentration of cotinine is equal to 0.44ng/ml, a total ban would eliminate exposure
completely. However, if both taxes and regulation are introduced in the model (column 5 of
Table 4) results change. The effects of excise taxes are comparable to the one estimated in

20
Given that ignoring serial correlation in a differences-in-difference framework can lead to wrong standard
errors (Bertrand et al., 2004), we check the robustness of our results by clustering the standard errors at state
level only, letting the covariance structure across years to be free. The standard errors do not change much as a
result. Our main analysis is therefore robust to serial correlation. Results available on request.
Passive Smoking 14

column (2). Regulations, on the other hand, appear to have no overall effect. The 95%
confidence interval for the effect of bans ranges from -0.05 to 0.07. Even if the effect is at the
lowest part of that interval, the effect of regulations would be small. At most, a two standard
deviation change in regulations eliminates about 0.02 ng/ml of exposure.

This appears to contradict previous epidemiological studies of bans, see for instance Hopkins et
al (2001) for a review, and Travers et al. (2003) and Siegel et al (2004) for more recent
contributions. The contradiction is, however, only apparent. Most of the epidemiological work
finds that a smoking ban reduces the concentration of ETS in the places where the restrictions
apply, but do not measure it directly in non smokers so they do not address the question of
displacement. Second, when exposure is measured at the individual level, the study designs are
often simple, relying on cross-sectional data or time series evidence. When we do not control for
state or year effect, we also find a negative and significant effect of smoking bans (Table 4,
column (3)).

As shown in Figure 4, exposure to ETS has dramatically decreased over the period of analysis.
However, the results in Table 4 seem to indicate that very little is due to changes in regulations.
The effect of taxes, although statistically significant, is not large enough to explain much of the
decline in cotinine levels. These results are however not incompatible with the descriptive
evidence. The identification of these effects relies in fact on year to year changes and it may be
that the effect of state interventions are more pronounced in the long run. Moreover, the average
decrease in cotinine could also be due to other factors such as a rise in education levels or in
health awareness.

4.2 Policy Impact for Light and Heavy Exposure to Passive Smoking
In the previous paragraph we show the effect of taxes and smoking regulation on the average
non-smoker (Table 4). We now present results for non-smokers in other parts of the exposure
distribution. Table 5 displays the effect for individuals at 25%, 50%, 75 and 90% percentile of
the distribution of cotinine concentration.

[Table 5]
Passive Smoking 15

The effect of taxes is important for individuals with light exposure at the 25 percentile, but is not
different from zero in a statistical sense for any other points in the distribution. Smoking
regulations have no effect except for light exposure, where a total ban in all public places would
decrease cotinine levels by 0.21ng/ml. This appears to be a large number given that the average
level of cotinine for this group is only 0.035ng/ml. Given the variation in the data, it means that a
one standard deviation increase in regulation leads to a reduction in cotinine levels of 0.015ng/ml.

The evidence suggests that regulations and changes in excise taxes failed to have an effect in
those who are the most exposed and would gain most from a reduction in ETS, as shown in
Section 3.2.

4.3 Policy Impact by Age Group and Family Smoking Status

It may well be that taxes and regulation affect non-smokers differently, according to their age and
whether they live with smokers. We therefore investigate the specific effect of state interventions
on different age groups and family smoking status (Table 6). 21

In Table 6 we distinguish between four different age groups. The first age group is from 4 to 8,
an age where children are mostly either at home or in school or day-care, and supervised by an
adult. At that age, it is unlikely that any peers would be smoking. These individuals are therefore
exposed either to ETS at home, where parents or other adults in the household smoke, or in
public places. The second age group ranges from 9 to 12, an intermediate age group between
early childhood and adolescence. The third age group ranges from 13 to 20. Exposure for these
individuals would come from parents and also from peers. Finally, we group all individuals aged
21 or above into group 4. We have experimented with different cut-off ages, in particular with
young and elder adults, and have found similar results. We therefore pool the two groups together
to gain in precision. Finally, given that we do not have enough infants in the sample to run a
separate analysis on them, we group children up to eight years of age.

[Table 6]

The first row of the first column of Table 6 displays the results relative to the overall sample
already reported in Table 4 for an easier comparison of the results. Column 2 considers the sub-

21
While Table 6 reports the results from an OLS regression, we do not find significant differences in the coefficients
when we re-weight the sample to take into account a possible change in the composition of the sample over time. Our
results are also robust to serial correlation (see note 20).
Passive Smoking 16

sample of individuals belonging to non-smoking families, and finally column 3 displays the
results for non-smokers sharing a household with one or more smokers.

We start by analysing individuals of all ages. Non smokers living in a non-smoking family
(column 2) benefit from an increase in excise taxes. A doubling of the excise tax leads to a
reduction of about 0.10ng/ml, which represent a third of total exposure for this group.
Regulations have, on the other hand, on average no effect on exposure. The exposure to passive
smoking of non smokers living with smokers is not significantly affected by tax changes. There
is, on the other hand, some evidence that bans on smoking decreases exposure in this group. A
total ban would decrease cotinine levels by 0.90ng/ml, a 40% reduction.

We now look at the effect by age groups (Rows 2 to 5, Column 1 of Table 6) We first analyse the
effect of taxes on the level of cotinine in non smokers. Children aged eight or younger (row 2,
Column 1) benefit from increases in excise taxes. Doubling the excise tax would reduce cotinine
concentrations by as much as 0.56ng/ml. Similar results are found for next two age groups, eight
to twelve years old (row 3, Column 1 of Table 6), and thirteen to twenty (row 4, Column 1 of
Table 6). It is interesting to note that the effect of excise taxes gradually decreases with age: the
effect is -0.56 for the first age group (age less than 8), -0.43 for the second (age 8-12), and -0.39
for the third (age 13-20). As we argue in the conceptual framework section, small children are the
age group that may have little choice but to stay with their parents or carer, and therefore benefit
the most from a reduction of the number of cigarettes smoked. The last group is the adult non
smoking population aged 21 and over. We find that on average taxes have no effects.

We now decompose the results by family smoking status (Column 2 and 3, Table 6). As we
conjectured above, the observed effect of taxes is mainly due to children living with smokers
(row 2, column 3 of the table) and to teen-agers in non smoking families. Doubling the excise tax
reduces cotinine levels of children aged less than 8 in smoking households by almost 1 ng/ml.
This suggests that excise taxes are effective at reducing smoking within the household.

For teen-agers (age 13-20), the effect of taxes is mainly due to individuals living with non
smokers (row 3, column 2 of the table), although we also find a negative (but not significant)
coefficient for teen-agers living in smoking families. At this age these children are likely to spend
time with smoking friends, who may be more sensitive to changes in excise taxes than adults.
Indeed, for the age group 21 and over, we find no effect of taxes for individuals in non smoking
Passive Smoking 17

households, and even a positive (though marginally significant) effect for those in smoking
households. In this group, excise taxes weakly increase cotinine levels: a one percent increase in
excise taxes leads to 0.45 percent increase in cotinine levels. This result is in line with the
hypothesis that cigarettes smoked in social moments are likely to be more difficult to be cut.
Higher prices induce smokers to cut down on cigarettes at home and not on the cigarettes smoked
with friends. This leads on one side to the observed reduction of the exposure of children in
smoking families, but does not induce a decrease of the exposure of non smoking adults engaged
in social activities with smokers. A tentative explanation is that non smoking adults may spend
more time with their smoking partners, when the number of cigarette smoked decreases. The
resulting exposure depends then on the relative strength of these two factors.

We now analyse the effect of regulation by age and then by family smoking status. Regulations
have no significant effect on tobacco exposure in children, except for the first age group (four to
eight years old). Bans on smoking in public spaces appear to increase their cotinine levels by
0.6ng/ml, but the effect is only significant at the 10% level.
Decomposing the effect of increasing regulations by family smoking status, the only noticeable
effect is for adults in smoking households (Row 5, Column 3, Table 6): going from no regulation
to a total ban reduces cotinine levels by about 2.1 ng/ml. Regulations may have no effects on non
smokers from non smoking households, because they may less go to public places where they
would be exposed to smoke and may therefore be less affected by stricter regulations. Members
of smoking households, on the other hand, may have less of an a priori against exposure. Tighter
regulations may therefore benefit them more.

In summary, excise taxes have an important effect on reducing the exposure of children living in
families where at least one parent smokes. This suggests that smokers are more likely to cut
down on cigarettes smoked within the household rather than those smoked in social occasions.
Bans on smoking on the other hand seem to be effective only for adults who live with smokers.
This group of non-smokers are likely to be exposed to ETS both in private and public places.
Tighter regulations in public places will contribute to an overall reduction of nicotine exposure in
adults. On the other hand, stricter regulations seem to lead to an increase in the cotinine levels in
young children.

4.4 Passive Smoking in Different Public Places

Until now we have referred to cigarette smoking regulations regardless of the place where these
regulations are enforced. Smoking bans may in fact apply to very different places. Table 7
Passive Smoking 18

displays the effect of taxes and regulation on passive smoking considering separately different
places where regulation may be enforced. In particular, we have distinguished between places
where individuals spend their leisure time, and called them “going out” (i.e. restaurants,
recreational and cultural facilities), and public transportation, shopping malls, workplaces, and
schools22.

[Table 7]

The first row of Table 7 reports the coefficient of (log) excise taxes. The other rows of the table
report the regression coefficients of regulation in different places. 23 Column (1) of Table 7 refers
to the overall sample of non smokers. The other columns of the table refer to different age groups.
A number of interesting results emerge from this analysis. Most remarkable is the difference in
sign between the coefficient of what we have defined as “going out”, meaning the leisure
activities carried out outside home, and the regulations coefficients relative to the other public
places where regulations can be enforced (column 1, Table 7).

The regression coefficient of regulation in places where individuals spend their free time
engaging in social and cultural activities (“Going out”) is positive. Regulations in these places
significantly increase exposure of non smokers: going from no regulation to a total ban increases
on average cotinine levels by about 1.5 ng/ml, whereas a one standard deviation change in state
regulation increases cotinine levels by 0.13ng/ml. This can be interpreted as the existence of a
substitution effect between leisure activities in public places, where regulation can be enforced,
and in private places, where no restriction to smoking can be enforced.

The findings about regulation in public transportation, shopping malls, and schools are different.
The regression coefficients relative to regulation in these public places are on average negative
indicating a negative relationship between a tighter regulation and exposure to environmental
tobacco smoke. Going from the absence of regulation to a total ban reduces on average cotinine
levels by about 0.4 ng/ml in public transport and in schools, and by about 0.6 ng/ml in shopping
malls. In workplaces, the effect of tighter smoking regulations on tobacco exposure of non
smokers is not significantly different from zero. It seems therefore that there is no evidence of an
effect of bans on non smokers’ exposure. However, the precision of the estimates does not
exclude the fact that a ban in such places could decrease exposure. This is not necessarily in
22
See Appendix A for a more detailed description of the regulation data.
23
Our analysis is robust to serial correlation (see note 20).
Passive Smoking 19

contradiction with the association found in the epidemiology literature between bans and a
reduction in the number of cigarettes smoked by continuing smokers, the increases in quitting,
and the reduction in smoking prevalence. 24 This could in fact be due to our study design, as we
control for state and time dummies and rely on differential changes in regulation between states
and time period for identification. We therefore eliminate potential confounding by state and year
effects. It should also be noted that the lower point of the confidence interval implies a reduction
of about 0.16ng/ml for a total ban.

Columns (2) to (5) of Table 7 distinguish between four different age groups. As noted above,
children of the first age group considered (4 to 8 years old) are mostly either at home or in school
or day-care, and supervised by an adult. Exposure at this age comes therefore either from adults
living in the same household, or from public places. The regression coefficient relative to
regulation in recreational public places is positive and large: going from no regulation to a total
ban increases on average cotinine levels by about 6 ng/ml, whereas an increase of one standard
deviation in state regulations would increase cotinine levels by 0.5ng/ml. The substitution
between public and private recreational places seems therefore to heavily affect children of this
age group. On the other hand, these children are also those who benefit more from tighter
regulations in public places other than recreational ones like public transport and shopping malls.
Adult individuals that can no longer smoke in public transport or while shopping, will expose the
children less to ETS. Going from no regulation to a total ban reduces on average cotinine levels
by about 1.5 ng/ml in public transports, and by about 2.5ng/ml in shopping centres. This last
result seems to indicate that small children that need constant adult supervision are those that can
benefit the most from tighter regulations in such public places. The effect for the second and third
age groups (8-12 and 13-20) is of similar sign but of smaller magnitude. For these individuals,
the effect of regulations in public recreational places is about half of that observed for younger
children (going from no regulation to a total ban increases on average cotinine levels by about 3.6
ng/ml compared to an increase of 6.2 ng/ml for younger children). The substitution between
public and private recreational places is therefore observed also for these two age groups but it is
smaller than that observed for very young children. Older children may be more independent
from their parents and may go out on their own. Finally, the regression coefficient relative to
public recreational activities is positive but not significant for individuals aged 21 or above
(group 4). A plausible explanation of this is that adult non smokers spend their recreation time
with other adults both in public and private places, and are therefore exposed to ETS either in one

24
See for instance Hopkins et al (2001), and Levy and Friend (2003).
Passive Smoking 20

or the other place. What is relevant is the overall amount of ETS. It makes no difference for this
age group whether smokers choose to smoke in public or private places.

Tighter regulations in public places other than recreational ones (i.e. transportation, shopping
malls, and schools) seem on average to lead to a reduction in tobacco smoke exposure of non
smokers. This is particularly the case for small children who are most likely spend time in these
public places accompanying older individuals. This interpretation is supported by the finding
about adults. We observe, in fact, that whenever a private alternative to the public place exists,
we do not observe a reduction in nicotine exposure as a consequence of a stricter regulation. A
first example is the regression coefficient of regulation in public transport (column 5, Table 7)
relative to individuals older than 21. It is positive though not significant. This may be due to the
fact that this group of individuals may in fact use the car as a private alternative to the public
transportation. We therefore observe for this age group the public/private substitution effect. A
second example is the coefficient of the regulation in shopping malls. Though generally negative,
it is positive, though not significant, for the age group 8-12, which most likely is the one that
spend less time engaging in shopping activities in commercial centres.
Finally, the effect of tighter regulation in school, though on average significant, is not
significantly different from zero when we consider separately the different age groups.

4.5 Distributional Effects of Smoking Regulations and Taxes

Finally, we investigate whether state interventions affect differently individuals according to their
socio-economic status. In many countries, public health authorities seek not only to improve the
health of the population, but also to reduce health inequalities across socio-economic groups. We
assess the effect of smoking regulations and changes in excise taxes by household income groups.
We split our sample in three income groups of equal size and estimate separately the effect on
passive smoking. The results are presented in Table 8.

[Table 8]

For the lowest income group, the effect of taxes is not significant, while the effect of
regulations is positive and significant. A total ban would increase exposure by 0.3 ng/ml. For
intermediate levels of income, taxes have a significant and negative effect, while bans appear
to have no effect. Finally, for non-smokers in high income households, introduction of
smoking regulations decreases (weakly) the exposure to tobacco smoke. These results
suggest that smoking regulations have a distributional effect, increasing the exposure and
Passive Smoking 21

putting at risk the health of poorer section of the population while it benefits individuals in
higher socio-economic position. The consequence of strengthening smoking regulations
would be a widening in health disparities across socio-economic groups.

5 Conclusion
The effect of passive smoking is of increasing public concern. Although the economic literature
has evaluated the effect of government intervention on smoking intensity or prevalence, there has
been, so far, no direct evaluation of these measures on non-smokers.

In this paper we characterize the extent of exposure to environmental smoke, and evaluate the
effect of changes in excise taxes and bans on passive smoking. We use a direct measure of
passive smoking which has not been used in the economic literature, the concentration of cotinine,
a metabolite of nicotine, in body fluids of non smokers. This allows us to precisely identify the
effect of state intervention on non-smokers.

We find that increasing taxes on cigarettes reduces on average exposure to cigarette smoke of
non smokers. The effect of state excise taxes also varies across demographic groups. We find that
taxes have a strong effect on young children living with smokers but little effect on non smoking
adults. This suggests that smokers cut down on the cigarettes they smoke at home but not those in
social activities with other adults.

Using information on the implementation of the Clean Air Act across time and different US
states, we also find that smoking regulations have on average no effect on exposure. We show
that this latter result is not due to a lack of statistical power to detect a precise effect but rather to
the fact that regulations have contrasting effects depending on where they are imposed and
depending on which group of the population is affected. 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. The level of cotinine in children considerably
increases as a result of bans in recreational public places, while decreases if tighter bans are put
in place in public transport or shopping malls.

A third and important finding is that state interventions through taxes and bans only affect non
smokers who have light exposure to tobacco smoke. Given the non-linear effect of exposure on
Passive Smoking 22

health, it implies that state interventions, including taxes, have had a moderate impact on health
outcomes. Finally, we find that smoking regulations increases exposure of poorer individuals,
while it is beneficial to individuals in higher socio-economic position. The rise in the number of
regulations observed over the nineties is likely to have increased health inequalities related to
passive smoking.

Our results question the usefulness of bans in reducing smoking exposure for non smokers. More
precisely, we show that policies aimed at reducing exposure to tobacco smoke induce changes in
behaviors which can offset these policies. It is therefore of crucial importance to understand how
smoking behaviors are affected by regulations. So far, the literature has not gone far enough in
studying smoking behavior to be able to evaluate their effect on non smokers. It is not enough to
show that smokers react to prices or taxes. Information on which particular cigarette is cut down
during the day, where smokers smoke and with whom are also relevant. There are complex
interactions at play and considerable heterogeneity in their effects across socio-demographic
groups. Using a biomarker such as cotinine concentrations is a very direct way of evaluating the
overall effect of interventions and the induced changes in behaviors.

On the policy side, it seems therefore important when designing public policies aimed at reducing
tobacco exposure of non smokers to distinguish between the different public places where bans
are introduced. Displacing smoking towards places where non-smokers live is particularly
inefficient. It may also increase health disparities across socio-economic groups and in particular
in children. Therefore, total bans may not be the optimal policy. A better policy may be to allow
for alternative places to which smokers can turn to. It would benefit children but harm non
smoking adults. There are several reasons why one may want to protect children. They constitute
a vulnerable group with little choices to avoid contamination. This age group is particular prone
to tobacco related diseases and poor health in childhood has lasting consequences not only for
future health but also for the accumulation of human capital (Case et al, 2005).

Governments in many countries are under pressure to limit passive smoking. Some pressure
groups can be very vocal about these issues and suggest bold and radical reforms. As often, 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. Up to know there is little guidance
on how to design optimal policies to curb passive smoking. This paper fills this gap.
Passive Smoking 23

Appendix A: Smoke free regulations data

The information contained in this appendix, are drawn from the codebook for the “Tobacco
Control Policy and PrevalenceData: 1991-2001”25, compiled by researchers in the Department of
Health Behavior at the Roswell Park Cancer Institute (RPCI) in Buffalo, New York, in
conjunction with researchers at the MayaTech Corporation in Washington, DC.26
Eleven different locations were regulations were enacted were identified: Government worksites,
Private worksites, Child care centers, Health care facilities, Restaurants, Recreational facilities,
Cultural facilities, Public transit, Shopping malls, Public schools, and Private schools. And for
each of these locations has been measured the degree of restrictions enforced in the various years
(1991-2001).

General Location Restriction Decisions

The following “standard coding scheme” was employed for the majority of locations of interest
(including: government worksites, private worksites, health care facilities, restaurants, public
transit, and shopping malls).

0 No provision/not meet a restriction

1 Restrict smoking to designated areas or require separate ventilation with exemptions for
locations of a certain size (e.g. restaurants with a seating capacity of less than 50)
2 Restrict smoking to separately ventilated areas or a ban with exemptions for certain
locations where only a restriction applies
3 Ban at all times

For locations other than those mentioned in the table above, different coding schemes were used.
In the following we report the coding schemes that were used for each of them.

1. Child Care Centers coding scheme

0 No provision/not meet a restriction

1 Restrict smoking to designated areas
2 Restrict smoking to separately ventilated areas or a ban when children are
present with exemptions
3 Ban when children are present (commercial daycare)
4 Ban at all times when children are present (explicitly including home-based)
5 Ban at all times (explicitly including home-based)

2. Recreational Facilities coding scheme

0 No restriction
1 Restricts smoking to DSAs in gyms or arenas
2 Restricts smoking to DSAs in both gyms and arenas
3 Restricts smoking to DSAs in all recreational facilities
4 Bans smoking in gyms or arenas and restricts to DSA(s) in other recreational

25
http://www.impacteen.org/tobaccodata.htm
26
Coding for public schools, private schools and cultural facilities were developed by MayaTech from the state
smoke-free air law coding provided by RPCI.
Passive Smoking 24

area(s)
5 Bans smoking at all recreational locations

3. Cultural Facilities coding scheme

0 No restriction
1 Restricts smoking to DSAs in fewer than 3 cultural areas
2 Restricts smoking to DSAs in 3-5 cultural areas
3 Restricts smoking to DSAs in more than 5 cultural areas
4 Restricts smoking to DSAs in all cultural facilities
5 Bans smoking at all cultural locations

4. Schools coding scheme

0 No provision/not meet a restriction

1 Restrict smoking to designated areas
2 Restrict smoking to separately ventilated areas or a ban when children are
present with exemptions
3 Ban when children are present (school buildings)
4 Ban at all times when children are present (buildings and grounds)
5 Ban at all times (buildings and grounds)

For details about the choices made in interpreting the language of the laws and regulations case
by case in the different States we refer to the official codebook drawn by Impacteen
(http://www.impacteen.org).

We have aggregated these different locations in a number of ways. First, we have constructed a
general measure of restriction, considering an average of all the locations. In a second time, we
have aggregated the different public locations in: 1. recreational activities (“going out”) which
includes restaurants, cultural and other recreational public places; 2. public transport; 3. shopping
malls; 4. workplaces, which includes both governmental and private workplaces; 5. school,
which includes childcare centres, and both public and private schools.
Passive Smoking 25

REFERENCES

Adda J and F Cornaglia (2005) “Taxes, Cigarette Consumption and Smoking Intensity“,
forthcoming, American Economic Review.

American Cancer Society. Cancer Facts and Figures 2003. Atlanta, GA.

ASH (2005) “Going Smokefree: the case for all pubs and clubs”. http://www.ash.org.uk/

Becker G and K Murphy (1988) “A Theory of Rational Addiction”, Journal of Political

Economy, 96, 4, 675-699.

Becker G, Grossman M and K Murphy (1994) “An Empirical Analysis of Cigarette Addiction”,
American Economic Review, 84, 3, 396-418.

Bertrand, M, Duflo E and Mullainathan S (2004) “How Much Should We Trust Difference in
Differences Estimates?”, Quarterly Journal of Economics 119(1), pp. 249-275.

Case, A Fertig A and C Paxson (2005) “The Lasting Impact of Childhood Health and
Circumstance”.Journal of Health Economics, 2005, v. 24, pp. 365-389.

Chaloupka F (1991), “Rational Addictive Behavior and Cigarette Smoking”, Journal of Political
Economy, 99, 4, 722-742.

Chaloupka F and K Warner (2000), “The Economics of Smoking”, Handbook of Health

Economics, J Newhouse and A Cuyler editors.

DeCicca P, Kenkel D, and A Mathios (2002), “Putting Our the Fires: Will Higher Taxes Reduce
the Onset of Youth Smoking?”, Journal of Political Economy , 110, 1, 144-69.

DiNardo Lemieux Fortin Econometrica (1996), "Labour Market Institutions and the Distribution
of Wages, 1973-1992: a Semiparametric Approach", Econometrica, 64, 1001-1046

Environmental Protection Agency –US (1992), “Respiratory Health Effects of Passive Smoking:
Lung Cancer and Other Disorders”, Washington, DC: EPA, Office of Air and Radiation, EPA
publication 600/6-90/006F.

Evans W, Farrelly M and E. Montgomery (1999) “Do Workplace Smoking Bans Reduce
Smoking?”, American Economic Review, 89, 4, 728-747.

Evans W and J Ringel (1999), “Can Higher Cigarette Taxes Improve Birth Outcomes?”, Journal
of Public Economics, 72, 1, 135-154.

Gergen P J, Fowler JA, Maurer KR, Davis WW, and M D Overpeck (1998) “The Burden of
Environmental Tobacco Smoke Exposure on the Respiratory Health of Children 2 Months
Through 5 Years of Age in the United States: Third National Health and Nutrition Examination
Survey, 1988 to 1994” Pediatrics, February 1, 101, 2, 8e-8.
Passive Smoking 26

Hackshaw AK, Law MR, and NJ Wald (1997), “The accumulated evidence on lung cancer and
environmental tobacco smoke”, BMJ, 315, 980-988.

He J, Vupputuri S, Allen K, Prerost MR, Hughes J, Whelton PK (1999), “Passive smoking and
the risk of coronary heart disease—a meta-analysis of epidemiologic studies”, N Engl J Med ,
340, 920-6.

Hopkins DP, Briss PA, Ricard CJ, Husten CG, Carande-Kulis VG, Fielding JE, Alao MO,
McKenna JW, Sharp DJ, Harris JR, Woollery TA, Harris KW (2001), “Reviews of the Evidence
Regarding Interventions to reduce Tobacco Use and Exposure to Environmental Tobacco
Smoke”, American Journal of Preventive Medicine, 20, 16-66.

Howard G, Wagenknecht LE, Burke GL (1998) “Cigarette smoking and progression of

atherosclerosis: the Atherosclerosis Risk in Communities (ARIC) study”, JAMA, 279, 119-124.

IARC (2004) “Tobacco smoke and involuntary smoking, Monographs on the evaluation of
carcinogenic risks to humans”, Vol 83, Lyon, France.

Jarvis M et al. (2000), “Children’s exposure to passive smoking in England since the 1980’s:
cotinine evidence from population surveys”, British Medical Journal, 321, 343-5.

Jarvis M, Tunstall-Pedoe H, Feyerabend C, Vesey C and Y Salloojee (1984), “Biochemical

markers of smoke absorption and self reported exposure to passive smoking”, Journal of
Epidemiology and Community Health, Vol 38, 335-339

Jarvis M, Tunstall-Pedoe H, Feyerabend C, Vesey C and Y Saloojee (1987), “Comparison of

tests used to distinguish smokers from non smokers”, American Journal of Public Health, Vol
77, 1435-8.

Jarvis M, Feyerabend C, Bryant A, Hedges B, and P Primatesta (2001), “Passive smoking in the
home: plasma cotinine concentrations in non-smokers with smoking partners”, Tobacco Control
10(4), 368 – 374.

Kriz P, Bobak M, and B Kriz (2000), “Parental smoking, socio-economic factors, and risk of
invasive meningococcal disease in children: a population based case-control study. Archives of
Disease in Childhood”, 83, 117-21.

Lam T, Leung GM, and LM Ho (2001), “The effects of environmental tobacco smoke on health
services utilization in the first eighteen months of life”, Pediatrics, 107, 6.

Law MR, Morris JK, and NJ Wald (1997), “Environmental tobacco smoke exposure and
ischaemic heart disease: an evaluation of the evidence”, BMJ, 315, 973-980.

Levy DT, and KB Friend (2003), “The effects of clean indoor air laws: what do we know and
what do we need to know?”, Health Educ Res, 18(5), 592–609.

Mannino DM, Moorman JE, Kingsley B, Rose D, Repace J., (2001) “Health effects related to
environmental tobacco smoke exposure in children in the United States”, Archives of Pediatric
Adolescent Medicine, 155, 36-41.
Passive Smoking 27

NHS Scotland (2005), “International Review of the Health and Economic Impact of the
Regulation of Smoking in Public Places”
www.healthscotland.researchcentre/pdf/internatioonalreviewfullreport.pdf

Otsuka R (2001), “Acute effects of passive smoking on the coronary circulation in healthy young
adults”, JAMA, 286, 436-441.

Orzechowski and Walker (2001) “The Tax Burden on Tobacco. Historical Compilation”, Volume
36.

Pirkle J L, Flegal K M, Bernert J T, Brody D J, Etzel R A, and K R Maurer (1996), “Exposure of

the US population to environmental tobacco smoke: the Third National Health and Nutrition
Examination Survey, 1988 to 1991”, JAMA, 275, 1233-1240.

Siegel M, Albers AB, Cheng DM, Biener L, Rigotti NA (2004), “Effect of Local Restaurant
Smoking Regulations on Environmental Tobacco Smoke Exposure Among Youths”, American
Journal Public Health, 94, 321-325.

Strachan DP and DG Cook (1997), “Parental smoking and lower respiratory illness in infancy
and early childhood”, Thorax, 52, 905-914.

Travers M et al. (2003), “Indoor air quality in hospitality venues before and after implementation
of a clean indoor air law”, Western New York, MMWR, 53, 1038-1041.

Whincup P, Gilg J, Emberson J, Jarvis M, Feyerabend C, Bryant A, Walker M, Cook D (2004),

“Passive smoking and risk of coronary heart disease and stroke: prospective study with cotinine
measurement”, BMJ.

Whitlock G, S MacMahon, S Vander Hoorn et al. (1998), “Association of environmental tobacco

smoke exposure with socioeconomic status in a population of 7725 New Zealanders”, Tobacco
Control, 7, 276-280.
Passive Smoking 28

Table 1 - Descriptive Statistics

Whole Individuals in Individuals in
sample smoking families Non smoking families

# of observations 29687 5770 23897

Average level of cotinine (ng/ml) 0.44 1.47 0.26
(1.02) (1.59) (0.75)
Proportion with detectable cotinine
84% 99% 79%
measure (>0.035ng/ml)
Proportion with cotinine>1ng/ml 14% 46% 5%
Proportion with cotinine>5ng/ml 1% 4% 0.5%
Average age 33.5 22.7 35.7
Age range 4-90 4-90 4-90
sex (% male) 46 46.8 45.8
% white 74 72 74
% black 12 18 11
Note: Standard deviations in parenthesis. The whole sample consists of all non-smoking individuals who have a
valid cotinine measure lower than 10ng/ml.

Table 2: Descriptive Statistics. Excise Taxes and Regulations

Average Level Range Within State
Standard dev.
Log tax 3.43 0.97-4.62 0.27
Average Regulation 0.79 0-2.63 0.22
Bans Going-out 0.76 0-2.67 0.25
Bans public transportation 1.24 0-3 0.31
Bans shopping mall 0.27 0-3 0.31
Bans workplace 0.70 0-3 0.28
Bans schools 0.85 0-2 0.27

Table 3: Health and Exposure to Tobacco Smoke: Odds-Ratio of Poor Health

Children Adults
Asthma Wheezing Stroke Chronic
In Chest Bronchitis
[0-0.1] 1 1 1 1

[0.1-0.2] 1.06 1.17 1.12 1.02

Cotinine
(0.17) (0.14) (0.22) (0.13)
Level
[0.2-1.0] 1.07 1.18* 1.00 1.45**
(ng/ml)
(0.14) (0.12) (0.19) (0.16)
[1.0-10.0] 1.62** 1.41** 1.70** 1.56**
(0.22) (0.15) (0.41) (0.23)
Robust standard errors in parenthesis. The regression also controls for age, age square, sex and race. ** significant at
5%, * significant at 10%
Passive Smoking 29

Table 4 - Effects of Taxes and Regulations on Passive Smoking.

Dependent variable: cotinine. Average Cotinine Level: 0.44ng/ml
(1) (2) (3) (4) (5)
Log Tax -0.097** -0.14** -0.14**
(0.034) (0.04) (0.06)
Regulations -0.16** -0.032 0.01
(0.028) (0.032) (0.03)
Controls:
Year Dummies X X X
State Dummies X X X
Age, sex, race X X X X X
Robust standard errors adjusted for clustering at state and year level in parenthesis. ** significant at 5%, * significant
at 10%

Table 5: Effect of Taxes and Smoking Regulations for Different Percentiles in the Cotinine
Distribution.
Percentile Mean 25% 50% 75% 90%
Cotinine level 0.44ng/ml 0.035ng/ml 0.10ng/ml 0.35ng/ml 1 ng/ml
Log Tax -0.14** -0.17** -0.007 -0.02 0.01
(0.06) (0.06) (0.03) (0.03) (0.01)
Regulations 0.01 -0.07** 0.003 0.02 -0.001
(0.03) (0.03) (0.015) (0.01) (0.008)
Controls:
Year Dummies X X X X X
State Dummies X X X X X
Age, sex, race X X X X X
Robust standard errors adjusted for clustering at state and year level in parenthesis. ** significant at 5%, * significant
at 10%
 Passive Smoking 30

Table 6 - Effect of Taxes and Regulation on Passive Smoking. Dependent variable: cotinine

(2) (3)
(1)
Individuals in Individuals in
All Individuals
Non smoking families smoking families
Average Cotinine
0.44ng/ml 0.26ng/ml 1.47ng/ml
Level
Log Tax Regulations Log Tax Regulations Log Tax Regulations
All ages -0.14** 0.01 -0.098** -0.004 -0.05 -0.29*
(0.06) (0.03) (0.04) (0.03) (0.28) (0.17)

Age <8 -0.56** 0.21* 0.056 0.057 -0.97** -0.37

(0.19) (0.11) (0.07) (0.04) (0.49) (0.36)

Age 8-12 -0.43** 0.067 0.05 -0.03 -0.94* -0.10

(0.20) (0.12) (0.06) (0.03) (0.55) (0.31)

Age 13-20 -0.39** 0.09 -0.41** 0.08* -0.08 0.14

(0.12) (0.06) (0.09) (0.05) (0.34) (0.18)

Age 20 + 0.02 -0.047 -0.067 -0.02 0.67* -0.69**

(0.06) (0.04) (0.05) (0.03) (0.40) (0.33)
Regressions controls for age, sex, race, state of residence and year of survey. Robust standard errors adjusted for
clustering at state and year level in parenthesis. ** significant at 5%, * significant at 10%.
Passive Smoking 31

Table 7 - Effect of Taxes and Regulation on Passive Smoking, by place of enforcement.

Dependent variable: cotinine.
(1) (2) (3) (4) (5)
All ages Age<8 Age 8-12 Age 13-20 Age 20+
Log Tax -0.12** -0.58** -0.38* -0.36** 0.01
(0.06) (0.20) (0.21) (0.12) (0.07)
Regulation Going out 0.52** 2.08** 1.15** 1.22** 0.30
(0.17) (0.51) (0.59) (0.33) (0.23)
Regulation Public Transport -0.13** -0.50** -0.67** -0.34** 0.008
(0.05) (0.18) (0.16) (0.10) (0.06)
Regulation Shopping Mall -0.21** -0.85** 0.15 -0.42** -0.15**
(0.06) (0.22) (0.25) (0.16) (0.07)
Regulation Workplace -0.005 - - - -0.13
(0.05) (0.10)
Regulation Schools -0.13** 0.14 -0.15 -0.10 -
(0.05) (0.22) (0.19) (0.12)
Regressions controls for age, sex, race, state of residence and year of survey. Robust standard errors adjusted for
clustering at state and year level in parenthesis. ** significant at 5%, * significant at 10%.

Table 8: Distributional Effects of Taxes and Bans

Household Income: Household Income: Household Income:
[0,33%] [33%,66%] [66%, 100%]
Log Tax -0.12 -0.24** -0.07
(0.09) (0.10) (0.05)
Regulations 0.10** 0.07 -0.07*
(0.05) (0.05) (0.04)
Controls:
Year Dummies X X X
State Dummies X X X
Age, sex, race X X X
Robust standard errors adjusted for clustering at state and year level in parenthesis. ** significant at 5%, * significant
at 10%
Passive Smoking 32

Figure 1: Cotinine Level by Number of Cigarettes Smoked in the Household

3
Cotinine Level, ng/ml

0
0 10 20 30 40

Total Number of Cigarettes Smoked in Household

Passive Smoking 33

Figure 2: State Excise Taxes, by US State 1988-2002.

Alabama Alaska Arizona Arkansas California Colorado Connecticut DC

100

50

Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa

100
State Excise Tax (Cents)

50

Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota

100

50

Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico
100

50

New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island
100

50

South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington
100

50

0
1990 1995 2000 1990 1995 2000 1990 1995 2000 1990 1995 2000 1990 1995 2000
West Virginia Wisconsin Wyoming
100

50

0
1990 1995 2000 1990 1995 2000 1990 1995 2000

Year
Passive Smoking 34

Figure 3: Level of Restrictions on Smoking, by US State, 1991-2001.

Alabama Alaska Arizona Arkansas California Colorado Connecticut DC

3
2
1
0
Level of Restriciton on Smoking

Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa

3
2
1
0

Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota

3
2
1
0

Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico
3
2
1
0

New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island
3
2
1
0

South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington
3
2
1
0
1990 1995 2000 1990 1995 2000 1990 1995 2000 1990 1995 2000 1990 1995 2000
West Virginia Wisconsin Wyoming
3
2
1
0
1990 1995 2000 1990 1995 2000 1990 1995 2000

Year
Passive Smoking 35

Figure 4: Average Cotinine Concentration in Non-Smokers.

1
Cotinine Concentration

.8

.6
ng/ml

.4

.2

0
1988 1992 1996 2000
Year

Figure 5: Fraction Individuals with Cotinine larger than 1ng/ml

.2
% Cotinine > 1 mg/ml

.15

.1

.05

1985 1990 1995 2000 2005

Year
Passive Smoking 36

Figure 6: Average Cotinine Concentration in Non-Smokers – Non Smoking

Households

.4
Cotinine Concentration

.35
ng/ml

.3

.25

.2
1988 1992 1996 2000
Year

Figure 7: Average Cotinine Concentration in Non-Smokers – Smoking Households

2
Cotinine Concentration
ng/ml

1.5

1
1988 1992 1996 2000
Year
Passive Smoking 37

1 Introduction................................................................................................................. 2
2 Conceptual Framework and Methodology.................................................................. 4
2.1 Effects of Tax Changes and Smoking Bans........................................................ 4
2.2 Cotinine as a Proxy for Smoking Intake ............................................................. 6
2.3 Methodology ....................................................................................................... 7
3 The Data and Descriptive Statistics ............................................................................ 9
3.1 Exposure to Passive Smoking............................................................................. 9
3.2 Health Effects of Passive Smoking................................................................... 10
3.3 Excise Taxes and Smoking Restrictions to Tobacco Exposure ........................ 10
3.4 Trends in Passive Smoking............................................................................... 12
4 Empirical Results: Passive Smoking and State Intervention .................................... 13
4.1 Whole Sample of non smokers ......................................................................... 13
4.2 Policy Impact for Light and Heavy Exposure to Passive Smoking .................. 14
4.3 Policy Impact by Age Group and Family Smoking Status ............................... 15
4.4 Passive Smoking in Different Public Places..................................................... 17
4.5 Distributional Effects of Smoking Regulations and Taxes............................... 20
5 Conclusion ................................................................................................................ 21
Appendix A: Smoke free regulations data ........................................................................ 23
REFERENCES ................................................................................................................. 25