Vous êtes sur la page 1sur 6

Economic Modelling 40 (2014) 284–289

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.elsevier.com/locate/ecmod

Economic stress and cigarette smoking: Evidence from the


United States☆
Rajeev K. Goel ⁎
Department of Economics, Illinois State University, Normal, IL 61790-4200, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper uses pooled data on U.S. states for the post-MSA period to estimate the demand for cigarettes, with the
Accepted 15 April 2014 main contribution lying in considering the effects of economic stress/uncertainty. Different measures of economic
Available online 14 May 2014 stress – standard deviations and averages of unemployment and property prices – are considered. Greater
economic stress is found to lower cigarette smoking across various specifications. Other findings largely support
JEL classification:
the literature on cigarette demand — price effects are negative, border price effects are positive and the effect of
D12
I10
income is negative.
Published by Elsevier B.V.
Keywords:
Cigarettes
Smoking
Economic stress
Uncertainty
Smuggling
United States

1. Introduction employment. This reduction in potential earnings can lead to reduced


spending, including lowering of cigarette purchases. In the context of
In recent years numerous investigations have been conducted to the investment-uncertainty linkage (see Dixit and Pindyck (1994)),
estimate the demand for cigarettes and the United States has been the greater economic uncertainty can be seen as impacting the investment
focus of a considerable share of this research (see Chaloupka and decisions in their health (i.e., cigarette purchases) by smokers. Cigarette
Warner (2000), Goel and Nelson (2008) and U.S. Department of retailers might also reduce inventories and keep fewer brands on hand
Health and Human Services (2000)). Researchers have examined (variety) in periods of greater economic uncertainty. This would in-
various aspects with data at different levels of detail. Part of this focus crease the transactions costs of search for smokers preferring particular
has been to satisfy policy makers' continuing and evolving interest in brands and this cost increase would induce them to lower purchases.
smoking control in the light of new health insights over time. The result Economic stress might also indirectly affect smoking through psycholog-
is that some determinants of smoking behavior (mainly price and ical effects related to greater economic anxiety and tension. In this case,
income effects) are relatively well understood. However, a few aspects smoking might actually increase especially among habitual (as opposed
remain unexplored and the present study focuses on one such area — to casual) smokers. For instance, greater nicotine intake might affect
namely, the effect of economic stress on smoking. the anxiety associated with stress (Pomerleau and Pomerleau, 1991).
Greater economic stress can directly induce shifts in cigarette pur- Furthermore, greater economic stress might induce greater substitution
chases by encouraging the reallocation of resources away from ciga- by smokers. This substitution might include switching to other, cheaper
rettes by smokers and potential smokers (i.e., by altering the discount tobacco products (e.g., chewing tobacco, and generic cigarette brands) or
rates of smokers and potential smokers). In times of greater economic to contraband cigarettes (e.g., illegally smuggled cigarettes).1 Thus, the
uncertainty, employed smokers are worried about potential job losses, overall effect of economic stress is a priori unclear and the formal analy-
whereas unemployed smokers face bleaker prospects of finding sis below will reveal the relative strengths of various influences.

☆ Helpful comments by a referee and by participants at the 2013 Academy of Economics


1
and Finance meetings are appreciated. Research support of the Katie School of Insurance is An indirect channel of influence of economic uncertainty might be that greater uncer-
gratefully acknowledged. tainty increases illegal behavior such as corruption (see Braun and Di Tella (2004), Goel
⁎ Tel.: +1 309 438 2360; fax: +1 309 438 5228. and Ram (2013)), which in turn increases the availability of contraband cigarettes, causing
E-mail address: rkgoel@ilstu.edu. a switch away from legal cigarette purchases.

http://dx.doi.org/10.1016/j.econmod.2014.04.009
0264-9993/Published by Elsevier B.V.
R.K. Goel / Economic Modelling 40 (2014) 284–289 285

Economists have generally been lagging their counterparts in other issue by including the lowest cigarette price in a geographically border-
disciplines in recognizing the role stress might play in smoking and ing state (BorPRICE). For instance, in the case of Florida, BorPRICE would
the present research will contribute in this regard. For instance, be the lowest price in Alabama and Georgia.6 The expected coefficient
Pomerleau and Pomerleau (1991) note that “integration of behavioral, on this variable is positive — higher border price boosts a state's
physiological, and biochemical research …should lead to a better under- cigarette sales as smokers and smugglers cross state borders to take ad-
standing of stress and smoking” (p. 599), (also see Barnes and Smith vantage of cigarette tax differentials. The border-price variable is fine-
(2009), Ruhm (2000, 2005), Siahpush and Carlin (2006), Tyas and tuned by including a dummy variable (Producer) for the six main tobac-
Pederson (1998), and Xu and Kaestner (2010)).2 co growing states (Georgia (GA), Kentucky (KY), North Carolina (NC),
This paper considers different measures of economic stress (specifi- South Carolina (SC), Tennessee (TN), and Virginia (VA)). These states
cally, averages and deviations in unemployment rates and in property generally have low cigarette excise taxes, making them attractive tar-
prices) in a fairly standard cigarette demand specification employing gets for cross-state cigarette smugglers.
state-level U.S. data for the post-MSA (1998) period. This research Turning to the main focus of this research, as discussed above,
gains added significance in light of the recent turbulence in financial economic stress can directly and indirectly impact the smoking behav-
markets and can also broadly be seen as adding another dimension to ior of smokers and potential smokers. Economic stress is accounted for
the empirical research on the economics of uncertainty (see Dixit and by including different measures: (i) UNEMdev is a three-year moving
Pindyck (1994), Goel and Ram (2001) and Pindyck and Solimano standard deviation of state unemployment rate; (ii) PrpPRICEdev is a
(1993)). The main finding is that greater economic stress reduces ciga- three-year moving standard deviation of the state-level index for single
rette demand and this result stands up to alternate specifications. family housing prices7; (iii) UNEMma is a three-year moving average of
state unemployment rate; and (iv) PrpPRICEma is a three-year moving
2. Empirical setup and data average of the index for single family housing prices.8 Various authors
have considered these measures (i.e., moving averages and moving
Empirical cigarette demand estimation studies almost universally standard deviations of macro variables) to capture economic uncertain-
control for price and income, with many frequently accounting for bor- ty in other contexts, including the impact of uncertainty on research and
der smuggling effects (see Advisory Commission on Intergovernmental development spending, corruption and investment (see Goel and Ram
Relations (1985), CDC (2010)). To this mix, one or two “unique” vari- (2001, 2013), Pindyck and Solimano (1993)). Other things being the
ables are added depending upon the focus of a particular investigation same, larger values of these measures would imply greater economic
(for broad overviews of the literature see Chaloupka and Warner stress. While the broader cross-disciplinary research has recognized the
(2000) and U.S. Department of Health and Human Services (2000)). relation between stress and smoking (see Pomerleau and Pomerleau
Given this overall framework, our contribution is in examining the ef- (1991), Ruhm (2005) and Tyas and Pederson (1998)), formal insights
fect of economic stress or uncertainty on state-level cigarette demand into the role of economic stress have been lacking. Greater economic un-
in the United States over the years 1998–2009. The choice of the time certainty could cause a reallocation of resources by some smokers and
period under study follows the period after the landmark Master Settle- potential smokers, in addition to some psychological effects that could
ment Agreement in 1998 between tobacco companies and U.S. states also affect the propensity to purchase cigarettes/smoke.
that imposed far-reaching restrictions on the marketing and sale of to- Our (balanced) pooled sample of the forty-eight contiguous U.S.
bacco products (see Viscusi and Hersch (2009)). states over 1998–2009 has 576 annual observations.9 Details about
The formal estimated equation, emphasizing the role of economic the various variables used, including definitions, statistics and sources,
stress or uncertainty, takes the following form (with subscripts i and t, are provided in Table 1. Among the measures of economic stress in
respectively, denoting a state and year)3 our analysis, PrpPRICEdev (3-year moving standard deviation of single-
family housing price index) was the highest in Arizona (2005) and the
  lowest in Connecticut (2007); while UNEMdev (3-year moving standard
CONSit ¼ f PRICEit ; BorPRICEit ; Produceri ; INCOMEit ; Economic stressijt
deviation of the unemployment rate) was the highest in Nevada (2009)
ð1Þ and the lowest in Montana (1998), (multi-state tie); PrpPRICEma (3-year
moving average of single-family housing price index) ranged from a high
i = 1, 2, …,48 in Oregon (2008) to a low in California (1998); and UNEMma (3-year
t = 1998, …,2009 moving average of the unemployment rate) had a high in Michigan
j = UNEMdev, PrpPRICEdev, UNEMma, PrpPRICEma.

Here CONSit denotes per capita cigarette consumption in the ith state
in year t. Following related literature, we take the available data on
state-level cigarette sales to denote cigarette consumption (Baltagi 6
The effects of border prices can be further fine-tuned by considering additional loca-
and Levin, 1986).4 PRICE is the retail price per 20-pack of cigarettes, tional dimensions such as the relative populations of border counties.
7
According to the Statistical Abstract of the United States (2011, Table 728), “The (prop-
inclusive of federal, state and (in some states) local taxes.5
erty price) index reflects average price changes in repeat sales or refinancings on the same
The issue of cross-border smuggling of cigarettes, both casual and or- properties…the index represents the annual percentage change in home values in the
ganized, has been recognized in the literature (Baltagi and Levin, 1986; fourth quarter of the year shown relative to the fourth quarter of the previous year. The
also Joossens et al., 2000 and Lovensheim, 2008) and we address this information is obtained by reviewing repeat mortgage transactions on single-family prop-
erties whose mortgages have been purchased or securitized by either Fannie Mae or
Freddie Mac.”
8
One could also use deviations in state-level consumer price indices to get at a similar
2
A well-researched related strand of the literature examines the effects of health risks measure of economic uncertainty. However, the available housing price index data seem
on smoking behavior (see Chaloupka and Warner (2000), Viscusi (2000)). Further, some more consistent at the state level.
9
scholars have examined the role of uncertainty in decisions to quit smoking (see Jones Observations from 1996 and 1997 were lost in creating the three-year moving stan-
(1994)). dard deviations of the various measures of economic stress. The choice of the data span
3
With individual level data, one can also assess the reverse linkages from smoking to is somewhat arbitrary. However, it covers the period of considerable economic upheaval
stress (see Siahpush and Carlin (2006)). and also follows the Master Settlement Agreement between tobacco companies and U.S.
4
The equation of cigarette sales with cigarette consumption implicitly assumes the ab- states in 1998 that led to far reaching and comprehensive restrictions on the marketing
sence of hoarding. Survey data on smokers would be able to provide insights into smoking and sale of tobacco products (see Viscusi and Hersch (2009)). Further, Alaska and Hawaii
habits of smokers. had to be dropped from the analysis because these states do not have geographically con-
5
All monetary variables are converted into real terms in the analysis. tiguous neighboring states.
286 R.K. Goel / Economic Modelling 40 (2014) 284–289

Table 1
Variable definitions, summary statistics and data sources.

Variable Definition Source


(Mean; std. dev.)

CONS Per-capita cigarette sales (in 20-packs) Orzechowski and Walker


(77.26; 29.45)
PRICE Retail price of cigarettes (cents/20-pack) Orzechowski and Walker
(346.52; 94.59)
BorPRICE Minimum cigarette price in a geographically-contiguous state (cents/20-pack) Orzechowski and Walker
(309.52; 73.13)
INCOME Per-capita disposable income ($) Bureau of Economic Analysis
(28,604.57; 6,185.90)
UNEMdev 3-Year moving standard deviation of unemployment rate (%), (0.633; 0.588) Statistical Abstract of the United States
UNEMma 3-Year moving average of unemployment rate (%), (4.782; 1.130) Statistical Abstract of the United States
PrpPRICEdev 3-Year moving standard deviation of single-family housing price index (first quarter 1991 = 100), (11.248; 9.383) Statistical Abstract of the United States
PrpPRICEma 3-Year moving average of single-family housing price index (first quarter 1991 = 100), (178.827; 46.149) Statistical Abstract of the United States
FedTax Dummy variable taking the value of 1 for years of federal cigarette tax changes (2000, 2002 and 2009), zero otherwise, (0.25; 0.43) Orzechowski and Walker
StateTax State cigarette tax (cents/20-pack), Orzechowski and Walker
(69.28; 57.66)
Producer Dummy variable identifying major tobacco growing states (GA, KY, NC, SC, TN, and VA), (0.12; 0.33)
CONS−1 One-year lagged per-capita cigarette sales (in 20-packs), (78.72; 29.38) Orzechowski and Walker

Notes: All monetary variables are converted in real terms in the analysis (denoted by prefix R) using the CPI (1982–1984 = 100). All observations are at the state level for the years 1998–
2009. Alaska and Hawaii are excluded due to an absence of border states in these cases.

(2009) and a low in Virginia (2000). Thus, the ranges of values of the un- fit of the various models estimated is quite decent as shown by the R2s
certainty variables are well dispersed across the span of our sample. and related statistics.
Appendix A provides corresponding information on the correlations.

3. Results 3.1. Baseline results

The estimation results are presented in Tables 2, 3, 4 and 5. Table 2 Two sets of baseline models, with and without border effects, are
includes baseline models, estimated using random-effects or fixed- presented in Table 2. The idea is to estimate basic cigarette demand
effects, depending upon the results of the Hausman test (reported models and to cross-check extant findings in the literature with more
towards the bottom of the tables). The fixed-effects model accounts recent data. The Hausman test recommended the fixed-effects model
for state-specific effects, while the random effects model allows individ- in column 2.1, while, due to the presence of a dummy variable
ual effects to vary. Consideration of these aspects allows one to control prohibiting the use of a fixed-effects model, random effects estimation
for state-specific nuances that might be impacting cigarette demand. results are reported in column 2.2.
The Hausman test allows one to choose which model to use, given the
model specification being considered (see Baltagi (2011)). The overall

Table 3
Economic stress and cigarette smoking: Effects of alternate measures of economic stress.
(Dependent variable: CONS).

Table 2 3.1 3.2


Economic stress and cigarette smoking: Baseline models. RPRICE −32.03** −27.20**
(Dependent variable: CONS). (14.1) (12.6)
RBorPRICE 13.63** 15.86**
2.1 2.2
(4.8) (6.2)
RPRICE −22.33** −32.46** RINCOME −0.30** −0.13**
(15.8) (14.3) (6.9) (3.0)
RBorPRICE 14.71** PrpPRICEdev −0.03
(5.4) (0.7)
RINCOME −0.33** −0.35** UNEMdev −1.78**
(9.1) (9.7) (2.6)
Producer 16.88* PrpPRICEma −0.10**
(1.8) (7.4)
UNEMma −3.10**
N 576 576 (7.3)
F-value/Wald χ2 445.7** 972.2** N 576 576
R2 0.30 0.38 F-value 195.4** 242.6**
Hausman test 3.56 R2 0.36 0.41
(p-Value of χ2) (0.17) Hausman test −88.0 −88.3
RE or FE FE RE (χ2)
RE or FE FE FE
Notes: See Table 1 for variable definitions. All monetary variables were converted in real
terms (denoted by prefix R) using the CPI (1982–1984 = 100). Constant included but Notes: See Table 1 for variable definitions. All monetary variables were converted in real
not reported. terms (denoted by prefix R) using the CPI (1982–1984 = 100). Constant included but
FE model could not be estimated in Model 2.2 due to the dichotomous nature of Producer. not reported.
t-statistics for FE (fixed effects) models and z-statistics for RE (random effects) models Asymptotic assumptions of Hausman test not satisfied and default FE model reported.
(absolute value) reported in parentheses, with ** and *, respectively, denoting statistical t-statistics (absolute value) reported in parentheses, with ** and *, respectively, denoting
significance at the 5% and 10% levels. statistical significance at the 5% and 10% levels.
R.K. Goel / Economic Modelling 40 (2014) 284–289 287

Consistent with the literature, own-price effects are negative and Table 5
significant, while border price effects are positive (see Chaloupka Economic stress and cigarette smoking: Consideration of habit-formation effects of
smoking.
and Warner (2000), U.S. Department of Health and Human Services (Dependent variable: CONS).
(2000); and Baltagi and Levin (1986) for a seminal related study).
While the (negative) coefficient on RPRICE in Model 2.1 is larger in 5.1 5.2

absolute value than that in Model 2.2, the coefficient estimates on RPRICE −24.94** −24.49**
the income variable in the two models are very similar. In terms of (8.2) (8.3)
RBorPRICE 20.28** 20.43**
magnitude, the price elasticity of demand is inelastic and relatively
(5.8) (6.2)
stable around − 0.5, while the income elasticity is − 0.9 (Model RINCOME −0.23** −0.17**
2.1). This finding is in line with elasticities for recent years (Gallet (3.9) (3.6)
and List (2003), Goel and Nelson (2008) for summaries of earlier es- PrpPRICEdev 0.0001
timates; also see Cebula et al. (2014)). While findings regarding the (0.0)
UNEMdev 0.35
effects of income are mixed in the literature (U.S. Department of
(0.5)
Health and Human Services, 2000), the negative income effects are PrpPRICEma −0.04**
consistent with the notion that more affluent smokers are either (2.5)
better aware of ill-health effects of smoking or are switching to sub- UNEMma −0.14
(0.2)
stances other than cigarettes. The border price effects are positive
CONS−1 0.53** 0.46**
(i.e., higher neighbor state price increase a state's cigarette sales). (9.4) (7.1)
Producer has the expected sign (i.e., greater cigarette sales in tobacco N 480 480
producing states, ceteris paribus), but lacks statistical significance at Wald χ2 1056.4** 1169.9**
the 5% level. Sargan test of overidentifying restrictions 151.1** 155.3**
(χ2)

Notes: See Table 1 for variable definitions. All monetary variables were converted in real
3.2. Economic stress and smoking terms (denoted by prefix R) using the CPI (1982−1984 = 100). Constant included but
not reported. Results from Arellano–Bond dynamic panel estimation are reported, with
the inclusion of a one-year lag of CONS as an additional regressor.
Table 3 adds the various measures of economic stress to Model z-statistics (absolute value) reported in parentheses, with ** and *, respectively, denoting
2.2, while dropping Producer, which was found to have statistically statistical significance at the 5% and 10% levels.
insignificant effects at the conventional levels of significance. The
Hausman test's asymptotic assumptions were not satisfied in both
Models 3.1 and 3.2, so we report results from fixed effects models
in Table 3.
The effects of economic stress, as captured by the deviation/average
percent increase in the 3-year moving average of the state unem-
in unemployment rates (UNEMdev/UNEMma) or the deviation/average
ployment rate would lower per-capita cigarette sales (consump-
in property prices (PrpPRICEdev/PrpPRICEma) are negative. A one
tion) by about three packs (Model 3.2). The relative magnitudes of
the stress measures in Model 3.1 with standard deviations are small-
er than the respective coefficients in Model 3.2 with moving aver-
ages. Of the four measures of economic uncertainty, only the effect
of PrpPRICEdev lacks statistical significance. Intuitively, greater eco-
Table 4 nomic stress or greater uncertainty might induce the shifting of
Economic stress and cigarette smoking: Robustness checks allowing for endogenous
some resources away from cigarettes (either by inducing cessation
cigarette price.
(Dependent variable: CONS).
or substitution to other products); dissuading some potential
smokers or greater stress could be psychologically altering the pro-
4.1 4.2 pensity to light up. The level of unemployment has previously been
RPRICE −46.72** −37.48** found to not significantly affect smoking (Goel, 2008), and now its
(12.8) (10.7) deviation is shown to have a negative impact. We have formally
RBorPRICE 27.53** 25.02**
demonstrated that economic stress matters in that it can be seen as
(7.0) (7.0)
RINCOME −0.22** −0.11** adding to the stress of smokers and potential smokers.
(4.7) (2.5)
PrpPRICEdev −0.03
(0.7)
3.3. Robustness checks
UNEMdev −2.12**
(3.0) To verify the robustness of the results in Table 3, we perform some
PrpPRICEma −0.09** checks, including (i) allowing for simultaneity between cigarette
(6.2)
consumption and cigarette prices; and (ii) considering the habit-
UNEMma −2.84**
(6.4) persistence effects of smoking.
N 576 576
Wald χ2 56,264.8** 67,441.3**
3.3.1. Allowing for possible endogeneity of cigarette prices
R2 0.41 0.44
First-stage F-value 624.3** 622.9** To test the robustness of the findings from Table 3, we account for
Hausman test −7.2 −5.7 the possible two-way causality between cigarette consumption and
(χ2) prices in Table 4. In other words, cigarette demand might impact cig-
RE or FE FE FE
arette prices. To that effect, two additional instruments ((i) state
Notes: See Table 1 for variable definitions. All monetary variables were converted in real cigarette tax rates (StateTax); and (ii) a dummy variable (FedTax)
terms (denoted by prefix R) using the CPI (1982–1984 = 100). Constant included but identifying the years of federal cigarette tax increases in the sample
not reported. RStateTax and FedTax were used as additional instruments for RPRICE.
Asymptotic assumptions of Hausman test not satisfied and default FE model reported.
(2000, 2002 and 2009)) are employed for real cigarette price
t-statistics (absolute value) reported in parentheses, with ** and *, respectively, denoting (RPRICE). These instruments make intuitive sense given the substan-
statistical significance at the 5% and 10% levels. tial proportion of taxes in cigarettes prices. The first-stage F-values,
288 R.K. Goel / Economic Modelling 40 (2014) 284–289

being statistically significant, lend formal support to our instrument effects of economic stress has largely been missing. The present research
choice by passing the weak instruments test.10 Since the asymptotic aims to somewhat “plug” this hole in the literature by examining the ef-
assumptions of the Hausman test are not satisfied (reported towards fects of economic stress on cigarette demand. To that effect, alternate
the bottom of the table), the fixed-effects results are reported measures of economic stress are employed in a “standard” cigarette
Models 4.1 and 4.2. demand specification — (i) a three-year moving standard deviation of
The results regarding the effects of economic stress support the unemployment; (ii) a three-year moving standard deviation of housing
findings from Table 3 — all related coefficients are negative and, ex- prices; (iii) a three-year moving average of unemployment; and (iv) a
cept for the coefficient on PrpPRICEdev, they are statistically signif- three-year moving average of housing prices, to capture the degree of
icant. Greater economic stress results in a lowering of cigarette economic stress or uncertainty. Similar measures of economic uncer-
demand, due likely to some smokers shifting funds to other needs tainty have been employed in other contexts in cross-national studies
by reducing smoking and some potential smokers not initiating (Pindyck and Solimano, 1993).
smoking. The effects of cigarette price, border price and income The findings are quite robust in that we consistently find that greater
on cigarette demand are quite similar to what was reported in economic stress tends to reduce cigarette demand (Tables 3 and 4).
Table 3. Nearly all measures of economic stress show negative influences on
smoking, and most are statistically significant. This finding supports
3.3.2. Considering habit-persistence in smoking the notion that in times of greater economic uncertainty, cigarette
Given the habit-forming nature of cigarettes, we account for this demand goes down as some smokers reduce consumption (either by
aspect by including the lagged dependent variables (i.e., cigarette con- smoking less or by switching to substitute tobacco products) and
sumption) as an additional regressor (see Baltagi and Levin (1986); some potential smokers are dissuaded. In the context of the uncertain-
also Johdo (2009) for a broader discussion).11 Since the lagged depen- ty–investment linkage (see Dixit and Pindyck (1994)), smokers and
dent variable is likely to be correlated with the errors, the dynamic potential smokers can be viewed as reluctant to invest resources in
panel estimation due to Arelleno–Bond is employed in this context smoking in the face of greater uncertainty. This finding appears to be
and the related results reported in Table 5 (for details, see Arellano novel in the related literature. Some scholars have found a pro-cyclical
and Bond (1991)). The fit of both models in the table is decent as association between smoking and business cycles (see Ruhm (2000,
shown by the statistically significant Wald's chi-square. 2005), Xu and Kaestner (2010)), while we are able to address a relat-
The coefficient on the lagged dependent variable (CONS − 1 ) is ed,12 but separate, dimension of uncertainty and find that greater uncer-
positive and significant, with a magnitude of 0.5 in each model. tainty reduces smoking.
This finding supports earlier results in the literature (see Baltagi The other results largely support the extant literature — price
and Levin (1986); and Goel and Nelson (2008) for a review). This effects are negative, border-price effects are positive and the effect
implies that cigarettes are indeed habit-forming and that early inter- of income is negative (for literature reviews see Chaloupka and
ventions to check smoking among the youth and to prevent smoking Warner (2000), Goel and Nelson (2008) and U.S. Department of
initiation would be recommended. Health and Human Services (2000)). In terms of magnitude, a ten
In regard to the effects of economic stress, Table 5 shows a difference percent increase in the (real) cigarette price leads to a five percent
between the standard deviation-based measures (Model 5.1) and mov- decline in cigarette demand, while a similar increase in (real per
ing average-based measure (Model 5.2). In particular, the two measures capita) income would lead to a nine percent decline in cigarette
in Model 5.2 are negative, with PrpPRICEma also statistically significant. sales. Finally, we find evidence to support the habit-forming effects
On the other hand, the standard-deviation-based measures in Model 5.1 of cigarettes (Table 5).
fail to attain statistical significance. In terms of other results, the previ- In the context of policy implications of the findings, a few new in-
ous findings are supported — i.e., negative (and significant) own-price sights emerge. First, as macroeconomic policies are targeting the
and income effects, and positive effects of border prices. The concluding smoothening of economic upheavals, especially in recent years,
section follows. such policies might inadvertently be increasing cigarette smoking.
While we cannot argue against the prudence of such initiatives, law-
4. Concluding remarks makers would be better off at least recognizing the spillover effects
on health in terms of increases in smoking. Second, the inelastic cig-
This paper uses pooled data on U.S. states over the most recent de- arette demand maintains the tax revenue-generating potential of
cade to estimate the demand for cigarettes. While numerous studies cigarette excise taxes, while at the same time limiting the cessation
have examined various aspects of cigarette demand (see Chaloupka potential (see Lanoie and Leclair (1998)). Depending on which side
and Warner (2000), Goel and Nelson (2008) and U.S. Department of of the fence a particular lawmaker is sitting, this might be good or
Health and Human Services (2000) for literature reviews), the effects bad news. Third, the habit-forming effects of cigarettes recommend
of economic stress (or economic uncertainty) have been largely policies geared towards preventing smoking initiation and curbing
ignored. Economic stress can directly and indirectly affect smokers' smoking among the youth. The other policy implications are not
propensity to purchase cigarettes and smoke, while at the same time new, but nevertheless noteworthy — cross-state cigarette price effects
it can also affect the behavior of potential smokers to initiate smoking. argue for excise tax policy coordination and other smuggling-control
Whereas the broader cross-disciplinary literature has recognized the initiatives across states. Finally, this line of research can benefit from
stress–smoking linkage (Pomerleau and Pomerleau, 1991; Siahpush additional insights from micro-level data on individual smokers.13
and Carlin, 2006; Tyas and Pederson, 1998), formal research on the However, until appropriate corresponding information is available, we
would have to contend with analyses based on some level of
aggregation.

10
However, the Sargan overidentification test was satisfied only for Model 4.2 (i.e., the
corresponding chi-square was 13.3 (and statistically significant) in Model 4.1, but 1.4
(and statistically insignificant) in Model 4.2). Alternate combinations of instruments for
12
Model 4.1, including Producer and a dummy variable identifying geographic smoking re- For instance, it is possible for economic uncertainty (i.e., variations in unemployment
strictions, did not yield significant improvements. The reason for the differences in the and/or property prices in our case) to be large even in periods of economic upturns.
13
two cases is that the instruments were differently correlated with the moving-average For instance, in a related aspect, the likelihood of smoking cessation was found to be
based and the standard-deviation based measures of economic stress. lower in individuals facing greater financial stress (Siahpush and Carlin (2006); see also,
11
In our sample, the correlation between CONS and CONS−1 was 0.97. Barnes and Smith (2009) and Jones (1994)).
R.K. Goel / Economic Modelling 40 (2014) 284–289 289

Appendix A. Correlation matrix

CONS PRICE BorPRICE INCOME PrpPRICEdev PrpPRICEma UNEMdev UNEMma

CONS 1.00
PRICE −0.58 1.00
BorPRICE −0.34 0.85 1.00
INCOME −0.41 0.77 0.75 1.00
PrpPRICEdev −0.29 0.32 0.34 0.29 1.00
PrpPRICEma −0.42 0.65 0.64 0.61 0.42 1.00
UNEMdev −0.18 0.25 0.24 0.37 0.16 0.19 1.00
UNEMma −0.19 0.31 0.24 0.15 0.09 0.12 0.46 1.00

Notes: See Table 1 for variable definitions. N = 576.

References Johdo, W., 2009. Habit persistence and stagnation. Econ. Model. 26, 1110–1114.
Jones, A.M., 1994. Health, addiction, social interaction and the decision to quit smoking.
Advisory Commission on Intergovernmental Relations (ACIR), 1985. Cigarette Tax
J. Health Econ. 13, 93–110.
Evasion: A Second Look. Advisory Commission on Intergovernmental Relations,
Joossens, L., Chaloupka, F.J., Merriman, D., Yurekli, A., 2000. Issues in the smuggling of
Washington, DC.
tobacco products. In: Jha, P., Chaloupka, F. (Eds.), Tobacco Control in Developing
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo
Countries. Oxford University Press, Oxford, pp. 393–406.
evidence and an application to employment equations. Rev. Econ. Stud. 58, 277–297.
Lanoie, P., Leclair, P., 1998. Taxation or regulation: looking for a good anti-smoking policy.
Baltagi, B.H., 2011. Econometrics, Fifth ed. Springer, Heidelberg.
Econ. Lett. 58, 85–89.
Baltagi, B.H., Levin, D., 1986. Estimating dynamic demand for cigarettes using panel data:
Lovenheim, M.F., 2008. How far to the border?: the extent and impact of cross-border
the effects of bootlegging, taxation and advertising reconsidered. Rev. Econ. Stat. 68,
casual cigarette smuggling. Natl. Tax J. 6, 7–33.
148–155.
Orzechowski, Walker, 2012. The Tax Burden on Tobacco: Historical Compilation.
Barnes, M.G., Smith, T.G., 2009. Tobacco use as response to economic insecurity: evidence
Orzechowski and Walker, Arlington, VA.
from the national longitudinal survey of youth. B.E. J. Econ. Anal. Policy 9 (Article 47).
Pindyck, R.S., Solimano, A., 1993. Economic instability and aggregate investment. In:
Braun, M., Di Tella, R., 2004. Inflation, inflation variability, and corruption. Econ. Polit. 16,
Blanchard, O.J., Fischer, S. (Eds.), NBER Macroeconomics Annual 1993. MIT Press,
77–100.
Cambridge, MA, pp. 259–317.
Cebula, R.J., Foley, M., Houmes, R., 2014. Empirical analysis of the impact of cigarette
Pomerleau, O.F., Pomerleau, C.S., 1991. Research on stress and smoking: progress and
excise taxes on cigarette consumption: estimates from recent state level data. J.
problems. Addiction 86, 599–603.
Econ. Financ. 38, 164–180.
Ruhm, C.J., 2000. Are recessions good for your health? Q. J. Econ. 115, 617–650.
Centers for Disease Control and Prevention (CDC), 2010. Tobacco Control State Highlights
Ruhm, C.J., 2005. Healthy living in hard times. J. Health Econ. 24, 341–363.
2010. U.S. Department of Health and Human Services, Atlanta.
Siahpush, M., Carlin, J.B., 2006. Financial stress, smoking cessation and relapse: results
Chaloupka, F.J., Warner, K.E., 2000. The economics of smoking. In: Culyer, A.J., Newhouse,
from a prospective sample of an Australian national sample. Addiction 101, 121–127.
J.P. (Eds.), Handbook of Health Economics, vol. 1. North-Holland, Amsterdam, pp.
Tyas, S.L., Pederson, L.L., 1998. Psychosocial factors related to adolescent smoking: a
1539–1627.
critical review of the literature. Tob. Control. 7, 409–420.
Dixit, A.K., Pindyck, R.S., 1994. Investment under Uncertainty. Princeton University Press,
U.S. Department of Health and Human Services, 2000. Reducing Tobacco Use: A Report of
Princeton, NJ.
the Surgeon General. Centers for Disease Control and Prevention, Atlanta, GA.
Gallet, C.A., List, J.A., 2003. Cigarette demand: a meta-analysis of elasticities. Health Econ.
Viscusi, W. Kip, 2000. Smoking risk policy: a special issue of the journal of risk and uncer-
12, 821–835.
tainty. J. Risk Uncertain. 21, 159–160.
Goel, R.K., 2008. Unemployment, insurance and smoking. Appl. Econ. 40, 2593–2599.
Viscusi, W. Kip, Hersch, J., 2009. Tobacco regulation through litigation: the Master Settle-
Goel, R.K., Nelson, M.A., 2008. Global Efforts to Combat Smoking. Ashgate Publishing,
ment Agreement. National Bureau of Economic Research, NBER Working Paper No.
Aldershot, UK.
15422, October 2009.
Goel, R.K., Ram, R., 2001. Irreversibility of R&D investment and the adverse effect of
Xu, X., Kaestner, R., 2010. The business cycle and health behaviors. National Bureau of
uncertainty: evidence from the OECD countries. Econ. Lett. 71, 287–291.
Economic Research, Working Paper # 15737.
Goel, R.K., Ram, R., 2013. Economic uncertainty and corruption: evidence from a large
cross-country data set. Appl. Econ. 45, 3462–3468.

Vous aimerez peut-être aussi