Journal of Health Economics 1 (1982) 121-145. North-Holland Publishing Company THE POTENTIAL FOR USING EXCISE TAXES TO REDUCE SMOKING* Eugene M. LEWIT UMDNJ - New Jersey Medical School, Newark, NJ 07103, USA National Bureau of Economic Research, Cambridge, MA 02138, USA Douglas COATE Rutgers University, Newark, NJ 07102, USA National Bureau of Economic Research, Cambridge, MA 02138, USA Received August 1981, final version received February 1982 We examine the potential for reducing cigarette smoking through increases in cigarette excise taxes by estimating the price elasticity of demand for cigarettes. Using information on individual smoking behavior from the 1976 Health Interview Survey, we estimate the adult price elasticity of demand for cigarettes to be -0.42. We fmd that price has its greatest effect on the smoking behavior of young males and that it operates primarily OQthe decision to smoke rather than via adjustments in the quantity of cigarettes smoked. An excise tax increase would discourage smoking by successive cohorts of young adults, 2nd those reduced smoking levels would be reflected in aggregate smoking as these cohorts mature. 1. Introduction In the last decade, soaring health care expenditures in the U.S. have been associated with only small improvements in health levels. AS a result, many health care observers have concluded that the primary potential for improving health and moderating the growth in health care costs lies in preventive medicine and in encouraging individuals to alter unhealthy behaviors [Fuchs (1974), Zook, Moore and Zeckhauser (1981)]. Cigarette smoking, regarded for over 25 years as a significant contributor to poor health, is perh;lFs the best example of an unhealthy behavior leading to *An earlier version of this paper was presented at an AEA/HERO Session at the AS% Annual Meetings in Denver, CO, Sept. 1980. Research for this paper was supported by grant no. HS-03738 from the National Center for Health Services Research to the UMDNJ - New Jersey Medical School. We have benefited from the helpful comments of Michael Grossman, Marvin Kristein, Alan Monheit, and Robert Shakotko. Data was graciously provided by June Sears of the Tobacco Tax Council, Robert Miller of thr: Department of Agriculture, and Robert Fuchsberg of the National Center for Health Statistics. Able research assistance was provided by Harry Barrett, Anne Colle, and Brian Woolley. None of the above mentioned individuals share any ‘responsibility for the views expressed in this paper. In addition, this is not an oficial National Bureau of Economic Research publication, and any opinions expressed are not those of the National Bureau of Economic Research. 0167-629682 $02.75 0 1982 North-Holland
Transcript
Journal of Health Economics 1 (1982) 121-145. North-Holland Publishing Company
THE POTENTIAL FOR USING EXCISE TAXES TO REDUCE SMOKING*
Eugene M. LEWIT UMDNJ - New Jersey Medical School, Newark, NJ 07103, USA
National Bureau of Economic Research, Cambridge, MA 02138, USA
Douglas COATE Rutgers University, Newark, NJ 07102, USA
National Bureau of Economic Research, Cambridge, MA 02138, USA
Received August 1981, final version received February 1982
We examine the potential for reducing cigarette smoking through increases in cigarette excise taxes by estimating the price elasticity of demand for cigarettes. Using information on individual smoking behavior from the 1976 Health Interview Survey, we estimate the adult price elasticity of demand for cigarettes to be -0.42. We fmd that price has its greatest effect on the smoking behavior of young males and that it operates primarily OQ the decision to smoke rather than via adjustments in the quantity of cigarettes smoked. An excise tax increase would discourage smoking by successive cohorts of young adults, 2nd those reduced smoking levels would be reflected in aggregate smoking as these cohorts mature.
1. Introduction
In the last decade, soaring health care expenditures in the U.S. have been associated with only small improvements in health levels. AS a result, many health care observers have concluded that the primary potential for improving health and moderating the growth in health care costs lies in preventive medicine and in encouraging individuals to alter unhealthy behaviors [Fuchs (1974), Zook, Moore and Zeckhauser (1981)]. Cigarette smoking, regarded for over 25 years as a significant contributor to poor health, is perh;lFs the best example of an unhealthy behavior leading to
*An earlier version of this paper was presented at an AEA/HERO Session at the AS% Annual Meetings in Denver, CO, Sept. 1980. Research for this paper was supported by grant no. HS-03738 from the National Center for Health Services Research to the UMDNJ - New Jersey Medical School. We have benefited from the helpful comments of Michael Grossman, Marvin Kristein, Alan Monheit, and Robert Shakotko. Data was graciously provided by June Sears of the Tobacco Tax Council, Robert Miller of thr: Department of Agriculture, and Robert Fuchsberg of the National Center for Health Statistics. Able research assistance was provided by Harry Barrett, Anne Colle, and Brian Woolley. None of the above mentioned individuals share any ‘responsibility for the views expressed in this paper. In addition, this is not an oficial National Bureau of Economic Research publication, and any opinions expressed are not those of the National Bureau of Economic Research.
0167-629682 $02.75 0 1982 North-Holland
122 E.M. Lewit aul D. Coate, Using excise taxes to reduce smoking
substantial health care costs. The direct (health-care) and indirect costs of smoking were estimated to have been nearly $30 billion in 1976 [Lute and Schweitzer (1978)j.
Public and private sector initiatives to discourage smoking have included the dissemination of information from the 1964 Surgeon General’s Report on smoking and health, anti-smoking television advertising tied to cigarette commercials (the Fairness Doctrine), a ban on all cigarette advertising on radio and television, and the labelling of cigarettes with warning messages. Warner (1977) presents evidence that these policies may have had a substantial effect on smoking and estimates that 1975 per capita cigarette consumption would have been 20 to 30 percent higher in the absence of these anti-smoking policies. 1 In fact, the quantity of tobacco consumed per adult in the United States is now at its lowest level of the century because of a 27% decline in the number of adult smokers since 1965 and an increase in the market share of low tar cigarettes. It is also true, however, that the rate of decline in cigarette consumption has been quite slow with consumption per adult down only 3% in the past decade.
In this paper, we examine the potential for reducing cigarette smoking through increases in excise taxes, a public policy that could lead to permanent reductions in smoking. During the past three decades, the federal excise tax has not been used as a policy tool to discourage smoking despite the large and growing federal share of health care expenditures and the large number of studies linking cigarette smoking and poor health.2 In fact, the federal policy of holding the excise tax constant at eight cents per pack since 1952 translates into a substantial reduction in the tax in real terms. There is evidence, however, that state and local governments have used their cigarette excise taxes to discourage smoking. The considerable anti-smoking publicity associated with the Surgeon General’s Report in 1964 was followed by 23 state and local tax increases compared with no more than a dozen in any of the preceding 14 years [Warner (1977)]. State and local taxes have continued to increase over time in many states; however, the ability of state and local governments to raise their own cigarette taxes is limited because of the presence of cigarette bootlegging from low to high tax areas [Inter- governmental Perspective ( 1978)].3
‘The effects of these policies are also discussed in Hamilton (1972), Ippolito, Murphy and Sant (1979), Klein, Murphy and Schneider (1981), Lewit, Coate and Grossman (1981), and Warner (1981). Not all of these studies agree on the relative impact of different government policies on smoking behavior.
‘There have been several attempts to increase the federal exicse tax in recent years because of concern over the health effects of cigarette smoking [Miller (1976)].
3The economies of several states depend importantly on the growing of tobacco and the production of cigarettes. At the state level, these states maintain very low exise taxes. At the federal level, the economic interests of these states and the tobacco industry have apparently succeeded in blocking an; increases in the federal tax since 1952, despite the Surgeon General’s Report in 1964 and the anti-smoking campaign waged by government and voluntary groups since then.
EM. L+ewit and D. Coate, Using excise taxes to reduce smoking 123
2. Previous estimates of the price elasticity of demand for cigarettes
The impact of excise tax changes on cigarette demand depends on the extent to which changes in excise taxes are reflected in cigarette prices4 and on the responsiveness of cigarette demand to price. The literature shows a broad range of cigarette price elasticity estimates. Studies completed since 1970 using U.S. data have yielded estimates ranging from -0.4 to - 1.3 [Miller (1970), Mann (1971), Padillo (1971), Hamilton (1972), Schrabel (1972), Kellner (1973), Miller (1974), Fujii (1980), Klein, Murphy and Schneider (1981)]. These estimated elasticities are generally high enough to suggest that excise rates can have a substantial impact on cigarette consumption. For example, Warner (1977) borrows from these findings to attribute a substantial portion of the decline in cigarette consumption which took place between 1963 and 1972 to the significant increase in cigarette prices during the period. There is reason to believe, however, that such conclusions may not be reliable guides for policy-makers with an interest in reducing health care costs through excise tax induced reductions in cigarette smoking. This is true for two reasons: (1) the cross-section estimates are biased upwards because cigarette consumption is inaccurately measured, and (2) the time series estimates are not stable because of the high correlation between cigarette price, income, and trend variables, and furthermore, reflect a short- run response to changes in price rather than the long-run or permanent response which is of interest to policy-makers.
Cross-section studies are generally relied on to provide price elasticity estimates because multicollinearity among the independent variables is usually less of a problem than in time series estimation and because the estimates are considered in most cases to represent the long-run or complete response of quantity demanded to changes in price. Cross-section studies of cigarette demand, however, have generally provided elasticity estimates that are biased upward in absolute value because the unit of observation has been the state and the dependent variable has been tax-paid cigarette sales per capita. For many states, tax-paid sales do not reflect actual consumption. This disparity results from the fact that excise taxes vary substantially between states while state cigarette markets are not completely distinct or separable. Therefore, smuggling or bootlegging of cigarettes from low tax to high tax states occurs and as a result, tax-paid sales are a biased measure of consumption. Using sales data causes the elasticity of demand to be biased because in high-tax (price) states own consumption is underestimated by sales and in low-tax (price) states own consumption is overestimated. Thus, the response of cigarette demand to price is exaggerated
41f cigarette supply is perfectly elastic, excise tax changes will be fully rekcted in cigarette prices. This would appear to be a reasonable approximation for most moiest tax changes. in fact, Barzel (1976) presents evidence that cigarette prices have risen by more than the amount of state unit tax increases.
124 E.M. Wit and D. Coate, Using excise taxes to reduce smoking
In this paper, we eliminate this bias by using information on individual smoking behavior to estimate the price elasticity of demand for cigarettes. The use of individual data also allows us to examine the effect of price on both the decision to smoke and the quantity of cigarettes smoked. The data set we employ is the 1976 Health Interview Survey (HIS) which contains self- reported information on the smoking behavior of a large sample of individuals in different tax (price) locations.5 In addition, the data contains information on an array of individclal and household characteristics suff:cient to allow for the estimation of a well specified demand relationship.
3. Specification of the demand for cigarettes equation
In specifying the demand function that we estimate in this paper, we assume that the demand for cigarettes is a linear function of price, income and taste variables where health and other considerations can enter the relationship via the taste component. Specifically, the function is
The dependent variable in the equation can either be the amount smoked by the ith individual in the jth locality or a dichotomous -*ariable indicating whether the individual is a smoker. The independent variables include the ‘average’ price (Pi) of cigarettes in the jth locality; a vector (Xij) of individual and household characteristics including family income, family size, education, age, sex, martial status, health status, and race; a vector (RJ) of region and city size characteristics; and a random disturbance team (Q). Some of the X, variables are measures of an individual’s command over resources and others are proxies for ‘taste’ variables. Almost all of them have been shown in previous studies to be related to differences in the propensity to smoke cigarettes lJ%tional Clearinghouse for Smoking and Health (197611. The
:gion of residence and size of place of residence variables are included to partially control for cross-sectional differences in the cost of living which are not otherwise reflected in our price or income measures.
We estimate the demand for cigarettes equation over all individuals (smokers and non-smokers) as well as over smokers only. Implicit in our single-equation approach is the assumption that variation in the supply curves across sample locations due to differences in excise tax rates identifies the demand curves we specify.
3.1. Average cigarette prices
Differences in state and local cigarette tax rates in the U.S. are
%dividuals may underreport their consumption of cigarettes substantially in sxveys [Warner (197811, but we argue below that the effect of this underreporting on estimated price elasticities is likely to be negligible.
E.M. L.&t and D. Coate, Using excise taxes to reduce smoking 125
substantial and account for almost all the variation in the market price of
cigarettes. In 1976, state taxes equalled 247; of the retail price of cigarettes on average. Because of the state tax differences, the ‘average’ price of cigarettes in 1976 was 57.3# a pack in Massachusetts and 36.6# a pack in North Carolina. In addition, local taxes can substantially increase cigarette prices in certain markets. The most notable of these in 1976 were in New York City (8g! per package) and in Chicago (56 per package).
Information on retail cigarette prices, excise tax rates, sales taxes and a composite cigarette price is available from the Tobacco Tax Council (1980). The Council calculates an ‘average’ retail price per state by taking a weighted average of reported retail prices plus applicable sales taxes of cigarettes sold by carton lot, by the single pack over-the-counter, and by the single pack through vending machines. The weights are the national proportions of cigarettes sold in these ways.6 This composite retail price is the basis for the price measure used in this study. In order to determine the prrvailing cigarette price for each observation in the HIS, we located each of the 430 Primary Sampling Units (PSU’s) in the HIS on a map. The composite cigarette price for each state was assigned to each PSU depending on state of location. If there was any local cigarette tax applicable to the site, the local tax rate was added to the composite retail price. When there was more than one price attributable to a particular PSU, prevailing price was determined by using population weights of the various geographic components within the PSU to calculate an average composite price for the PSU.
3.2. Cigarette price differentials and bootlegging
Because of differences in state and local excise taxes on cigarettes, substantial price differentials exist across geographic boundaries. Since cigarettes are relatively easy to transport across these boundaries, they are not infrequently purchased in low-tax areas for resale or personal consumption in high-tax areas, Much of this ‘buttlegging’ activity has taken the form of large-scale smuggling for resale which has become an important law enforcement and tax collection problem in certain high-tax states. There is reason to believe that many cigarettes processed in this manner are sold
‘In all sates, the unit price of cigarettes is lowest when they are purchased in carton lots while substantial mark-ups are associated with purchasing cigarettes by the single pack. In 1976, 56% of all ciguettes were purchased in carton lots while 29% of sales were of over-the-counter single packs and 15% single packs in vending machines. Because the price of cigarettes within a locality may vary according to the way in which they are purchased, the use in a demand equation of the weighted average price v.ariable as a measure of the price faced by actual and potential smokers is preferred to using the price actually paid by smokers. This IS because heavier smokers have a greater incentive to purchase cigarettes cheaply. To the extent that they economize on the purchase price of cigarettes, the price paid is a function of the quantity demanded and the coefYicient of an actual price paid variable in a demand equation would be biased due to the reverse causality.
126 E.M. L,ewit and D. Coate, Using excise taxes to reduce smoking
through regular distribution channels at prices approximating the fully taxed retail level [Advisory Commission on Intergovernmental Relations (1977)]. This is because large-scale smuggling requires distribution through a large network of retail dealers, and the sale of these cigarettes at substantial discounts from fully taxed retail price would greatly facilitate their detection by law enforcement oficials. 7 To the extent that cigarettes smuggled on a large scale are sold at fully taxed retail prices, individuals in a PSU where large-scale smuggling exists can still be assumed +n *al’ +ha ~~~v~iling retail iv y”J CIlv yrv U... price for cigarettes as reported by the Tobacco Tax Council (1980). Accordingly, large-scale smuggling should not bias estimates of the price elasticity of demand for cigarettes obtained from our methodology.
There is also evidence of small-scale smuggling for resale at perhaps less than full retail price, but because, for the majority of smokers, the transactions costs assodated with relying on small-scale smugglers are likely to be high, it is unlikely that the presence of small-scale smuggling substantially biases our estimates of the average price of cigarettes.
A potential problem arises, however, in the case of individuals who reside in areas bordered by lower-price (tax) areas. In these cases, smokers and would-be smokers have the opportunity to purchase cigarettes at less than their own area retail price if they are willing to travel for that purpose or, perhaps more importantly, if they travel into the lower-price areas for other reasons. In many areas of the United States, the substantial amount of across excise-tax boundary commuting for work and recreation affords ample opportunity for the ‘incidental’ purchase of cigarettes in adjacent low-price areas by many residents of high-price .areas and large differences in per capita tax paid cigarette sales along many of these boundaries suggests this practice is not uncommon.*
Generally then, the ‘average’ price of cigarettes reported by the Tobacco Tax Council will overstate the actual cigarette prices faced by individuals in high price areas which border lower p&e areas. III these cases, the use of this ‘average’ price variable can lead to biased estimates of the price elasticity of demand for cigarettes. To correct for this problem, a procedure was
‘In addition, cmecdotal evidence suggests that largescale smuggling has become dominated by ‘organized tie*. This would tend to discourage the price competition among suppliers of bootlegged tigaret!es necessary to cause a decline in the price of smuggled cigarettes [Advisory Commission on Intergovernmental Relations (1977)].
*For e;nampie, in !978 per capita tax paid cigarette sales were 500/, lower in New York City which imposed a local $$ per pack cigarette tax than in the rest of New York State. Similarly, annual tax paid cigarette sales in New Hampshire were 278.8 packs per capita in 1977, while in neighboring Massachusetts the comparable figure was 118.9. State cigarette taxes were 9# a pack higher in Massachusetts than in New Hampshire in that year. If we assume that the cigarettes purchased in New Hampshire beyond the national average of 133.6 packs per capita were consumed in Massachusetts, per capita ‘consumption’ in Massachusetts would rise to 139 or approximately to the national average. This would imply that as much 2s 15% of the cigarettes smoked in Massachusetts might have been purchased in New Hampshire.
EM. Lavit and D. Coate, Using excise taxes to reduce smoking 127
developed to identify and eliminate from our sample observations in PSU’s where because of the possible existence of ‘incidental’ bootlegging average prices reported might not accurately reflect prices faced by consumers. A 20 mile wide band was drawn completely around each PSU and the prevailing retail price of cigarettes within this band determined for each PSU.’ A ‘restricted’ sample of the HIS data set was then obtained by deleting from the full sample individuals in PSU’s where the own average price was greater than the price within the 20 mile band. This restricted sample should then consist almost solely of individuals who face cigarette prices equal to those oT their own state and accurately represented by the Tobacco Tax Council price series. In this paper, demand estimates are presented for the full HIS sample and the restricted sample.
4. Data
The 1976 I-Ieaith Interview Survey (HIS) was a nationwide survey which collected data weekly by household interview for the purpose of determining the health status of the U.S. civilian non-institutionalized population. The 1976 HIS sample is comprised of 28,033 individuals between the ages of 20 and 74 from 430 survey sites (PSU’s) nationwide. The survey population is representative of the population of the United States. All the variables included in the analysis, with the exception of cigarette prices, are reported on the public use data tapes purchased from the National Center for Health Statistics (NCHS). Average cigarette prices were calculated for each PSU in the HIS based on data from the Tobacco Tax Council (1980) and placed on the data tape under an arrangement with NCHS which preserved the confidentiality of the respondents in the Survey. Arter editing, the data set contained 19,266 observations with information on the smoking behavior of individuals 20-74 years of age. lo The edited data set contained 416 PSU’s in the entire sarrplc and 242 PSU’s in the restricted samp!e.
5. Results
Definitions, means, and standard deviations of the three dependent variables and the independent variables are presented in table 1. The summary statistics are presented for the entire HIS sample and for the restricted sample. Inspection of the sample means reveals that the two
9As noted previously, the purchase of cigarettes in adjacent areas at lower prices is largely a function of travel costs. Twenty miles was arbitrarily chosen as the distance beyond which price differentials would not substantially affect cigarette purchases for own consumption.
r”Editing consisted primarily of eliminating observations for which little information on smoking behavior was available. In addition, the 1976 HIS included a small number of observations in PSU’s rhat NCHS could not identify. Since we could not calculate appropriate average prices for these observations, they were eliminated from our working sample.
Tabl
e 1
Des
crip
tions
, m
eans
and
sta
ndar
d de
viat
ions
of
depe
nden
t an
d in
depe
nden
t va
riab
les
for
the
entir
e ed
ited
1976
HIS
sam
ple
and
the
rest
rict
ed sa
mpl
e.
Var
iabl
es
Des
crip
tion
Mea
ns (s
tand
ard
devi
atio
ns)
Entir
e sa
mpl
e R
estr
icte
d sa
mpl
e
(a) S
mok
ers &
(b)
Sm
oker
s (c
) Sm
oker
s & (
d) S
mok
ers
non-
smok
ers
only
no
n-sm
oker
s on
ly
Dep
ende
nt v
aria
bles
C
IGD
AY
SMO
KER
CIG
SMO
KER
Inde
pend
ent
vari
able
s PR
ICE
INC
OM
E
Num
ber
of c
igar
ette
s sm
oked
pe
r da
y by
sm
oker
s and
no
n-sm
oker
s. D
umm
y va
riab
le th
at is
equ
al
to 1
whe
n th
e in
divi
dual
is
a sm
oker
. N
umbe
r of
cig
aret
tes s
mok
ed
per
day
by s
mok
ers.
Com
posit
e pr
ice i
n ce
nts
per
pack
of c
igar
ette
s a c
onsu
mer
fa
ces i
n hi
s im
med
iate
are
a.
Fam
ily in
com
e (m
easu
red
cont
inou
sly
in th
ousa
nds
of d
olla
rs) c
ompu
ted
by a
ssig
ning
SSO
O to t
he lo
wes
t re
port
ed in
terv
al, $
40,0
00 to
the
hi
ghes
t re
port
ed in
terv
al, a
nd
mid
poin
ts t
o th
e fo
llow
ing
clos
ed
inco
me
inte
rval
s:
$l,O
OO
- 1,9
99
$6,0
00-
6,99
9 2J
KK
L a9
99
7,O
iW 9
,999
3,
OcK
L 3,9
99
10,0
0&14
,999
4,
OtW
4,9
99
15,0
0&24
,999
5*
000-
5,9
99
7.40
1 (1
2.13
0)
0.36
7 (0
.482
)
- 49.5
50
(5.4
74)
15.3
30
(11.
276)
- 7.
264
(11.
927)
- 0.
364
(0.4
8 1)
20.1
44
- (1
1.99
2)
49,4
80
47.9
84
(5.5
93)
(5.2
05)
15.2
38
15.0
84
(11.
015)
(1
1.15
1)
- - 19.9
40
(11.
736)
47.7
72
(5.4
19)
14.9
60
(10.
856)
i3
W
(OLI’O) O
EUO
(18TO) L80’0
kn4’0) 020’0
(IEZ’0) LSO’O
(Z81’0) i4Xl.O
(OLZ’0) 6LOW
&sro~ szuo
(Ia-01 m-0
a3&zlfYvd3s
“papmu,
R~lU
allna Ey ssi?p pquI0
SQ
%yad~I
‘,pawm
das, JO J
XD
IOA
!~,
‘6pa
!J18m
Ja
Aa
u, ‘J
WM
Op
~,
S8 sn
lels sv
od
aJ
a
euy
u8x
a
ua
qm
&IO
1 p
mb
a
#trql
Sa
jq8!l8A
d
Um
rnp
Sn
J8J
S 189.l8y
U
-al8IIK
iJ S!
aa
u~
w8x
a
ua
qA
i 1 01 pm
ba
SE
18ql a
[q8!l8A
h
uIn
a
sled
j?L59 sJ
8ax
WE
sred
j7g-s~
m
a4
p~
_sz
SJ
8d
K-9
SJ
8aA
PZ
-OZ
:s~8AJqu!
%uplo~oJ a
ql
05 sm~
od
py
p
u8
pm
alu
!
Isap
a
ql
al
smd
S
L %
u#?p
w8
Lq
palndum
smL
u! a%y
a3moA
Ia
2l32 TN
(13MO
(IIM
37YW3d
3w
(LIE-O) EII’O
(SZE ‘0) O
ZI’O
(srtd (ZLZ’0)
8po’O
080-O
(EZE’O)
61 I-0
kzz’o) zsu0
(6ZE’0) EZI’O
(8L2’01 p80’0
C66VO
) SLVO
(86V0)
9ps’0 klos01 f 8V
O
(86V0)
6PSO
-sa1en
pel%
Io
oq
3s q
sy
wy
l sS
a1 ln
q
uo
pm
np
a
18~0~
au
tos
qlp
h
saa
ym
xa
s! ss8p
pa
ll+u
o
!&ja
g)
uo
!wn
pa
c&la~~o3+md
amos saq
aaupxa u
oy
pp
8 u
!: ua
qm
1 len
ba
p
u8
fb7oDaa)
aiq@
a@p
8 s! aa
uy
mx
a
aq
l u
oy
pp
8
u! u
ar(M
I pnb X
7oxa3) aSap
atu
os
sey
aa
u-a
aq
l u
oy
pp
e u
! ua
qk
A I
pn
ba
UsH
a m
vd
rows
*YI a
Quo s! a
au
yu
2xa
u
aq
m
I p
nb
a
Qiw
oNa3)
uo!wnpa r-0~
ou seq aa
uy
xa
n
aq
m
1 @x
3
pu
8 6~
pz+
oZ
?ai83
uo
yn
pa
am
seaw
18q
1 saIq
8peA
b
um
ma
(61VH
) (9SV
91) (9ZV
t’I) (W
-91) 186’ob
Z9fw7
9LI’IP 9zSw
boz-0) P
PO
'O
(6f Z-0) 190-o
b61’0) IW
O
(IKO) z90’0
ma3
(L#cO) s9uo
(08Z.O)
980-O
@K
-o) S90’0
(LLZ’O) P80’0
(8Sf ‘0) ISI'O
(Z8V0) s9r0
9703a3
(99E’O
) 6SI'0
(99f -0) 6SI’O
(18V0)
WE.0
(L80’0) 800-O
(8Sf ‘0) ISI’O
703sa3
mfa3
3NoN
a3
(L8VO) 98E’O
(OLO’O) 500’0
(88V01 16f’0
(690’0) SO
O-0
&SO-O) LOO.0
Tabl
e 1
(con
tinue
d)
Mea
ns (
stan
dard
de
viat
ions
)
Var
iabl
es
Des
crip
tion
Entir
e sa
mpl
e Re
stric
ted
sam
ple
(a) S
mok
ers
& (
b) S
mok
ers
(c)
Smok
ers
& (
d) S
mok
ers
non-
smok
ers
only
no
n-sm
oker
s on
ly
GOOD
FAIR
POO
R
Dum
my
varia
bles
tha
t eq
ual
1 w
hen
exam
inee
re
ports
hea
lth a
s ‘g
ood
‘fai?
or
‘poo
r’ re
spec
tivel
y; o
mitt
ed
cate
gory
is
‘exc
elle
nt*
heal
th.
0.40
0 0.
411
0.39
8 0.
407
(0.4
90)
(0.4
92)
(0.4
89)
(0.4
9 1)
0.12
1 0.
121
0.12
2 0.
125
(0.3
26)
(0.3
26)
(0.3
27)
(0.3
30)
0.03
9 (0
.195
) BL
ACK
D
umm
y va
riabl
e th
at e
qual
s 1
whe
n ex
amin
ce i
s bl
ack.
O
THER
D
umm
y va
riabl
e th
at e
qual
s 1
whe
n ex
amin
ee i
s ne
ither
bla
ck n
or w
hite
.
0.09
2 0.
106
0.08
1 0.
092
(0.2
89)
(0.3
08)
(0.2
73)
(0.2
89)
0.01
3 0.
017
0.01
2 (0
.115
) (0
.129
) (0
.109
)
FAM
SIZE
LABO
R
NEA
ST
NC
ENTR
AL
SOU
TH
NO
NC
ITY
LCIT
Y
MC
KY
Num
ber
of p
erso
ns i
n ex
amin
ee’s
fam
ily.
Dum
my
va.ri
able
that
equ
als
1 if
exam
irrkc
is fe
mal
e an
d in
the
lab
or f
orce
. D
umm
y va
riabl
es t
hat
equa
l 1
whe
n ex
am&
e liv
es in
th
e N
orth
east,
N
orth
cent
ral
or S
outh
res
pect
ivel
y;
omm
itced
cla
ss is
re
siden
ce i
n W
est.
Dum
my
varia
bles
tha
t eq
ual
1 w
hen
exam
inee
doe
s no
t liv
e in
an
SMSA
(N
ON
CIT
Y);
lives
in a
n SM
SA w
ith
a po
pula
tion
over
3
mill
ion
(LC
ITY)
; or
liv
es in
an
SMSA
with
a
popu
latio
n be
twee
n 1
and
3 m
illio
n (M
CITY
); om
itted
cla
ss is
resid
ence
in
an
SMSA
with
pop
ulat
ion
ks
than
1 m
illio
n.
2.94
7 (1
.917
) 0.
219
(0.4
14)
0.11
4 (0
.319
)
0.27
9 (0
.449
)
0.31
8 (0
.466
)
0.32
7 (0
.469
)
0.12
3 (0
.328
)
0.20
3 (0
.402
)
3.04
2 (1
.970
) 0.
218
(0.4
13)
0.11
4 (0
.317
)
0.27
6 (O
&7]
0 32
8 (0
.469
)
0.32
0 (0
.467
)
0.12
1 (0
.326
)
0.20
9 (0
.407
)
132 E.M. Lewit and D. Coate, Using excise taxes to reduce smoke ‘!a
samples are not substantially different. Not unexpectedly, the mean price in the restricted sample is lower than the mean price in the entire sample and by extension the mean price in the excluded PSU’s. There are, however, a substantial number .of observations from high priced PSU’s in the restricted sample and the range of cigarette prices for the restricted sample (3$.8# to 57.6$ per pack) is nearly as large as the range (35.8# to 62.2pl per pack) for the entire sample.
The most substantial difference between the two samples is in their geographic representation of the nation. The restricted sample contains proportionately more observations from the West (29%) than the total sample (18%) and fewer observations from the Northeast (11% in the restricted sample vs. 24% in the total sample). This is largely because the Northeast contains many small political divisions with varying tax rates while the West is dominated by larger geographic areas with more uniform tax rates. Representation of the other regions is approximately equal in both samples.
Estimates of the demand for cigarettes (quantity smoked by smokers and non+mokers) and the smoking participation rate for both the total and the restricted sample were obtained by OLS regression and by a variance- components GLS procedure which allowed for PSU-specific random disturbances to account for any omitted cite-specific variab1es.l’ Becaues of the large sample size, estimates of the variance-components svere not obtained directly from the data: rather, an iterative procedure was used wherein GLS estimates where obtained for different values of p, the ratio of the variance of the cite-specific error component to the variance of the total model error term. Estimated standard errors of the price coefficients were larger in the GLS regressions than. in the OLS regressions, but estimated regression coefficients were generally not sensitive to the choice of p and the maximum value of the likelihood function was obtained from the OLS estimates. Accordingly, only the OLS estimates are presented in table 2.
There are substantial differences in the estimated price effects in the two samples - the price coefficients and corresponding elasticities at the mean are almost twice as large in the restricted sample as in the total sample. IMoreover, the price coefficient only achieves statistical significance at the 5% level of a one-tailed test in the quantity smoked - regression for the total sample [column (a)] while it is more robust in both the quantity smoked and
- smoking participation rate regressions [columns (b) and (d)] in the restricted sample.
The estimated coefficients of the other independent variables are generally not sensitive to the sample chosen.12 These findings are 6;onsistent with
“The GLS procedure followed closely that of Fuller and Battese (1974) and Wallace and Hussain (1969).
“Except for income we do not discuss in this paper the effects of the other independent variables on smoking.
Tabl
e 2
OLS
reg
ress
ion
coef
ficie
nts
of t
he d
eman
d fo
r ci
gare
ttes
in t
he e
ntire
19
76 H
IS s
ampl
e an
d th
e re
stric
ted
sam
ple
(r-s
tatis
tics
in
pare
nthe
ses
belo
w c
oefli
cien
ts).
Inde
pend
ent
varia
bles
Qua
ntity
sm
oked
by
smok
ers
Smok
ing
parti
cipa
tion
rate
Q
uant
ity
smok
ed b
y sm
oker
s an
d no
n-sm
oker
s (C
ZGD
A I’)
(SMOKER)
(CIGSMOKER)
(a) E
ntire
(b
) Res
trict
ed
(c) E
ntire
(d
) Res
trict
ed
(e) E
ntire
(r)
Res
trict
ed
sam
ple
sam
ple
sam
ple
sam
ple
sam
ple
sam
ple
PRZCE
INCOME
EDNONE
EDHSG
EDSCOL
EDCOLG
EDGS
AG
E
FfiM
ALE
WID
OW
ED
NEV
ER
-0.0
33
-0.0
63a
(- 1.
708)
(-2
.729
) 0.
03Y
0.
038O
(4
.417
) (3
.299
) -
3.55
5’
-4.2
96a
(-3.5
50)
(-3.3
63)
-0.9
73.
-0.7
598
(-4.3
92)
(-2.6
21)
- 1.
95T
- 1.
612”
(-6
.757
) (-4
.323
) -
3.74
5”
-3.3
10”
(- 10
.536
) (-7
.226
)
- 5.
255”
-
4.28
3”
(- 13
.085
) (-8
.168
) -o
.114
a -0
.112
” (-
17.4
98)
(- 13
.207
) -4
.112
a -
4.03
w
( .,I*
20.4
37)
(- 15
.374
) -0
.222
-0
.225
(-0
.632
) (-0
.484
) -
2.02
8’
- 2.
39Y
(-7
.063
) (-6
.302
)
-0.0
01
(- 1.
538)
7.
84E-
04b
(2.2
11)
-0.1
05*
( - 2
.624
) -
0.04
6”
(- 5.
239)
-0
.091
a ( -
7.8
88)
-o.1
668
(-11.
735)
-0.2
178
(- 13
.604
) -
0.00
6”
(-21.
860)
-0
.129
” (-
16.0
80)
-0.0
11
(-0.7
76)
-0.0
71”
(-6.2
10)
-o.oo2b
(-2.252)
7.59
E-04
(1
.618
) -0
,116
b (-2
.253
) -
0.64
7”
(-4.0
28)
- 0.
087”
(-5
.799
) -0
.165
” (-8
.941
)
-0.1
92’
(-9.0
50)
- 0.
006°
(-
17.1
87)
-0.1
32’
(- 12
.487
) -0
.012
(-0
.654
) -
0.08
3”
(-5.4
37)
-0.0
15
(-0.5
01)
o.07
38
(5.0
36)
-7.1
17”
(-3.5
49)
- 0.
088
(-0.2
56)
- 0.
430
(-0.9
43)
- 1.
383b
(-2
.212
)
- 2.
798”
(-
3.73
9)
0.00
3 (0
.277
) -
4.39
2m
(- 12
.990
) -
1.04
5 (-
1.52
4)
- 1.
818”
(-3
.785
)
-0.0
43
(- 1.
188)
0.
074”
(3
.823
) -
9.54
6”
(-3.7
08)
0.43
1 (0
.957
) 0.
223
(0.3
79)
- 0.
225
( - 0
.277
)
- 1.
742
(- 1.
834)
0.
018
(1.2
07)
-4.1
02’
(-9.2
51)
- 0.
886
(-0.9
66)
-2.3
05’
(-3.5
72)
Tab
le 2
(co
nti
nu
ed)
Inde
pen
den
t va
riab
les
Qu
anti
ty s
mok
ed b
y sm
oker
s S
mok
ing
part
icip
atio
n r
ate
Qu
anti
ty s
mok
ed b
y sm
oker
s an
d n
on-s
mok
ers
(CIG
DA
Y)
(SM
OK
ER)
(CIG
SMO
KER
)
(a)
En
tire
(b
) R
estr
icte
d (c
) E
nti
re
(d)
Res
tric
ted
(e)
En
tire
(1
) Res
tric
ted
sam
ple
sam
ple
sam
ple
sam
ple
sam
ple
sam
ple
DIV
ORC
ED
SEPA
RATE
D
GO
OD
FAIR
POO
R
BLAC
K
OTH
ER
FAM
SIZE
LABO
R
4.95
3”
4.86
1”
(123
65)
(9.6
90)
3.53
2”
4.14
ga
(6.2
79)
(5.1
68)
i.943
8 !.1
94P
(6
.047
) (4
.846
)
1.39
P
1.34
8”
(4.8
09)
(3.5
77)
1.74
2”
1.99
3”
(3.7
96)
(3.3
30)
- 2.
49ga
-
2.28
2’
(-8.
051)
;-
5.36
7)
-3.9
12a
- 3.
368”
(-
5.24
6)
(-3.
892)
-0
.041
-0
.108
(-
0.78
5)
(-
1.58
0)
0.86
80
0.50
7 (3
.560
) (1
.597
)
0.17
9a
(11.
260)
0.10
2”
(4.5
56)
0.02
6”
(3.4
61)
0.04
3”
(3.7
21)
(z$ 0.01
8 (1
.429
) -0
.124
” (-
4.16
9)
-0.0
03
(--
1.26
6)
0.04
1”
(4.1
98)
0.17
6”
(8.6
98)
O.lW
(4
.324
)
0.02
9”
(2.8
95)
0.05
2”
(3.4
28)
0.05
9b
(2.4
54)
0.00
8 (0
.459
) -0
.100
” (-
2.85
6)
-0.0
05
(-
1.82
2)
0.03
1b
(2.4
57)
2.65
5”
(4.8
52)
3.05
6”
(3.9
00)
1.65
3”
(5.4
08)
1.39
6”
(2.9
75)
2.29
8”
(3.0
58)
- 6.
660”
(-
14
.139
) -
6.03
6”
(-4.
054)
(KZ
)
0.54
4 (1
.329
)
2.81
4”
(4.0
77)
3.01
6”
(2.7
76)
1.76
9”
(4.4
3 1)
0.
998
(1.6
49)
2.23
Sb
(2.3
14)
- 6.
078”
(-
9.31
7)
- 5.
636”
(-
3.37
9)
-0.0
49
(-0.
456)
0.07
4 (0
.138
)
NEA
ST
NCE
NTR
AL
SOU
TH
NONCITY
LCZT
Y
MCZ
TY
R2
Pric
e
0.965
’ (3
.136
) 0.
336
(1.2
90)
0.48
3 (1
.801
) -0
.398
(-
1.85
6)
0.23
2 (0
.855
) 0.
283
(1.1
21)
15.1
21’
(15.
672)
0.
069
19,2
68
-0.2
21
(0.1
26)
0.95
Sb
(2.1
96)
0.22
4 (0
.730
) 0.
598
(1.8
82)
-0.3
31
(-1.2
09)
- 0.
083
(-0.2
05)
0.39
9 (1
.243
) 17
.326
” (1
3.95
3)
0.06
9 11
,052
’ -0
.416
(0
.158
)
0.02
8b
(2.3
10)
0.00
5 (0
.460
) 0.
012
(1.0
88)
-0.0
16
(- 1.
832)
0.
015
(1.3
80)
0.01
4 (1
.401
) 0.
765”
(1
8.68
3)
0.06
7 19
,268
-0
.135
(0
.086
)
0.02
8b
(1.5
94)
0.00
5 (0
.420
) 0.
017
(1.3
55)
-0.0
10
(-0.8
95)
0.00
7 (0
.410
) 0.
020
(1.5
66)
0.81
9”
(16.
334)
0.
069
11,0
52
- 0.
264
(0.1
22)
1.13
7b
(2.2
92)
0.53
9 (1
.269
) 0.
526
(1.2
09)
-0.4
57
(- 1.
301)
-0
.271
(-0
.619
) 0.
03 1
(0.0
77)
21.3
55’
(13.
161)
0.
078
7,07
9 -
0.03
7 (0
.077
)
1.01
0 (1
.429
) 0.
193
(0.3
86)
0.47
6 (0
.929
) -0
.603
(-1
.412
) -0
.781
(-
1.18
3)
-0.0
53
(-0.1
02)
21.8
32’
(11.
277)
0.
075
4,02
6 -0
.103
(0
.087
)
%at
istic
ally
sig
nific
ant
at 1
% o
n tw
o-ta
iled
test.
bS
tatis
tical
ly s
igni
fican
t at
5%
on
two-
taile
d te
st.
‘App
roxi
mat
e st
anda
rd
erro
rs a
re r
epor
ted
in p
aren
thes
es
belo
w t
he e
stim
ated
pric
e el
astic
ities
. Th
e st
anda
rd
erro
rs a
re b
ased
on
a Ta
ylor
se
ries
linea
rizat
ion
of th
e el
astic
ities
and
are
mea
nt t
o be
illu
strat
ive.
136 E.M. L&wit and D. Coate, Using excise taxes to reduce smoking
measurement error in the full sample price variable which biases the full
E.M. L,ewit and D. Gate, Using excise taxes to reduce smoking 137
nature of the underreporting. If all smokers underreport their consumption by a constant proportion, then all coefficients in a regression of quantity smoked on price would be biased downward by the same proportion. Estimated market elasticities calculated at the sample means will not be biased, however, because mean consumption will be underreported by the same proportion as the estimated coefficients and the errors will cancel out. Even if the underreporting is not exactly proportional, the use of the sample means to calculate the elasticities will tend to compensate for the downward bias in the estimated coefficients. Hence, it is unlikely that our finding that the participation elasticity exceeds the quantity smoked elasticity is an artifact caused by measurement error.
The income elasticity of 0.08 is also small relative to previous estimates. This point is taken up in the concluding section of the paper. In contrast to price, income appears to impact cigarette demand primarily by inlluencing the number of cigarettes consumed by smokers rather than by affecting the smoking participation rate.“’
5.1. Cigarette demand by different age groups
Because price appears to impact cigarette demand primarily through its affect on the decision to smoke or not, demand equations were estimated across the following age groups in the restricted sample: 20-25 years, 26-35 years, and over 35 years. The price coefficients and elasticities at the means from these estimated demand equations are presented in table 3. These estimates are from regressions which included all the independent variables included in the regressions in table 2. While we do not present or discuss the effects of these other independent variables on smoking, it should be realized that all the estimated price effects presented in this paper control for the effects if these other variables. The estimated price effects from the smoking participation regression are estimated by a FIML logit procedure, the preferred estimation procedure when the dependent variable is dichotomous merlove and Press (1973)J.16
“The income elasticity of the smoking participation rate at the means is 0.03 and the income elasticity of the quantity smoked by smokers is 0.06.
‘@The smoking participation rate equations were estimated by both OLS and FIML logit procedures for all the age and sex subsamples of the restricted sample reported in tables 3,4 and 5. In all cases, the price effects obtained from the OLS regressions were almost identical to those obtained when the FIML logit procedure was used. We present the results of the logit estimates because they are preferred on theoretical grounds. The entire HIS sample and the entire ‘restricted’ sample were too large to be accommodated by our logit program. Those results, reported in table 2, were obtained by OLS regression. Amemiya (1981) has indicated that when the dependent variable lies between 0.3 and 0.7 (ours is about 0.361, OLS estimates are roughly ‘equivalent’ to logit and probit estimates, and given the similarlity between our OLS and logrt estimates in the subsamples, we fsel confident that the OLS estimates presented in table 2 are nearly identical to what logit estimates would be. We did not use a Tobit procedure to estimate our CIGDAY regressions because we did not have access to a computer program that could accommodate large data sets.
138 E.M. Lewit and D. Coate, Using excise taxes to reduce smoking
Table 3 Regression price coefficients and elasticities, means and standard deviations in the
restricted 1976 HIS sample according to age and smoking status.“
Variables 20-25 years 26-35 years Over 35 years
(A) Regression Coeftcients
1. Demand for cigarettes by smokers and non-smokers (CZGDAY) (OLS) PRICE -0.125b -0.081
‘Standard errors of regression coelllcients are in parentheses below coefficients. Approximate standard errors are reported in parentheses below the estimated price elasticities.
bStatistically significant at 5% on two-tailed test. “ay/&c’s evaluated at the mean of SMOKER are reported in brackets alongside
coefRcients. dThe reported chi-square is twice the difference in the log-likelihood of the model from
the likelihood based on the intercept only. ‘Mean of PRICE in the smokers-only sample.
EM. bit and D. Gate, Using excise taxes to reduce smoking 139
The differences in the price coefficients by age are not statistically significant at 5%; however, the cigarette demand equations for smokers and non-smokers yield a price elasticity of - 0.89 for the 20-25 years age group, a figure twice as large in absolute value as the price elasticity for the other age groups. Furthermore, the smoking participation price elasticity of -0.74 for the 20-25 year olds accounts for a great portion of the aggregate elasticity of this age group. This is not to say that the results for the older age groups do not show some price sensitivity. The price coefficient is statistically significant in the demand for cigarette equations for the over 35 year old group and nearly significant for those 26-35 years old. The price elasticity of demand. at the means for both these groups of about -0.45 is not insubstantial. In the case of the 26-35 year olds, price appears to operate primarily through smoking participation while the decomposition of the aggregate price effect for the older age group is less conclusive.
The results reported here that price has its greatest effect on the smoking behavior of younger people and that it operates primarily via the decision to begin smoking regularly rather than on the quantity smoked is consistent with the notion that cigarette smoking is in some sense an addictive behavior [Krasnegor (1979a)] initiated primarily before age 25 [National Clearinghouse for Smoking and Health (197611. They are also consistent with the findings of Lewit, Coate and Grossman (19&l? who used another large micro sample, the Health Examination Survey 19661970, and a methodology similar to that described in this study to examine cigarette demand by 12-17 year olds. They report a total price elasticity for the quantity smoked by teenage smokers and non-smokers of -1.4 and a smoking participation elasticity of - 1.2. Thus, the pattern of larger price elasticities in the younger age groups and the attribution of price effects primarily to the smoking participation decision is confirmed in another sample.
5.2. Cigarette demand by age and sex
Further insight into the effects of price on cigarette demand can be gained by looking in some detail at the price elasticities of different age and sex groups. In previous specifications, sex differentials in cigarette demand were represented by a dummy variable. The price elasticity results presented in table 4 for males and table 5 for females are based on separate regressions for each sex in each age group. Thus, price effects can now differ by sex as well as by age.
The results indicate that cigarette demand by females is generally not sensitive to price (table 5), while males (table 4) appear more sensitive to price than was implied by previous results where male and female price coefficients were constrained to be equal. For 20-25 year olds, for example,
140 E.M. Zkwit and D. Coate, Using excise taxes to reduce smoking
Table 4 Regression price coefficient, and elasticities, means and standard deviations for males
according to age and smoking status in the restricted 1976 HIS sample.
Variables 20-25 years 26-35 years Over 35 yeaes
(A) Regression Co@cienrs 1. Demand for cigarettes by smokers and ,non-smokers (CZGDAY)(OLS) PZUCE - 0.236b - 0.066
at means (0.313) (0.214) (0.152) R2 0.156 0.101 0.063 Sample size 294 591 1,229
(B) Means and Standard Deviations (in parentheses) of Price and Dependent Variables CZGDAY
SMOKER
PRICE
CZGSMOKER
PRICE (SMOKER)’
(1;:;) 0.45 (0.5Q 47.84 (5.14) 17.98
(10.38) 47.22 (5.54)
9.85 (12.59)
0.49 (0.50) 47.6!3 (5.13) 19.92
(10.95) 47.39 (5.411)
(1:::) 0.39
(0.49) 48.05 (5.25) 23.39
(13.19) 47.78 (5.50)
‘Standard errors of regression coefficients are in parentheses below coefficients. Approximate standard errors are reported in parentheses below the estimated price elasticities.
bStatistically significant at 5% on two-tailed test. ‘i$@x’s evaluated at the mean of SMOKER are reported in brackets alongside
coefficients. %atistically &nificant at 1% on two-tailed test. @The reported model chi-square is twice the difference in the Ic.g-likelihood of the
model from the likelihood based on the intercept only. ‘Mean of PRICE in the smokers-only sample.
E.M. L.ewit and D. Coate, Using excise taxes to reduce s<:toking
Table 5
141
Regression price coefficients and elasticities, means and standard deviations for females according to age and smoking status in the restricted 1976 HIS sample.
Variables 20-25 years 26-35 years Over 35 years
(A) Regression Coejicients
1. Demand for cigarettes by smokers and non-smokers (CIGDAY) (OLS) PRICE - 0.036 - 0.083
*Standard errors of regression coefiicients are in parentheses below coenicients. Approximate standard errors are reported in parentheses below the estimated price elasticities.
bdy/Ws evaluated at the means of SMOKER are reported in brackets alongside coefficients.
“The reported model chi-square is twice the difference in the log-likelihood of the model from the likelihood based on the intercept only.
dMean of PRICE in the smokers-only sample.
142 EM. hit and D. Gate, Using excise taxes to reduce smoking
the price elasticity for quantity smoked by smokers and non-smokers was -0.89 for both sexes combined, - 1.40 for males, and -0.30 for females (the price coefficient for females is not significantly different from zero), and as regards smoking participation, the differences between the sex-specific price
* coefficients are statistically significant at 5%. Even for males, however, the effect of price on smoking behavior varied
with age. The difference between the price coefftcients of smoking participation for young males and for males over 35 years old as a group is statistically sign&ant at 5%. Moreover, the coefficient of price in the regression for men aged 26-35 years fails to achieve statistical significance and price elasticities at the mean are much smaller than for the younger group. Price does, however, seem to act to reduce smoking by males more than 35 years old (table 4). Here the impact of a price is split between changes in the smoking participation rate and changes in the quantity smoked by smokers. It appears, therefore, that to some extent price (tax) increases may have a beneficial effect on the health of older males, a group that has experienced the greatest health losses due to cigarette smoking [U.S. Public Health Service (197911.
6. Implications of the research
In this paper, we have attempted to assess the potential for using excise taxes to reduce smoking by measuring the price elasticity of demand for cigarettes. Excise tax increases will discourage smoking to the extent that excise tax increases are passed on to smokers in the form of higher retail cigarette prices, and Braze1 (1976) has presented evidence that retail cigarette prices have more than reflected state excise tax increasesl’ Our empirical results have indicated (1) that the price elasticity of demand for cigarettes is -0.42, (2) that the decision to begin smoking regularly by males less than 25 years old is price elastic, and (3) that price effects appear to be larger for males than for females.
These results have implications for any future attempts by the Federal government to influence cigarette demand through excise tax policy. The short-run impact of an excise tax increase would be small. For example, if the federal excise tax was doubled to 16 cents a pack, and if the tax increase was completely passed on to the consumer, then the average retail price of cigarettes would increase by about 11.5% (using the 1981 average retail price as a reference). Accordingly, applying our estimated price elasticity of -0.42, cigarette consumption would fall by about 4.8%. The fall- off in demand would result from an approximate 3% decline in smoking
“Of course, increases in the retail price of cigarettes which result from increases in the costs of growing, manufacturing or marketing cigarettes will discourage cigarette smoking to the same extent as tax increases.
E.M. Lewit and D. Coate, Using excise taxes to reduce smoking 143
participation and 1.2% decline in the quantity of cigarettes smoked by smokers. In the long run, however, the impact of such a tax increase could be much more substantial. Our results imply that a doubling of the excise tax would lead to a 16% decline in smoking participation by males 20-25 years old. Furthermore, it may be that over time the response by young females to price changes could approach that of young males.‘* Thus an excise tax increase, if maintained in real terms, might continue to discourage saoking participation by successive generations of teenagers and young adults and gradually impact tE5 smoking levels of older age groups as the smoking-discouraged cohorts move through the age spectrum.rg
Concern has been voiced about the fact that smokers can compensate for reductions in the number of cigarettes they consume by switching to higher tar and nicotine brands, by inhaling more deeply, or by reducing idle bum [for example, see Krasnegor (1979a, b), and Gori and Bock (1980)]. Any of these compensatory behaviors could greatly reduce the health benefits which might be obtained from a reduction in the number of cigarettes smoked because of higher taxes. Within this context, Harris (1980) has analyzed a step-function tax on cigarettes graded by tar/nicotine content. His analysis, however, only considered the effect of taxation on established smokers and not the effect of taxes on smoking participation rates, as in this paper. To the extent to which a large benefit would result from an excise tax increase maintained in real terms over the long run as a consequence of a decrease in smoking participation per se, the exact nature of smokers’ compensatory behavior is of reduced importance and reservations about an excise tax policy because of potential compensation by smokers lose ground.
Some final words are reserved for a discussion of the income elasticity of demand for cigarettes. The size of the income elasticity has had important implications for evaluations of federal government policies to discourage smoking [Hamilton (1972), Ippolito, Murphy and Sant (1979), Klein, Murphy
“It is beyond the scope of this paper to determine why price seems to affect male smoking behavior but not female smoking behavior or to explain differences in aggregate smoking behavior by the two sexes. It should be noted, however, that until the mid-1970’s, trends in smoking participation of males and females in the United States were markedly different. Male smoking rates peaked in the mid-1960’s at 51 percent of the male population and have fallen continously since then to 37 percent in 1979, the latest year for which llgures are available. Female smoking rates peaked later than male rates and were relatively constant through 1976, the year in which the HIS data used in this study were collected. More recent data, however, show a decline in female smoking, with participation rates down to 28 percent from the 33 percent plateau of the late sixties to mid-seventies [U.S. Public Health Service (1980)]. At least to sopne extent then, one aspect of female smoking is becoming similar to that of males.
“It is possible, however, that the price elasticity of demand for cigarettes might vary over time.’ For example, it is conceivable that the elasticity has fallen over the past few decades as those smokers who would be most responsive to any factor discouraging smoking have stopped smoking in response to the anti-smoking campaign. Therefore, any discussion of the long-run effect+s of an excise tax increase must acknowledge the possibility that the price elasticity could change in the future.
144 E.M. Lewit and D. Coate, Using excise taxes to reduce smoking
and Schneider (1981)]. Our estimated income elasticity of demand, 0.08 (table 2), is about one-tenth the size of the estimate obtained by Ippolito, Murphy, and Sant (1979) from time series data and about one-tenth the size of Hamilton’s (1972) estimate which was obtained from a cross-state analysis of tax paid sales data. As a result, the models of both Hamilton (1972) and Ippolito, Murphy and Sant (1979) tend to attribute more of the secular variation in per capita. cigarette consumption to variations in income than would a model inccrporating our estimated income elasticity. For example, Hamilton (1972) estimates that annuai cigarette consumption would have increased from 3506 cigarettes per capita in 1953-55 to 4482 cigarettes per capita in 1968-70 due to the substantial increase in income during that period. Since consumption only increased to 3868 cigarettes per capita in 1968-70, he credits the health scare and anti-smoking advertising under the Fairness Doctrine with substantially reducing smoking below the level it otherwise would have been. To the extent to which Hamilton’s estimate of an income induced increase in smoking may have been too large, he may have overstated the impact of the health scare. Moreover, his analysis of the diflerential effects of the Fairness Doctrine and advertising ban policies may also require reconsideration if they too rely substantially on an overstated income effect. Of course, it is possible that the income elasticity of demand for cigarettes has declined substantially in recent years for many reasons including the health scare and related government programs. In fact, Klein, Murphy and Schneider (1981) assess the impact of government cigarette policies within the context of a time series model wherein the income elasticity of demand for cigarettes falls as income rises. As a result, their estimated income elasticity for 1976 is closer to our estimate than the estimates cited immediately above.20
“l!n the Klein, Murphy and Schneider (1981) model, income can influence th demand for cigarettes in two ways: by affecting the demand for tobacco and by afkcting the $ roportion of tobacco consumed as cigarettes. They assume that below a critical level of income (551 1929 dollars) all tobacco is consumed in forms other than prerolled cigarettes, where level all tobacco is consumed as cigarettes. Accordingly, as income rises over time cigarette demand that is due to switching declines as there are a declining cigarette smoking tobacco smokers who can switch to cigarettes. In the limit, smokie cigarettes exclusively, they estimate the income elasticity of demand for tobacco) to be 0.47.
References Advisory Commission on Intergovernmental Relations, 1977, Cigarette bootlegging: A state and
federal responsibility (Washington, DC). Amemiya, Takeshi, 1981, Qualitative response models: A survey, Journal of Economic Literature
19,1483-1536. Barzel, Yoram, 1976, An alternative approach to the analysis of taxation, Journal of Political
Economy, 1177-l 198. Fuchs, Victor, 1974, Who shall live? Health, economics and social choice (Basic Books, New
York).
E.M. Lmit and D. Coate, Using excise taxes to reduce smoking 145
Fuller, Wayne A. and George E. Battese, 1974, Estimation of linear models with c rossed error
structure, Journal of Econometrics 267-78. Fujii, Edwin T., 1980, The demand for cigarettes: Further empirical evidence and its
implications for public policy, Applied Economics 12,479-489. Gori, Gio B. and Fred G. Rock, eds., 1984 A safe cigarette?, Banbury report no. 3 (Cold Spring
Harbor Laboratory, New York). Hamilton, James L., 1972, Advertising, the health scare, and the cigarette advertising ban,
Review of Economics and Statistics 54,401-411. Harris, Jeffrey E., 1980, Taxing tar and nicotine, American Economic Review 70,3OO-311. Intergovernmental Perspective, 1978, Congress, states move to curb cigarette bootlegging. Ippolito, Richard A., Dennis R. Murphy and Donald Sam, 1979, Consumer responses to
cigarette health information, Bureau of Economics, Federal Trade Commission staff report (U.S. Government Printing Office, Washington, DC).
Kellner, Irwin L., 1973, The American cigarette industry - A reexamination, Ph.D. thesis (New School for Social Research, New York) unpublished.
Klein, Benjamin, Kevin M. Murphy and Lynne Schneider, 1981, Government regulation of cigarette health information, Journal of Law and Economics 24, 575-612.
Krasnegor, Norman A., ed., 1979a, Cigarette smoking as a dependence process, NIDA Research monograph no. 23 (U.S. Government Printing O&e, Washington, DC).
Kcaanegor, Norman A., ed., 1979b, The behavioral aspects of smoking, NIDA Research monograph no. 26 (U.S. Government Printing Office, Washington, DC).
Lewit, Eugene, Douglas Coate and Michael Grossman, 1981, The effects of government regulation on teenage smoking, Journal of Law and Economics 24,545-569.
Lute, Bryan R. and Stuart 0. Schweitzer, 1978. Smoking and alcohol abuse: A hmparison of their economic consequences, New England Journal of Medicine 298,569571. !
Mann, Charles Kellogg, 1971, Taste, technology, and the government: The c+e of tobacco, Ph.D. thesis (Harvard University, Cambridge, MA) unpublished.
Miller, Robert H., 1970, The demand for cigarettes in the United States, Talk phesented at the 23rd National Tobacco Workers Conference, Jan. 15 (College Park, MD). d
Miller, Robert H., 1974, Factors affecting cigarette consumption, Talk presented at National Tobacco Tax Association Annual Meeting (Kiamesha Lake, NY).
Miller, Robert H., 1976, Cigarettes: Tax developments and sales trends, Talk presented at the 50th Annual Meeting National Toba-?o Tax Association (Atlanta, GA).
National Clearinghouse for Smoking and Health, 1976, Adult use of tobacco - 1975 (U.S. Department of Health, Education, and Welfare, Center for Disease Control, Washington, DC).
Nerlove, Mark and S.J. Press, 1973, Notes on the log-linear or logistic model for the analysis of qualitative socioeconomic data (The Rand Corporation, Santa Monica, CA).
Padilla, Arthur H., 1971, The elfect of price fluctuations on cigarette consumption in the U.S., Tobacco, 31-33.
Schnabel, Morton, 1972, An oligopoly model of the cigarette industry, Southern Economic Journal 38,325335.
Tobacco Tax Council, 1980, The tax burden on tobacco (Richmond, VA). U.S. Department of Agriculture (Economic Research Service), 1981, Tobacco situation. U.S. Public Health Service, 1979, Smoking and health, DHEW publication no. (PHS) 7950066
(U.S. Government Printing Office, Washington, DC). U.S. Public Health Service, 1980, The health consequences of smcking for women (Department
of Health and Human Services, Washington, DC). Wallace, T.D. and Ashiq Hussain, 1969, Use of error components models in combining cross
section with time series data, Econometrica 37, 56-68. Warner, Kenneth E., 1977, The effects of the anti-smoking campaign on cigarette consumption,
American Journal of Public Health 67,645-650. Warner, Kenneth E., 1978, Possible increases in the underreporting of cigarette consumption.
Journal of the American Statistical Association 73, 314-318. Warner, Kenneth E., 1981, Cigarette smoking in the 1970’s: The impact of the antismoking
campaign on consumption, Science 211,729731. Zook, Christopher J., Francis D. Moore and Richard Zeckhauser, 1981, Catastrophic health
insurance a misguided prescription, The Public Interest 62,66-81.
