Are Filipino Smokers More Sensitive to Cigarette Prices due to the Sin Tax Reform Law?: A Difference-in-difference Analysis
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
DLSU Business & Economics Review 28(2) 2019, pp. 10–25
RESEARCH ARTICLE
Are Filipino Smokers More Sensitive to Cigarette
Prices due to the Sin Tax Reform Law?:
A Difference-in-difference Analysis
Myrna S. Austria
De La Salle University, Manila, Philippines
myrna.austria@dlsu.edu.ph
Jesson A. Pagaduan
Asian Development Bank, Philippines
Abstract: Employing a two-part estimation model using the Family Income Expenditure Survey before (2009) and after
(2015) the tax reform, our study assessed the impact of the Philippine Sin Tax Reform Act (2012) on cigarette consumption
and the responsiveness of cigarette consumption to price changes. The results are consistent with existing studies that
cigarette consumption is price inelastic. The demand, however, has become less inelastic in the Philippines over the period
2009 to 2015, indicating a more responsive cigarette demand to price increases. Of the total effect of cigarette price increase
on demand, the decrease in consumption by smokers (smoking intensity) accounts for much of the decline in cigarette
consumption, rather than the decrease in the number of cigarette users (smoking prevalence). The increase in excise tax due
to the tax reform has been effective on lowering cigarette consumption in the country and in making cigarette demand more
responsive to price increases. Specifically, the tax reform has reduced the number of cigarettes purchased by smokers more
than the number of cigarette users.
Keywords: excise tax, elasticities, smoking prevalence and intensity, difference-in-difference analysis
JEL Classifications: D120, H200
One of the significant legislation during the Aquino the tax structure, and removed the price classification
Administration was the Sin Tax Reform Act of 2012 freeze. Prior to the reform, tobacco taxation in the
(Republic Act No. 10351, 2012). The tax reform is country followed a complex four-tiered tax system
primarily a health measure as well as a governance using a tax base freeze at 1996 price levels. Since
measure. It addresses public health issues related the excise tax was not indexed to inflation, prices
to alcohol and tobacco consumption along with the of tobacco products in the country were among the
structural weaknesses of the country’s tax system on cheapest in the world despite the increases in excise
alcohol and tobacco products. tax over the years (Quimbo et al., 2012). In contrast,
For tobacco, the law significantly increased the the tax reform provides a two-tiered system effective
excise tax on tobacco and tobacco products, simplified January 2013 with a gradual shift to single and uniform
Copyright © 2019 by De La Salle UniversityAre Filipino Smokers More Sensitive to Cigarette Prices due to the Sin Tax Reform Law? 11
rate taxation starting 2017, after which the rate will be is not as elastic as the demand for other consumer goods
increased by 4% every year effective January 2018. (Tennant, 1950), there is a consensus in the empirical
The current system is considered simpler and more literature that tobacco consumption falls in response to
efficient in raising tobacco taxes. an increase in the price of tobacco because of a decrease
Depending on the cigarette classification, the in smoking prevalence (i.e., decrease in the number
increase in excise tax varies from 108% to as high of individuals who smoke), because of a decrease in
as 341%. Given the significant increase, the price of smoking intensity (i.e., decrease in the consumption
cigarettes also substantially went up. After almost six by those who use the tobacco products), or because of
years of implementation, what has been the impact a combination of the two possible outcomes (IARC,
of the Sin Tax Reform on the demand for cigarettes? 2011; World Bank,1999).
This study aims to: (i) estimate the price and income There were almost no micro-level studies on the
elasticities of the demand for cigarettes after the tax impact of tax and price on tobacco consumption in low-
reform; (ii) determine the impact of the tax reform on and middle-income countries up until the publication of
cigarette consumption; (iii) determine the impact of the the World Bank’s (1999) Curbing the Epidemic report.
tax reform on the price responsiveness of the demand Since then, however, there has been a growing body
for cigarettes; and (iv) recommend policies for future of tobacco demand studies for developing countries
tax reform in the country. (IARC, 2011). The World Bank review revealed that,
This empirical study offers two major contributions ceteris paribus, a 10% price increase would reduce
on tobacco economics and taxation. First, this is the tobacco consumption by about 8% in less-developed
first analysis that evaluate empirically the impact of the countriesand about 4% in advanced economies (Jha
tax reform on cigarette demand in the Philippines after & Chaloupka, 2000).The thorough synthesis in IARC
it was implemented starting 2013. The most recent (2011)concluded that price elasticity of demand for
study on the demand for cigarettes in the country was tobacco products for low- and middle-income countries
done in 2012 by Quimbo et al. Second, this is the varies over a wide range between -0.2 and -1.0.
first study on the demand for cigarettes in the country In the Philippines, there is a dearth of empirical
that used a two-part estimation strategy, estimating evidence on tobacco demand elasticities either using
separately the components of the total price elasticities, individual- and household-level data or even aggregate
namely the price elasticity of smoking prevalence and data. The most recent is the study by Quimbo et
the price elasticity of smoking intensity, both of which al. (2012), which used cross-sectional household
are key parameters in assessing the impact of the policy survey data taken from the nationally representative
reform. The findings and recommendations of the 2003 FIES. The study found that cigarette price has
study will be of invaluable help to the government in a negative and statistically significant impact on
designing the next sin tax policy reform. household cigarette consumption, both for the overall
The paper is structured as follows. The next sample and across income groups. The estimated
section reviews the literature on tobacco demand and price elasticity for the full sample is -0.87, which
taxation. It is followed by a discussion of the data and is close to the upper bound of the range obtained in
methodology, after which the results are presented and studies based from low- and middle-income countries
discussed. The final section presents the conclusion (Chaloupka,Hu, Warner, Jacobs, & Yurekli, 2000;
and recommendations. Guindon, Perucic, & Boisclair, 2003; IARC, 2011).
There is a consensus among policymakers that
Literature Review raising tobacco taxes reduces cigarette consumption.
In fact, among the tobacco control measures, “raising
Due to the adverse health and economic tobacco taxes is the most effective and cost-effective
consequences of tobacco consumption, several strategy for reducing tobacco use” (World Health
studies both in developed and developing economies Organization, 2015, p. 26). This has led to a number of
have examined empirically the extent of the impact empirical studies which examined the effectiveness of
of tobacco price increases on smoking, including the tobacco taxation in cutting cigarette use, one of which
effectiveness of raising tobacco taxes as part of tobacco is by Kevin Callison and Robert Kaestner(2013). Using
control strategy. Although demand for tobacco products data from the U.S. Current Population Survey Tobaccopricedata
price datafrom
fromthethesurvey.
survey.If Ifnotnotaccounted
accountedfor, for,thetheendogeneity
introduce endogeneity
considerable ininself-reported
self-reported
bias in the price price
price datadatamay
elasticity mayestimates. These measures may also be su
introduce considerable introduce
bias price
inconsiderable
data
the from
price bias
the
elasticity in
survey. the price
If
estimates. not elasticity
accounted
These estimates.
for,
measures the
mayThese measures
endogeneity may also be price
in self-reported subjecd
introduceconsiderable
introduce considerablebias biasininthetheprice may give
priceelasticity
elasticity rise to potential
estimates.
estimates. endogeneity
Thesemeasures
These measures maymay ofalso
also thebeprice variable
besubject
subject duealso be subject
to the self-reported nature of th
to measurement/reporting errors since in these household expenditure surveys, it is typical
to measurement/reporting to measurement/reporting
introduce
errors sinceconsiderable
in these errors bias since
in theinexpenditure
these elasticity
price household expenditure
estimates. These surveys,
measuresit ismay
typical tha
alsomabe
totomeasurement/reporting
measurement/reportingerrors errorssince price
sinceininthesethese data from
household
household the survey.
expenditure
expenditure Ifhousehold
not accounted
surveys,
surveys, it it
is is for, the
typical
typical
surveys,
endogeneity
that
that
it is in
typical that
self-reported price data
one family member reports total household expenditures on tobacco and quantity purchased.
12 M.S. Austria, et al purchased. W
one family memberone family
reports
introduce tototal member
household
considerable
reports
measurement/reporting total household
expenditures
bias inquantity
the errorsonsince
price
expenditures
tobacco
elasticityin these
and on tobacco
household
quantity
estimates.
and quantity
expenditure
purchased.
These measures surveys,
Wemay it subjec
also be is typ
onefamily
one familymember
memberreports
reportstotal
totalhousehold
household expenditures
expenditures onon tobacco
tobacco andand quantity purchased.
purchased. We
addressed the endogeneity issue by employing two-stage least squares (2SLS) and twoWe
addressed the endogeneity addressed one the
issue endogeneity
familyby member
employing issue
reports by
totalemploying
two-stage household
least two-stage
expenditures
squares (2SLS) leaston squares
andtobacco (2SLS)
two-step and two-step
and quantity purchas
addressedthetheendogeneity
Use addressed
Supplements, endogeneity
the issuebybyemploying
issue
study employed to measurement/reporting
aemploying
novel two-stage
two-stage
paired least
least
two-stage errors
squares
squares since
(2SLS)
(2SLS)
least in
squares these
andand household
two-step
two-step
(2SLS) and expenditure
two-step surveys,
efficient it is typical tha
efficient generalized method of moments (GMM) estimators.
difference-in-difference (DID) efficienttechnique efficient
tomethod
estimategeneralized
addressed method
generalized
the endogeneity ofestimators.
momentsissue (GMM)
method by estimators.
of employing
moments (GMM) estimators.
efficient
efficient generalizedmethod
generalized
generalized
methodofofmoments
moments(GMM) one
(GMM) family ofmember
moments
estimators.
estimators.
(GMM)
reports total household expenditures two-stage
on tobaccoleast squares purchased.
and quantity (2SLS) andWt
the association between recent large tax increases and New Estimates Using the 2015 FIES
Newthe Estimates
efficient Using the 2015
generalized method FIES ofUsing
moments
cigarette consumption.
NewEstimates
New EstimatesUsingUsing
New Estimates
Results
thethe2015
2015 reveal
FIES
Using
FIES that
2015
increases
addressed theFIES New Estimates
endogeneity issue by employing the(GMM)
2015
two-stageestimators.
FIES least squares (2SLS) and two-ste
For the baseline model using the 2015 FIES, we estimated the following cross-sect
in cigarette taxes are associated with small decreases For the baseline Formodelthe baseline
using model using the 2015the FIES, we cross-sectiona
For the baseline efficient New
model Estimates
using
generalized themethodUsing
2015 the
FIES,
of wethe
2015
moments FIES2015 FIES,
estimated
(GMM) the we estimated
following
estimators.
following
cross-sectional
ForFor the
the baseline
baseline model
model using
using
in cigarette consumption and that it will takemodel: the
the 2015
2015 FIES,
FIES,
sizable we we estimated
estimated thethe following
following cross-sectional
cross-sectional
estimated the following cross-sectional model:
tax increases, on the order of 100%, to decrease
model: model: adult For the baseline model using the 2015 FIES, we estimated the following cross-s
model:
model: New Estimates Using the 2015 FIES (1)
smoking by as much as 5%. log��� � � �� � �� log��� � � �� log��� � � � �� ��� � �� (1)
model: log�� � � � � � log�� � � � log�� � � � � � � �(1) (1)
log�� ��� �� �For� ��the baseline
��� log�� �� �model
���log�� using
� � �the � � 2015 FIES,
�(1)�� � we estimated
�� ����(1) �� �� the� following cross-sectiona
log��
log�� � ���� ������� ���log��
� log�� ��
� ��� � log��
� log�� � �� �� � ����������������� �
Data and Methodology �
�
model:where �� denotes where log��� �denotes� �� � �the quantity
� � � ��of cigarettes
� � the quantity of cigarettes consumed by household �, measured as the numb
� log�� log�� � � � �consumed
�� ��� � ��
where �� denotes the quantity of cigarettes consumed by household �, measured as the number o
Our primary
where data where �� denotes the
source is the quantity
2015consumed
and 2009 by household
of cigarettes consumed , measured
by household as the
�, measured number of as the � packs of
number (20
where � denotes
���denotes thethe quantity
quantity ofofcigarettes
cigarettes consumed
packs (20byby household
household
sticks per measured
�, �,measured
pack); � is asasthe
the thenumber
average number
price ofof of cigarettes in household �’s region; (1) �� i
Family Income and Expenditure Survey packs (FIES)
(20 log��sticks� � per
� � pack);
� � log�� �is� the
� average
log�� � � �price � � of � cigarettes
�
��sticks per pack); � is ofthe
� average price of �cigarettes � in�� household �’s region;as�the
�quantity � is nu
the
� � � � �
packs (20 sticks per pack); where �� denotes
is the average the
price of cigarettes
cigarettes consumed
in household �’s byregion;
household � is �,
themeasured
packs(20
packs
provided by(20sticks
sticksPhilippine
the perperpack);
pack);��� � is
thetheaverage
isStatistics average priceofofcigarettes
price
Authority
annual household in inhousehold
cigarettes inhousehold
household
income; ’s�’s
�’s
�� is region;
region;
aregion;
vector��of � iscontrol
is� isthe the annual
the �
variables household
�
consisted of household
(2011,annual
2017).household
As in Quimbo et al. (2012), annual
the demandhousehold
packs�� is(20 income;
income;
a sticks per �
of pack); is a�vector
� is vector
isvariables ofofcontrol
the consumed
average control price variables
variables
ofofcigarettesconsisted
consisted andofof household
in household region;and
annual household
annual household
income;���is� isa avector
income; vector
income;
where � � denotes
ofcontrol
control
ofhousehold
vector
the quantity
variables
variables
control
of cigarettes
consisted
consisted of�household
ofand� household
household
consisted
and andby householdhousehold �, measured as the�’snumber o
analysis is subject to a number of limitations. First, head’shousehold
characteristics; and � � head’s
������ � characteristics;
� �is
a normally and
distributed disturbance
� �is
household head’s characteristics; and�
�isis�average � ������
adisturbance
� normally � a normally
distributed distributed disturbance term
annual household income; a� is a vector of control indisturbance
variables consisted of househ
�
the unit of analysis
household
household head’s
head’s thehousehold
ischaracteristics;
household.
characteristics;
head’sWhile
and�
and� � �
characteristics;
packs
� �
it can
������
������ (20 � be
� �is�isa and�
�sticks per
anormally
normally
� ������
� pack); � the
�� is
distributed
distributed
normally
disturbance price distributed
termof
term cigarettes disturbancehousehold term �’s region; �� is th
with�constant mean
term and
with variance.
constant mean and variance.
argued that demand for cigarettes is anwith individual
with constant mean annual and constant
household
variance. mean and variance.
householdhead’s The
income; characteristics;
�� is aX consists
vector vector and� � �
ofof ������ �variables
variables
control
� �is
thatacontrol
normally fordistributed
consisted theof householddisturbanan
with
and with
not constant
aconstant
household mean
mean andvariance.
and
choice,variance.
the lack of availability The vector � consists of variables that control for the household as well as house
of individual-level data on cigarette The vectorconsumption The vector
withhead’s
consists constant
of
household
�
variables consists
mean of
andcontrol
that
as well
variables
variance. as
for the
household
that control
household
head’s
for the characteristics
household as well as household
The Thevector
vector� �consists
consistsofofvariables
variables
�
household
that
that controlforforthe
control characteristics;
thehousehold
household
correlated asas
with and�
wellwell �as
� as
cigarette ������ � � �is as
household
household
consumption
well as household
a normally such distributed
asasage, disturbance term
constrains us to use data at the household head’s
level.characteristics correlated withcigarette consumptionsuch age, sex, educat
head’sapproaches
characteristics head’scorrelated
characteristics
The sex,
vector correlated
withcigaretteeducational
� consists withcigarette
of
consumptionsuch variables consumptionsuch
attainment, that
as and employment
control
age, theashousehold
sex,foreducational age, sex,as educationa
status, well as ho
Hence, we characteristics
head’s
head’s followed similar
characteristics correlatedwithcigarette
correlated undertaken
with constant
withcigarette mean
consumptionsuch
consumptionsuch and variance.
asasage, age,sex, sex, educational
educational
attainment, andwhich
employment are status,
all which
categorical are all categorical
variables. The variables.
age of The
the age of the house
by various studies in the tobacco taxation literature
attainment, andcharacteristics
employment
attainment, and employment head’s
status, which are allstatus,
household
which variables.
correlated
categorical
head is(18–29,
are all categorical
withcigarette The age variables.
consumptionsuch
of the The age
household of
asaswell
the household
age,as sex, edu
attainment,
suchattainment,
as those and andemployment
employment
mentioned status,
in status,
the whichareare
which
preceding all The
section,
head
vector
allcategorical
categorical
is coded into
consists
�variables.
variables. The
four categories
of
The variables
ageageof ofcoded
thethethat into
control
household
household fourfor categories
the household (18–29,
30–45, 46–59, and 60 and above, for which we c
househol
and in particular,Bishop, headLiu,
is coded andintoMeng head
four is coded into
(2007)
attainment,
categories
30–45,
four
and
(18–29,
46–59,
categories
employment
30–45, 46–59,
and
(18–29,
status,
60which
and above,
and 30–45,
60 andare 46–59,
above,
for
and
all categorical
which
for which
we
60 andvariables.
weabove,chosefor
Thewhich
age ofwethechoseho
headis iscoded
head codedintointofour
fourcategories
categories(18–29,
(18–29, head’s
30–45, characteristics
30–45,46–59, 46–59,and and 6060and
the correlated
above,forwithcigarette
andabove,
last category for as which
which the webasewechose consumptionsuch
chose
group) and education aschose
age,
into sex, educationa
Page 6
and John (2008).Likewise, the households in the
three Page 6 of 29
two periods (2015 and 2009) in the FIES datahead
attainment, is coded
areand into categories
employmentfourstatus, (none/primary,
categories
which(18–29,
are all
Page
Page
30–45, secondary,
46–59,
6categorical
6 of of
2929
and 60tertiary,
Page
variables.
6 and
of 29above, for
The age which w
of the househol
not identical. The paucity of a longitudinal dataset for which we chose the latter as the base group). Sex
Pag
head is
which could have tracked the cigarette consumption coded into and
four employment status
categories (18–29, are 46–59,
30–45, both dummies
and 60 andindicating
above, for which we chos
patterns of households before and after the sin tax whether the household head is male and has a job,
Page 6 of 2
reform is another constraint. As we argue in this respectively. To account for households’ risk attitude,
study, in the absence of panel data, pooled cross we included a dummy variable indicating the positive
sections can be very useful for evaluating the impact expenditure on any form of insurance. We also control
of a certain event or policy (Wooldridge, 2009). the household’s family size and urbanicity of the
Lastly, there is no available household-level data household’s regional location.
on cigarette prices. Instead, we used province-wide To account for the potential endogeneity of the
average prices of cigarettes taken from the Survey of price variable arising from the self-reported nature
Retail Prices for the Monthly Consumer Price Index of the price data as well as measurement/reporting
(CPI) produced by the PSA. This may give rise to errors, we employed 2SLS and two-step efficient
potential endogeneity of the price variable due to the GMM estimation with regional fixed effects as the
self-reported nature of the price data from the survey. instruments. The disturbance terms of different
If not accounted for, the endogeneity in self-reported individuals within the same region are likely to be
price data may introduce considerable bias in the correlated. The two-step efficient GMM estimator
price elasticity estimates. These measures may also generates estimates of coefficients as well as standard
be subject to measurement/reporting errors since in errors which are robust to both serial correlation and
these household expenditure surveys, it is typical cluster-specific heteroscedasticity (Hayashi, 2000).
that one family member reports total household There are efficiency gains in using the two-step GMM
expenditures on tobacco and quantity purchased. estimator relative to the conventional 2SLS estimator,
We addressed the endogeneity issue by employing and this lies from the use of the optimal weighingAre Filipino Smokers More Sensitive to Cigarette Prices due to the Sin Tax Reform Law? 13
matrix, the overidentifying restrictions of the model, (Wooldridge, 2009). By pooling random samples
and the relaxation of the i.i.d. assumption (Baum, drawn from the same population but at different points
Schaffer, & Stillman 2010). in time, the sample size is increased which results in
more precise estimators and test statistics with more
Elasticities of Smoking Prevalence and Intensity power. Under impact evaluation studies, typically, two
There has been a long tradition of using two-part cross-sectional data sets, collected before and after
econometric models of cigarette demand developed the occurrence of the event, are used to determine the
by Cragg (1971) when using individual-level data effect on economic outcomes. The common technique
(IARC, 2011). This framework is designed to model applied to such impact evaluation analyses is the
smoking prevalence and smoking intensity separately. difference-in-difference (DID) framework, which
The two stages represent the two sequential decisions systematically measures impacttheevaluation difference analyses in the outcome is the difference-in
impact evaluation analyses is the difference-in-difference (DID) frame
an individual faces in consuming tobacco products, variable of interest across groups before and after the
impact evaluation analyses impact isevaluation the difference-in-difference
analyses is measures
systematically (DID)
the difference-in-difference
thewhich framework,
difference in the which (DID)
outcome
namely the decision toimpact
whether consumeanalyses
evaluation or not, and the occurrence
is systematically of an the
difference-in-difference
measures event. For instance,
(DID)
difference inframework,
the outcome a study by Kiel
variable of interest across
among those who have decided to consume tobacco,
systematically measures and theMcClain
difference(1995)
systematically in theand
measures estimated
outcomethe difference
after the the impact
variable
occurrence of the
in interestthat
ofoutcome a new
across
an event.For groups
variable ofbefore
instance, interest
a stu
the decision on how muchsystematically
to consume. measuresThe thefirstdifference ingarbage
stepand after thetheoutcome incinerator
occurrence variable
of anhadofevent.For
interest
on housing across
instance, groups
values
a study before
inbyNorth Kiel and McClain (19
is usually modeled using and after models
the occurrence of an
and event.For
after instance,
the occurrence ofausing
study by
an event.For Kiel and McClain
instance, a study (1995)by Kiel estimated
andhad McCo
andnonlinear probability
after the occurrence of an event.For And
instance,
the impact
over, Massachusetts
a study
that a new by Kiel the and
garbage
impactMcClain athat
DID
incinerator
a analysis
(1995) new
hadestimated
for
garbage cross incinerator
on housing values in No
such as logit and probit specifications due to the sections pooled across various years.
the impact that a newthegarbage impact incinerator evaluation
that newhadanalyses
a Massachusetts on housing
garbage is athe
incinerator values hadin on North
difference-in-difference
forhousing Andover, values (D
binary nature of the first
the decision.
impact that The a second
new garbage step,Massachusetts
incinerator Similarhad
using toonathe approach
housing
DID analysis values for inusing
ofcross
CallisonNorth
sections
DID
and analysis
Andover, Kaestner
pooled
cross sections
across various years.
p
meanwhile, is modeled using ordinaryMassachusetts least squares (2013)
using a DID and Kiel
analysis
systematically
Massachusetts for usingandmeasures
cross aMcClain
sections
DID Similar the(1995),
pooled
analysis across
difference
toforthe we
cross constructed
various
in
sections
approach years.
the outcome
ofpooled
Callison variable
across
and variouof in
Kaestn
(OLS) techniques. The Massachusetts
resultingusing price a DID analysis for cross
elasticity sections
a Similar
two-year to the pooled acrossofvarious
independently
approach pooled
Callison years.andcross-section
Kaestner (2013)ofand Kiel and M
from the first stage is known as the Similar to the approach ofand
andSimilarCallison
after the and
to occurrence
the Kaestner
approach of of
an (2013)
event.For
Callison andofandKiel
instance, and aMcClain
Kaestner study
(2013) by(1995),
Kielcross
and and
Kiel
Similar to price elasticity
the approach of Callisontheand
we constructed
2009 Kaestner
a two-year
2015
(2013) FIES.
we and
independently
The
constructed
Kiel equation
and aMcClain
two-year
pooled
interest
independently
(1995),
cross-section
in pooled
of the 2009 and 20
of prevalence, while the resulting elasticity from measuring the causal impact of the Sin Tax Reform
we constructed a two-year weindependently
the impacta pooled
constructed that a cross-section
two-year new garbageof
independently the 2009cross-section
inincinerator
pooled andhad 2015 onFIES. housing
of the The ofva
2009
the second stage is known as the price
we constructed elasticity
a two-year ofequation
independently Act ofininterest
pooled 2012 givenequation
is measuring
cross-section
in ofbythethe2009 of interest
causal andimpact
2015 measuring
FIES.
of the Sin Thethe causal impact
Tax Reform Actin 20
th
intensity. The total price elasticity of tobacco equationdemand
of interest in measuring
equation the
Massachusetts causal impact
of interest using
in aofDID
measuring
log�� the� �analysis
Sin� Tax
the causal
� Reform
for�impact
cross� Actin
sections
of
��the 2012 �is
�pooled
Sin Taxgivenacross
Reform by vA
equation � � �is ��� �� � ��� � �� � �
is derived by combining the oftwointerest
pricein elasticities.
measuring the causal impact
log�� of the Sin Tax Reform Actin 2012
� � � �� � �� ��� � �� �� � �� ��� � �� � �� log��� � � �� ��� � log���
given by
Other studies have employedlog�� sample selection models
log�� � � � �� � �� ��� �log�� �� ���Similar
��������� �� to�� the
���
��� approach
� � �� of
���log��
� ��� �Callison
�� ���
�� ��� � �� and �Kaestner
� log�� �
� log�� �� (2)�� ���
(2013) and
� � � �� � �� ��� � �� �� � �� ��� � �� � �� log��� � � �� ��� � log��� � � �� log�� (2) � � � ���� ����� log�� �� �
such as Heckman’s (1979) two-step sample selection � �� log��� � � �� ��� � log��� � � � �� ��� � ��
we� constructed a two-year independently pooled
� � � � (2)
cross-section of the
correction model. Known as the Heckit�model, this � �� log�� � � �� ��� � log��
�� log��� � � �� ��� � log��� � � � �� ��� � ����
� � �log�� �� �� � �� ����� ���
� ��� log�� � �� ��� � ��
approach corrects the self-selection problem in the equation� ofthe interest In Equation(2),
in measuring
� the parameter
theiscausal impact
� of interest is �� ,
of the Sin TaxinteraRefo
In Equation(2), parameter of interest �� , the coefficient of the
second stage of the two-part model by including
In Equation(2), theIn Equation(2),
parameter of interest
In Equation(2), thethe parameter ofofinterest
, the coefficient
is �parameter interest of is and
the , the
the
�interaction between
coefficient of th
the inverse mills ratio as anIn additional
Equation(2),variable
the parameter intheofyear
interest
dummy ,log��
is ��variable �the��and
the coefficient
� � ���
year
� �of �dummy
the the
� ��� � �variable
interaction
treatment
���
� �between
� �� variable � �����.
the
����This�� treatment
is
� log��
the
varia
� � � est
DID ��
coefficient of the interaction between the year dummy
the second equation. We employed the these two-part
year dummy variable the���year and the
dummy treatment
variable variable
���the and ��.theThis is the
treatment DID
ofvariable estimator,
Sin��. TaxThis whichis theAc
the year dummy variable ��� and measures variable d15 and themeasures
treatment causal
variable impact
dT. This the
is the Reform
econometric techniques to generate the estimates for the treatment variable
the causal ��.
impact This
of the is the
� Sin DID estimator,
Tax �Reform
�� log�� � � �� ��� Actwhich
� (2012)on
log��� � � cigarette
� �� ��� consu � ��
measures the causal impact DID estimator,
of the the
measures Sincausal which
Tax Reform measures
impact Act of the
the
(2012)on causal
Sin Tax
impact
cigarette
Reformthe
of the
consumption.
Actparameter
(2012)on In�cigarette
the
smoking prevalence and intensity
measures the elasticities.
causal impact of the natural Sin Tax Reform Actliterature,
(2012)on naturalcigaretteexperimentconsumption. literature, In the
�
� is o
Sinexperiment
Tax Reform Act (2012)on the parametercigarette �� is consumption.
often called the average tr
natural experiment literature, Innatural
the natural
the Inexperiment
parameter
experiment Equation(2),�� is literature,
literature,
because the
often
it parameter
the
measurescalled the
parameterthethe parameter
ofeffect
interest
average
�� of is the
often � , called
is treatmentthe coefficient
�treatment effect
or ave
thepolicy
Measuring the CausalnaturalImpact of the 2012
experiment Sin Tax
literature, the because
parameter �� is often called ofthethe average treatment effect
is itoften
measures called thetheeffect average treatment
treatment or policy
effectonbecause average outcomes.
Reform Law on Cigarette Consumption: Difference-
because it measures theiteffect the ityear
of the dummy
treatment variable
or policy on and
average the treatment
outcomes. policyvariable This
Priceis
because it measures the effect of theHas treatment
because
measures
or policythe
measures effect
onofaverage
the
Has effect
of theof���
Responsiveness
outcomes.
the treatment
treatment of or orpolicy
Cigarette on average
Demand
on ��.tooutcom
in-Difference (DID) Analysis Responsiveness Cigarette Demand to Price Increases Changed After
Hassuggest
Responsiveness
average
of Has
Cigarette
outcomes.
measures Demand the causal toofPrice impact of Demand
Increases the Sin Tax
Changed ReformAfter Act
the (2012)on
2012 Sincig
Both theoretical andHas
empirical evidence
Responsiveness of Cigarettethat Demand to
Responsiveness
Price Increases Tax Cigarette
Reform Act?:
After A Chow’s to Price Test Increases
Approach Changed
Tax Reform Act?: A Chow’s Changed Test Approach the 2012 Sin
tax-induced price increase would decrease the demand
Tax Reform Act?: A Chow’s HasTaxResponsiveness
Test Approach
natural
Reform experiment
Act?: A Chow’s of literature,
Cigarette the
Test Approach
Another Demand
parameter
important to �Price
question � isthat often we called
uncovere the
for cigarettes. Hence, weTax
attempted
Reform to test the
Act?: hypothesis
A Chow’s Test Approach
Increases Changed After the 2012 Sin Tax in
Another important question that we uncovered Reformour empirical analysis
that the tax-induced price increase after the 2012 Another Sinimportant question Another
because that weimpact
uncovered
important
it measures thequestion
ofeffectin of
the our theempirical
that
2012 Tax analysis
we uncovered
treatment
Sin or policy
Reform in involves
our
on average
Acton empirical
the theresp
ou
Another important
Tax Reform Act has a negative effect on cigaretteimpact question that Act?:
we A Chow’s
of uncovered
the 2012 Sin in Test
our Approach
Taxempirical
Reform analysis Acton the involves
responsivenessthe of cigarette co
impact of athe 2012 Sinimpact
Another
TaxHas Reform
important
the Acton
question
the Tax responsiveness
thatActon we uncovered
of cigarette consumption of tociga
consumption. Towardsimpactthis end,of wethe constructed
2012 Sin Tax two-
Reform Acton the
ofResponsiveness
responsiveness
2012changesSin
of
ofinCigarette
Reform
cigarette
cigarette
Demand
prices.
consumption
the to Price
responsiveness
Accordingly, Increases
to the we construct
Cha
in our empirical analysis involves the impact of
year independently pooled cross section by poolingchanges in cigarette prices. Accordingly, we constructed a two-year independently
changes in cigarette prices. 2012
changes Sin
TaxinTax
Accordingly,Reform Reform
cigarette we Act?: Act
constructed
prices.
section A Chow’s onpooling
athe
Accordingly,
by Test responsiveness
two-year the independently
Approach
we2009 constructed
and 2015 ofa pooled
two-year
FIES. To crossindep
deter
the 2009 and 2015 FIES, changeswhich are collected
in cigarette prices. before
Accordingly,
section webyconstructed
pooling thea2009two-year andto2015independently pooled
FIES.inTocigarette determine cross whether the total pric
cigarette consumption changes prices.
and after, respectively, the Sin Tax Reform section Actby(2012)
pooling theAccordingly,
2009 andby2015
section by pooling thevery
2009useful
and 2015 FIES.
section weFIES.
Another
pooling theTo2009
demand
constructed determine
important and
has changed
a two-year whether
question
2015 that
FIES.
due the we
toTo total
the
independently pricereform
uncovered
determine
sin
of taxthe
elasticity
in our
whether law,empiofwetoe
the
was enacted. Pooled cross sections can be demand hasTo determine
changed due to whether
the sinthe tax total
reform pricelaw,elasticity
we estimated following mod
demandorhas
pooled
changed duedemand to the
cross
sin
section
taxchanged
ofreform
by Sin
law,towe
pooling
theestimated
thereform2009
the following
and the 2015
model:
for evaluating the impact
demandof ahas certain
changed eventdue to thepolicy impacthas the 2012
sin tax reform law, we estimated the following model:
due Tax
sin tax Reform Acton
law, we responsiveness
estimated the follow of
changes in cigarette prices. Accordingly, we constructed a two-year i
section by pooling the 2009 and 2015 FIES. To determine
Page 9whether
of 29
Page 9 of 29
demand has changed due to the sin tax reform law, we estimated the fo14 M.S. Austria, et al
nalyses is the FIES. To determine whether
difference-in-difference (DID)theframework,
total price elasticity
which primarily designed to capture the structural change in
of demand has changed due to the sin tax reform law, the parameter of interest.
wethe
s the difference in estimated
outcome the following
variable model:
of interest across groups before
Results and Discussion
of an event.For instance, a study by Kiel and McClain (1995) estimated
log�� � � � �� � �� ��� � �� log��� � � �� ��� � log��� � � �� log��� � � �� ��� � log��� �
log�� � � �� � �� ��� � �� log��� � � �� ��� � log��� � � �� log��� � � �� ��� � log��� �
��
w � �garbage
� �� � incinerator
�� ��� � �� log�� had on �� ��� �� log��
� � �housing values �� � � �� ��� � log��� �Descriptive Statistics
log��� �Andover,
in��North
log��� ������������ �� � �
3)
�
��� �� ���
� ���log�� log�� � log�� � �� ���
���� ��log�� �� ��������
log�� �� ��� ��� � log��� � The key descriptive statistics for our
� � analysis (3) 3) samples are
DID �� ��� � log��for � � �sections
cross �� log��pooled � � � ��across ��� � � log��
various �� � � �� ��� � ��
years. 3)
� � �� ��� � �� 3) � 3) presented in Table 1. There were 38,400 and 41,544
� �� ��� � �� � households independently sampled in the FIES
pproach of Callison and where the variables
Kaestner (2013) are andasKiel defined 3) above. In(1995),
and McClain Equation (3), �� measures the difference between
where the variables are as defined above. In Equation
where the variables are as defined above. In 2009 and
Equation (3),2015, respectively.
�� measures There between
the difference are recognizably
e� ���
the variables
� log��� � �are ��as defined
log�� � � �above. ��� �In log��Equation� (3), �� measures the difference between
ove.
ear asIndefined
Equation
are independently (3),
(3),
above. average
�In
pooled measures
�
� measures cigarette
cross-section
Equation
�
(3),the
the �of� difference
consumption
difference
the
�
measures2009 of
between
and
the between
households
2015
differenceFIES. inaverage
The
between significant changes
2009 and 2015 for reasons other than changes in in household income and tobacco
cigarette consumption average cigarette
of households consumption of households in 2009 and
consumption 2015 for reasons other than
over the seven-year period. changes in In 2015,
above.
age In Equation
cigarette consumption (3), ��ofmeasures households the in difference
2009 2015 for in
3) between
and 2009other
reasons andthan changes in
eholds
measuring
nsumptionthe in 2009 and
causal
of households price,
2015
impact for income,
of reasons
the and and
other
Sinother other
than
Tax Reform factors.
changes
Actinother inThe
2012 year
is given dummy
by in variable ��� captures tobacco control
fewer households had tobacco expenditures than in
2015 forinreasons
2009 2015
price,
for reasons
than
income, changes and other in than
price, changes
factors. income,
The year dummy variable ��� captures tobacco control
useholds
, income,in and 2009and and other
other 2015
factors. for reasons
The
factors. year other
The dummy than
year changes
variable
dummy in ��� captures
variable d15 tobacco 2009; the proportion of tobacco-consuming households
control
he ���
�other year �dummy
�factors.
� �� �The ��variable
��� measures
year � �other
���
� ��dummy �captures
log�� than�� �
variable the
tobacco Sin Tax
� ���captures
���� �control
log��Reform Act(2)
� � tobacco
(2012)
controlthat hasbeen implemented over the seven-
. In Equation (3),captures
�� measures tobacco controltobacco
measures measures
other than the other Sinthan
Tax theReform Sin Act (2012) declinedthatby 12 percentage
hasbeen implemented points
over from 65% to 53%.
the seven-
The year dummy
ures other than the variable Sin Tax Reform ���thecaptures difference between
Act (2012) that control hasbeen implemented over the seven-
rm Act Sin (2012) Tax
that hasbeenReform
year period. Act This
implemented (2012)
is a that
necessary
overimplemented has been
step in
the seven- over the seven- implemented
singling out the impact Notwithstanding
of the reform. Thethis
parametersizeableof decline, household
�the � log��
Tax Reform
�� � � � ���
Act (2012)
� log�� � � � �other
that hasbeen
� ��� than
�year � �� changes
period. This is a necessary step in singling out the impact of the reform. The parameter of
olds in
form Act
period.
2009 and
(2012)
This
2015
is aover that the
necessary
for
hasbeenreasons
seven-year
stepimplemented period. over This in
the isseven-a necessary step in expenditures on tobacco and, specifically, on cigarettes,
�in singling out the impact of the reform. The parameter of
na necessary
singling out step the in interest
impact
singling of
singling out the impact is
the
out � ,
reform.
the
� the
impact coefficient
The of
ofisthe parameter
the of
reform. the
of interaction
The
reform. The parameter parameter between
of
of the year dummy ��� and the price
picked up considerably by 62% and 53%, respectively,
pest year
in is dummyout
singling variable
the impact���ofcaptures the interest
reform. tobacco The , the coefficient
��parameter
control of of the interaction between the year dummy ��� and the price
the �� , the of
parameter coefficient
interest
interest isof��the , the interaction
coefficient
coefficient between
of of the
the the year dummy
interaction
interaction between
between ��� and after adjusting for inflation. Tobacco expenditures as
the price
interactionofbetween
oefficient the interactionvariable
the yearbetween dummyThis
log���. ���parameter
the and dummy
year the pricemeasures
��� and the change
the price in the price elasticity of demand from 2009
Act (2012) that the
hasbeenyear dummy andover
variable the price
log���. variable
This parameter log(P). a proportion
This the change in the price
measures of total household
elasticity of demand expenditures
from 2009 also rose
eble
le interaction
log���.
��� thebetween
and This parameter
treatment theimplemented
year
measures
variable dummy��. theThis ���
change
the
is and
the
seven-
in DIDthe price
the price elasticity
estimator, whichof demand from 2009
sres the change
parameter inparameter
measures the before
price
the change measures
elasticity
the in taxthe of the
demand
reform
price changeto from2015
elasticity in2009the price elasticity
post-tax
of demand reform.
from 2009 We by a percentage point
of hypothesized that ��� is negative and in 2015 from its value in 2009.
ingling outchange
the impact of the elasticity
reform. The
before parameter
thethe taxtax of
reform Consequently, household �
that income expanded and by nearly
sures
e theofthe
pact tax
the Sinreform demand
Tax in tothe 2015
Reform
from
price 2009
post-tax
Act (2012)on
before
of demand
reform. cigarette Wefrom 2009to to
reform
hypothesized
consumption.
2015
2015 post-tax
post- reform.
In the��� is negative and
that
We hypothesized �� is negative
-tax toreform.
rm statistically
Wetaxhypothesized
2015 post-tax reform.
reform.We We that ��� is that
significant,
hypothesized
hypothesized negative is,that
cigarette
that ��� is
and isconsumption
negative
negative and and 4%. Tobacco
of households has become expenditures
more responsive accounted for 1% and 2%
eractionreform.
ost-tax betweenWe thehypothesized
year dummy statistically �����andis the
that price and
significant, that is, cigarette consumption of theofhousehold’s
households has become
annual more responsive
income in 2009 and 2015,
tically significant,
rature, the parameter statistically
that � is,� cigarette
is oftensignificant,consumption
called thatnegative
� the is, cigarette
of households
average treatmentconsumption
has become
effect more responsive �
e consumption
nt, that is, cigarette to price has
of households
consumption increases
become
of after
households morethe reform.
responsive
has become Tomoredetermine
responsive the statistical significance of � , we used
respectively. Our demand analysis�focuses on cigarettes
the change in the ofpricehouseholds
elasticityhas has
ofto become
demand
price more from
increasesmore 2009afterresponsive the reform. to priceTo determine
�
the statistical significance of �� , we used
ette ice consumption
effect of the treatment
increases after of households
theorreform.
policy onTo become
average
determine outcomes. the responsive
statistical the significance of ��� , as
wethey
used account for more than 90% of households’
To determine the increases Chow’s
statistical aftertest, the
significance which
fter the reform. To determine the statistical �significance of �� , we used
reform. isof ��To
primarily
, determine
we designed
used �to statistical
capture the structural change in the parameter of
significance �
of �Chow’s� , iswenegative
used expenditures on tobacco products.
x. To reform.
f Cigarette
w’s determine
test,
We hypothesized
which Demand istheprimarily
statistical
to Price
that
significance
Increases
designed
test,
of ��which
Changed
to capture ,Chow’s
�the
and
weAfter used
structural
test, which
is primarily
the 2012
change
designed is to capture the structural change in the parameter of
Sinin the parameter of
nedprimarily
is to capture designed interest.
the structural
to capture change in the parameter
the structural change of in the parameter of
onsumption
signed of households has become interest.
more responsive
est.how’s to Test capture
Approach the structural change in the parameter of
Table 1
determine �� , we used
ant questionthethat statistical
Meanwe uncovered significance
Household ourofempirical
inIncome �and Expenditures
analysis involves on Tobacco the Products (PhP)
Results and Discussion
dn to capture the Actonstructural Results and Discussion
Tax Reform the change
responsiveness
Results in the and parameter
of cigarette
Discussion of consumption to
esults and Discussion Variable
Descriptive
Results and Discussion Statistics Mean
Descriptive Statistics
Results
ces.
riptive and Discussion
Accordingly,
Statistics we constructed a two-year independently pooled cross 2009 N 2015 N
cs The key descriptive statistics for our samples are presented in Table 1. There were 38,400
The key descriptive statistics for our samples are presented in Table 1. There were 38,400
2009 Theand key2015descriptive Annual
FIES. To household
determine
statistics for our income
whether
samples theare total price elasticity
presented in Table 1. of There were 38,400 195,811.50 38,400 247,555.60 41,544
samples are and
presented 41,544
riptive statistics for our samples are presented were
r our in Tablehouseholdsindependently
1. There in Table 38,400 1. There sampledwerein38,400the FIES 2009 and 2015, respectively. There
Proportion of households and 1. 41,544 with tobacco expenditures (%)
householdsindependently sampled in the FIES 65.00
2009 and 38,400
2015, respectively. 53.00There 41,544
elts
41,544 toand
for our Discussion
samples are presented
the householdsindependently
sin tax reform law, we in Table
estimated
sampled There
the thewere
infollowing FIES38,400 model:
2009 and 2015, respectively. There
sampled in the FIES
ldsindependently are
Household
sampled2009 recognizably
inand the2015,FIESsignificant
expenditures onand
respectively.
2009 tobacco changes
2015,There productsin household
respectively. There income and tobacco consumption
2,180.08 24,962 over the 4,314.67 22,095
yecognizably
sampled insignificantthe ShareFIES of 2009 and are recognizably
2015, respectively. significant
There changes in household income and tobacco consumption over the
household
changes expenditures
in household income on tobaccoand tobacco products in overall over the
consumption
in household
nificant changes income seven-year
and tobacco
in household period.
income consumption
andIn 2015,tobacco over fewerhouseholds
the
consumption overhadthetobacco expenditures 1.87 than in24,962 2009; the 2.88 22,095
expenditures (%) seven-year period. In 2015, fewerhouseholds had tobacco expenditures than in 2009; the
ur samples
s in household
n-year are
period. In presented
income2015,and in Table 1. There
tobacco consumption
fewerhouseholds were 38,400 over the
onhad tobacco expenditures
Page 9 of 29than in 2009; the
ouseholds had tobacco
In 2015, fewerhouseholds Household
proportion expenditures
expenditures
had oftobacco
tobacco-consuming incigarettes
thanexpenditures 2009; the households
than in 2009; declined
the by 12 percentage 2,106.21 24,962
points from 65% to 3,927.89 22,095
mpled in the FIES
had Household2009 and 2015,
expenditures proportion
respectively. on cigarsof tobacco-consuming
There households declined by 12 percentage points from 65% to
rhouseholds
ortion of tobacco-consuming tobacco expenditures householdsthan declined in 2009; by 12thepercentage points from 65% to 9.95 24,962 311.98 22,095
useholds
co-consuming declined by53%.
households Notwithstanding
12 percentage
declined bypoints
12 this
from
percentage sizeable
65% points decline,
to fromhousehold
65% to expenditures on tobacco and, specifically,
Household expenditures 53%.points on chewing tobacco
Notwithstanding 33.59 22,095
household
ouseholds
Notwithstanding income
declined and
this tobacco
bysizeable
12 percentage consumption
decline, household over
from 65% this
the
expenditures to sizeable decline, household expenditures on tobacco and, specifically,
on tobacco and, specifically, Page 10 of 29 41.20
cline,
g thishousehold
sizeable decline, Household
expenditures household expenditures
on tobaccoexpenditures on other
and, specifically, tobacco products
on tobacco and, specifically, 63.91 24,962 22,095
Page 10 of 29
eholds had
decline, householdtobacco expenditures
expenditures than inand,
on tobacco 2009; the
specifically, Page 10 of 29
Note: All figures are reported in nominal terms.
Page 10 of 29 The Page
mean10
expenditures
of 29 are calculated for the subsamples for which household expenditures
on tobacco is nonzero.
holds declined by 12 percentage points from 65% to of 29
Page 10
Source: Authors’ calculations using data from the Family Income and Expenditure Survey (FIES) provided by the Philippine Statistics
Authority (PSA).
e, household expenditures on tobacco and, specifically,
Page 10 of 29Are Filipino Smokers More Sensitive to Cigarette Prices due to the Sin Tax Reform Law? 15
Price Elasticity of Demand for Cigarettes— Consistent with economic theory and studies in
New Evidence the literature, poor households are relatively more
We presented novel elasticity estimates in Table 2 responsive to cigarette price increases than richer
and Table 3 using the 2015 and 2009 FIES, respectively. households (see, for instance, Barkat, Chowdhury,
Our elasticity estimates provide support to the Nargis, Khan, & Kumar, 2012; Townsend, Roderick,
theoretical and empirical consensus that cigarette & Cooper,1994). Cigarette demand is price elastic for
consumption declines when cigarette price increases. households in the lowest income group (-1.254) and
We found a negative and statistically significant impact inelastic for the relatively richer households (-0.968,
of cigarette price on consumption, with the estimated -0.869, and -0.598).Consequently, deprived households
overall price elasticity equal to -0.93, suggesting are more responsive to income increases than the well-
that cigarette consumption is price inelastic. Hence, off. Estimated income elasticities decline as income
given a 10%-increase in average cigarette prices, increases. The increasing trend in income elasticities
demand declines by 9.3%, everything else constant. is also reflected across income groupings.
Historically, tobacco products typically exhibit Using two-part econometric techniques, we
relatively inelastic demand due to their addictive nature estimated separately the two components of price
and the unavailability of close substitutes. For many elasticity of demand for cigarettes, that is, price
low- and middle-income countries where cigarettes are elasticities of smoking prevalence and intensity.
generally less affordable than in advanced countries,
A key step in this exercise is to check whether
elasticity estimates lie between -0.2 and -0.8(see
Warner,1990; Blecher & van Walbeek, 2004, 2009).
selecting only households with positive cigarette
A comparison with the estimates for 2009 shows consumption in the regressions introduces sample
that cigarette demand has become more responsive selection bias. Table 4 presents the estimates of
to price increases. This increase in cigarette demand the Heckman model, indicating that the estimated
elasticity could be attributed to various factors such inverse mills ratio in column (1) is statistically
as the permanent increase in cigarette prices brought significant and positive. Therefore, selecting
about by the significant rise in excise taxes from the only smoker households to be included in the
reform as well as the increasing presence of close regressions and, to the same effect, ignoring those
substitutes such as electronic (e-) cigarettes. households with zero cigarette consumption,
Our estimated income elasticities, meanwhile, fall would result to sample selection bias. Thus, to
in the lower estimate at 0.56, indicating the positive and
avoid sample selection bias, we included also the
statistically significant relationship between income
and cigarette consumption. Hence, a 10%-increase in
small proportion of the sample with zero cigarette
average income will yield a 5.6% increase in cigarette consumption. Since we log-transformed the
demand, everything else constant. Compared to 2009, cigarette consumption variable as the dependent
the estimates show that the responsiveness of cigarette variable, all zero-valued observations will be
demand to income increases significantly went up after missing values. We resolved this issue in two
the reform. Consistent with the findings of Ulep (2015), ways: (i) setting these observations equal to zero;
cigarettes in the Philippines became less affordable and (ii) employing the Heckman sample selection
after the reform as shown by the increase in relative correction method. The estimates are robust across
income prices (RIP). This means that the proportion the two approaches.
of income required to purchase cigarettes rose, making In Table 5,we present the estimates of prevalence
demand more responsive to income increases. and intensity elasticities as the average marginal effects
Our estimated income correlates suggest that of the two-part estimation technique. The results reveal
households with household heads who have jobs but that the elasticity of smoking intensity dominates the
did not finish college are more likely to consume elasticity of smoking prevalence, suggesting that of
cigarettes. Our estimates also confirm the hypothesis the total effect of cigarette price increase on demand,
that risk-averse households—those with expenditures it is the decrease in consumption by smokers (smoking
on any form of insurance—are less likely to have intensity) that account for much of the decline in
expenditures on cigarettes. cigarette consumption rather than the deterioration in16 M.S. Austria, et al
Table 2
Estimates of Overall Price Elasticity of Demand for Cigarettes - 2015
(1) (2) (3) (4) (5)
Income Deciles
VARIABLES Overall 1st-3rd 4th-6th 7th-9th 10th
ln(cigarette price) -0.927*** -1.254*** -0.968*** -0.869*** -0.598***
(0.0406) (0.0623) (0.0800) (0.0953) (0.189)
ln(HH income) 0.557*** 0.742*** 0.656*** 0.538*** 0.362***
(0.0136) (0.0293) (0.0294) (0.0354) (0.0674)
HH age 18-29 0.0322 0.143** 0.104 -0.0902 0.253
(0.0449) (0.0629) (0.0835) (0.101) (0.280)
HH age 30-45 0.0498* 0.198*** 0.0936** -0.0132 -0.121
(0.0262) (0.0464) (0.0445) (0.0488) (0.116)
HH age 46-59 0.0802*** 0.117** 0.120*** 0.0967** -0.0361
(0.0243) (0.0460) (0.0430) (0.0419) (0.0764)
dummy HH is male 0.128*** 0.111** 0.0743 0.113*** 0.113
(0.0254) (0.0486) (0.0454) (0.0427) (0.0792)
HH education: none to primary 0.246*** 0.217*** 0.218*** 0.199*** 0.229**
(0.0282) (0.0670) (0.0488) (0.0467) (0.101)
HH education: secondary 0.207*** 0.164** 0.147*** 0.110*** 0.250***
(0.0265) (0.0677) (0.0481) (0.0405) (0.0752)
dummy HH has a job 0.0313 0.0988* -0.00820 0.0341 0.0877
(0.0281) (0.0571) (0.0515) (0.0454) (0.0834)
dummy household has insurance -0.0266 0.110** -0.0709** -0.0521 -0.00908
(0.0207) (0.0460) (0.0310) (0.0350) (0.114)
ln(family size) -0.106*** -0.426*** -0.161*** 0.00656 0.261***
(0.0197) (0.0319) (0.0343) (0.0402) (0.0740)
dummy household from urban 0.0606*** 0.0177 0.0207 0.169*** -0.0748
(0.0197) (0.0374) (0.0316) (0.0348) (0.0683)
Constant 1.805*** -0.0650 0.155 5.826*** 8.656***
(0.248) (0.655) (1.125) (0.967) (1.508)
Observations 19,662 6,478 6,591 5,242 1,351
R-squared 0.083 0.108 0.035 0.041 0.063
*** = significant at 0.1%, ** = significant at 1%, * = significant at 5%. Robust standard errors in parentheses. Dependent Variable:
ln(number of pack of cigarettes consumed). HH is household head.
Notes: We test for endogeneity of the cigarette price variable for the overall sample and each four subsamples, and find that the null of
exogeneity is rejected. Hence, generalized method of moments (GMM) estimation is used where the instruments employed are regional
fixed effects.
Sources: Authors’ calculations using data from the Survey of Retail Prices of Commodities for the Generation of CPI and Family
Income Expenditure Survey (FIES), both provided by the Philippine Statistics Authority (PSA).You can also read
