Introduction
The
significance of smoking as a contributor to preventable illness
and premature death is now widely accepted and proactively
addressed by governments globally.
Annually,
according to the Department of Health, smoking alone causes around 79,000
deaths in England, with a predicted cost to the National Health Service (NHS)
of £2.5 Billion in 2015.
Over the past decade,
a broad range of government and health group interventions have
been implemented and evaluated. These have been especially targeted
at young people due to the perpetuation of smoking into adult life,
which arises the issues of the long-term effects associated with health
loss and medical costs.
There is a broad range of
literature surrounding this topic which explores many different aspects of the
specifics that affect smoking behaviour and the most effective elements in
reducing demand.
The primary purpose of
this paper is to observe the effect of cigarette taxation on cigarette
consumption within the United Kingdom (UK). Many of the studies within the
literature have taken place in Asia or the United States of America; there have
only been a small number of relevant papers that use up-to-date data within the
UK. There is, therefore, a need for further analysis to evaluate the
differences specifically within the UK market. This paper will evaluate the
significance of increasing taxation on cigarette consumption and whether these
results coincide with findings from other countries. A logical explanation of
any differences that may exist will also be presented. The model was estimated
using an OLS estimate to regress the cigarette consumption in the UK against
multiple explanatory variables.
The results from the OLS
regression show that as cigarette prices increase by 10%, the level of annual
cigarette consumption per person decreases by 21.4%. This estimate supports the
ideas of economic demand and supply theory, which is reflected in the other
empirical evidence. Alongside this, other literature pieces reviewed in the
research show that as price increases, demand decreases. The size of the
estimate can be attributed to several economic theories; it is therefore
directionally correct and consistent with the other literature pieces reviewed.
The main limitation of
the empirical model is the possibility of heteroscedasticity, autocorrelation,
and multicollinearity. However, it is commonly accepted that such problems
arise in time-series data, and therefore several tests will be used to
eliminate them as practicably as possible. The tests used to deal with these
problems are the White Test, robust standard errors, Breusch-Watson test, and
multicollinearity procedures.
The remainder of the
paper will be organized as follows: Section 2 will discuss the other literature
pieces; Section 3 describes the data sources and includes a thorough
description of the variables and their relation to the model; Sections 4 and 5
provide an outline of the empirical model and evaluate the results; Section 6
discusses the limitations surrounding the project and what areas could be
improved; and lastly, Section 7 summarises the key findings and
recommendations.
Literature Review
The Effects of Smoking
Smoking affects everyone who is around
the smoker, both emotionally and physically. Many people die prematurely due to
smoking-related illnesses, which are preventable by not smoking, seen in Hu et
al. (2002). People may also be affected by the sight of their loved ones
suffering due to a smoking-related illness; this can also cause psychological
problems later in life, which reduces their ability to work productively. This
causes a loss of earnings to the individual and tax revenue to the economy as
there would be a reduction in productivity and any loss of skills that the
individual may have. The health services, particularly in the UK, are under
immense levels of pressure because of the real term decrease in their health
budget that the health service receives annually. In 2017 the Institute of
Economic Affairs (IEA) discovered that the government spends £3.6 billion on
smoking-related diseases per year. The individuals that need medical attention
due to smoking are putting the NHS under more stress than before and possibly
stopping other individuals from receiving important surgeries.
Deforestation is a significant problem
as well, with many forests being cut down for manufacturers to grow more
tobacco. In 2011 the World Health Organization (WHO) estimated that the area
used to grow tobacco is approximately 4,200,00 hectares, which makes up about
1% of all land globally; however, this land could be used for essential goods
instead. Additionally, the distribution of millions of cigarette packets across
the world creates a lot of air pollution through transportation. Furthermore,
the chemicals that are needed to grow tobacco are very harmful to the
surrounding environment and often leak into the water system, which destroys
ecosystems. This devastation is further perpetuated by the immense number of
cigarette butts that are thrown onto the floor, according to Novotny (2015).
Second-hand smoking has historically
been a significant cost to individuals and previously unrecordable. However,
once this was discovered and deemed substantial, it led to a smoking ban in
public places in 2007, as well as people not being able to smoke in the car
with an individual who is under 18 years of age.
Governments have a difficult task as
they need to weigh up the expenses involved in responding to the cost to the
NHS and the loss of productivity of the economy. However, governments do take
into consideration that the people who die prematurely from smoking-related
illnesses do not draw from their pension or other governmental resources, thus
reducing the government expenditure on pension. According to the IEA, it is
estimated that the government saves £9.8 Billion a year from the premature
deaths of smokers. Additionally, it is estimated that there is a net saving of
£14.7 billion per year from premature smoking-related deaths. This data shows
that it is in the government’s best interest financially to maintain the
current situation.
Addictiveness of Smoking
One of the reasons why cigarettes are
overly addictive is because of the nicotine that alters the chemical balance in
individuals’ brains. According to the NHS website, the chemical alterations in
the brain change the smoker’s mood and concentration levels. Additionally, the
act of smoking regularly can become a habit that is very difficult to break.
Therefore, many individuals are not addicted to smoking itself, but to the act
of smoking and the social inclusion that it brings to their working lives.
Cigarettes are addictive, and therefore many academics believe
that addictive goods do not strictly adhere to the laws of economic behaviour.
However, Becker et al. (1988) state that people behave rationally by maximising
their utility for the good as they would for a normal good. Despite this, some
conditions need to be considered with addictive goods that make them differ
from normal goods with significant external factors. The main external factors
that Becker et al. (1988) mention are age and stressful events, which cause the
demand for addictive goods to increase. It is noted that normal goods will not
be as significantly affected by these factors. Historically, age has been found
to influence individuals to consume more cigarettes in their younger years due
to society and peer pressure. Conversely, in later years evidence has shown
that people will consume more addictive goods as the negative aspect of dying
early is not as relevant as before. However, it is worth noting that the older
population has not had the benefit of education on the dangers of smoking to
the same extent as the younger generation. Nevertheless, it is evident that
stressful events lead individuals to crave addictive goods more to help deal
with their problems.
How Different Consumers Are
Affected
Unsurprisingly, price increases affect
different socioeconomic groups differently. The groups affected most are
younger people (<24 years old), low-income individuals (<$25,000 a
year), and people with little educational qualifications (people who exit
education prior to A-Levels or equivalent). Franz (2008) states that younger
people are affected more by price change making them more price responsive than
older people. This is also seen in Bader et al. (2011), though it is worth
noting that the sample size used was very small. Franz (2008) states that older
individuals are also significantly affected by price increases due to the
reduction in income in later life.
Wasserman et al. (1991)
discovered that married people smoke less than un-married people, but this
could be correlated and not causal. Harris and Chan (1999) uncover that as age
increases from 18-29 the price elasticity dropped significantly from -0.831 to
-0.095. This reduction shows that younger individuals are greatly affected by a
change in cigarette prices compared to other age groups.
Sharbaugh
et al. (2018) contend that the lowest income
individuals were affected the most, which is as expected due to individuals
having such a low disposable income budget. This small amount of disposable
income means that there will be a trade-off between other goods that are much
more useful; this is similar to
Chaloupka
et al.
(2002). Goodchild et al. (2016) created a simulation model to test how low,
lower-middle, upper-middle, and high-country-income groups would react to a $1
increase in excise tax to all countries. The data found that the country in the
low-income group reduced their annual consumption by 32%, which was
significantly higher than all the other countries’ income groups. This shows
that the lower-income group is the most responsive to the price changes as they
alter their annual consumption the greatest. However, this does not include any
illegal purchases that take place, which would be especially prominent within
countries with low-income groups; therefore, this large decrease in annual
consumption may not be as large as it states. As expected
Chaloupka
et al. (2002) discovers that higher prices prevent relapse among past smokers.
According to Becker et al. (1988);
Chaloupka
(1991),
heavy abusers of addictive goods drastically change consumption according to
the price. Wasserman et al. (1991) find that individuals with higher education
consume fewer cigarettes than someone with a lower education; this is also seen
in Bader et al. (2011);
Chaloupka
et al. (2002).
Lee et al. (2005) look at the
significance of the 5 New Taiwan Dollar tax increase on domestic and imported
goods, cigarettes, and cigars. This tax increase works out as approximately a
20% increase in price. Before the tax increase, only 7-10 minutes of work was
required to earn enough to buy a pack of cigarettes. Lee et al. (2005) use a demand
model that estimates the price and expenditure elasticity of cigarettes. As
expected, the results were that the consumption of cigarettes was reduced by
18%; however, the cigarettes could have been illegally imported, and therefore
the reduction of 18% would most likely be a few percent lower. The possibility
of illegal importing has not been factored into the model especially with big
illegal imports coming from neighbouring countries. This large tax hike has
shown that this will dissuade non-smokers from beginning to smoke in the first
place, as seen in
Sharbaugh
et al. (2018).
Keneggarpanich et
al. (2016) show that after a 9.7% cigarette tax increase, 48% of consumers
decreased their cigarette consumption or altered their preference from premade
to roll-your-own cigarettes. This increase in tax forces consumers to change
their preferences if they still want to consume the same number of
cigarettes.
If instead, the direct
fiscal cost outweighs the benefit, which is, in this case, enjoyment of
smoking, then the consumer will stop altogether. However,
Stehr
(2005) states that large tax hikes can also cause individuals to order
cigarettes online or travel to other states at a relatively lower price. As of
the start of 2007, 67% of states had delivery laws on cigarettes to reduce the
problem of individuals buying cigarettes from different states for a cheaper
price (
Chriqui
et al. (2008)). If an individual has
purchased a tax-free item online in the United States of America, they are
supposed to report the purchase to the state tax agent who will then work out
the tax that is on the good and directly charge it to the individual. However,
this is not strictly enforced, and as a result, many sales often are hidden and
not reported. The main problem with this is that it is in the United States of
America, where different states have different taxes on cigarettes; however, in
the UK, the tax rate on cigarettes is flat in all counties.
Stehr
(2005) also goes on in his paper to mention that effective policies need to be
paired with a tax increase for there to not be a large amount of tax evasion;
an example of this would be a more severe punishment for smuggling cigarettes
over the border or through customs. The UK cigarette black market in 2016-2017
was valued at £2.5 Billion by (Her Majesty’s Revenue and Custom) HMRC, which
shows that there is a significant black market for cigarettes. This is more
than likely due to the extremely high tax rate compared to the rest of the
(European Union) EU. According to the tobacco manufacturers’ association, the
UK has the highest rate of taxation within the EU, which leads to more
individuals buying cigarettes illegally to get around the high prices.
Additional research by governments and tobacco firms to find different ways in which
this black market can be eliminated is key to increasing the level of revenue
by the government to pay for the health costs of smoking.
Is Cigarette Tax Progressive
or Regressive
Regressive taxation is where the amount
of tax is a relatively higher percentage of a low-income individual’s
disposable income than a high-income individual, according to
Vickrey
(2008). Therefore, regressive taxation impacts the
poor more than the rich. As Colman (2004) states, individuals who are not as
wealthy are more price-sensitive than wealthy individuals, and therefore the
tax increase would have a more negative effect on the former. Similarly,
Remler
(2004); Goldin (2011) find that the traditional
economists’ view is that cigarette taxes are very regressive as the tax
significantly affects the poor, whereas the rich are less affected.
In contrast, progressive taxation is where
the “amount of tax paid as a proportion of the tax base rises with that base” (
Vickrey
, 2008, P1). However, under specific
behavioral
economic models, taxes to a small number of
smokers can be progressive when using “extreme elasticity estimates” (Colman,
2004, P4).
Additionally,
Remler
(2004) finds that as lower-income individuals are more price responsive, as
aforementioned, they would be more likely to cut back on how many cigarettes
they smoke or quit altogether. If the difference in the lower-income consumers’
tax expenditure is lower than the difference in the higher-income consumers’
tax expenditure, the tax would be progressive. However, this is unlikely to
occur as it is unrealistic, given real-world consumer behaviour. Also,
alongside tax increases, government policies may be connected; for example,
counter-advertising on cigarette packaging may cause a large deterrence in
smoking, which has not been accounted for in the model above.
Price Manipulation
Only a select few cigarette companies
control the market; this is called an oligopoly. Price and product manipulation
within markets can cause many problems when governments attempt to raise prices
by raising taxation. The firms can manipulate their product, which will lead to
the firm not needing to reduce their price.
The cigarette firms are divided into
high-end and low-end manufacturers consumed by high-income and low-income
individuals, respectfully. Becker (1994) finds that many cigarette firms will
manipulate the price even when the government has not put another tax on
cigarettes. This is shown by Hiscock et al. (2017): high-end firms will pass on
the tax and add a small amount on top, which is disguised as a hidden increase.
This increase causes the price of cigarettes to exceed the value of the tax
increase. The firm can raise its price artificially, as cigarettes are
addictive and have an inelastic demand. This is shown as a 10% price increase
in cigarettes that have been directly passed onto consumers and leads to an
average 4% reduction in consumption (Bader et al. 2011). Consumers will still
buy the good to a certain extent; this occurs predominantly in the higher-end
cigarette manufacturers due to the high inelasticity (Wasserman et al. 1991).
Alternatively, Hiscock et al. (2017)
state that in serving lower-end markets, cigarette firms will absorb the tax
increase as the elasticities are different in the medium and high-end markets,
as portrayed by Goodchild et al. (2016). If the demand is inelastic and the
supply is elastic, the cigarette tax will be passed directly to the consumer,
and the tax increase would come out of the consumer surplus. However, if the
supply is inelastic, then the tax increase will come out of the cigarette
firms’ (producer) surplus. If the cigarette submarket is inelastic, then the
cost is either directly absorbed by the producer or passed to the consumer in
another way. Many of these ways are seen in Hiscock et al. (2017), and they
include reducing the weight of tobacco packs or reducing the number of cigarettes
in a pack by one to combat the tax increase. The consumer will still buy the
cigarettes for the same price, but there will be one or two fewer cigarettes
per pack. Additionally, to add value, firms will bundle certain products
together, especially with roll-your-own cigarettes, to make it seem as if the
consumer is saving money.
This is, however, a necessity for the lowest-priced products. As a result, the
cheapest cigarette brand will get most of the lower-income market, as shown in
Hyland et al. (2005). This is due to the ease of substitutability within the
market. Low-income consumers will generally buy the cheapest variant of
cigarettes even if that means purchasing illegally, which often happens (Hyland
et al. 2005). As a result, people are less likely to quit if there is an easily
accessible and cheap product black market. The black markets are more easily
reachable in Asia, according to Lee et al. (2005).
Legally there are many different substitutions
for a cigarette pack of 20, such as roll-your-own cigarettes and E-Cigarettes.
The substitutes for cigarettes can affect cigarette consumption in varying
degrees. Over time, the most popular substitute for pre-rolled cigarettes has
altered from roll-your-own cigarettes to E-Cigarettes. This is seen in
Caponnetto
et al. (2013); it was found that smokers who do
not want to quit smoking value E-cigarettes as a good substitute for
cigarettes.
Summary of the Literature
The literature investigating cigarette
consumption covers a variety of different areas, not only economics. All the
findings from the literature are conclusive that an increase in taxation leads
to a decrease in consumer consumption. The level of the reduction in
consumption, however, depends on the socio-economic background of the
individual. Firms can get around the problems of increasing tax rates by
manipulating their cigarette packs, and as a result, their product can remain
cheap and competitive.
Data Description and Sources
The data for the daily consumption was
sourced from the Office of National Statistics (ONS). It is comprised of data
gathered between 1980-2018. However, the data was only collected on the even
years between 1980-2000; therefore, the odd years in that date range were
calculated as an average of the year before and the year after. The daily
consumption was then multiplied by the number of days in that year and
considered leap years. The average price per 20 cigarettes was from the ONS and
between 1987-2017. There was not another reliable before 1987. The GNI PPP
(Gross National Income-based on Purchasing Power Parity, using the Atlas
Method) was from the World Data Bank, which was converted from US Dollars to
Pounds, spanning from 1962-2018. The percentage of the rural population was
also from the World Data Bank and has data from 1960-2018. However, only data
from 1980-2018 was needed, as the average price restricted the years able to be
used. The tax amount was sourced from the tobacco statistics tables by HMRC.
The tax amount was per 1000 sticks; it was then equated to the same number of
cigarettes per pack, which is 20 by dividing it. This data is from 1978-2018. I
researched all policies on the government’s legislation website and selected
all articles relating to tobacco between 1980-2018. If the policy affected the
quantity or price of cigarettes, it was regarded as a relevant policy.
Due to certain data limitations, the
average price sample is from 1987-2017, which leaves 31 annual observations.
All data used is based on the United Kingdom, which includes Northern Ireland,
England, Scotland, and Wales.
Table
1
- List of Variables
Meaning
|
Source
|
|
C
|
The
annual cigarette consumption per person in the UK.
|
ONS
|
P
|
The
price level after it has been decorrelated from taxation in the UK.
|
ONS
|
GNI
|
Gross
National Income per person in the UK.
|
The
World Data Bank
|
Rural
|
The
percentage of individuals who live in rural areas in the UK.
|
The
World Data Bank
|
Pol
|
The
number of policies that affected tobacco consumption/ production excluding
taxation
|
GOV
Legislation
|
Tax
|
The
amount of taxation on cigarettes in the UK in Pounds Sterling.
|
HMRC
|
ln C
t
= β
1t
+ β
2
P
t
+ β
3
LnGNI
t
+
β
4
Rural
t
+ β
5
Tax
t
+
ε
t
(1)
t
(years) where t = 1, 2, 3 … 31
The data
collected created a cigarette demand model, with one dependent variable and
five independent variables. The dependent variable (
Ct
)
is the annual consumption of cigarettes per capita. This is measured by
calculating how many cigarettes are smoked a year and dividing it by the
population to get an average number of cigarettes smoked per capita at time t.
The independent variables chosen are the price of cigarettes in the UK, the GNI
(Gross National Income) per capita for the UK, the percentage of the rural
population in the UK, the number of policies created by the UK and EU (European
Union) that would affect cigarettes, and the amount of tax on cigarettes. (
Pt
) Price is the average price of a 20 pack of
cigarettes at time t; this has been taken away from tax at time t to eliminate
the collinearity between the two variables. (
GNIt
) GNI PPP (Gross National Income
per capita) is the value of a country’s income divided by the population of the
country at time t. (
Ruralt
)
Rural is the percentage of people in the United Kingdom that live in a rural
environment at time t. (
POLt
)
is the number of policies created in time t that would influence cigarette
prices either supply or demand, the lagged term will be a period for the
effects to take place. (
Taxt
)
Tax is the amount of tax that is added to the cost of a pack of cigarettes.
This was initially per 1000 cigarettes; however, it was divided by 50 to align
with the 20 per cigarette pack price at time t.
Other Relevant Variables
E-cigarettes have
been excluded as the data started in 2011, which would not allow enough data
points to create a reliable time series data regression. The E-Cigarette data
is an important variable as it would be able to show the change in E-cigarette
consumption and its potential impact upon cigarette consumption. Another
substitute omitted is cigars. After research, it was discovered that cigars
were not a close enough substitute to cigarettes as originally thought. Other
potential substitutes included nicotine gum which Shahan et al. (2000) find to
be a weak substitute for cigarettes.
Model and Methodology
The paper
hypothesizes that cigarette taxation has a significant effect on cigarette
consumption. The model used shares many similar variables to Yeh et al. (2017).
However, in their empirical specification and analysis, they looked at the
annual cigarette consumption per capita using data from 2005-2014, and they
looked at the 28 European countries. My model looks at the annual cigarette
consumption per capita only within the UK, which is similar to Yeh et al.
(2017). Conversely, the model differs in some explanatory variables, these
being the number of policies made by the EU; the UK regarding tobacco; and the
amount of tax paid on a 20 pack of cigarettes. This differs from Yeh et al.
(2017) who use the MPOWER measurement (effective interventions to reduce the
demand for tobacco) and the cigarette prices of Eastern European countries. The
variables in my model are now policies and taxation to make it more specific to
the UK. The policy variable was included as the policies created by the EU and
UK have direct and indirect effects on both the price and quantity supplied of
tobacco within the UK. Therefore, this variable would have a significant effect
on consumption. Taxation on cigarettes also affects consumption by altering the
total price. However, the firms can choose whether they should artificially
increase the total price on top of the taxation, therefore these variables are
different.
Formula (1) is a log-linear
demand model, which has been regressed using OLS with robust standard errors.
The OLS will identify if there is a correlation between the explanatory
variables (Price before tax, GNI, Rural, Tax) and the dependent variable
(annual cigarette consumption per capita). The OLS tests will first be run;
then, several other robustness tests, including the Durbin-Watson (DW) test,
which tests for autocorrelation, and the White Test, which tests for
heteroscedasticity, will be run.
Adapted demand model formulae:
Formula (2)
ln C
t
= β
1t
+ β
2
LnPt + β
3
LnGNI
t
+ β
4
Rural
t
+ β
5
LnTax
t
+
ε
t
(1)
t
(years) where t = 1, 2, 3 … 31
Formula (3)
ln C
t
= β
1t
+ β
2
P
t
+ β
3
LnGNI
t
+
β
4
Rural
t
+ β
5
Tax
t
+
β
6
POL
t
+
ε
t
(1)
t
(years) where t = 1, 2, 3 … 31
Formula (4)
ln C
t
= β
1t
+ β
2
P
t
+ β
3
LnGNI
t
+
β
4
Rural
t
+ β
5
Tax
t
+
β
6
L1POL
t
+
ε
t
(1)
t
(years) where t = 1, 2, 3 … 30
Formula (5)
ln C
t
= β
1t
+ β
2
LnP
t
+ β
3
LnGNI
t
+
β
4
Rural
t
+ β
5
LnTax
t
+
β
6
POL
t
+
ε
t
(1)
t
(years) where t = 1, 2, 3 … 31
Formula (6)
ln C
t
= β
1t
+ β
2
LnP
t
+ β
3
LnGNI
t
+
β
4
Rural
t
+ β
5
LnTax
t
+
β
6
L1POL
t
+
ε
t
(1)
t
(years) where t = 1, 2, 3 … 30
Analysis and Results
Table
2
- OLS for annual cigarette consumption per
person
Regressions |
||||||
|
(1) |
(2) |
(3) |
(4) |
(5) |
(6) |
Independent Variables |
|
|||||
β 1 |
8.1767*** |
6.7275*** |
8.1680*** |
8.1366*** |
6.729*** |
6.5784*** |
|
(0.2336) |
(0.2323) |
(0.2330) |
(0.2986) |
(0.2370) |
(0.2425) |
P |
-0.0223 |
|
-0.0242 |
-0.0223 |
|
|
|
(0.0156) |
|
(0.0259) |
(0.0276) |
|
|
LnP |
|
-0.0385 |
|
|
-0.0425 |
-0.0459 |
|
|
(0.0578)
|
|
|
(0.0590)
|
(0.0581)
|
Rural
|
0.0075
|
0.0377***
|
0.0077
|
0.0082
|
0.0378***
|
0.0387***
|
|
(0.0081)
|
(0.0067)
|
(0.0094)
|
(0.0109)
|
(0.0067)
|
(0.0070)
|
Ln(
GNI)
|
0.0358***
|
0.1089***
|
0.0360***
|
0.0383**
|
0.1087***
|
0.1223***
|
|
(0.0139)
|
(0.0224)
|
(0.0105)
|
(0.0132)
|
(0.0230)
|
(0.0234)
|
Tax
|
-0.0604***
|
|
-0.0578**
|
-0.0599***
|
|
|
|
(0.0157)
|
|
(0.0226)
|
(0.0221)
|
|
|
LnTax
|
|
-0.0838*
|
|
|
-0.0796
|
-0.0794
|
|
|
(0.0433)
|
|
|
(0.0472)
|
(0.0430)
|
Pol
|
|
|
0.0006
|
|
0.0005
|
|
|
|
|
(0.0014)
|
|
(0.0020)
|
|
L
1.Pol
|
|
|
|
-0.0001
|
|
-0.0007
|
|
|
|
|
(0.0015)
|
|
(0.0019)
|
N
|
31
|
31
|
31
|
30
|
31
|
30
|
R
2
|
0.9772
|
0.9643
|
0.9774
|
0.9764
|
0.9664
|
0.9651
|
R
2
adj
|
0.9737
|
0.9599
|
0.9729
|
0.9714
|
0.9572
|
0.9579
|
DW
|
2.36
|
1.70
|
2.35
|
2.36
|
1.68
|
1.89
|
N=Number of Observations
*=10% significance level
**=5% significance level
***=1% significance level
Robust standard errors in parentheses
DW = Durbin-Watson test
By using the estimated Equation (1) and
computing the OLS with robust standard errors, I aim to observe how significant
a change in taxation affects smoking behaviour, which is measured by the
consumption level. The regression shows that price and taxation are negatively
correlated. This was expected as all literature pieces support this
relationship.
The two tests differ as regression 1 is
log to levels, and then regression 2 is log-to-logs. The elasticity of demand
for regression 1 is slightly higher. The regression shows that the estimations
are both inelastic. This means that as the price increases, the quantity
demanded decreases. The estimates are close to perfectly inelastic demand (=0)
where consumers will always buy the good, no matter if the price increases
significantly or not. However, there are several issues with the elasticity
estimates, specifically, the addictive nature of the cigarettes combined with
the aggregated cigarette consumption data. Additionally, the estimates assume
that everyone consumes a non-zero amount of the good; we know this is not the
case with cigarettes.
The remainder of the tests is slightly
different. Regression 1, 3, and 4 are very similar, as Policies (Pol) were
added and Policies were lagged over one period, respectfully. Regressions 2, 5,
6 were log-to-logs. In regression 5, the variable Policies were added, and in
regression 6 the Policies were lagged over one period. The log–level regression
is slightly more negatively correlated according to the Durbin-Watson test. The
log-log regressions are slightly more positively correlated over time, as seen
in the table.
Table
3
- Comparison of regressions
High
Elasticity of Demand
|
Low
Elasticity of Demand
|
Author
|
|
Global
|
-0.25
|
-0.5
|
Goodchild
et al. (2016)
|
Taiwan
|
-0.14
|
-2.23
|
Lee
et al. (2005)
|
Europe
|
-0.50
|
-1.23
|
Yeh
et al. (2017)
|
UK
|
-0.0223
|
-0.0459
|
My
estimate
|
The use of control variables is important in the model, as they help separate
the effects to show the relationship between important variables and the
dependent variable. Within the model, there are several controlled independent
variables to investigate and observe the relationship between the price and
cigarette consumption. These independent variables are Price, GNI, Rural and
Policies, which all help to explain the model, but also help separate the
effect out to see the true value of the relationship.
Yeh et al. (2017) find that a 10% level
of increase in cigarette price leads to an average decrease of 7% on the level
of cigarettes consumed. However, in regression 1, my model regression shows
that a 10% increase in cigarette prices leads to a 21.4% decrease in cigarette
consumption, which is a lot larger than expected after researching many other
pieces of literature. This can be attributed to the fact that in Asia, the
United States of America, and Eastern Europe the black markets are more
prominent. This is because the supply chains can easily move illegal products
around the countries. Asia produces a large quantity of tobacco, which is then
sold at cheaper prices to counterfeiters who will mix the tobacco with other
products; this effectively reduces the amount of tobacco contained to create a
cheaper alternative. Within Eastern Europe, the prices of cigarettes are
considerably lower, and therefore an increase in the price by a percentage is
not as impactful. The United States of America has different cigarette taxation
levels depending on the state, which means individuals can travel and easily
buy cigarettes cheaper in other states. This is not possible in the UK, as seen
by
Stehr
(2005). In Goodchild et al. (2016), a
simulation of a $1 increase reduced low country income groups consumption by
32%, which is higher than my model’s estimation. This is probably since
countries with low-income groups are more likely to reduce their consumption
legal consumption and buy illegally instead. This is because their disposable
income is limited, and they are more price sensitive. Therefore, the simulation
estimate is more likely to be closer to my model’s estimation. Lee et al.
(2005) find that an increase of approximately 20% in the price of taxation
leads to an 18% reduction in the number of cigarettes consumed. With such a
significant price increase, and using other literature pieces as a guideline,
we would expect to see a greater decrease in consumption than 18%. However, as
Taiwan is very close geographically to China, which produces the largest
quantity of tobacco in the world, the country has easy access to the Chinese
black markets. Taiwanese people may already buy from China and not from Taiwan
at all, which would mean the increase in taxation would only affect wealthier
individuals who still bought cigarettes legally. The price increase would lead
to a smaller decrease in consumption than expected. The estimation by
Keneggarpanich
et al. (2016) differs from the rest of the
literature as it uses roll-your-own cigarettes data, which has a completely
different subset of users. Generally, if you roll your own cigarettes, then you
are more conscious of money, as it is significantly cheaper; therefore, such
people are more price sensitive. As we see in his estimation, a 9.7% increase
in cigarette taxation causes 48% of consumers to either decrease their
consumption or switch what type of cigarettes they consume from premade to
roll-your-own. The test conducted differs from most other available literature
pieces as it does not look at cigarette consumption in isolation but looks at
the change between subsets as well. This is likely to have a similar estimation
to my model, where most people switch from premade to roll-your-own, and then a
smaller estimated percentage of 15/20% stop smoking altogether. This shows that
there is a significant substitution effect when the price of cigarettes is
increased: many individuals will start to use roll-your-own cigarettes.
There are several limitations, such as
the small data set; this could have been altered if the data from the Office of
National Statistics was measured quarterly rather than annually. This may lead
to outcomes that may exacerbate the estimation. There is a lack of accurate
data for the black market in the UK; this means that estimates are higher than
they should be. There was also a lack of data for E-cigarette sales before
2014, which reduced the number of variables in my model. To avoid
heteroscedasticity, I ran tests to reduce the serial correlation using the
Durbin-Watson test. By doing the Durbin-Watson test for this formula, the value
was 2.36, which is a very weak negative autocorrelation; this means that if the
value before dropped, there would be a greater likelihood that the value would
rise in the following year. To lower the heterogeneity of the residuals, the
White Test will be used. Additionally, the robust standard errors will reduce
the heteroscedasticity in the model. The statistical software used to perform
the regression and further tests were Stata.
Addictiveness Estimates
While using Keeler et al. (1993)
estimation with rational addiction, as price increases by 10% this causes a 28%
decrease in the demand for cigarettes. However, the regression without rational
addiction showed a 35% decrease in consumption when the price increased by 10%.
To relate this to my model, I observed the difference between the two estimates
from Keeler et al. (1993) as a 25% reduction in consumption when the addiction
model is applied. With this estimate applied to the regression run by my model,
a 10% increase in price would cause a 16.05% decrease in consumption, factoring
in rational addition. My original regression without addictiveness applied was
a 10% increase in price, resulting in a 21.4% reduction in consumption. The
difference is approximately 5.3%, which shows that addicted individuals are
less likely to reduce their consumption when the price increases.
Socio-Economic Estimates
Following
Sharbaugh
et al. (2018), as cigarette prices increase by approximately 4% from the 0.25$
increase, cigarette consumption decreases by 0.6%. If it is equated to a 10%
increase in price, then the consumption would decrease by 1.5%. In the youngest
category,
18-24 year
old’s, a 10% increase causes a
decrease in consumption of 3.75%, which is more than twice more than
Sharbaugh
et al. (2018) original estimation. Using this
basis for an estimation on my model, the youngest individuals would reduce
their consumption by 54%. Whilst this reduction is significant, the average
price of cigarettes in the United States of America is considerably lower than
in the UK. This price difference causes the consumption to vary more drastically;
the assumptions are that the price differences do not affect the consumption
levels significantly. The estimation by
Sharbaugh
et
al. (2018) is unrealistically large and the decrease in consumption will
predominantly be consumers switching to substitutes products, such as
E-Cigarettes or roll-your-own cigarettes.
The lower-income individuals
(<$25,000 a year) have little sensitivity to the price increase as a 10%
increase leads only to a 0.25% decrease in consumption; however, all the other
income groups have an average decrease in consumption of 1.7%. The difference
in cigarette consumption when there is a 10% increase in price results in the
lower-income individuals not changing their consumption as much as the other
income groups do. The change in consumption in different income groups could be
attributed to the difference in what income level is counted as at the lower
income.
Financial Implications of a
Reduction in Consumption
My
model estimated that in 2017 the UK made approximately £16.7 billion in tax
revenue from cigarettes. This was done by multiplying the percentage of smokers
in the UK by the population in 2017. Then this calculation was multiplied by
the taxation per pack of cigarettes.
However, when the cigarette tax is
increased by 10%, the cigarette tax revenue decreases by 20% to £13.2 billion.
Additionally, an estimated 2.7 million people stopped smoking from the tax
increase. My model estimated that a 10% price increase leads to approximately
2.7 million people reducing their cigarette consumption, which was 21.7% of
smokers in the UK; this was approximately 14.5 million people. The new lower
number of smokers multiplied by the taxation per pack of cigarettes which has
increased by 10% (assuming that taxation is directly passed on to consumers)
will show the new tax revenue. It is difficult to consider the number of
individuals that are made unemployed within the tobacco industry in response to
the reduction in cigarette consumption.
In
2017 the government’s tobacco control plan estimated that the cost to the
economy (NHS, economy, environmental, other second-hand smoking factors) was
approximately £11 billion a year. Given a reduction of smokers by 2.7 million,
this would reduce the total economic cost to approximately £8 billion. Dividing
the cost of the economy by how many smokers there were in 2017, then
multiplying that number by the new level of smokers, gives an approximate cost
to the economy.
However,
the effects of those 2.7 million people stopping smoking are not instant, and
many health problems would persist. Furthermore, many factors are not
considered, such as the environmental impact, the productivity levels of not
smoking every 15 minutes, and the improved concentration from not smoking. The
number of individuals who will buy from the black market instead of shops after
the tax increase is not currently calculatable.
Limitations
The limitations of
the model were that the data for E-Cigarettes, which was originally included
within the model did not have enough data points to put into the model.
Additionally, there was not any reliable data before 1987 for average cigarette
prices. The missing six years reduced the whole model by six observations. I
did write an email to the Office of National Statistics, unfortunately they did
not have any additional data that would be relevant. Additionally, the data on
the size of the black market for cigarettes is another limitation, as other
research papers in Asia have lots of data to predict the size of the black
market and to then take it into account within their model.
Conclusion
In
conclusion, the estimation of the model in the project has a slightly higher
than expected reduction in cigarette consumption. However, this can easily be
explained by economic reasoning. My model regression shows that a 10% increase
in total cigarette prices leads to a 21.4% decrease in the consumption of
cigarettes consumed. The paper aimed to analyse the relationship between cigarette
taxes and smoking behaviour, which in this case is the level of consumption.
Specifically, the focus was on the effects of taxation within the UK this was
done by looking at time series data between 1987-2017.
The OLS test was used to see if there was a significant relationship between
cigarette taxation and cigarette consumption. The regression shows that there
is a negative relationship between cigarette taxation and cigarette
consumption. The results are that the cigarette tax has a negative correlation
with cigarette consumption; consequently, as cigarette taxation increases,
cigarette consumption decreases. The challenges that arose within testing were
the problems with collinearity and the heteroscedasticity of the model. To
rectify this issue, I used several tests and reformulated certain variables to
limit the collinearity. Additionally, to deal with the heteroscedasticity,
robust standard errors were used, which limited the heteroscedasticity in my
model. In addition, the White Test was run to ensure that the
heteroscedasticity in my model was not significant. The final challenge was the
fact that the data was collected annually; therefore, there were not many
observation points, which is difficult in time series models and could
potentially alter the results.
The results gathered from the data and the model echo the reading of other
literature. However, further empirical research is required for gathering more
data points and discovering other relevant variables. Although policies were
seen to be insignificant in my model, it is difficult to value each policy in
importance. Many policies, including the ban on smoking inside buildings, have
had substantial effects on consumption along with the ban on cigarette
advertisement. To further this study, a more detailed look into pairing
policies with taxation and seeing the effects of it would explain the model
better. Additionally, it would also be useful to compare the results with a
similar country where taxation is the same throughout the whole country and the
black-market influence similar to that in the UK. There could be more data for
relevant substitutes throughout the dataset which would allow the effects of
the cigarette substitutes to be observed and to see if it is significant on the
consumption levels. These adaptations would give more of a holistic view of the
effects of cigarette taxation on consumption and whether simply just taxation
is enough in combatting the problem of smoking.
References
Bader, P.,
Boisclair, D. and
Ferrence
, R. (2011). Effects of
Tobacco Taxation and Pricing on Smoking
Behavior
in
High-Risk Populations: A Knowledge Synthesis.
International Journal of
Environmental Research and Public Health
, [online] 8(11), p. Available at:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3228562/
Becker,
G.S., Grossman,
M.
and Murphy, K.M. (1994).
An
Empirical Analysis of Cigarette Addiction
. Cambridge, Mass. National Bureau
of Economic Research
Becker, G.S.
and Murphy, K.M. (1986).
A Theory of Rational Addiction
. Chicago:
Center
For The Study Of The Economy And The State,
University Of Chicago
Caponnetto
, P., Campagna, D.,
Cibella
,
F.,
Morjaria
, J.B., Caruso, M., Russo, C. and
Polosa
, R. (2013). Efficiency and Safety of an electronic
cigarette (ECLAT) as Tobacco Cigarettes Substitute: A prospective 12-Month
Randomized Control Design Study.
PLoS
ONE
,
[online] 8(6), p. Available at:
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0066317
Chaloupka
, F. (1991). Rational Addictive
Behavior
and Cigarette Smoking.
Journal of Political Economy
, 99(4), pp.722–742
Chaloupka
, F.J., Cummings, K.M., Morley, C. and Horan, J.
(2002). Tax,
price
and cigarette smoking: evidence
from the tobacco documents and implications for tobacco company marketing
strategies.
Tobacco Control
, 11(Supplement 1), pp. i62–i72
Chriqui
, J.,
Ribisl
, K.,
Wallace, R., Williams, R., O’Connor, J. and
Arculli
,
R. el (2008). A comprehensive review of state laws governing Internet and other
delivery sales of cigarettes in the United States.
Nicotine & Tobacco
Research
, 10(2), pp.253–265
Colman, G.J.
and
Remler
, D.K. (2008). Vertical equity consequences
of very high cigarette tax increases: If the poor are the ones smoking, how
could cigarette tax increases be progressive?
Journal of Policy Analysis and
Management
, 27(2), pp.376–400
Durbin,
J.
and Watson, G.S. (1952).
Testing for serial correlation
in least squares regression
. Cambridge: University Of Cambridge. Department
of Applied Economics
Franz, G.A.
(2008). Price effects on the smoking behaviour of adult age groups.
Public
Health
, 122(12), pp.1343–1348
Goldin, J.
and
Homonoff
, T. (2013). Smoke Gets in Your Eyes:
Cigarette Tax Salience and Regressivity.
American Economic Journal: Economic
Policy
, 5(1), pp.302–336
Goodchild,
M.,
Perucic
, A.-M. and Nargis, N. (2016). Modelling
the impact of raising tobacco taxes on public health and finance.
Bulletin
of the World Health Organization
, 94(4), pp.250–257
Harris, J.E.
and Chan, S.W. (1999). The continuum-of-addiction: cigarette smoking in
relation to price among Americans aged 15-29.
Health Economics
, 8(1),
pp.81–86
Hiscock, R.,
Branston, J.R., McNeill, A.,
Hitchman
, S.C.,
Partos
, T.R. and Gilmore, A.B. (2018). Tobacco industry
strategies undermine government tax policy: evidence from commercial data.
Tobacco
Control
, [online] 27(5), pp.488–497. Available at:
https://tobaccocontrol.bmj.com/content/27/5/488
Hu, T.W.
(2002). Effects of cigarette tax on cigarette consumption and the Chinese
economy.
Tobacco Control
, 11(2), pp.105–108
Hyland, A.
(2005). Higher cigarette prices influence cigarette purchase patterns.
Tobacco
Control
, 14(2), pp.86–92
IEA (2017).
INSTITUTE
OF ECONOMIC AFFAIRS
. [online] Institute of Economic Affairs. Available at:
https://iea.org.uk/
Keeler,
T.E., Hu, T.-W., Barnett, P.G. and Manning, W.G. (1993). Taxation, regulation,
and addiction: A demand function for cigarettes based on time-series evidence.
Journal
of Health Economics
, 12(1), pp.1–18
Kengganpanich
, M.,
Termsirikulchai
, L.
and
Benjakul
, S. (2009). The impact of cigarette tax
increase on smoking
behavior
of daily smokers.
Journal
of the Medical Association of Thailand =
Chotmaihet
Thangphaet
, [online] 92
Suppl
7,
pp.S
46-53. Available at:
https://pubmed.ncbi.nlm.nih.gov/20232561
Lee, J-M.
(2005). Effect of cigarette tax increase on cigarette consumption in Taiwan.
Tobacco
Control
, 14(1),
pp.i
71–i75
NHS (2018).
Statistics
on Smoking -
England ,
2018 [PAS] - NHS Digital
.
[online] NHS Digital. Available at:
https://digital.nhs.uk/data-and-information/publications/statistical/statistics-on-smoking/statistics-on-smoking-england-2018/content
Novotny,
T.E.,
Bialous
, S.A., Burt, L., Curtis, C., Luiza da
Costa, V.,
Iqtidar
, S.U., Liu, Y., Pujari, S. and
Tursan
d’Espaignet
, E. (2015).
The environmental and health impacts of tobacco agriculture, cigarette
manufacture and consumption.
Bulletin of the World Health Organization
,
[online] 93(12), pp.877–880. Available at:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4669730/
Office of
National Statistics (2020).
Home - Office for National Statistics
.
[online] Ons.gov.uk. Available at:
https://www.ons.gov.uk/
Palgrave
Macmillan (Firm (2018).
The new Palgrave dictionary of economics.
London, United Kingdom: Palgrave Macmillan
Remler
, D.K. (2004). Poor Smokers, Poor Quitters, and Cigarette Tax
Regressivity.
American Journal of Public Health
, 94(2), pp.225–229
Sharbaugh
, M.S.,
Althouse
, A.D.,
Thoma
, F.W., Lee, J.S., Figueredo, V.M. and
Mulukutla
, S.R. (2018). Impact of cigarette taxes on
smoking prevalence from 2001-2015: A report using the
Behavioral
and Risk Factor Surveillance Survey (BRFSS).
PLOS ONE
, 13(9), p.e0204416
Stehr
, M. (2005). Cigarette tax avoidance and evasion.
Journal of Health
Economics
, 24(2), pp.277–297
The World
Bank (2020).
Annual Report 2020
. [online] World Bank. Available at:
https://www.worldbank.org/en/about/annual-report
The World
Health Organization (2015).
WHO global report on trends in prevalence of
tobacco smoking 2015
. [online]
apps.who.int
. World Health
Organization. Available at:
https://apps.who.int/iris/handle/10665/156262
Wasserman,
J., Manning, W.G., Newhouse, J.P. and Winkler, J.D. (1991). The effects of
excise taxes and regulations on cigarette smoking.
Journal of Health
Economics
, 10(1), pp.43–64
White, H.
(1980). A Heteroskedasticity-Consistent Covariance Matrix Estimator and a
Direct Test for Heteroskedasticity.
Econometrica
,
48(4), p.817
Yeh, C.-Y.,
Schafferer
, C., Lee, J.-M., Ho, L.-M. and Hsieh, C.-J.
(2017). The effects of a rise in cigarette price on cigarette consumption,
tobacco taxation revenues, and of smoking-related deaths in 28 EU countries--
applying threshold regression modelling.
BMC Public Health
, [online]
17(1). Available at:
https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-017-4685-x
Copyright Statement
©
Harvey
James Grew.
This article is licensed under a Creative Commons Attribution 4.0 International
Licence (CC BY).