Technical ReportPDF Available

Understanding and measuring tax avoidance and evasion: A methodological guide.

Authors:

Abstract and Figures

This Methodological Guide updates and expands the 2001 World Bank Toolkit on how to estimate the scope of tobacco smuggling. It draws on the results of numerous empirical studies that tested the applicability of five methods described in the Toolkit while critically evaluating new methods that emerged in response to the evolving nature of illicit tobacco trade, the policy debates surrounding the issue, and the development of new technologies.
Content may be subject to copyright.
HANA ROSS, PhD.
MARCH 2015
Acknowledgements
The author would like to thank the following individuals
for their assistance in reviewing drafts of this document:
Johanna Birckmayer, PhD, MPH, Frank Chaloupka, PhD,
Grieve Chelwa, PhD, David Merriman, PhD, and
Guillermo Paraje, PhD.
The guide is available at:
http://tobaccoecon.org/publications/reports/
http://tobacconomics.org/research/
Design by Parrilla Design Workshop
MARCH 2015
Tobacconomics is a collaboration of leading researchers
who have been studying the economics of tobacco control
policy for nearly 30 years. The team is dedicated to helping
researchers, advocates and policymakers access the latest
and best research about what’s working—or not working—
to curb tobacco consumption and the impact it has on
our economy. As a program of the University of Illinois at
Chicago, Tobacconomics is not aliated with any tobacco
manufacturer. Visit www.tobacconomics.org or follow us on
Twitter www.twitter.com/tobacconomics.
The Economics of Tobacco Control Project is hosted
by the South African Labour and Development Research
Unit at the School of Economics, University of Cape Town
in partnership with the American Cancer Society, the Bill &
Melinda Gates Foundation and the African Capacity Building
Foundation. The aim of this project is to expand current
research eorts in the economics of tobacco control and to
enhance the knowledge of economic and tax issues among
tobacco control advocates and policymakers to strengthen
support for tobacco tax and price increases in sub-Saharan
Africa. Visit www.tobaccoecon.org.
Ross H (2015). Understanding and measuring tax avoidance
and evasion: A methodological guide. Prepared for the
Economics of Tobacco Control Project¸ School of Economics,
University of Cape Town and Tobacconomics, Health Policy
Center, Institute for Health Research and Policy, University
of Illinois at Chicago.
-1-
Contents
Executive Summary 2
Introduction 4
Definitions 5
Chapter 1: Theory of Tax Avoidance and Tax Evasion and Its Impact 7
1.1 Overview
1.2 Theory of Tax Avoidance and Tax Evasion
1.3 Impact of Tax Avoidance and Tax Evasion
Chapter 2: How to Measure the Scope of Tax Avoidance and Evasion 12
2.1 Survey of Tobacco Users
2.2 Examination of Cigarette Packs
2.3 Compare Tobacco Sales and Consumption (gap analysis)
2.4 Econometric Modeling
2.5 Other Methods
2.6 Summary and Recommendations for Estimating
the Scope of Tax Avoidance and/or Evasion
Chapter 3: Assessing the Quality of the Estimates 32
Conclusions and Summary 47
References 48
-2-
Executive Summary
This Methodological Guide updates and expands the
13-year old World Bank Toolkit on how to estimate the
scope of tobacco smuggling. It draws on the results of
numerous empirical studies that tested the applicability
of five methods described in the Toolkit while critically
evaluating new methods that emerged in response to
the evolving nature of illicit tobacco trade, the policy
debates surrounding the issue, and the development of
new technologies.
This guide covers not only smuggling of tobacco
products, but also tax evasion related to illicit tobacco
trade and legal tax avoidance, since the methods to
estimates these phenomena are often intertwined. It
also provides guidance on how to assess the quality
of existing estimates, including those oered by the
tobacco industry. Unlike the World Bank Toolkit, this
guide does not cover the topic of how to reduce tax
avoidance and tax evasion and instead refers the reader
to the WHO FCTC Protocol to Eliminate Illicit Trade
in Tobacco Products.
   begins
by defining key variables. It explains the dierence
between “smuggling”, “illicit trade”, “tax evasion” and
“tax avoidance”, and describes behaviors that constitute
either tax evasion or tax avoidance. Tobacco products
that avoid/evade all or a portion of the required taxes
are defined as “low-tax” products, while the rest of
tobacco products fall into the category of “full-tax”
products.
   lays the theoretical
foundation for the methods and describes the impact
of tax avoidance and tax evasion on the supply of and
demand for tobacco products, on tax revenue, on the
price elasticity of demand, and on the aordability of
tobacco products. We provide insight into the economic,
social, and political determinants of tax avoidance
and tax evasion, including factors such as price/tax
dierences, the tax structure, the costs of obtaining
low-tax products, the costs of overcoming legal and
regulatory obstacles, informal distribution networks,
and the level of corruption. This chapter demonstrates
how tax avoidance and tax evasion impact average
cigarette prices, the price elasticity of tobacco demand,
brand proliferation, and other tobacco control policies.
It concludes that tax avoidance/evasion may reduce,
but do not eliminate the eectiveness of tobacco tax
increases in reducing tobacco use and raising revenues.
The motivation for a government to address tax
avoidance/evasion is directly linked to the size of the
revenue loss while the eectiveness of the interventions
depends on the level of enforcement. The level of
enforcement is directly linked to the level of investment
to combat tax evasion/avoidance.
   is central to the
Guide and describes eleven methods of measuring
the scope of tax avoidance and tax evasion: surveys of
tobacco users, examination of cigarette packs obtained
from smokers, examination of discarded cigarette
packs, examination of cigarette packs obtained from
retail, comparison of sales with consumption (Gap
Analysis), econometric modeling, comparison of tax
paid sales with estimated consumption, comparison
of actual and projected tobacco tax revenue, key
informant interviews, monitoring tobacco trade, and
analyzing seizures of illegally transported tobacco.
The chapter begins with a brief overview of the main
principles of conducting research, which are applicable
to all methods. Then the Guide provides a step-by-step
description of the specifics of each the method, starting
with those most frequently discussed in the literature,
followed by approaches that are unique to certain
market conditions and approaches suering from
multiple weaknesses. Each method has a background
section that links it to theory, a list of pros and cons,
and a recommendation when a particular method
should be used. The application of each method is
demonstrated by examples.
The chapter concludes that there is not a single method
that will produce a definitive estimate, because all of
them have advantages and disadvantages. Since the
weakness of a particular approach can be exacerbated
by specific market conditions, it is important to use
local specific knowledge and creativity when applying
these methods. Given the complexity of tobacco tax
avoidance and evasion and the methods’ limitations, it
is important to triangulate the estimates of the scope
of the problem using dierent methods. Many studies
apply the same method over time in order to capture
changes in the scope of tax avoidance/evasion rather
than generating a single point estimate of its scope.
smuggling
illicit
trade
tax
evasion
low-tax
full-tax
econometric
modeling
tax paid sales
consumption
monitoring
tobacco
trade
illegally
transported
tobacco
 

updates and expands the 13-year-old World Bank
Toolkit on how to estimate the scope of tobacco
smuggling. It draws on the results of numerous
empirical studies that tested the applicability
of five methods described in the Toolkit while
critically evaluating new methods that emerged in
response to the evolving nature of illicit tobacco
trade, the policy debates surrounding the issue,
and the development of new technologies.
This guide covers not only smuggling of tobacco
products, but also tax evasion related to illicit
tobacco trade and legal tax avoidance, since the
methods to estimates these phenomena are often
intertwined. It also provides guidance on how to
assess the quality of existing estimates, including
those oered by the tobacco industry. Unlike the
World Bank Toolkit, this guide does not cover
the topic of how to reduce tax avoidance and tax
evasion and instead refers the reader to the WHO
FCTC Protocol to Eliminate Illicit Trade in
Tobacco Products.
Executive Summary
 

begins by defining key variables. It explains the
dierence between “smuggling”, “illicit trade”,
“tax evasion” and “tax avoidance”, and describes
behaviors that constitute either tax evasion or tax
avoidance. Tobacco products that avoid/evade all or a
portion of the required taxes are defined as “low-tax”
products, while the rest of tobacco products fall into
the category of “full-tax” products.
lays the theoretical foundation for the methods and
describes the impact of tax avoidance and tax evasion
on the supply of and demand for tobacco products,
on tax revenue, on the price elasticity of demand,
and on the aordability of tobacco products.
We provide insight into the economic, social,
and political determinants of tax avoidance and
tax evasion, including factors such as price/tax
dierences, the tax structure, the costs of obtaining
low-tax products, the costs of overcoming legal
and regulatory obstacles, informal distribution
networks, and the level of corruption. This chapter
demonstrates how tax avoidance and tax evasion
impact average cigarette prices, the price elasticity
of tobacco demand, brand proliferation, and other
tobacco control policies.
This chapter concludes that tax avoidance/evasion
may reduce, but do not eliminate, the eectiveness
of tobacco tax increases in reducing tobacco use and
raising revenues. The motivation for a government to
address tax avoidance/evasion is directly linked to the
size of the revenue loss while the eectiveness of the
interventions depends on the level of enforcement.
The level of enforcement is directly linked to the level
of investment to combat tax evasion/avoidance.
1 

-3-
Such an approach addresses some methodological
weaknesses of the methods and is useful for evaluating
the impact of policies and other factors with a possible
impact on tax avoidance/evasion.
 
guides the reader through a series of studies and
analyses the quality and reliability of their estimates
while taking into account the agendas of those who
fund and/or conduct research on tobacco tax avoidance/
evasion. It outlines a set of criteria and then applies
them to evaluate eight studies. It concludes that studies
supported by the tobacco industry cannot be trusted
due to lack of transparency and the use of potentially
contaminated data. The industry estimates are
consistently and substantially higher compared to those
produced by independent researchers.
 ...
this Methodological Guide recommends using multiple
methods that suer from the minimum weaknesses,
executing them according to the principles of rigorous
research, and triangulating the results in order to cross-
validate the estimates and minimize the methodological
limitations of individual methods. Such an approach
will result in methodologically sound and objective
quantitative estimates of tobacco tax avoidance and tax
evasion.
smuggling
illicit
trade
tax
evasion
low-tax
full-tax
econometric
modeling
tax paid sales
consumption
monitoring
tobacco
trade
illegally
transported
tobacco


this Methodological Guide recommends using
multiple methods that suer from the minimum
weaknesses, executing them according to the
principles of rigorous research, and triangulating
the results in order to cross-validate the estimates
and minimize the methodological limitations of
individual methods. Such an approach will result in
methodologically sound and objective quantitative
estimates of tobacco tax avoidance and tax evasion.
is central to the Guide and describes 11 methods
of measuring the scope of tax avoidance and tax
evasion: surveys of tobacco users, examination of
cigarette packs obtained from smokers, examination
of discarded cigarette packs, examination of cigarette
packs obtained from retail, comparison of sales with
consumption (gap analysis), econometric modeling,
comparison of tax paid sales with estimated
consumption, comparison of actual and projected
tobacco tax revenue, key informant interviews,
monitoring tobacco trade, and analyzing seizures of
illegally transported tobacco.
The chapter begins with a brief overview of the
main principles of conducting research, which are
applicable to all methods. It then provides a step-by-
step description of the specifics of each the method,
starting with those most frequently discussed in
the literature, followed by approaches that are
unique to certain market conditions and approaches
suering from multiple weaknesses. Each method
has a background section that links it to theory, a
list of pros and cons, and a recommendation when a
particular method should be used. The application of
each method is demonstrated by examples.
The chapter concludes that there is not a single
method that will produce a definitive estimate,
because all of them have advantages and
disadvantages. Since the weakness of a particular
approach can be exacerbated by specific market
conditions, it is important to use local specific
knowledge and creativity when applying these
methods. Given the complexity of tobacco tax
avoidance and evasion and the methods’ limitations,
it is important to triangulate the estimates of the
scope of the problem using dierent methods. Many
studies apply the same method over time in order
to capture changes in the scope of tax avoidance/
evasion rather than generating a single point estimate
2 

guides the reader through a series of studies and
analyses the quality and reliability of their estimates
while taking into account the agendas of those
who fund and/or conduct research on tobacco tax
avoidance/evasion. It outlines a set of criteria and
then applies them to evaluate eight studies.
It concludes that studies supported by the
tobacco industry cannot be trusted due to lack of
transparency and the use of potentially contaminated
data. The industry estimates are consistently and
substantially higher compared to those produced by
independent researchers.
3 

of its scope. Such an approach addresses some
methodological weaknesses of the methods and is
useful for evaluating the impact of policies and
other factors with a possible impact on tax
avoidance/evasion.
-4-
Introduction
The purpose of this document is to update the
World Bank Toolkit #7 “Understand, Measure, and
Combat Tobacco Smuggling”1 published more than
13 years ago.
Toolkit #7 was a part of a series of methodological
guides for conducting tobacco control research.
It described five methods on how to measure the
scope of illicit trade in tobacco and provided guidance
on how to address the issue, including various policy
options. Toolkit # 7 influenced not only the research
community, but also NGOs and policy makers. Its
methods have been applied and tested in numerous
empirical studies and the lessons learned from those
eorts informed this Methodological Guide.
The tobacco industry has also learned from the Toolkit
#7. They learned how their illicit activities can be
detected, what to do in order to disguise them, and how
to use (and misuse) the methods to provide alternative
estimates of illicit trade. This helped the industry to
position itself as a stakeholder in the debate about
combating illicit trade while oering solutions that are
in most cases damaging to public health.
The methods described in Toolkit #7 have been
applied and tested in many empirical studies.2 Lessons
learned from those studies led to the refinement and
improvement of the methods, and pointed to new ways
to study tax avoidance and tax evasion that take into
account the evolving nature of illicit tobacco trade,
the policy debates surrounding the issue, and the
development of new technologies.
This Methodological Guide captures this newly gained
knowledge. Application of the methods described
in the Guide will result in methodologically sound
and objective quantitative estimates of tobacco tax
avoidance and tax evasion. The Guide also provides
assistance with assessing the quality of existing
estimates which are useful for educating policymakers
and the general public while countering results
generated to manipulate the public opinion.
We hope that the methods described in this Guide will
motivate the research community to use them,
build upon them, and test new ones in order to advance
our understanding of the scope of tax avoidance and
tax evasion.
Unlike World Bank Toolkit #7, this Guide does not
provide guidance on how to reduce tax avoidance and
tax evasion. IARC Handbooks, Volume 142 and the
WHO FCTC Protocol to Eliminate Illicit Trade in
Tobacco Products3 provide a comprehensive overview
of that topic.
The Guide begins by defining the key variables of
interest and explains how they fit into the commonly-
used terminology relevant for tax avoidance and tax
evasion. Chapter 1 provides the theoretical foundation
for the Guide and describes the impact of tax avoidance
and tax evasion on the supply of and demand for
tobacco products. Chapter 2 provides an overview of
the methods of measuring the scope of tax avoidance
and tax evasion, and Chapter 3 discusses how to
assess the quality of various estimates. The last section
summarizes the main points and provides some
concluding remarks.
-5-
Definitions
The tobacco control community, policy makers, and
the general public are mostly familiar with expressions
“smuggling” or “illicit trade” when the issue of not
paying all tobacco taxes is being discussed. However,
the complexity of the phenomenon calls for use of more
precise terminology. Being familiar with and using
the proper terms when debating this issue will help to
advance the discourse and to determine the correct
approaches to measuring the scope of the problem.
The Merriam-Webster Dictionary defines smuggling as
importing or exporting secretly, contrary to the law, and
especially without paying duties imposed by law.
The term illicit tobacco trade is defined by Article
1 of the WHO Framework Convention on Tobacco
Control4 as a practice or a conduct prohibited by
law which relates to production, shipment, receipt,
possession, distribution, sale or purchase of tobacco
products, including any practice or conduct intended
to facilitate such activity. Therefore, the term “illicit
tobacco trade” covers all illegal activities related to the
tobacco trade, not just the circumvention of tobacco
taxes.
illicit trade can occur anywhere along the
tobacco supply chain, from manufacturing, through
distribution, to the retail stage. Diversion from legal
to illegal trade typically occurs before the point where
taxes are assessed, particularly when diversion is
motivated by tax evasion.
circumvention of taxes is classified as either tax
avoidance (legal methods of circumventing tobacco
taxes) or tax evasion (illegal methods for circumventing
tobacco taxes).
tax avoidance includes legal activities and purchases
in accordance with customs and tax regulations, most
of which include the payment of some tobacco taxes,
and are done mostly by individual tobacco users (e.g.,
cross-border shopping, duty-free shopping, Internet
and mail/phone purchases), but tobacco companies also
engage in it (e.g., changing some product features or its
production process in order to reduce tax liability).
cross-border shopping refers to individuals
purchasing tobacco products which are not intended
for resale in a lower price (or lower tax) jurisdiction
(country, district, Native American reservation, etc.).
If these purchases are within the limits imposed by the
customs regulations, tax may have been legally avoided.
The purchases in excess of the limits constitutes illegal
tax evasion.
duty-free shopping involves the purchase of limited
amount of tax-free tobacco products in duty free shops
(e.g., at airports, on-board planes and boats, etc.).
internet and mail / phone purchases may result
in either tax avoidance or tax evasion depending on the
law applicable in the product’s destination. If tax on a
product is not paid in its destination, but the purchase
is not forbidden by the law, the tax has been avoided.
If the buyer or sellers are legally obligated to pay taxes
on such purchases, this transaction is qualifies as tax
evasion. Unlike the sellers, the buyers often do not
realize that they are committing an illegal act.
tobacco industry reformulation,
repositioning, forestalling takes advantage
of the country’s tax system in order to reduce its
tax payments. In a country with a multi-tiered tax
structure, a company can change some features of
its products in order to reclassify them to a lower tax
tier. For example, the industry can reduce the size of
cigarettes (in a specific tax regime with dierent tax
rates based on product length or weight; reformulation)
or their prices (in an ad valorem tax regime when tax
rates vary with price; repositioning) and pay less taxes.
Where allowed by law, the industry can pre-purchase
tax stamps before a tax increase and use them after
the new tax rate is in place to reduce their tax liability
(forestalling).
tax evasion consists of illegal activities intended to
avoid paying some or all taxes. It includes smuggling
cigarettes across borders, selling genuine cigarettes
that were manufactured illegally, selling counterfeiti or
illicit white cigarettesii, or selling or buying cigarettes via
Internet, phone or mail without paying the appropriate
taxes.
i Counterfeit cigarettes are cigarettes manufactured without authorization of the rightful owners of the trademarked brand, with intent to deceive
consumers and to avoid paying duty
ii Illicit white cigarettes are brands manufactured legally, but distributed to large extend via illegal supply channels for the purpose to evade taxes.
-6-
smuggling of cigarettes involves both small and
large scale operations. Small scale smuggling occurs
when the quantity of cigarettes moved across the
border is in excess of the allowable limits (but is still
relatively small) and/or when products purchased in
another jurisdiction are intended for resale without
paying appropriate taxes in the destination jurisdiction
(although some taxes are usually paid in the country of
origin). This activity is also called bootlegging. Large
scale smuggling involves large quantity of products
and generally results in avoiding all taxes. It involves
disguising or hiding products and moving them by
means of often expensive operations orchestrated
by criminal networks. It may take advantage of
“in-transit”iii regimes and tax-free zones, and often
transports counterfeit cigarettes, or genuine cigarettes
with counterfeit tax stamps, or illicit white cigarettes.
It is associated with various kinds of fraud, including
intentional mislabeling of cigarettes as other products
to evade taris, falsification of the true country of
origin of a shipment in order to gain preferential taris,
and overt evasion of Customs duties and taxes.5
illicit manufacturing means the production of
tobacco products without complying with applicable
laws such as licensing, tax law, and other government
regulations that govern the manufacture of tobacco.
This also includes underreporting of actual production
quantities by licensed manufacturers with the dierence
between reported and actual production being diverted
through illegal channels. Counterfeit tax stamps are
often applied to illegally manufactured products. The
destination of the illegally manufactured cigarettes can
be the domestic or a foreign market.
counterfeiting is a form of illicit manufacturing
that involves using a trademark without the approval
of the trademark owner. Counterfeit cigarettes often
bear counterfeit tax stamps and are distributed through
criminal networks.
illicit whites (also called “cheap whites”) are
cigarettes manufactured by legitimate business
enterprises, but a large share of the production is
sold illegally, usually outside the jurisdiction where
they are produced.
Selling or buying cigarettes via Internet, telephone, or
mail-order usually involves vendors based in low-tax
states or in tax-exempt locations and the buyer based in
jurisdictions where prices are higher compared to those
oered via these channels. These sales constitute tax
evasion if the seller and/or the buyer is legally obligated
to pay taxes on such purchases in accordance with the
law applicable to the destination jurisdiction.
While manufactured cigarettes comprise the majority
of tobacco goods channelled via illicit trade, tobacco
leaf and other tobacco products (and possibly
e-cigarettes) may also be the subject to tax avoidance
and tax evasion.
In this report, low-tax products are products that
escape paying some or all of the taxes on them, either
via tax avoidance or via tax evasion. They are the result
of either tax avoidance or tax evasion, and can be either
legal (if tax is avoided) or illegal (if tax is evaded). Full-
tax products are products that pay all taxes as intended
by the tax law/regulations. These are legal products.
iii In-transit regime applies to goods that cross the territory of another country on their journey between the departure and the final destination country
-7-
1.1 Overview
Tax avoidance and tax evasion have been studied by
scholars from many disciplines. Economists, political
scientists, criminologists, and other social scientists have
examined them from a wide range of perspectives with
the main focus on measuring the scope of tax avoidance
and tax evasion, analyzing the motivation for engaging in
them, developing measures to curb them, and studying
how these behaviors aect and are aected by the
political process.
A theoretical framework can provide insight into the
economic, social, and political determinants of tax
avoidance and tax evasion, including the role played
by price/tax dierences, the tax structure, the costs of
obtaining low-tax products, the costs of overcoming
legal and regulatory obstacles, the role of informal
distribution networks and the grey economy, the
level of corruption, and the involvement of global and
new firms. Theory is also the point of departure for
measuring the scope of tax avoidance and tax evasion
and their impact on the overall supply of and demand
for tobacco products, on tax revenue, on the price
elasticity of demand, and on the aordability of tobacco
products. Some studies have oered possible solutions
to the problem by analyzing various aspects of the illicit
cigarette supply and governance strength, including the
role of tax administration, the level of law enforcement,
anticorruption eorts, the certainty, swiftness and
severity of punishment if convicted, the advantages and
disadvantages of using administrative rather than criminal
sanctions, and the level of coordination and collaboration
among dierent authorities within the government.
The scientific community also analyzes the degree
to which tax avoidance/evasion influences and is
influenced by political processes, public policy
formulation, and international negotiations. The
vested interests of various stakeholders such as the
tobacco industry, governments, and the tobacco control
community, can interfere with the line of inquiry into
tax avoidance/evasion including the study design, the
choice of methods, the objectivity of the research results
and their presentation.
This section will describe the theoretical models
underlying empirical studies measuring the scope
and impact of tax avoidance and tax evasion. It will
not discuss the link between the theory and measures
designed to curb tax avoidance/evasion. IARC
Handbooks, Volume 142 and the WHO FCTC Protocol
to Eliminate Illicit Trade in Tobacco Products3 provide
a comprehensive overview of that topic.
1.2 Theory of Tax Avoidance and Tax
Evasion
According to economic theory, customers (i.e., current
or potential tobacco users) allocate their income among
full-tax (i.e., legal) tobacco products, low-tax tobacco
products (i.e., products that avoided/evaded some or all
tax; can be both legal and illegal), and other goods and
services. Consumers choose how much of their income to
allocate to each category on the basis of relative monetary
prices, perceived quality, ease and costs of purchase,
expected legal costs associated with purchasing illegal
products, social norms, and other relevant variables.
Tobacco users treat low-tax cigarettes as (potentially
imperfect) substitutes for full-tax tobacco products and
consider their full price when determining quantity
demanded. The full price consists of the amount of
money the buyer pays to the seller in exchange for the
product (i.e., monetary price), the costs of convenience
of obtaining the product (e.g., time needed to get the
product, travel distance, purchasing experience), and the
risk associated with the transaction and consumption of
the product. The non-monetary component of the full
price represents transaction costs. For example, the point
of sale can be a well-kept store near the place of residence
(lower transaction costs) or a dark alley in an unsafe part
of town (higher transaction costs). Those who purchase
illegal cigarettes may face legal sanctions and uncertainty
about the quality of the product, (e.g., they might not
be able to distinguish between genuine and counterfeit
products).
Given the higher transaction costs of illicit cigarettes,
their monetary (i.e., sale) price must be lower compared
to legal cigarettes, unless the perceived quality of illicit
cigarettes is higher or a particular brand is not supplied
via legal channels. The degree of substitution between
legal and illegal products also depends on availability of
a particular brand, individual taste and income.
CHAPTER 1
Theory of Tax Avoidance and Tax Evasion and Its Impact
-8-
The price dierences between legal and illicit cigarettes
can be observed in many markets. For example,
individuals oering low-tax cigarettes in the United
Kingdom in 1999 were selling them for £1.00 less
compared to full-tax cigarettes sold in recognized outlets,
yet 17% of adult smokers still preferred to buy their
cigarettes in stores.6 This indicates that their transaction
costs were equal or higher than £1.00 assuming that
they perceived low-tax and full-tax products as close
substitutes. In 2013, a pack of Camel cigarettes
smuggled to New York City from Virginia was bought
for $8, while the fully-taxed Camel cigarettes cost
around $12 per pack.7 In some markets, however, the
price of lower-taxed cigarettes can be higher. In Viet
Nam, for example, the price of the smuggled brand
555, manufactured in the United Kingdom, was higher
than the locally produced 555, because the smuggled
cigarettes were perceived as being of higher quality.8
Potential suppliers of low-tax tobacco, motivated both
by the expected profit margin and the expected total
amount of profit, choose the quantity supplied (which
can be zero) and price based on interaction of supply
and demand. This interaction is aected by the cost of
manufacturing and/or obtaining low-tax cigarettes,
transportation and distribution costs, and costs
associated with the illegal or semi-legal nature of the
operations, competitive conditions, and other variables.
The theory holds that the greater the perceived
consumer’s net benefit and the greater the supplier’s
estimated profit, the greater the probability that an
individual or a company will engage in tax avoidance
and tax evasion. The size of the profit determines
the way low-tax products are supplied to the market.
Small-scale smuggling, or bootlegging that generally
oers lower profit is negatively related to the distance to
travel, the opportunity costs of time spent obtaining the
products (i.e. foregone salaries), but positively related
to the relative price dierences between adjacent
geographical areas.9 Factors related to overcoming
the legal and regulation obstacles play an important
role in the decision to supply the market via large
scale smuggling that has the potential to generate larger
profit.2,10 Large scale tax evasion is usually present in
countries where corruption is high, the control of
the authorities is lax, and commodities other than
tobacco are also being smuggled.11 Since corruption
is usually more pervasive in low- and middle-income
countries, these countries are at greater risk for large-
scale smuggling activities.12 These countries often have
weaker governance and tax administration, which reduce
the costs of supplying low-tax products. Large scale tax
evasion is usually associated with criminal networks.13
1.3 Impact of Tax Avoidance and
Tax Evasion
The impact of tax avoidance and tax evasion on the
overall supply and demand for tobacco products is an
important issue since it is related to the eectiveness of
tobacco taxes both as a revenue generating mechanism
and as a public health intervention. This impact can be
classified as the impact on average cigarette prices, on
brand proliferation, on tobacco industry investments,
and on other tobacco control policies.
Even though cigarette markets are not fully competitive,
and low-tax and full-tax cigarettes are not perfect
substitutes14, there can be some competition that could
result in lower average cigarette prices and, therefore, in
higher consumption.15
Evidence suggests that the consumption of tobacco
products is higher than it would be in the absence of tax
avoidance/evasion. Joossens and colleagues12 estimated
that in 2007 the global average cigarette price was about
3.75% lower due to illicit cigarette trade and predicted
that this price dierence was responsible for about
164,000 premature deaths a year. The impact of tax
avoidance/evasion on cigarette demand varied by country
and it depended not only on the degree of tax avoidance/
evasion, but also on population characteristics, including
the degree of responsiveness to cigarette prices. Another
study found that the presence of tax avoidance/evasion
in the UK lowered the average cigarette price by about
11.6%, increased cigarette consumption by 5.0–8.2%,
and increased the tobacco death toll by 4,000–6,500
premature deaths a year.16 The taxes/prices dierence
across US states and the possibility to purchase low-tax
cigarettes on Native American reservations increased
consumption by 4.0–8.2% and smoking prevalence by
2.0–4.3% in the period of 1992–2002, but the degree
of impact varied with the distance from the residence to
a border with a lower-price state.17 This demonstrates
that transaction costs impact the full price of low-tax
cigarettes.
One highly debated issue is whether a tax/price increase
will change the degree of tax avoidance/evasion. Theory
suggests that a tax increase will lower the consumption of
tobacco products even in the presence of tax avoidance/
evasion since the prices of both full-tax and low-tax
cigarettes increase. The reasons for the low-tax cigarette
price increase are an upward shift in the demand for
low-tax cigarette and upward sloping supply curve (due
to higher marginal costs of supplying larger quantities
of these products). In addition, the suppliers of low-tax
products see an opportunity to gain extra profit while
keeping the price gap between full-tax and low-tax
-9-
cigarettes constant. A tax increase may also prompt
enhanced enforcement eorts in anticipation of higher
levels of tax evasion, which raises the cost of supplying
low-tax products, and their prices. On the other hand,
the competition between legal and illegal products could
result in a lower impact of a tax increase on the legal
products’ prices.12
In an eort to prevent tax increases, the tobacco
industry asserts that higher taxes and prices will motivate
customers to buy illegal products rather than smoking less
or quitting. The industry claims that the higher demand
for low-tax products will increase their supply as well as
the level of crime, and that there will be no decline in
tobacco use and tax revenue will be hurt.18
The overall impact of a tax increase in terms of tobacco
use and tax revenue is a matter of empirical evidence
since it depends on the price elasticity of tobacco
demand, the cross-price elasticity for the full-tax and
low-tax products, and their new full prices. Numerous
studies have concluded that higher taxes lead to
higher prices of cigarettes sold via legal channels.2 The
responsiveness of illegal cigarette prices to tax increases
has been studied less, but there is some evidence
that the prices of both legal and illegal cigarettes go
up after a tax increase.14,19 Research demonstrates
that an increase in cigarette taxes can lead to more
tax avoidance/evasion, but also to a decline in overall
cigarette consumption and higher tobacco tax revenue,
since the observed reduction in full-tax products after a
tax increase is only partially oset by substitution towards
low-tax cigarettes.9,20,21
In Sweden, for example, cigarette tax increased by
43% between December 1996 and August 1997, while
the share of illegal cigarettes consumption rose from
2.3% to 5.8% of total consumption between 1996
and 1998. However, the overall demand for cigarettes
also declined and the prevalence dropped by 19.1%
and 4.4% among men and women, respectively. The
largest decrease in cigarette demand was among youth
and young adults (16–24 years old), whose prevalence
fell by 25% and 17.4% among males and females,
respectively. In addition, tobacco tax revenue rose by
9% in 1997 compared to 1996.11,2
Canada had a similar experience when both cigarette
consumption and smoking prevalence dropped sharply
after significant cigarette tax and price increases in the
1980s and early 1990s, despite an increase in the share
of illicit cigarettes on the market. Per-capita cigarette
consumption declined by 43% from 1979 to 1993,
youth smoking prevalence (15–19 years old) fell by
47% from 1981 to 199123, and just between 1990 and
1993 the tobacco tax revenue grew by 13%.24
In France, on the other hand, a sizeable tax increase
that doubled its cigarette prices from 1991 to 1996 did
not increase tax avoidance/evasion, but lowered adult
smoking prevalence, which decreased from 40% in
1991 to 34% in 199725, and youth smoking prevalence
(12–18 years old), which went from 30% in 1991
to 25% in 1997.2 Tobacco tax revenue rose by 78%
during that period, while the share of illicit cigarettes
on the market stayed low around 2%.25 The relatively
low degree of tax avoidance/evasion was attributed
to a tightly controlled retail environment in which all
tobacco retailers must be licensed.
Similarly, a significant 1999 cigarette tax increase
in California that resulted in relatively large price
dierences with all its bordering states (including
Mexico) motivated only 5% of all smokers to purchase
tax-free cigarettes. The study demonstrated that a
cigarette tax increase can achieve the public health
objective of reducing smoking despite the presence of
tax avoidance/evasion.27
A study of the impact of a 83% tax increase in New York
City in 200821 found that the share of littered packs that
had an appropriate tax stamp fell from 55% prior to the
tax increase to 49% after the tax increase, but the overall
cigarette consumption in the city also felt from about
22.1 million to between 20.5 and 19.8 million packs per
month. The impact of the tax increase on taxable sales
was small, which resulted in a substantial increase in tax
revenue.
These examples demonstrate that tax avoidance/evasion
may reduce, but do not eliminate, the eectiveness
of tobacco tax increases in reducing tobacco use and
raising revenues.2
On the other hand, lowering taxes for the purpose of
reducing tax avoidance/evasion led to reductions in tax
revenues and higher tobacco use.15 When the Swedish
government reduced the cigarette tax in 1998 in an
eort to curb tax avoidance/evasion, the demand
for cigarettes measured by legal tax paid sales went
up, but tax revenue went down.15 As in Sweden, the
Canadian government responded to political pressure
to reduce cigarette smuggling and in 1994 reduced
cigarette taxes. This led to a 27% increase in per-capita
consumption between 1993 and 1998, higher smoking
prevalence among both youth and adults, and tax
revenue losses.23
Given the importance of the price elasticity of tobacco
demand when assessing the impact of a proposed
tax increase, researchers examined consumers’ price
responsiveness when faced with an opportunity to buy
low-tax products. They concluded that the presence of
tax avoidance/evasion leads to significant overestimating
-10-
of the price elasticity of demand when using tax paid
sales data.14,29,30 There is some evidence that the price
elasticity of tax paid sales has increased with the rise of
on-line shopping.31
Some studies have examined whether tax avoidance/
evasion disproportionally aects youth and the poor.
Theory predicts that low income smokers and youth
will more likely buy low-taxed tobacco products due to
their lower transaction costs — the value of their time is
lower compared to high income smokers. In addition,
both youth and the poor are more price sensitive
compared to the general population.32,15 On the other
hand, a minimum set of resources might be necessary
for a person to have access to low/untaxed cigarettes.
These resources are related, for example, to travel costs
or to costs of getting Internet access.33-35
The empirical results of the impact of tax avoidance/
evasion on these vulnerable populations are mixed
and likely influenced by a country’s specific context.
Wiltshire et al.36 found that the availability of cheaper
illicit cigarettes in socioeconomically deprived areas
of the United Kingdom undermined the desire of
many smokers to quit, thus undermining the potential
impact of tobacco tax policy on their consumption.
Moodie et al37 studied a cross sectional sample of 11-
16 year olds living in the UK in 2008 and found that
a quarter of ever-smokers claimed to have been oered
and 14% claimed to have purchased cigarettes or hand-
rolled tobacco that they believed were smuggled in the
previous 6 months. Those from lower social strata were
more likely to have been oered smuggled tobacco and
to have purchased tobacco products they believed were
smuggled. In Taiwan, low-income and poorly-educated
smokers were more likely to purchase smuggled
cigarettes.38 Using data from Canada, Gruber et al.39
concluded that cigarette smuggling disproportionally
aects low-income groups and, therefore, increases
smoking-related disparities. There is some evidence
that young smokers are more likely to engage in tax
avoidance, that they consume more cigarettes if they
consume illegal cigarettes, and that those who avoid
taxes are less likely to change their smoking behavior
in response to a tax increase.40,41 On the other hand,
studies from the US showed that those with higher
income33 and higher education33,35 are more likely to
engage in tax avoidance.
The supply of low-tax products has broadened
the choices for tobacco users and increased brand
proliferation in some markets. The international
trade journal World Tobacco reported in 1996 that
“smuggling has helped to promote some of the world’s
leading brands in markets which had remained closed
to foreign imports.”42 Traditionally, illicit cigarette
brands have been products of the multinational tobacco
companies, because these are easier to market and have
a price advantage over less-known brands.43 Marlboro,
for example, represented 66% of all seized cigarettes
worldwide in 2005.44 The supply of illegal international
brands has been an important component of British
American Tobacco’s market entry strategy in Africa45,
while the supply of contraband enabled access to closed
markets in many Asian countries in the 1980s and
1990s.46 Markets in Argentina, the Islamic Republic of
Iran, Lebanon, Bulgaria, and former Soviet Republics
have also been opened using a similar strategy.47-53
The legitimate brands still dominate the illicit US
market, since bootlegging accounts for most or nearly
all of the US illicit market. European illicit trade,
however, experienced a shift from genuine products to
illicit whites (brands such as Jin Ling and Classic) and
counterfeit products (primarily from China) since the
early 2000s. For example, the illicit white brand Classic
produced by the Imperial Tobacco in Ukraine was
the third most seized cigarette brand in the European
Union in 2008,54 and the majority of UK large seizures
in 2012 – 2013 were of illicit whites.55 However, there
is some evidence that the seizure data overestimate
counterfeit cigarettes. The major tobacco companies
in Europe are obligated to pay penalties for the seizure
of genuine products according to the agreements with
the EU, and since the companies themselves determine
the origin of a product, there are doubts regarding the
reported higher rate of counterfeit seizures.56
The existence of tax avoidance/evasion may interfere
with public health policies other than tax, such as youth
access laws, bans on cigarette advertising, and laws
pertaining to product labelling, ingredients disclosure,
and retail environment.15,57-60 In addition, the existence
of underground retailers can result in a competitive
disadvantage for legitimate retailers, increasing
their motivation not to comply with tobacco-control
laws.15 There are also concerns about the relationship
between illicit tobacco trade, public safety, and the
general level of corruption.61,62 The tobacco industry is
using evidence of avoidance/evasion in order to scare
governments not only from increasing tobacco taxes,
but also from implementing other tobacco control
policies such as warning labels, plain packaging, and
ban on flavouring.43,63
The degree of government eort to combat tax
avoidance and evasion is motivated by the potential tax
revenue gain. The higher per unit taxes and the larger
the size of the market, the greater the government
incentive is to invest in these activities. A system with
clear responsibilities and incentives for all parties
involved in tax administration and law enforcement
is important, because a lack of clarity can create
loopholes to be exploited by those who engage in tax
-11-
avoidance and tax evasion. For example, a change in
enforcement responsibilities between the US state and
US federal authorities in 1978 generated a loophole
in the tax audit procedure that led to more tax evasion
by underreporting of the number of cigarettes released
to distribution.64 A problem also arises if the dierent
government levels/agencies are not equally motivated to
enforce the law.
In gauging the eectiveness of public policies, it is
useful to know how compliance varies with the level
of enforcement activities. Theory holds that resources
should be allocated to law enforcement up to the level
where their marginal benefit is equal to their marginal
cost, and when enforcement is cost eective compared
with alternative approaches.91 Therefore it might be
ecient for society to tolerate some level of tax avoidance/
evasion if the additional cost of achieving no avoidance/
evasion exceeds the benefit.
To conclude this chapter, the relevant studies are
summarized in Table 1.
of enforcement activities. Theory holds that resources
should be allocated to law enforcement up to the level
where their marginal benefit is equal to their marginal
cost, and when enforcement is cost eective compared
with alternative approaches.65 Therefore, it might be
ecient for society to tolerate some level of tax avoidance/
evasion if the additional cost of achieving no avoidance/
evasion exceeds the benefit.
To conclude this chapter, the relevant studies are
summarized in Table 1.
Table 1
Theory of tax avoidance and tax evasion: summar y of resources
topic study / source
Full price and transaction costs Merriman et al., 2000; Coleman, 1998; Goel, 2008;
Lovenheim, 2007; DTZ Pieda Consulting, 2000; Goolsbee
et al., 2007; Liber et al, 2015; Joossens, 2003
Cost of supplying illegal products Becker, 1968; Levy, 2002; Thursby & Thursby, 2000;
Campaign for Tobacco-Free Kids, 2008; Merriman et al.,
2000; IARC, 2011; Joossens, 1998; Joossens, 1999
Competition & substitution between
legal and illegal products Joossens et al., 2000; Duy, 2006
Impact of tax avoidance/evasion on the
overall tobacco use Joossens et al., 2009; West et al., 2008; Lovenheim, 2007
Impact of tax increase on tax avoidance/evasion
(and on demand for tobacco products) IARC, 2011; Duy, 2006; Merriman et al., 2000; Merriman,
2002; Chernick and Merriman, 2013; Liber et al, 2015;
Gilmore et al, 2013; Wendleby & Nordgren, 1998; Joossens,
1999; Canadian Cancer Society, 1999; Baudier, 1997;
Comité Français d’Education pour la Santé, 1998; Emery et
al., 2002
Impact of tax decrease on tax avoidance/evasion
(and on demand for tobacco products) Joossens et al., 2000; Canadian Cancer Society, 1999
Impact of tax avoidance/evasion on price elasticity
of tobacco demand Duy, 2006; Baltagi and Levin, 1986; Licari and Meier,
1997; Galbraith and Kaiserman, 1997; Gruber et al., 2002;
Goel, 2004; Licari and Meier, 1997; Goolsbee et al., 2007
Impact of tax avoidance/evasion on
vulnerable population Joossens et al., 2009; Joossens et al., 2000; Hyland, et al.,
2005; DeCicca et al., 2010; Fix et al, 2014; Wiltshire et al.,
2001; Moodie et al., 2010; Gruber et al., 2002; Callaghan et
al., 2009; Cantreill et al., 2008
Impact of tax avoidance/evasion on
brand proliferation Joossens & Raw, 1998; World Customs Organization, 2007;
LeGresley, et al., 2008; Collin et. al., 2004; ; World Health
Organization, 2003; Gilmore and McKee, 2004a; Gilmore
and McKee, 2004b; Gilmore et al, 2007; Nakkash and Lee,
2008, Skafida et al, 2012; Walton, 2009; HC, 2014; Joossens
et al., 2014
Impact of tax avoidance/evasion on
tobacco policies other than tax
Joossens et al., 2000; Ribisl et al., 2001; Ribisl et al., 2006;
Stephens et al., 2005; Pappas et al., 2007; Fleenor, 2003;
Joosens and Raw, 1998; Fooks et al., 2014
-12-
The illegal nature of tax evasion and the possible
social stigma attached to tax avoidance make the task
of measuring the scope of these activities extremely
dicult. Yet various stakeholders are interested in
understanding the phenomena, their magnitude,
and the degree of market disruption they potentially
represent. Reliable quantitative measures of tobacco tax
avoidance and tax evasion can enhance public discourse
and policy making. This has motivated the development
of methods for estimating the scope of tax avoidance
and tax evasion.
A lack of reliable data is a major challenge, since those
engaged in tax avoidance and tax evasion do not keep
public records, are not willing to provide the data, and/
or are not interested in cooperating with researchers.
Enforcement authorities may have some data, but are
often bound by confidentiality. Therefore, those who
estimate the scope of tax avoidance/evasion either find
a way to creatively use the existing data that have been
collected for other purposes, or collect new data with
the main goal to assess the scope of tax avoidance/
evasion.
This section describes various methods employed to
estimate the magnitude of tax evasion and tax avoidance,
discusses their pros and cons, suggests when it is
appropriate to apply them, and provides examples of
studies that have employed these methods.
To begin with, we briefly review the main principles
of conducting research, which apply (with some
modifications) to all the approaches described below.
Following these steps will result in sound and well
documented studies.
1. Establish research goals, define the final product
and your target audience.
2. Select the method(s) you want to apply based
on the available resources (both financial and
human).
3. Develop a research protocol. This is a grand plan
of how to execute the study including the data
collection, data analyses, and the presentation of
the results. Seek feedback from your colleagues
and those experienced with the type of research
you want to conduct. You do not want to waste
your eort on a study that might be later criticized
due to a research design flaw.
4. Many studies require obtaining ethical clearance
before the projects starts, particularly if human
subjects are involved. This procedure varies
by country. If you are not certain whether you
need ethical clearance, enquire with the local
authorities. It is better to be safe, because you
might not be able to publish your results if you do
not follow the required procedure.
5. If needed, hire and train the research sta and
make sure they understand and know how to
follow all steps of the research protocol.
6. Conduct a pilot study of your research protocol
and fine-tune it based on the experience in the
pilot test. Once this step is completed there should
be no deviation from the research protocol unless
absolutely necessary, and if there are deviations,
these should be thoroughly documented.
7. Execute the research protocol according to
the principles of the obtained ethical clearance
and carefully document each step (e.g., non-
participation rate, the date, time, and place of the
interviews). Try to collect the data in a relatively
short period of time to avoid their contamination
by a policy or a market change. Take into account
any seasonal variation in cigarette consumption
(e.g., New Year’s Eve resolutions, summer travel
season, etc.).
8. Clean the data (e.g., analyze missing, incomplete
and inconsistent responses as well as outliers)
and analyze them using the appropriate software.
Products/purchases that cannot be definitely
categorized as low-tax or full-tax products
according to the set of established criteria need
to be marked as uncertain and excluded from
the main analysis. A separate analysis can be
performed to study whether the uncertain
products are distributed randomly or not and to
what extent this eects the main results. Carefully
document all steps of the data analysis so that the
results can be replicated.
9. Report results. In most studies you will report
some measures of a central tendency such as mean
(both weighted and unweight), and measures of
dispersion (e.g., the standard error or confidence
interval). Express the estimated scope of tax
avoidance and tax evasion as a percentage of the
total market, which is the market that consists of
both full-tax and low-tax cigarettes.
CHAPTER 2
How to Measure the Scope of Tax Avoidance and Evasion
-13-
10. Present the results in the context of the existing
literature/studies. Clearly articulate all weaknesses
of the data, the method applied and any problems
in its application. Ideally, propose how the study/
research/method can be improved in the future.
11. Disseminate the results.
12. If your goal is to evaluate the impact of a policy
change (e.g., a tax increase, higher level of
enforcement, an implementation of a tracking
and tracing protocol) or assess the trend in tax
avoidance/evasion over time, repeat the same data
collection at various time points and compare the
results. The best time to repeat a study is a few
months before and a few months after a policy
change. The timing is critical since repeating a
study too soon after a change will capture a short-
run response that could overstate the reaction
(e.g., tax avoidance) due to an initial response
to the new policy. A long-run, more permanent
response is better assessed a few months after the
change.
Various methods of quantifying tax avoidance/evasion
are described below starting with approaches most
frequently discussed in the literature (survey of tobacco
users, examination of cigarette packs, gap analysis,
econometric modelling), followed by approaches that
are unique to certain market conditions (comparison of
tax paid sales with estimated consumption, comparison of
actual and projected tobacco tax revenue, key informant
interviews) and approaches suering from multiple
weaknesses (monitoring tobacco trade, analysing the
seizures).
2.1 Survey of Tobacco Users
Background
Certain characteristics of tobacco packs (e.g., the
presence/absence of a tax stamp, a health warning, price
paid, etc) as well as their sources (e.g., a duty free store)
are good indicators of tax avoidance/evasion. Such data
collected either directly from tobacco users via surveys
and/or by inspecting tobacco users’ packs (a method
described in the next section) can help us determine
how many tobacco users consume low-tax products and
then estimate the extent of various forms of individual
tax avoidance, including cross-border shopping, direct
purchases, and duty-free purchases. Some tax evasion
can also be detected. Information on the quantity and
the frequency of these purchases will help to quantify
the share of low-tax products in total consumption.
Surveys can also collect data on the characteristics of
those purchasing and consuming low-tax products as well
as solicit a subjective opinion as to whether the full-tax
has been collected.
Principles
It is crucial that the survey design (including the sampling
plan and the sample weights) and the questionnaire
be reviewed by an experienced statistician/researcher,
otherwise the results could be uninformative.
First, decide the survey mode. Surveys can be conducted
by interviewing subjects face-to-face, by telephone, by
mailing in a questionnaire, or via the Internet. Computer-
assisted interviewing methods such as CAPI (computer
assisted personal interviewing), CATI (computer assisted
telephone interviewing), or CASI (computer assisted self-
interviewing), tend to improve data quality and appear to
encourage more complete reporting of sensitive behaviors
such as tax avoidance and tax evasion. However, the
computer-assisted methods are also more expensive.
The survey mode has implications for the accuracy and
representativeness of the data as well as for the cost of the
survey, with face-to-face being the most expensive.
Second, consult a statistician about the sampling frame
and sample size. Surveys can be expensive and a good
statistician will make sure that the money dedicated to
the survey is not wasted on collecting data with very little
explanatory power. Compromising on the sample size in
order to save money is possible to the extent that valid
results can still be obtained. Use sampling techniques
that produce an accurate representation of all tobacco
users. It is important that the geographical area surveyed
be representative of the tobacco market in the entire
country or the area of interest. Selecting neighborhoods
where low-taxed cigarettes are known to be prevalent will
generate biased estimates.66 Select the unit of observation,
which can be a tobacco consuming household or a
tobacco consuming individual. The household is a
common unit of observation and can provide information
on the overall consumption pattern of a family, but
it disguises the individual level behavior if only one
person is interviewed on behalf of the entire household.
Therefore, all family members who consume tobacco
should be interviewed if possible, or one of them can be
randomly chosen. Selecting only those who consume low-
tax products will not generate an accurate estimate of the
scope of tax avoidance/evasion.
Since participation in the survey is voluntary, those
carrying low-taxed cigarettes might be less likely
to participate due to fear of legal prosecution,
confiscation, or embarrassment. This will result in
underestimating the scope of tax avoidance/evasion.
Therefore, it is important to protect interviewees’
anonymity, particularly in countries where there is
a stigma attached to buying/using low-tax tobacco
products67 and/or there is a probability of being caught
and punished.68 To improve the response rate, the
questions on tax avoidance/evasion can be imbedded
-14-
into a larger survey that collects data on a broader
range of tobacco-related issues. Often, statisticians
recommend oversampling; that is, collecting data from
more than absolutely needed number of participants
since not all subjects will be willing to complete the
survey.
Third, develop a standard questionnaire that will be
administered to all survey participants. The set of local
specific criteria for identifying low-tax products will guide
the type of questions included in the survey, because
full-tax and low-tax packs will look dierent in dierent
jurisdictions. It is important to focus on objective criteria
for identifying tax avoidance/evasion such as the source
of the product (i.e., places of purchases), the distance
travelled to get the product, price paid, etc. Very low
price and suspicious purchase location, for example,
are all possible signs of tax avoidance/evasion and
will determine if you categorize the product as full-
or low-tax. You may need to combine several criteria
in order to determine the pack’s correct category. For
example, in some countries missing a tax stamp or a
very low price might not be a sucient sign of a low-tax
product. On the other hand, missing a tax stamp and
a very low price and place of purchase associated with
tax evasion (e.g., a street market) might be sucient
evidence to categorize a product as a low-tax product.
The information about the place of purchase can help
to distinguish between tax avoidance and tax evasion.
For example, tax avoidance occurred if the permissible
number of products was purchased in a duty free
store; tax evasion would be suspected if a product was
purchased outside the established retail system (e.g., on
a street) and had other signs of a low-tax product (e.g.,
very low price, incorrect health warning). You will also
need to collect information about the type and amount of
product purchased and the frequency of its use in order
to determine the share of low-tax products in the total
consumption. Of a particular interest is the frequency
of use of products that can be defined as low-tax. For
example, you will want to know how often a product with
certain characteristics is purchased, are there tobacco
products with dierent characteristics purchased as well,
how often are these products obtained from a particular
source, etc. Sample questionnaires that include questions
designed to estimate the scope of tax avoidance/evasion
can be found on the website of the International Tobacco
Control Policy Evaluation (ITC) Project (www.itcproject.
org/surveys), or on the website of the Tobacco Use
Supplement to the Current Population Survey (TUS-
CPS) (http://appliedresearch.cancer.gov/tus-cps/info.html).
Subjective questions related to the awareness of low-
tax product purchase, knowledge of sources of low-
tax products in the area, etc. are less desirable since
the survey participants might not be familiar with or
be able to distinguish between dierent types of tax
avoidance/evasion. Therefore, there is a danger that
they would, for example, report the same purchase
as a contraband and as a counterfeit, which would
overestimate the scope of tax evasion if these two
categories were added to calculate the total size of
illegal market. In addition, tobacco users may not be
aware of a low-tax purchase if the product was obtained
via a legitimate distribution channel.
The questionnaires should also collect data on social
and demographic characteristics and other aspect
of smoking behavior. This data will determine how
representative the sample is of all tobacco users. If the
sample you collect is not representative, weights can
sometimes be developed with the help of a statistician
to correct for this. Be mindful of survey fatigue – the
survey needs to collect essential information, but
cannot be so long that it discourages participation or
provides incentive to give incorrect answers just to get
the survey done. The number of questions will also
drive the cost of the survey.
Fourth, collect and analyse the data based on the
principles of conducting research outlined above. The
data analysis will consists of calculating the share of
smokers who possessed low-tax cigarettes and the share
of low-tax cigarettes consumed by those who possessed
them (since they might also consume full-tax products),
which will make it possible to estimate the share of
low-tax cigarettes among those surveyed and among
the entire population, after applying the appropriate
weights.
Advantages and Disadvantages
Surveys are one of the most direct methods of obtaining
estimates of the scope of tobacco tax avoidance and
of availability of low-tax products via various supply
channels. However, this method is relatively expensive.
It also relies on self-reported data and all self-reported
studies have validity problems. Participants could
under-report low-tax purchases due to social norms.
Even when the purchases are legal, consumers are known
to under-report purchases of cigarettes on surveys.69 If
the degree of underreporting of consumption, rather
than being random, is systematically greater among
heavier smokers, and heavier smokers are more likely to
purchase low-tax products,70 then surveys of smokers may
underestimate the amount of tax avoidance/evasion. The
method is prone to understate tax evasion since tobacco
users might be uncertain in some cases if the cigarettes
they purchased were legal or not (e.g., may not be able
to detect counterfeit cigarettes) and whether appropriate
taxes were paid. To mitigate some of the weaknesses of
this method, it is recommended to combine it with an
independent examination of cigarette packs, a method
described in the next section.
-15-
When This Method Should Be Used
Use this method when it is possible to determine
whether taxes were paid by analysing tobacco users’
purchasing behaviour and/or the self-reported features
of a cigarette pack. Since this method can be rather
expensive, it is important to secure sucient resources.
Conducting surveys and analysing the data can be time
consuming. Therefore, this method should not be used
if results are needed quickly.
2.2 Examination of Cigarette Packs
Background
This method is based on classifying packs as low-tax or
full-tax products given the law and regulations applicable
to the jurisdiction where they were found. The features
that allow this distinctions are the absence of the
correct tax stamp, an incorrect health warning, markings
of a duty-free store, missing price information (if
required by the law), low price, and some other features
of a pack required by the law. The data are objectively
recorded from the packs. Therefore this method belongs
to the category of observational studies. Packs can be
obtained from tobacco users, from retail outlets, or
collected on the street and in trash.
Principles
First, develop a sampling frame to make sure that
collected packs are representative of the entire population.
If this is not possible (e.g., due to budget constrain),
consult with a statistician to determine if there are ways
to correct for this, or how best to present the results if the
sample is not representative.
Second, train those who will be examining the packs to
become familiar with characteristics of low-tax products
so that they can determine if a pack is a low- or full-
tax product. Test these examiners with packs of known
origin to ensure that they received clear instructions
and that they are able to identify package successfully.
Independent experts (e.g., sta of a government lab)
can detect genuine and counterfeit products and/or tax
stamps.
Third, collect the data. Specifics of obtaining packs from
dierent sources are described below.
Fourth, inspect each pack and record the data. The
data collection can be organized by using a standard
questionnaire that captures characteristics of each pack
such as cigarette brand, pack size, presence of a tax
stamp, health warning, price, markings of a duty-free
store, and other pack markings that can determine if
the correct excise taxes have been paid and/or a possible
origin of the pack.
Fifth, categorize the packs as low- or full-tax based on the
set established local-specific criteria. The presence of a tax
stamp on a pack is the best evidence that taxes were paid.
However, tax stamps are not required in all countries or
jurisdictions, or they might have been removed or lost in
the process of opening the pack (for example, if they are
axed to outer cellophane wrap) or discarded before the
pack ends up on the street or a garbage bin. Therefore,
it is desirable to inspect an unopened pack35 or obtain
it for later detailed inspection that could generate more
definite answers. Help from independent experts is often
required to identify counterfeit tax stamps or counterfeit
cigarettes.71
Another sign of tax avoidance/evasion is an incorrect
health warning (e.g., health warning in another language
or text warning instead of pictorial health warning), a
very low price paid for the product or a product that was
obtained from a source known to be associated with tax
avoidance/evasion (e.g., another country/jurisdiction, a
street seller, etc.). You may want to combine several low-
tax products features if one of them is not sucient to
determine low-tax product with certainty. For example,
a missing tax stamp might not be sucient sign of tax
avoidance/evasion if the stamp can be easily removed in
the process of opening a pack. However, a missing tax
stamp and a very low price could be sucient evidence
of a low-tax pack. The particular combination of pack
features will be local specific. Packs that cannot be
classified as low-tax products with certainty should be
classified as uncertain and excluded from the study.
Six, analyse the data based on the principles of
conducting research outlined above. The data analysis will
consists of calculating the share of packs categorized as
low-tax cigarettes among all collected packs. If you have
a statistically representative sample, you can generate an
estimate of a share of low-tax cigarettes on the market by
applying the appropriate weights. If packs are obtained
from tobacco users, you need to account for the share of
low-tax cigarettes consumed by those who provided them
similarly to the way it was done in the survey of tobacco
users’ method. This will determine the share of low-tax
among those surveyed, and among the entire population
after applying the appropriate weights.
Advantages and Disadvantages
Since this is an observational study, it eliminates the
validity problems associated with self-reported data.
However, examining cigarette packs alone without
additional information from those who possessed
them or from a counterfeit expert cannot distinguish
between legal tax avoidance and illegal tax evasion with
the exception of packs obtained from retailers, where all
low-tax packs represent tax evasion. It can detect some
illegal tax evasion, but must rely on expert inspection
-16-
(in case of counterfeit products or counterfeit tax stamps)
or combine the data with consumers’ self-reports if the
pack is obtained from tobacco users (e.g., how a pack was
obtained, how much it costs). However, since the survey
of tobacco users relies on self-reported information, there
might be dierences between estimates generated by a
survey of tobacco users and by examination of cigarette
packs. Discarded packs from street/garbage bins or packs
obtained from stores usually provide results for limited
geographical areas. Therefore, the generalization of the
results is limited. However, a statistician can help to
design a sampling frame so that the data can sometimes
provide more information.
When This Method Should Be Used
Use this method when it is possible to determine
whether the correct taxes were paid by studying features
of a cigarette pack. Collecting packs and analysing the
data can be time consuming. Therefore, this method
should not be used if results are needed quickly.
The costs of this method will depend on the sources
of packs and the representativeness of the sample.
Collecting packs from tobacco users can be expensive,
but the costs can be cut substantially if inspecting packs
is an add-on into an existing survey. Collecting packs
from streets, garbage bins or retail stores in one city/
town is generally cheaper than collecting packs from
survey participants, but the generalization of such
results is limited. Obtaining a national representative
sample from these sources can be expensive.
2.2.1 Specifics of Obtaining Packs
from Tobacco Users
Cigarette packs can be obtained during a face-to-face
interview, by mail, or by intercepting smokers in public
places.
Obtaining packs during an interview requires following
the same steps described for the survey method. Ask
for all presently opened packs since some smokers
may have both a legal and an illegal pack open at the
same time.72 Packs can then be either inspected in the
presence of those being surveyed or at a remote site if
the subjects are willing to surrender their packs (usually
in an exchange for a reward). Obtaining a pack for
future detailed examination is desirable since it will
increase the precision with which packs are classified.35
Another option is to photograph all sides of the pack
for later inspection of the images.
If inspecting the pack is not feasible (for example, if
the survey is conducted via phone or via Internet), the
survey participants can send pictures of packs, or mail
them for inspection. The mail-back should be done at
no cost to those mailing the packs and can be motivated
by a reward. However, anonymity could be a concern
when mailing cigarette packs, if those who mail in their
packs receive a reward in exchange. In addition, mailing
cigarette packs is rather time consuming, making it likely
that those with a high opportunity cost of time (that
is those with high income) will not participate. These
individuals would be also less likely to consume low-
tax products, because their financial resources simply
make it not worth the risk. Therefore, this method can
both understate (concerns of anonymity) or overstate
(high-income smokers who consume few illicit cigarettes
will not participate) the scope of tax avoidance/evasion.
Despite these limitation, there are mail-in pack samples
that are reasonably representative of the smoking
population.35
A relatively new method is obtaining cigarettes packs
from smokers by intercepting them in public places and
oering them a new pack for the pack they are currently
smoking (pack swaps). This approach is a combination of
the method of obtaining cigarette packs from smokers
during a survey and the methods of obtaining discarded
packs from the streets. An advantage of this method is
that surveyed respondents may retain anonymity, but the
selection of the location of the intercepts is problematic,
as it is with any convenience sampling approach. Some
industry-funded studies have used this approach, but do
not provide sucient details about how it was executed.
We have not identified any peer–reviewed publication that
has applied this method.
Once collected, the packs data are entered into a
standardized form. This will facilitate the data analysis.
2.2.2 Specifics of Obtaining Discarded Packs
Packs are collected from a random sample of discarded
cigarette packs from the ground (litter) and/or from
garbage (properly disposed trash). The goal is to
obtain a representative sample of cigarette packs being
consumed in a relatively concise geographical location.
The representativeness of the sample is key in this
method since it will otherwise produce ambiguous
information about the population being studied.
First, select a geographical area of interest. The size of
the area will depend on the resources available for the
project. If possible, gain cooperation with local public
authorities in this area such as revenue departments,
police departments and sanitation departments. Getting
these authorities involved will be useful in case there
is a need to collect discarded packs from appropriately
disposed garbage and/or if the data collectors run into
any diculties when collecting trash on the streets.
Revenue authorities might be interested in the results
and can help with their dissemination.
-17-
Second, develop a statistically valid plan to collect a
representative sample of discarded cigarette packs being
consumed in the given area. The geographical area of
interest will be divided into sub-areas that completely
cover it, and then you will randomly select among
them to determine where the discarded packs will be
collected. Weights are often used during the process of
selecting sub-areas in order to account for population
commuting patterns and/or tourists’ presence. Trash
from commuters, visitors and tourist can make the
results more dicult to interpret since their litter
packs legitimately purchased in their jurisdiction/
country would be illegal if purchased in the study area.
Collecting packs in locations that are frequented by
visitors/tourists (e.g. tourist markets, football stadiums)
or by those with a higher propensity to use low-tax
tobacco products (e.g. near the border with a lower tax
jurisdiction) should be generally avoided.
Select routes in each selected sub-area along which
the discarded packs will be collected while taking into
account the possibility of finding littered cigarette
packs, the possibility of being able to physically walk the
entire route, and safety. Even though walking along all
streets in the selected sub-areas is desirable, dierent
studies have used dierent approaches in selecting
routes (e.g., walking perimeters of study areas21,
randomly selecting routes throughout the study area73,
and walking along all sidewalks within each study
area71). The route selection has implications for the final
estimates. For example, the perimeter approach is likely
to overstate the problem given that the proximity to
lower tax jurisdictions is one of the key determinants of
tax avoidance.
Third, begin collection of discarded packs. A typical
collection involves data collectors (a team of 2 or 3
people) to walk a certain distance (e.g., one mile) along
the selected route. All littered packs are collected and
well-documented. Cigarette packs can be put into pre-
labeled bags with the route location, date, time, and
names of the collection team members. Each pack must
be uniquely labelled.
If resources permit, conduct a separate survey of
appropriately discarded cigarette packs (e.g., in
garbage bins) within the same areas and compare
the results. This comparison provides information
about whether littered packs systematically dier
from properly disposed of empty packs. This method
generates unbiased estimates if these two groups are
not systematically dierent. Existing empirical evidence
suggests little dierence between littered packs and
properly disposed of packs.73,74
An important weakness of this method is the lack of
sample representativeness when data is collected in
a limited geographical area. This can be addressed
by expanding the scope of the study, which could
require substantial resources. A statistician can help
with a sampling frame that will generate an acceptable
representativeness of a sample given the available
resources.
Another concern is related to the sample being
contaminated by commuting patterns and tourists.
Researchers have dealt with these concerns in dierent
ways. Some used dierent weights when selecting
the pack collection sites, and some repeated the data
collection at dierent points in time focusing on
estimating the change in tax avoidance/evasion rather
than its scope.21
2.2.3 Specifics of Obtaining Packs
from Retail Outlets
This method is similar to the survey of tobacco users
except for the unit of observation, which is now a store
selling tobacco products. First, develop a sampling
frame and calculate the sample size taking into account
the density of tobacco retail outlets, which might be
easier to get in places that require licensing to sell
tobacco products. Alternatively, the area of interest
needs to be first surveyed to identify all outlets that sell
tobacco products. The information on the density and
type of outlets can be used to calculate the sampling
weights. It is important that the geographical area
surveyed is representative of the tobacco market in the
entire area of interest. Selecting neighborhoods where
low-taxed cigarettes are known to be sold or consumed
will generate biased estimates.
Second, develop a standard protocol for selecting a
retail outlet and for steps taken during the visit, with
the goal of purchasing low-tax product if available. The
availability of a low-tax product can be determined
during an interaction with the sales person by asking
for the cheapest tobacco product available and then
verifying with the sales person that this is truly the
cheapest product available.
Third, conduct the store visits. Collect information
about the store type, the size of the store, its location,
and the gender and approximate age (and race if
relevant) of the clerk administering the sale. This
data can then be analyzed to determine if the store
sample represents all tobacco retail outlets in the area,
and if there is a possible gender/age/race bias in the
willingness to sell low-tax products. If the sample is not
representative, weights can be developed with the help
of a statistician.
-18-
Fourth, calculate the share of stores that sold low-tax
products based on the set of low-tax product criteria
described above. Since these products were sold in
outlets not designated to sell these products (i.e., these
are not duty-free stores), all of them evade taxes and are
considered illegal.
This method cannot determine the scope of tax
evasion because it does not provide information about
the portion of total tobacco sales at that store that is
illicit, nor does it allow us to estimate the store’s share
of tobacco sales in the overall tobacco market. In
addition, this method will most likely underestimate
the share of stores that sell illegal products since
clerks might be less willing to sell these products to
an unknown person conducting the study as opposed
to a regular customer. In many places, it could be
rather challenging to get good data on the density and
type of tobacco outlets, particularly if street vendors
represent an important part of the distribution system.
Despite these limitations, it is one of the most direct
methods of obtaining estimates of availability of illegal
products via legal distribution channels and can provide
important information about the role of these channels
in supplying illegal low-tax products. If repeated over
time, the method can inform about changes in the role
retail outlets play in the distribution of illegal low-tax
products.
Examples for Survey of Tobacco Users and for
Examination of Cigarette Packs Methods
Guindon et al (2014)75 assessed the levels and
trends in tax avoidance/evasion in 16 countries using
longitudinal cohort survey data from the International
Tobacco Control Policy Evaluation Project (ITC)
2002 – 2011 and combining the survey method with
the examination of cigarette packs. Countries used
probabilistic sampling but the survey mode diered
by country (face-to-face, telephone interviewing,
web-based interviewing), with some countries using
mixed survey modes. Sampling weights accounting
for the survey mode and the survey non-response
were employed to generate nationally representative
estimates (with the exception of China and Mexico
where the surveys were conducted only in large cities).
The extent of tax avoidance/evasion was measured
using two approaches. One was based on self-reported
information about the source of a smoker’s last
cigarette purchase. Cigarettes that came from a Native
American reservationiv, out of state/province/country, a
duty-free outlet, a direct purchase (mail, telephone or
Internet), ‘someone else’ (such as an independent seller
or a military commissary) were classified as low-tax
products. The individuals who reported that their last
source for cigarettes was from a friend or a relative or
who reported not knowing or refused to answer were
excluded from the calculation. The second approach
was based on packaging information (self-reported
or observed during a face-to-face interview). Packs
that did not have a warning label, a tax stamp, and/
or a security ink required by the applicable law were
classified as low-tax products.
The authors calculated the share of low-tax purchases
in individual countries and how it evolved over time.
In high-income countries such as Canada, France
and the UK, this share was about 10%, but the share
was much higher in some low-income countries. For
example, up to 40% of all purchases in Malaysia were
classified as low-tax purchases. There was a decline
in tax avoidance/evasion in the UK, an initial large
increase—more than fourfold—in Canada followed
by a small decrease, and relatively stable levels of tax
avoidance/evasion in the USA. The sources of low-tax
cigarettes were very much country-specific, highlighting
the importance of country-specific contextual factors.
The study could not disentangle tax avoidance from
tax evasion and could not capture sales of low-tax
cigarettes in legitimate stores, counterfeit cigarettes, or
cigarette with counterfeit stamps. The major drawback
of the study is that the data did not document the
frequency and amount of purchases from low-tax
sources. This led to underestimation of the extent of
tax avoidance/evasion, because the data revealed that
those who buy cigarettes from low or untaxed sources
tend to be heavier smokers. The estimates based on
self-reports were substantially dierent compared to
the estimates based on pack inspection. This dierence
could have been driven by social desirability bias (when
respondents provide socially desirable answers) and/
or imperfect recall. This suggests that self-reported
information may not be always reliable.
Joossens et al. (2014)76 collected data on cigarette
packs during a 2010 household survey of adults
conducted in 18 European countries while also
examining cigarette packs in order to study tax
evasion. Dierent sampling methods across countries
took into account the local specific setting and
generated a representative sample of 18056 subjects
interviewed face-to-face (a computer-assisted personal
interview). Those classified as current smokers (5114
observations) provided information on the pack’s
provenance and price. Then they were asked to show
their latest purchased pack of cigarettes (manufactured,
hand-rolled or other types) to a trained interviewer who
recorded information about health warnings and a tax
stamp. A pack was identified as illicit if it had at least
one of the following characteristics: it was bought from
iv Reser ves sell low-tax cigarettes that are intended for its inhabitants only. However, they also sell to visitors
-19-
a known illicit source (e.g., from individuals selling
cigarettes at local markets or in the streets, delivery
service, door-to-door sale), had an inappropriate
tax stamp given its self-reported origin, had an
inappropriate health warning given its self-reported
origin, or its price was 70% below the lowest price of
a cigarette pack in the country as reported by WHO.
Packs with destroyed or removed tax stamps were
not classified as illicit due to the possibility that the
stamp was removed in the process of opening the pack.
The study could not detect counterfeit tax stamps or
counterfeit products; thus, it may have underestimated
the size of tax evasion. However a counterfeit
product was still classified as an illicit product if it
possessed other characteristics such as very low price,
inappropriate health warning, or was obtained from
a known illicit source. The smokers may have tried
to hide or may not have known the origin of the pack
generating a potential downward bias in the estimates.
In order to address this issue, the interviewers avoided
using words such as ‘smuggling’ or ‘illicit trade’ in order
to decrease the perceived sensitivity of the issue.
About a quarter of smokers were not willing to show
their pack. If the willingness to show a pack was
negatively associated with the probability of owning
an illegal pack, the study would underestimate the rate
of tax evasion. To study this possibility, the rate of tax
evasion was compared between those who showed the
pack and those who just described the pack but did
not show it. No substantial dierence between the two
groups was found. In addition, the multiple signs of tax
evasion were cross-validated, showing a high degree of
consistency. The likelihood of possessing illicit cigarette
was estimated using regression analysis that took into
account sex, age, education, the number of cigarettes
consumed per day, and the heterogeneity between the
18 European countries.
Results showed that about 6.5% of smokers in the
18 EU countries possessed a pack classified as illicit,
with the highest estimate in Latvia (37.8%) and the
lowest in Portugal (0.0%). The country level estimates
were compared with a study based on empty packs
collection from streets and public bins.77 Since the
empty packs method could not separate tax evasion
from tax avoidance, the KPMG (2010)77 estimates were
larger compared to Joossens et al.56 that focused only
on the tax evasion. This demonstrates the importance
of correctly dierentiating between legal tax avoidance
and illegal tax evasion.
Nagelhout et al. (2014)78 focused on legal tax
avoidance in Europe. They also used a cross-sectional
survey of adult smokers, but surveyed only five
European countries and relied only on self-reported
data on the frequency of cigarette purchases outside
the country in the last six months. They found that
cross-border cigarette purchasing is more common
in European regions bordering countries with lower
cigarette prices. For example, 24% and 13% of smokers
living in near a lower-priced border in France and
Germany, respectively, reported purchasing cigarettes
frequently outside their country. On the other hand,
only 2–7% smokers living in the non-border regions of
France and Germany reported frequent purchasing of
cigarettes outside the country. The data did not allow
for estimating the scope of tax avoidance due to missing
information on the share of cross-border purchases in
total cigarette consumption among those who engaged
in tax avoidance.
Fix et al (2014)35 combined survey data on smokers
with mail-in pack collection to estimate the prevalence
of cigarette packs that were not taxed by the US state in
which the participant lived.v A nationally representative
sample of the daily smoker cohort participating in
the 2009 and 2010 waves of the USA International
Tobacco Control United States Survey was asked
during a telephone interview to send an unopened
pack of their usual brand of cigarettes purchased at
their usual outlets. Those who agreed to send in a pack
were mailed a data collection kit, which included an
information sheet, cover letter, instructions, a short
questionnaire, a plastic zip-top bag for the pack and
a postage-paid return envelope. Participants received
US$25 in order to compensate them for their time
and eort. The response rate among those who
initially agreed to take part in the pack collections
was 79% and 75% in 2009 and 2010, respectively,
and the researchers were able to visually inspect 684
packs for the presence of a tax stamp. If there was no
stamp or the stamp did not match the participant’s
state of residence, the pack was classified as low-tax.
This selection criteria disqualified participants from
three US states (North Carolina, North Dakota
and South Carolina) that do not use tax stamps.
Self-reported usual brand and its pack Universal
Product Code (UPC) obtained during the telephone
interview was compared with the brand family and
UPC printed on the pack sent for analysis in order
to address concerns over whether a participant might
have reported smoking a more expensive brand, but
sent a less expensive brand. There was 97.2% and
92.6% agreement with respect to the brand variety
in 2009 and 2010 surveys, respectively, but a lower
agreement between the self-reported and the observed
UPC in both surveys. Further analysis found that the
majority of the mismatched UPCs were a result of the
participants making a mistake in reporting the UPC
digits over the phone.
v In the USA tobacco tax rates var y by state
-20-
The results showed that 20% of the packs in 2009
and 21% in 2010 were classified as low-tax with the
prevalence higher in states with higher-excise taxes.
Smokers who did not plan to quit were significantly
more likely to have sent a pack that was classified as
untaxed.
A particular strength of this analysis is that the data
collection was conducted in a similar fashion at two
dierent points in time from a nationally representative
sample of US smokers. The high rates of participation
and replication of findings over time suggest that this
type of data collection is feasible and relatively cost-
eective. However, the study has several limitations.
First, the method yields an estimate of tax avoidance
and tax evasion without being able to separate them.
This limitation could have been addressed if those
who were sending packs also provided information
about their purchase behavior (e.g., where was the pack
purchased). Second, the sample size was small. Third,
smokers who knowingly avoid taxes by purchasing
cigarettes from unlicensed tobacco outlets might be
less likely to answer a survey or send a cigarette pack
for inspection. The prevalence of packs that did not
show evidence of tax avoidance or evasion was higher
among those who sent a pack when compared with the
information provided over the phone, lending some
support to this hypothesis. The study was unable to
measure the distance between a participant’s residence
and the tobacco outlet from which the cigarette pack
sent for inspection was purchased. This limited the
possibility to test whether the proximity of lower-
priced sources is an important factor in motivating tax
avoidance/evasion behavior.
Scollo et al. (2014)79 evaluated changes in the
availability of illicit tobacco in small retail outlets
following the December 2012 introduction of plain
packaging in Australia. The sample of 303 stores was
obtained from randomly selected postcode-based areas
stratified by socioeconomic status in four large cities.
Fieldworkers who were demographically similar to
known users of illicit tobacco started to walk from a
predetermined starting point in each area and sampled
every eligible store they encountered. A minimum of
six stores per area were visited at the time when the
store was not too busy with other customers. In each
store, the fieldworker asked for a particular brand
of low-cost cigarettes in a small pack size so that the
retailer knew that the purchaser was interested in cheap
cigarettes. After the retailer retrieved the requested
pack, the fieldworker enquired whether a cheaper pack
of cigarettes was available and purchased the cheapest
pack oered. In a subset of 179 stores, the retailer
was also asked about the availability of chop-chop (an
illegal low-tax tobacco) in the area. The same stores
were visited six times during the study period: two
times prior to implementation of plain packaging, once
during the implementation period, and three times well
after the period of initial implementation. Fieldworkers
rotated across stores.
Collected packs were later examined to assess any
divergence from Australian packaging regulations,
and their prices were compared with tax liability and
recommended retail price (RRP) for the particular
brand and pack size. Prices that were more than 20%
cheaper than the RRP, packs cheaper than the tax
liability, and packs with incorrect packaging were
considered illegal products. The retailer responses
regarding the availability of chop-chop were aggregated
into three categories: (1) positive responses — oered
to sell or provided information where to get it;
(2) negative responses — did not oer to sell, did not
know of a source or confused chop-chop with roll-
your-own tobacco (RYO); and (3) suspicious — the
retailer behaved as if they were suspicious about the
fieldworker’s request. Logistic regression analyses
examined dierences between waves in the likelihood
of encountering a negative, positive, and suspicious
response.
The results showed that 13 (2.2%) of 598 packs
purchased pre-plain packaging were classified as low-
tax. Four packs (1.3%) of 297 were illegal in December,
the implementation month, and five (0.6%) of 878 in
the three collections following implementation. The
availability of chop-chop in small retail outlets was
low — it was oered directly only on six occasions
(0.6%), 5.8% enquiries resulted in information about
where to get it, while in 88.5% cases retailers either
did not know what chop-chop tobacco was, did not
know where it could be purchased, or they confused
unbranded tobacco with RYO tobacco. The authors
concluded that there was no change in availability of
illicit tobacco observed following the implementation of
plain packaging.
The study acknowledged its weaknesses: the survey
did not include specialist tobacconists and informal
sources. If retailers are more willing to sell illegal
products to known regular customers, this study would
have underestimated the size of the problem. In 5.2% of
cases the retailer became suspicious of the fieldworker
and did not give a response. The repeated surveys thus
mostly assess a change over time, which was critical for
evaluating the impact of plain packaging.
Merriman (2010)73 collected discarded packs in the
city of Chicago in 2007 to study how tax avoidance/
evasion varied with the level of tax and the distance
to lower tax neighbouring jurisdictions. Chicago was
-21-
selected for the study due to its close proximity to several
jurisdictions with dierent cigarette tax rates, and due to
the fact that the city requires that all packs have both the
state and the city/county stamp axed to each pack. This
feature assisted with classifying packs as low-tax or full-
tax. The total of 135 zones were randomly selected from
the sample of all city zones and zones in neighbouring
jurisdictions, with the probability of a zone selection being
weighted by population (100% weight) and employment
size (49% weight). In each selected zone, a one-mile data
collection route was selected based on being a safe place
to walk as well as an appropriate place to find littered
packs. This method generated 2,391 littered cigarette
packs. Since the tax stamp is attached to the cellophane
wrap, only 1,141 packs (47.7 %) with a wrap were
considered for the analysis. To control the quality of data,
the tax stamp information from each pack was recorded
twice by dierent researchers and the records were
cross-verified to detect any discrepancies. The packs with
mismatched records were retrieved and correct coding
was determined.
Data on commuting flows and the ratio of residents
versus non-resident workers was used to obtain
predictions of the share of the packs from dierent
jurisdictions that might be expected if there was no tax
avoidance/evasion. To address a potential bias of the
discarded pack method based on the notion that litterers
may be systematically dierent than non-litterers with
regard to tax avoidance behaviour, a separate survey
of appropriately disposed cigarette packs from some
of the same areas was also conducted. Further, the
distribution of cigarette brands sold in Chicago by legal
vendors was compared with the littered sample. Both
tests demonstrated that littered cigarette packs are
representative of cigarette consumption in the area. The
study found that 75% of packs found in within the city
boundary did not display the correct Chicago tax stamp,
but it could not distinguish between tax avoidance and
tax evasion.
Chernick and Merriman (2013) employed the
discarded pack method to study the impact of an
83% state tax increase in New York State in 2008.
To narrow the geographical focus of the study, the
researchers selected New York City (NYC) where the
combined state and city tax reached $4.25 per pack
after the tax increase. The tax rates in neighboring
states of New Jersey, Connecticut, and Pennsylvania
were $2.70, $3.00, and $1.35, respectively, which may
have provided incentives for tax avoidance. Discarded
packs were collected on the periphery of a randomly
selected sample of 30 census tracts in NYC. They
were assigned equal weights, but the census tracts were
selected proportionally to the number of residents and
the workers employed in them. The generated sample
of littered packs was therefore representative of packs
smoked by residents and those who worked in the area.
Since the tax stamp is attached to the cellophane wrap,
only the discarded packs with a wrap were considered
for the analysis. Packs without cellophane were also
collected to determine whether there was a systematic
dierence in the distribution of brands with and
without cellophane. The results show that the share of
littered packs that had an appropriate tax stamp fell
from 55% prior to the tax increase to 49% immediately
after the tax increase. The 49% share was essentially
unchanged in subsequent rounds of data collection
(three and 15 months after the tax increase). By
collecting data both before and after the tax increase,
the study estimated the impact of a policy change while
limiting the potential influence of unrelated factors
(e.g. the presence of tourist, the higher propensity to
avoid taxes among those who litter) on the estimate.
Presenting data from several rounds of pack collection
after the tax increase demonstrated the stability of the
tax avoidance measure over time while data collected
15 months after the tax increase captured long-term
market adjustments. As in Merriman73, the study
couldn’t distinguish between tax avoidance and tax
evasion.
2.3 Compare Tobacco Sales and
Consumption (Gap Analysis)
Background
This method, also called “gap analysis”, estimates tax
avoidance/evasion as the dierence between estimated
consumption of cigarettes at national and/or local
levels and tax-paid sales for the corresponding area.
It draws on data that most governments collect and
can be deployed with little methodological variation
across countries/regions. The estimates can serve as a
benchmark for existing estimates, and may allow for
a better understanding of the relation between policy
changes, including tax changes, and changes in low-tax
consumption.
The method is based on a simple arithmetical model.
The total market for cigarettes is defined as:
Q = QL + Q
Where Q is the total quantity of cigarettes consumed,
QL is the quantity of legal cigarettes consumed and
QI is the quantity of illicit cigarettes consumed. The
number of people in the population who smoke, i.e.,
the smoking population (PS), can be calculated by
multiplying the population (P) by smoking prevalence
(R):
PS = P x R
-22-
The smoking population (PS) multiplied by the average
consumption per smoker or smoking intensity (A) gives
us the size of the total market.
Q = PS x A
Substituting equation 3 into equation 1 and making QI
the subject of the formula gives us:
QI = (PS x A) - QL
The method assumes that the smoking population
(PS), the smoking intensity (A) and the size of the legal
market (QL) are known.
This method is primarily used to detect deviations
from the trend. For example, a sudden increase in the
gap following a tax increase would be evidence of an
increase in tax avoidance/evasion. It could also estimate
the magnitude of tax avoidance/evasion if it is possible
to safely assume that there was no tax avoidance/
evasion in a country at some period of time, for which
data is also available.
Principles
First, obtain several years of reliable data on physical
quantities of tax paid sales (QL). The best source
is government agencies responsible for tobacco tax
collection such as the Tax Administration and Customs
Department.
Second, obtain data on cigarette consumption (R and
A) from national representative surveys for the same
years as QL. If you are estimating tax avoidance/evasion
within a smaller geographical area (for example to
address tax avoidance/evasion of taxes that are levied
locally), use the appropriate local survey. Surveys
usually report tobacco use prevalence R and smoking
intensity A, but it may be necessary to adjust both
variables using weights so that they are representative of
the population of interest.
When using prevalence data, be sure to account for
non-daily smokers if they are reported separately
from daily smokers. If the survey covers only the
adult population, the adult smoking prevalence will
be multiplied by the total number of adult population
P to generate the number of adult tobacco users PS.
Estimate the smoking youth population by using youth
tobacco survey and follow the same approach as for
adults, only this time using the total youth population
instead of the adult population. Make sure there is no
age overlap between the adult and the youth survey.
If there is an age overlap, adjust the population size
accordingly.
Estimate smoking intensity for adults and the youth
separately. Smoking intensity needs to correspond to
the time frame of reported sales data QL. In most cases,
this time frame is one year, so you will need to estimate
an average yearly consumption of cigarettes based on
survey answers. Most surveys ask about daily or weekly
consumption, and these will need to be aggregated
to obtain yearly consumption. If daily consumption
is reported in ranges (e.g., 5 – 10 cigarettes/day),
use linear interpolation of the mid-distribution
function (using, for example, command iquantile in
Stata) to estimate the average daily cigarette use. Be
careful distinguishing between overall tobacco use,
manufactured cigarette use and roll-your-own cigarette
use. If this method is used for estimating manufactured
cigarette tax avoidance/evasion, use only manufactured
cigarette prevalence.
Third, multiply the smoking population PS by smoking
intensity A separately for adults and youth. Add the two
estimates to generate an estimate of the size of the total
market Q for the year in which the survey was taken.
Fourth, repeat the estimate of Q for all survey years.
Carefully investigate any changes in the wording of
survey questions or the sampling strategy over time,
since even small changes in survey procedures can
significantly aect reported consumption. Ideally,
the consumption data would have been collected
systematically over time. If this is not the case, consult
with a statistician as to how these changes may have
aected the survey results and adjust the estimates
accordingly. Document all steps and all assumptions
made when calculating total consumption.
Fifth, compare the tax paid sales QL with the estimated
consumption Q and study how the gap evolved over
time. This can be done by calculating the percentage
change in tax paid sales and the percentage change in
reported consumption over time. Estimates of changes
over time are more useful and reliable than estimates
of the scope of tobacco tax avoidance/evasion due to
inherent weaknesses of this method (see below). If there
is a period when it can be safely assumed that there was
no tobacco tax avoidance/evasion (e.g., due to historical
events), the gap between the tax paid sales QL and the
estimated consumption Q, if there is any, would have
been caused by other factors such as underreporting.
Any increase in this gap would measure the magnitude
of tobacco tax avoidance/evasion.
Advantages and Disadvantages
This method is transparent, replicable, and relatively
inexpensive since it relies on secondary data. However,
it cannot distinguish between tax avoidance and tax
evasion and cannot determine whether illicit cigarettes
-23-
are counterfeit or contraband. It is primarily used
to detect deviations from the trend, not to estimate
the scope of tax avoidance/evasion. However, the
magnitude of tax avoidance/evasion could be assessed
if there was a period with no tax avoidance/evasion, for
which data exists.
In many countries it is relatively easy to obtain reliable
statistics about tax-paid sales of tobacco products, but
there might be countries where this type of data is not
publically available. Even more problematic is the lack
of nationally representative survey data on tobacco
use for multiple years. In addition, the quality of the
data collection may be questionable in some countries,
especially in countries that lack resources for data
collection. Poor data quality will result in estimates that
are highly unstable over time and might be erroneously
interpreted as volatile changes in the extent of tax
avoidance/evasion.
Estimates of consumption from surveys can suer
from problems other than the reliability of the
survey data. The data might be contaminated, for
example, by consumption underreporting, recall
bias when the survey participants do not remember
correctly how many cigarettes they consumed, and
the problem of “rounding” when smokers report
smoking a pack or half a pack per day even though
the actual number of cigarettes consumed per day
was dierent (e.g., 23 instead of a pack or 7 cigarettes
instead of a half pack). It has been documented that
respondents consistently understate the quantity of
tobacco consumed when responding to surveys. An
adjustment for underreporting is possible if there are
independent estimates of the level of underreporting.
If the exact level of underreporting is unknown, one
can use several possible scenarios based on evidence
from other countries. However, underreporting is
often related to the social acceptability of smoking
and if smoking is becoming less socially acceptable,
underreporting of consumption may increase over time.
Therefore, assuming the same level of underreporting
over time might not be accurate. To the extent that
underreporting varies across countries (potentially
reflecting dierences in social norms about smoking),
dierences in the size of the gap between total market
size Q and tax-paid sale QL won’t necessarily reflect
cross-country dierences in avoidance/evasion. If there
is no tax avoidance/evasion, this method can be used to
estimate the degree of consumption underreporting.
Estimates of consumption Q can be distorted by
the presence of tourists and immigrants. If these
populations are buying tobacco products, but are not
included in the calculation, the consumption estimates
will be biased downwards, leading to lower estimates of
tax avoidance and tax evasion.
The comparison between cigarette consumption and
legal sales is further complicated by the presence
of roll-your-own (RYO) cigarettes that might not
be included in the ocial sales statistics, but are
reported as cigarette consumption during the survey.
Consequently, the comparison of consumption (that
may also include RYO cigarettes) with manufactured
cigarette sales would overestimate the level of tax
evasion/avoidance.
In the case that some tax-paid cigarettes were illegally
exported from a country, the gap between sales and
estimated consumption will provide a downward-biased
estimate of tax avoidance/evasion.
There can be some temporal biases in tax-paid sales
measures, as these generally reflect shipments from
factories wholesale rather than actual consumption. It
can be particularly profound if the industry is trying to
ooad cigarettes before a tax increase in an attempt to
reduce their tax liability.
When This Method Should Be Used
This method is well-suited for countries with reliable
and consistent estimates of tobacco consumption over
time and with unbiased records of tax-paid sales. The
estimates can be generated relatively quickly.
The best candidates for this methods are countries with
known period of virtually no tax avoidance/evasion
for which data exist. Otherwise, the baseline scope of
tax avoidance/evasion will need to be estimated using
another method and this method will used to estimate
changes in tax avoidance/evasion over time.
Example for Gap Analysis
This method has been applied in many countries, but
the most successful example is the United Kingdom.80
HM Revenue and Customs Department80 in the United
Kingdom has been employing gap analysis to estimate
the size of tax avoidance/evasion and the associated
tax revenue loss for several key commodities, including
tobacco (cigarettes and hand-rolled tobacco) since
2004. The sales data consists of tax paid sales and sales
in duty free stores, and are adjusted for legal cross-
border shopping. The legal cross-border purchases
are estimated based on the International Passenger
Survey (IPS) and commercial data about deliveries
of cigarettes to duty free stores on ferries. The sales
data are further corrected for stockpiling before a
tax increase by using a three-month average. Total
consumption is calculated using estimates of prevalence
of cigarette smoking, the average cigarette consumption
per smoker, and the size of the adult population.
-24-
This consumption is then adjusted for underreporting
by an “uplift factor”. This factor is a ratio of adjusted
sales and estimated consumption in a year which is
believed not to be aected by tax avoidance/evasion
(i.e., the fiscal year 1994–95). Since the factor is
greater than 1, the adjusted consumption is larger
than the consumption estimated based on the survey.
HM Revenue and Customs calculates two estimates
of consumption. The lower-bound estimate assumes
that the level of underreporting has not changed
since 1994/95 and uses the current smoking intensity
as reported by smokers. The upper-bound estimate
assumes an increase in underreporting over time and
uses smoking intensity as reported in 1994/95 even
though there is evidence of declining smoking intensity
since that year. The consumption is further adjusted for
underreporting of smoking prevalence using survey and
lab data on the share of the non-smoking population
that hides the fact that they smoke.
HM Revenue and Customs is transparent about
the weaknesses of the methodology and admits that
the estimates are subject to both random errors
due to sampling employed by the national survey
and systematic errors due to assumptions used to
derive the estimates (e.g., the degree of tobacco use
underreporting). Therefore, all results are presented
with upper and lower bounds, and a calculated
midpoint. The midpoint estimates over time are
interpreted as an indicator of long term trends rather
than a precise estimate of year-to-year changes. The
fiscal year 2012-13 midpoint estimate revealed that
taxes were not paid on 9% of cigarettes consumed by
the UK population (with the associated revenue losses
of £1.1 billion), a continuation of a declining trend
since 2001.55 HM Revenue and Customs continues to
review the methodologies in light of new information
and data, and revises the older estimates accordingly.
2.4 Econometric Modeling
Background and Principles
This method is used to infer tax avoidance/evasion on
the basis of estimated demand functions for cigarettes
using regression analysis and either micro or macro level
data. Demand is usually measured by ocial tax-paid
sales, which is estimated as a function of a set of variables
aecting demand, including variables measuring
incentives for tax avoidance and evasion. These
incentives, typically modelled as a function of price
dierences across jurisdictions, population density near
borders, the extent of cross-border or tourist trac,
Internet penetration, and other factors such as the level
of corruption, are expected to have negative impacts on
tax-paid sales.
Coecient estimates from the resulting models can be
used to predict what tax-paid sales would have been
if the variables reflecting the tax avoidance/evasion
incentives were set to zero, with the dierence between
predicted sales and actual sales measuring the extent
of tax avoidance/evasion. Since statistical estimates
have some margin of error, it is possible to generate
a confidence interval on estimates of the scope of tax
avoidance/evasion.
Advantages and Disadvantages
The method can assess the sensitivity of tax avoidance/
evasion to changes in variables that are hypothesized to
influence it. It can distinguish between tax avoidance
and tax evasion if researchers can find a variable that
impacts one type of low-tax consumption but not
others. For example, the price dierence between a
country average price and the average price for which
the tobacco industry is selling its cigarettes would
measure tax evasion since this dierence is likely not
related to price dierence with a neighboring state,
which would motivate legal cross border shopping (i.e.,
tax avoidance). This method can be inexpensive if it
uses the existing data, but quite expensive if the data
needs to be collected. It requires high quality, preferably
nationally representative data, excellent econometric
skills, a good command of the economic theory, and
creativity in developing the right measures of tax
avoidance/evasion. An experienced econometrician is
needed to handle issues related to statistical issues such
as econometric misspecification, low explanatory power
and omitted-variable and other biases.
When This Method Should Be Used
The method should be used in countries with high
quality data and when experienced researchers with
good econometric skills are available to conduct the
analysis. This could be a time consuming and expensive
method if the data needs to be collected first. Even
when the data exists, their preparation for the analysis
can take time and eort.
Examples for Econometric Modeling
The method has been used to assess the extent of legal
cross-border shopping, direct low-taxed purchases, and
illegal bootlegging in the USA and, to a limited extent, for
global and regional estimates.
Thursby and Thursby (2000)64 developed a model
of commercial smuggling (bootlegging) to estimate
the extent of tax evasion in the US using the state level
annual data from 1972–90 excluding states that were
hypothesized to be the source of commercial smuggling.
Since data on cigarette sales are on tax-paid or legal
sales, only a portion of a state’s cigarette consumption is
-25-
observed when there is commercial smuggling. The tax-
paid sales in the state was estimated as a function of retail
prices, state taxes, the cost associated with acquiring and
camouflaging smuggled cigarettes, as well as enforcement
while controlling for time trend, income, and the incentive
for tax avoidance. The cost of acquiring and camouflaging
smuggled cigarettes was measured by the dierence
between the state tax rate and the tax in North Carolina,
since this tobacco-growing US state was thought to be the
primary source of commercially smuggled cigarettes (i.e.,
tax evasion). The incentive for tax avoidance was captured
by the average retail price of cigarettes in adjacent states,
by the presence of military bases and Native American
reservations in the statev1, and by the ratio of average
tax in adjacent Canadian provinces to neighboring state
tax. The enforcement was measured by existence of
state penalties for tax evasion, the state membership in
a revenue enforcement association, the presence of a
discount or rebate for each legal sale for wholesalers and
by implementation of the US Contraband Cigarette Act
(CCA) in 1978. The time trend was included to account
for secular trend in cigarette use. The study found 3–4%
of all cigarettes sold in the US evaded taxes during the
1970s, and that the tax evasion increased in 1990 to
7.3%. The authors explained this increase by a change
in the balance of enforcement activities between the US
state and US federal authorities after passing the CCA,
which generated a loophole in the tax audit. The results
estimated by the model were compared with the estimate
of cigarette sales in excess of consumption in three US
states (NC, KY, VA) that were the source of commercial
smuggling, and estimates accorded reasonably well.
Merriman et al. (2000)9 used cigarette tax-paid
sales data for 1989–95, cigarette prices and frequency
of international travel from 23 European countries
to estimate the extent of small-scale smuggling
(bootlegging) and cross-border shopping (a
combination of tax avoidance and tax evasion). The
per capita cigarette sales was modeled as a function
of domestic price, income measured by GDP per
capita, the incentives for tax avoidance/evasion and
other variables, such as the degree of corruption in the
country. The incentives for tax avoidance/evasion were
modeled as dierence in price between the home and
destination countries and the total number of cross-
border travelers. The model also included a dummy
variable for each year and a dummy variable for each
country to correct for any factors that are constant over
time but vary by country (such as the cultural heritage
of the country) or are constant across countries but
vary over time (such as the state of knowledge about
how smoking aects health). The country dummies
also controlled for the average level of corruption in
the country that has been associated with the level of
large scale organized wholesale smuggling (tax evasion).
The study found that, in a typical European country,
the share of cigarettes acquired by bootlegging and/
or cross-border shopping accounted for about 3% of
domestic consumption.
Yurekli and Sayginsoy (2010)81 used econometric
modeling to study the extent of global large-scale
organized smuggling in 1999 using per capita legal sales
and trade data from 110 countries in seven regions.
They developed a variable that measured incentives
for tax evasion as a function of smugglers’ expected
profit, which is driven by the price dierences between
legally sold cigarettes and the cigarette world price.
The export prices of the US and UK cigarettes to a
country and its trading partners were used to proxy
the world price. The world price for a pack of US/UK
cigarettes calculated by dividing the value of exports
by the volume of exports ranged from US$0.15 to
US$1.09 per pack, depending on the importing
country. In all countries, the world price was lower
than the average retail price. This dierence and the
lack of anti-smuggling law enforcement eorts were
assumed to motivate tax evasion. Law enforcement
was proxied by the inverse country-specific level of
corruption. A static global demand model estimated
per capita legal cigarette sales (i.e., a measure of
consumption) as a function of the Purchasing Power
Parity (PPP), the PPP adjusted average retail price of a
cigarette pack, per capita income adjusted for PPP, the
smuggling incentives variable, the level of corruption
and additional variables capturing demographic and
geopolitical characteristics of a country. The model
was estimated using OLS (Ordinary Least Square)
with White’s heteroskedasticity-robust standard errors.
The dierence between the consumption estimated
by the model and the consumption predicted when
the smuggling incentive variable value was set to zero
provided an estimate of the scope of tax evasion in each
country. Aggregating these results showed that 3.4% of
global cigarette consumption in 1999 was smuggled,
which resulted in a 7.4% loss of tax revenue. The
method of Yurekli and Sayginsoy81 is suitable only when
data is available for a large number of countries.
2.5 Other Methods
2.5.1 Comparison of tax paid sales with
estimated consumption
This method is a variation of the gap analysis. It
compares the change in legal cigarette sales with the
predicted change in total cigarette consumption (i.e.,
legal and illicit) estimated using changes in cigarette
-26-
prices, in income, and the price and income elasticities
of demand. If the actual change in legal consumption
is dierent than the predicted change, tax avoidance/
evasion could be increasing or decreasing depending on
the direction of the change. For example, if actual legal
cigarette consumption decreases by 5%, and predicted
total consumption decreases by only 2%, this would
imply that an additional 3% of cigarettes (compared to
the previous level of tax avoidance/evasion) are likely
to have escaped paying tax. The simulation should be
done for multiple years to determine any systematic
pattern in the deviation between the predicted and
actual sales.
The method is rather simple, intuitive, replicable and
relatively inexpensive since it relies on secondary data.
However, it requires high-quality time-series data
and estimates of country-specific price and income
elasticities of cigarette demand, with these estimates
taking into account the presence of tax avoidance/
evasion. It cannot distinguish between tax avoidance
and tax evasion. Similar to the gap analysis, it is
primarily used to detect deviations from the trend,
not to estimate the scope of tax avoidance/evasion.
However, the magnitude of tax avoidance/evasion could
be assessed if there was a period with no tax avoidance/
evasion, for which data exists.
Example
Walbeek (2014)82 applied this method in South Africa
in order to investigate the industry claim that there
has been a sharp increase in the illicit market in recent
years. He compared the actual changes in tax-paid
cigarette sale with predicted changes in total cigarette
consumption for the period 1995 – 2012. The changes
in cigarette consumption were predicted using data
on cigarette prices, GDP, and previously published
price and income elasticity estimates. The upper and
lower limits of these elasticities were used to perform a
sensitivity analysis. The changes in the gap between the
sales and predicted consumption revealed a substantial
decrease in tax-paid sales compared to the model
prediction in 2000–2002, which would indicate an
increase in illicit trade. However, cigarette consumption
could have also been influenced by advertising
restrictions in 2000-2001 and comprehensive smoke-
free legislation in 2001. In 2003 – 2009, there was
no evidence that the illicit market has grown. On the
contrary, the model predicted that the illicit market
declined during this period. There was a spike in the
size of the illicit market in 2010 when it grew by about
10.2 percentage points (8.2 points – 12.2 points).
However, the spike was not the start of a trend. In 2011
the illicit market increased only marginally, and in 2012
it decreased by 0.6% points (-1.3 % points to 0.0 %
points). The study concluded that the industry claim of
a substantial increase in the illicit market in 2011 and
2012 is unfounded.
2.5.2 Comparison of actual and projected
tobacco tax revenue
This method is based on a comparison of budgeted and
actual excise tax revenue as reported by tax revenue
authorities over time. It assumes that if tax avoidance/
evasion is structural, tax authorities would have taken
it into account when projecting tobacco tax revenue.
This would make budgeting more dicult and one
would expect to see large deviations between the actual
and projected revenue. The ability of a tax authority to
accurately budget for tobacco tax revenue is measured
by the mean percentage error (MPE) and by the root
mean squared percentage error (RMSPE). The MPE
indicates whether forecasts/budgets are consistently too
high or too low, compared to the actual tax collection.
The RMSPE is a measure of dispersion similar to
standard deviation. A negative or increasingly negative
MPE value is consistent with an increase in tax
avoidance/evasion, assuming that the budget was done
correctly. RMSPE quantifies the magnitude of
the deviation, regardless of whether it is positive
or negative.
In order to judge the overall ability of a tax authority
to forecast revenue, this methods should be applied
to other taxed products whose tax revenue is also
budgeted, but subject to no or less tax avoidance/
evasion (e.g., alcohol). Comparison of MPE and
RMSPE for cigarettes with those of other products
can reveal any systematic dierences between revenue
budgeting for cigarettes and for other products.
For this method to work, the budgeted and actual tax
revenue must be available (i.e., reported by the tax
authorities) and independent of the producers’ pricing
decisions. That is, this method can be used only with a
specific or quasi-specific tax regime. It requires data for
relatively long periods of time (at least five years) and
cannot detect large once-o deviations that could be
caused by a sudden spike in illicit trade. This method
will detect increase/decrease in tax avoidance/evasion,
but it cannot estimate its scope and cannot distinguish
between tax avoidance and tax evasion.
Example
Walbeek (2014)82 investigated whether the alleged
increases in illicit cigarette trade significantly
undermined the South African Treasury’s capability
to accurately predict excise tax revenue. He studied
whether cigarette excise tax revenue had been below
-27-
budget in recent years (2000 – 2012), compared
to previous decades (1910 – 1999), by calculating
MPE and RMSPE for budget revenue deviation for
cigarettes, beer, and spirits. Data on budgeted and
actual excise revenue for beer, spirits and cigarettes
were taken from individual Auditor-General reports
and the Treasury’s Budget Reviews. The study found
that cigarette excise revenues were 0.7% below budget
for 2000 – 2012 on average, compared with 3.0%
below budget for beer and 4.7% below budget for
spirits. Higher predictability of cigarette excise tax
revenue indicated little change in illicit tobacco trade
during this period, contrary to the alleged increase
in illicit cigarette trade. However, the cigarette excise
revenue was not as predictable in 2009 – 2012 as in
the preceding period 2000–2008. The analysis detected
a structural break in 2009 when the actual cigarette
excise revenues were below budget in each of the
four years between 2009 and 2012, suggesting that
over this period the illicit market share has increased.
The shortfall of actual tobacco tax revenues (relative
to budget) peaked at 11.5% in 2010 but improved
in subsequent years. The study concluded that the
industry claim of a substantial increase in the illicit
market since 2010 is unfounded.
2.5.3 Key Informant Interviews
This method consists of obtaining information from
people who likely possess information on the subject
matter (key informants). These individuals may work
at various government agencies dealing with tax
avoidance/evasion (e.g., customs, law enforcement), at
academic institutions, private research companies, or as
public health advocates and investigative journalists with
particular interest in the issue. The information can come
directly from those involved in tax avoidance/evasion
when these individuals surrender relevant information
in the process of legal investigation. In some cases,
smugglers oer information voluntarily to journalists,
academics or government authorities on the condition
of anonymity.
People working in the distribution of tobacco products
are another possible source of information. Wholesalers
and retailers selling legal products might be aware of
their competition selling illicit cigarettes. Even those
selling low-taxed products are sometimes willing to
talk to researchers as long as they do not fear legal
consequences.
Customs and/or police authorities have data on both
legal tax avoidance (importing cigarettes within the
legal limits) and illicit cigarette seizures that could be
used to assess the trend of these activities over time (see
also Analysing Seizures method below).
The method can contribute to the understanding of
the modus operandi of tax avoidance and tax evasion
in a specific country or a region. The key informants
who have been dealing with tax avoidance/evasion
for a long time usually have a thorough and accurate
understanding of the nature and scope of these
activities. They are also familiar with changes over time
and responses to various measures and public policies.
Key informants should come from a variety of
disciplines and settings, which will be helpful in cross
verifying and contrasting the estimates. When making
the selection, assess the key informants’ motivation
for over estimating or underestimating the scope of
tax avoidance and tax evasion. For example, customs
authorities may be motivated to exaggerate the issue
in order to get more resources for their activities. On
the other hand, those advocating for a tax increase
may want to understate the scope of the problem. If
possible and safe, also arrange for direct interviews with
those directly involved in tax avoidance/evasion. The
selection of retailers could be more complicated (and
more expensive) if the results are to be representative
of a larger retail community, but a statistician can help
with a sample selection process and calculating weights
assigned to those who were selected.
Ideally, a standard questionnaire is administered to
all key informants. The questions must be clear so
that the same event/incident is not reported multiple
times under a dierent tax avoidance/evasion category.
Geographical areas and time periods must be clearly
defined to make sure that dierent people provide
estimates of the same events. The questions should
focus on the scope of tax avoidance and tax evasion,
but they may also collect information about the modus
operandi, knowledge of industry involvement, and
law enforcement and other government activities
undertaken to deal with the issue. In addition, the
information about the interviewees such as their
occupation, time on the job, number of people they
supervise, age, gender, and race will be useful in
developing the weight assigned to information from
dierent experts. For example, you will give more
weight to a customs ocer who has been in charge of
investigating cigarette smuggling for the last 10 years
compared to a police ocer just assigned to border
patrol. The rules for weight assigning need to be clearly
documented.
A thorough literature review, on-line searches, and an
analyses of newspaper articles and internal industry
documents can be important components of this
method, particularly when it is dicult to find a variety
of key informants on sensitive topics.
-28-
The method is relatively simple and requires the least
technical and statistical sophistication, except for
retailers’ interviews that aim at generating statistically
representative estimates. The cost of this method is low
relative to other methods. Information can be generated
relatively quickly and provide valuable background
and corroborating information. The method requires
good networking and people skills and the ability to
conduct a productive interview. An intimate knowledge
of local culture is very important when analyzing the
information.
The main drawback of the method is the subjectivity
of the estimates and their possible bias due to the
individual expert’s experience, position, interests and
exposure to the media. Many experts are familiar
with only certain aspects of tax avoidance/evasion and
therefore it might be dicult to get the full picture.
In addition, those influenced by the tobacco industry
have an incentive to report high levels of tax avoidance
and evasion in order to prevent tax increases and/
or adoption of other tobacco control measures. Law
enforcement ocers can be motivated to amplify the
problem in order to secure more resources. Tobacco
control advocates may want to focus on lower estimates
of tax avoidance in their eorts to support public
policies aimed at reduction of tobacco use, and on
the industry role in tax evasion. A comparison of the
estimates across countries is problematic due to cultural
and political dierences.
Given its weaknesses, this method should not be used
in isolation and is best accompanied by an alternative
methodology in order to cross-validate the results. It is
not recommended if tobacco tax avoidance/evasion is a
controversial or sensitive topic, because the objectivity
of information might be questionable.
Examples
There are no good examples of studies that applied this
method. Joossens and Raw (1998)43 used information
from experts working in three dierent organizations
(the European Confederation of Cigarette Retailers,
Her Majesty’s UK Treasury and the Swedish National
Police College) to classify 15 countries in the European
Union as high-smuggling countries (contraband market
share of 10% or more), medium-smuggling countries
(contraband market share between 5% and 10%), and
low-smuggling countries (contraband market share of
less than 5%). However, the authors do not provide
more details about how was this information obtained.
Market research companies such as Euromonitor
International or ERC Group publish annual country
level illicit trade data that rely on information from
trade associations, trade press, and trade interviews, but
the methodology of data collection is not described.
2.5.4 Monitoring tobacco trade
The method estimates the extent of large-scale tax
evasion activities by monitoring the dierence between
countries’ mirror records (pair-wise records of trade
partners) on imports and exports. It is based on the
hypothesis that the dierence between recorded exports
of an exporting country and recorded imports of the
receiving country is likely to reflect the amount of
products diverted to illegal markets while in transit.
The destination country of the diverted products will
remain unknown.
The method relies on data published either by the
United Nations Statistics Division or by the World
Trade Organization (WTO) and is very sensitive to data
quality. This quality reflects the capacity of national
agencies to generate these statistics, and this capacity
(and data quality) might be positively related to
countries’ income level.
Using this method is complicated by the existence of
dierent trade classification systems and their changes
over time. Some countries report export/import in
monetary values, and others in volumes. Volume data
are preferred as they are not subject to changes in
currency exchange. Even the volume statistics may pose
diculty if reporting switches from weight (e.g., kg)
to the number of cigarette sticks and the weight of one
stick is unknown. Furthermore, recorded trade data
do not always match correctly within a given month
or year. For example, if a cargo is recorded as exports
in November or December, it may not be recorded as
imports until January or February of the following year.
Given the intrinsic weaknesses of this method, it should
not be used for estimating the scope of tax evasion on a
country level. It cannot detect small-scale tax evasion,
tax avoidance, domestic manufacturing of illegal
cigarettes, diversion of cigarettes to a third country,
or counterfeit cigarettes. It has been used to generate
global estimates of large-scale cigarette smuggling,
but studies have pointed out many weaknesses in this
method. The method is useful for identifying the source
of illicit cigarettes and hubs from which illicit cigarettes
are being distributed.
-29-
Examples
Merriman et al. (2000)9 studied the trend in
aggregated global cigarette export and import from
1975 till 1996. They found that recorded cigarette
exports grew about five-fold while recorded imports
grew only slightly more than four-fold during this time.
In 1996, recorded exports exceeded recorded imports
by about 400 billion cigarettes, suggesting that perhaps
one-third of all recorded exports were not recorded as
imports by the trade partner. This number of cigarettes
represented about 6% of global cigarette consumption.
The authors recommended viewing this estimate
with caution, since large discrepancies between total
reported imports and exports exist for many products,
not only for cigarettes.
Yurekli and Sayginsoy (2010)81 reported that 5383
billion cigarettes were smoked globally and 832 billion
cigarettes, or about 15.5% of global consumption, were
exported in 1999. Only 661 billion cigarettes (about
79% of global exports) were recorded as imports
with no import records for the remaining 171 billion
cigarettes. This dierence was equal to about 3.2% of
global cigarette consumption. As in Merriman et al.9
(2000), the authors pointed out that such discrepancies
exist for many globally traded commodities, and that
this dierence does not necessarily indicate the level of
worldwide smuggling because of dierent export and
import coding systems across countries.
2.5.5 Analyzing Seizures of Illegally
Transported Tobacco
This methods measures only tax evasion and it is based
on local customs and/or police authorities’ reports on
illicit cigarette seizures. Globally, the World Custom
Organization (WCO) provides annual data on tobacco
seizures from its Customs Enforcement Network
(CEN). The authorities may know the likelihood that
illegal cargo is intercepted and observed changes in the
rate of illicit cigarette seizures could indicate changes in
the scope of tax evasion, other things being equal.
For example, if seizures of illicit cigarettes doubled
with little or no change in the level of enforcement,
one might conclude that the level of tax evasion
also doubled.
However, using seizure data to assess the scope of tax
evasion is problematic. First, the information may not
be complete or easily available, and it could be dicult
to establish its accuracy. For example, the submission
of information to the WCO CEN database is not
mandatory and the WCO reports warn that the CEN
database does not permit the assessment of tax evasion.
Second, the large seizures may not be representative
of the illicit market as a whole. Third, the amount of
seizure depends heavily on the level of enforcement. If,
for example, the budget of law enforcement authorities
increases, seizures may increase as well without any
change in the scope of tax evasion. Nevertheless, the
amount of seized cigarettes provides the lower bound
of the scope of tax evasion. Studying the seizures can
provide information about the composition of the
illicit market by analyzing, for example, the share of
counterfeits among all seized cigarettes. It is important
that the counterfeit products are determined by an
independent expert.
There are no examples of studies published in the peer
review literature that applied this method to estimate
the scope of tax evasion. In Europe, Joossens and Raw
(2008)10 analyzed seizures to study the changes in illicit
market. They found that the amount of cigarettes seized
in Europe was negatively related to the Memorandum
of Understanding between governments and various
tobacco companies, to the strength of various anti-
smuggling measures, and to legal actions brought
against the tobacco industry. They concluded that the
size of the illicit market is to large part controlled by the
tobacco industry.
2.6 Summary and Recommendations
for Estimating the Scope of Tax
Avoidance and/or Evasion
Various methods of estimating the scope of tax
avoidance/evasion exist. No single methodology will
produce a definitive estimate since all of them have
advantages and disadvantages. For example, some
methods will capture a mix of tax avoidance and tax
evasion without being able to distinguish between
them, others will not be able to separate the impact
that tourism and/or commuting patterns have on the
estimates. Since the weakness of a particular approach
can be exacerbated by specific market conditions, it is
important to use specific local knowledge and creativity
when applying these methods.
Given the complexity of tobacco tax avoidance and
evasion, the multiple ways to engage in them, and the
methods’ limitations, it is important to triangulate
the estimates of the scope of the problem. Generating
multiple estimates using dierent methods will
cross-validate results and minimize methodological
objections.
The results obtained from multiple methods should
be carefully compared taking into account the fact
that dierent methods could measure dierent
-30-
phenomena. For example, estimates generated by
methods that cannot separate tax avoidance from tax
evasion should be at least as large as (or larger than)
estimates generated by methods that capture only tax
avoidance. The dierence in estimates between these
two methods could indicate the size of tax evasion. It
is extremely important to carefully document all steps
when conducting research so that studies can be peer-
reviewed and their result can be replicated.
Many studies apply the same method over time in
order to capture changes in the scope of tax avoidance/
evasion rather than generate a point estimate of its
scope. Such an approach is useful for evaluating the
impact of policies and other factors with a possible
impact on tax avoidance/evasion. Measuring the
change rather than the scope also addresses some
methodological weaknesses of the methods, even
though this approach may not generate the estimate of
the size of the problem. Repeating the same method
over time when no changes relevant for the scope of tax
avoidance/evasion occurred can be another useful way
to cross validate the results.
This Methodological Guideline presents the most
recent, commonly used methods to quantify tax
avoidance/evasion, but there might be other methods
available and new approaches can be invented
taking advantage of new technologies and advanced
techniques. Constantly changing market conditions will
present opportunities for creative researchers to develop
and test new methods. For example, new tracking and
tracing systems employing an online coding system
will allow researchers to use mobile audit devices to
distinguish between products that avoid or evade taxes.
Designing studies around the distribution network is
another possibility.
Table 2 summarizes all methods measuring the scope of
tax avoidance and tax evasion described in this chapter.
-31-
Table 2: Over view of Methods that Measure the Scope of Tax Avoidance and Evasion
principles advantage disadvantage when to use example
Survey of
tobacco users
Collecting self-reported
data on packs’ features
and their sources
from a statistically
representative sample of
the population
Direct method of
estimating the scope of
tax avoidance/evasion
and availability of low-
tax products
Underestimates tax evasion;
problems of validity;
potential bias due to social
stigma and underreporting
Description of packs
features sucient to
determine tax avoidance/
evasion; sucient budget
for a representative sample
Guindon et al.
(2014); Stoklosa
and Ross, 2014;
Nagelhout et al.
(2014)
Exam of
cigarette
packs
obtained
from smokers
Collecting packs
from a statistically
representative sample
of smokers during an
interview or by mail
Direct and objective
method of estimating
the scope of tax
avoidance/evasion
Tax evasion cannot be
detected without self-
reported info from smokers
and/or lab inspection;
possible selection bias
Packs features allow to
determine tax avoidance/
evasion by visual inspection;
sucient budget for a
representative sample
Joossens et al.
(2014); Fix et al
(2014); Stoklosa
and Ross, 2014
Exam of
discarded
cigarette
packs
Collecting a random
sample of littered
cigarette packs from
streets or from garbage
Direct and objective
method of estimating
the scope of tax evasion;
can be less expensive
than surveys
Cannot distinguish tax
avoidance from tax evasion;
estimates relevant only for
narrow geographical areas;
dicult to account for
tourists/commuters
Packs features allow to
determine tax avoidance/
evasion by visual inspection
Merriman
(2010); Chernick
and Merriman
(2013)
Exam of
cigarette
packs
obtained
from retail
Collecting packs from a
random sample of retail
outlets
Direct and objective
method of estimating
the availability of illicit
products via legal
channels
Cannot estimate the scope
of tax evasion; cannot
detect tax avoidance; lab
inspection needed to detect
counterfeits
Packs features allow for
determining tax evasion by
visual packs’ inspection;
sucient budget for a
representative sample
Scollo et al.
(2014)
Compare
sales with
consumption
(gap analysis)
Subtracting tax-paid
sales from consumption
estimated from surveys
Transparent, replicable,
and relatively low
cost method that
uses secondary data;
estimates can be
generated relatively
quickly
Lack of reliable survey data;
consumer underreporting,
tourist purchases and RYO
cigarettes can bias the
results; better at estimating
the change rather than the
scope
Reliable and consistently
collected tobacco use data
exist over long period of
time; sales data are available
for the same time period
HM Revenue and
Customs (2011)
Econometric
Modeling
Estimating the demand
for tobacco products as
a function of incentives
for tax avoidance/
evasion using regression
analysis
Can detect various
types of tax avoidance/
evasion; can model
impact of policies
Sensitive to data quality;
technically demanding
High quality data and
an econometrician
are available
Thursby and
Thursby (2000);
Merriman et al.,
2000; Yurekli and
Sayginsoy, 2010
Comparison
of tax paid
sales with
estimated
consumption
Compares trend in tax
paid sales with trend
in total consumption
predicted using changes
in prices, income, and
known price/income
elasticities of demand
Simple and intuitive
method
Cannot distinguish tax
avoidance from tax evasion;
better at estimating the
change rather than the
scope
High quality data and
an econometrician are
available; estimates of
price/income elasticities of
demand are available
Walbeek (2014)
Comparison
of actual and
projected
tobacco tax
revenue
Comparison of
budgeted and actual
excise tax revenue for a
long period of time
Simple and intuitive
method; can detect
changes in tax
avoidance/evasion
Cannot estimate the scope
of tax avoidance/evasion;
cannot distinguish tax
avoidance from tax evasion;
cannot detect one time
deviation from a trend
Tax revenue prediction and
actual revenues for various
products is available over
time; country uses only
a specific tax; authorities
consider tax avoidance/
evasion when generating tax
revenue estimates
Walbeek (2014)
Key
informant
interviews
Systematic collection
of information from
experts
Little technical skills
required; low costs; rela-
tively quick assessment
of the situation
Subjectivity of the estimates;
may generate bias results
Low budget; low technical
skills; information needed
quickly
Joossens and Raw
(1998)
Monitoring
tobacco trade
Monitoring the
dierence between
countries’ mirror
records on imports and
exports
Can detect smuggling
hubs
Cannot estimate the scope
of tax avoidance/evasion
for individual countries;
captures only large-scale tax
evasion
Global estimate of a trend
in large-scale tax evasion is
needed
Merriman et al.
(2000); Yurekli
and Sayginsoy
(2010)
Analyzing
seizures
of illegally
transported
tobacco
Obtaining data on
cigarette seizure during
certain time period for
the whole country
Can generate the
minimum scope of tax
evasion. Can inform on
the composition of the
illicit market
Underestimate the scope
of tax evasion; sensitive to
enforcement eort
Complete data on seizure
are publically available
Joossens and Raw
(2008)
-32-
The agendas of those who fund and/or conduct
research on tobacco tax avoidance/evasion may have
an influence on the methodology, presentation, and
interpretation of the results. The tobacco industry
may be interested in exaggerating the extent of tax
avoidance/evasion in order to oppose tobacco tax
increases and other tobacco control policies such
as health warning labels or plain packaging.83 Law
enforcement agencies and policymakers may want to
minimize the issue as this may indicate problems of
eciency or corruption. Alternatively, law enforcement
agencies, the World Customs Organization,
departments responsible for tax collection, and
companies selling tracking and tracing technology
(e.g., SICPA) could be interested in highlighting
the issue in order to secure more resources for their
activities. Tobacco control activists may either prefer
higher estimates (e.g., to point out the manufacturers’
role in tax evasion) or lower estimates (e.g., to minimize
concern about the unintended eects of tobacco
control measures) depending on the issue at stake.
Given the inherent diculties estimating the scope of
tax avoidance/evasion, and the motivation of various
stakeholders to either overestimate or underestimate
the size of the problem, it is very important to assess the
quality of the estimates in such studies.
This chapter will first lay out criteria for assessing the
quality of various studies (Table 3) and then provide
examples of studies that in general meet, partially meet,
or do not meet those criteria. The criteria do not have
equal weight when determining the overall quality of a
study, and not all of them are relevant for all methods.
We ordered them somewhat arbitrarily, but tried to
follow the logical structure of a typical research article.
Therefore, the criteria do not need to be applied in
the order presented in the table. Using various study
examples, we will demonstrate how these criteria can
be used to critique and interpret results generated by
various research eorts.
The reminder of the section will review eight studies in
the light of the criteria described in Table 3.
CHAPTER 3
Assessing the Quality of the Estimates
-33-
Table 3
Criteria for Assessing the Quality of Estimates
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed Published in a peer-reviewed journal; and/or
explicitly refers to a peer-review process; and/
or it is an ocial document of a reputable
international or government organization.
No reference to a peer-review process; and/or
specific terms under which study was prepared
are not disclosed; and/or disclaimer about using
the results at your own risk.
2. Funding Funding acknowledged. Funding not disclosed or acknowledged.
Funding entity has no potential conflict of
interest with respect to the subject of the study.
Funding entity has a potential conflict of
interest with respect to the subject of the study.
3. Grounded in theory Study distinguishes between various types of
tax avoidance/evasion, and clarifies which types
are the subject of the study.
Study doesn’t distinguish between tax
avoidance and tax evasion; it is not clear which
type of avoidance/evasion is being measured.
Takes into account any relevant factors that
could influence the scope of tax avoidance/
evasion.
Fails to account for factors that could influence
the scope of tax avoidance/evasion.
4. Transparency and replicability Methods and data are adequately described
so that the results can be replicated; data is
publicly available or can be made available
upon request.
Methods and data are not adequately
described; the results cannot be replicated
using the information provided in the study;
data is not publicly available.
Assumptions are clearly stated.
Assumptions are not stated or not stated clearly.
5. Generalizability of results Sample size and sampling design are well
described and allow for generalization of results
to the entire country/region/population.
Sample size and sampling design are not
adequately described; sample size is too small
to allow for generalization of results.
The sample is selected objectively. The sample selection is biased.
Sample attrition and non-response is described
and taken into account; there is an attempt to
establish the representativeness of the sample.
Sample suers from large attrition and/
or high non-response rate and there is no
attempt to correct for this or to establish the
representativeness of the sample.
6. Objective criteria preferred over
subjective criteria
Low-tax purchases are identified based on a set
of objective criteria such as place of purchase,
product price, etc.
Low-tax purchases are identified by
respondents’ self- report.
Self-reported low-tax purchases are cross-
verified using objective criteria.
There is no attempt to cross-verify the self-
reported information using objective criteria.
7. Measurements are defined correctly Survey questionnaire distinguishes between
dierent tax avoidance/evasion categories.
Survey questionnaire doesn’t clearly distinguish
between dierent tax avoidance/evasion
categories; categories may overlap and the same
event might be counted multiple times.
Conversion of cigarette sticks to/from weight
measure is transparent and based on a well-
established conversion factor.
Conversion of cigarette sticks to/from weight
measure is not transparent or is not justified.
8. Identification of counterfeit products Identification of counterfeit products is
performed by an independent researcher or lab.
Identification of counterfeit products is
performed by a party with a vested interest in in
the results.
9. Presentation of results Estimates are presented as a range or with
confidence intervals that account for the
statistical properties of the sample and/or
various assumptions used in generating the
estimate.
Results are not presented as a range or with
confidence intervals. Results are not robust
with respect to assumptions made.
The size of the illicit market is expressed as a
share of the total market.
The size of the illicit market is expressed as
a share of the licit market. This makes the
problem look bigger.
10. Cross-validates a point estimate using
multiple methods or measures change
over time using the same method
Uses multiple methods and/or corroborating
information to cross-verify the estimates.
Estimates the scope of tax avoidance/evasion
at one point in time without using multiple
methods to cross-verify the results.
Estimates changes in tax avoidance/evasion over
time using the same method.
Corroborating evidence used to cross-verify
results cannot be trusted based on criteria
presented in this table.
11. Acknowledgement of methodological
weaknesses
Points to possible weaknesses of the applied
methodology/data and assesses the implication
of these shortcomings for the estimates.
Weakness of the applied methodology/data are
not acknowledged/discussed.
-34-
Example 1
The first study is from France where Lakhdar
(2008)84 applied three approaches to assess the size
of legal cross border shopping (tax avoidance) and
illegal cigarette smuggling (tax evasion) after a series
of tobacco tax increases that led to a 44.7% increase
in cigarette prices from 2002 to 2004. First, he used a
simulation model to predict cigarette sales in France
over time (1999 – 2006) as if all regions experienced
the same decline in consumption as reported by
the region with the lowest decline in cigarette sales.
The assumption was that the region with the lowest
decline has not been aected by smuggling or cross-
border shopping, because it experience such a small
drop in sales. The predicted sales were then compared
with the actual sale and the gap was attributed to tax
avoidance/evasion. Sales data came from the tobacco
industry because it possesses regional sales data.
Second, he employed gap analysis and compared
the ocial sale of manufactured and hand-rolled
cigarettes with the estimates derived from national
surveys that captured the consumption of the same
products in one year (2005). The comparison was
done on a regional level in order to assess the impact
of the proximity to a border with a country with
lower priced cigarettes. Third, the study collected all
cigarette packs properly disposed of and processed
in one waste management plant in a Paris suburb
at two dierent points in time in order to identify
the countries of origin of foreign tobacco entering
France based on the brand, the language, the heath
warning messages (if any), or other features. The
author pointed out the lack of representativeness of
this sample and considered this part of the study to
be exploratory.
The first method estimated that cigarettes equivalent
to 14 – 17% of legal sales in the period of 2004 –
2006 did not pay taxes in France, while the second
method estimated that cigarettes equivalent to 20%
of legal sale were not paying taxes in France in 2005.
The collection of cigarette packs at a waste collection
centre showed that foreign cigarettes accounted for
18.6% of the sample in 2005 and 15.5% in 2006.
All three methods resulted in a very similar estimate
of the size of the illicit market.
Based on the first method, the study concluded that
the substantial tobacco tax increases in 2003 and
2004 led to an increase in cross-border shopping (tax
avoidance) and cigarette smuggling (tax evasion), but
the study could not distinguish between them. Since
the majority of tax avoidance/evasion occurred near
the borders, the study speculated that the problem is
primarily related to cross-border purchases (which
can be both legal and illegal). The study discussed
alternative approaches to measuring tax avoidance/
evasion and their implication for the estimates. The
authors admitted that they did not consider distance
to the low-price border as a factor motivating cross-
border shopping and acknowledged the weakness of
the survey-based consumption estimate that most
likely suered from underreporting. The estimates
were presented as a range and the study was peer
reviewed with no competing interests declared.
Table 4 summarizes how Lakhdar (2008)84 fits the
criteria for assessing the quality of the estimates.
The main strength is the use of multiple methods
to assess the scope of tax avoidance/evasion. The
study has some weaknesses, but most of them
are acknowledged. Therefore, the study can be
categorized as well-executed, and its results can be
trusted with the caveats highlighted by the author and
in Table 4.
-35-
Example 1.Table 4
Assessing Lakhdar (2008)84
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed Published in a peer-reviewed journal.
2. Funding No competing interests were declared;
The author’s institution has no potential
conflict of interest with respect to the subject
of the study.
No funding acknowledged in the
Acknowledgement, but most likely funded by
the author’s institution.
3. Grounded in theory Study acknowledges that it measures a
combination of tax avoidance (cross-border
shopping) and tax evasion (illegal smuggling).
The distance to the state border is not
taken into account, but this weakness is
acknowledged.
The impact of only one factor (tax increase) is
considered.
4. Transparency and replicability Methods and data are adequately described;
some data is publically available; results could
be replicated.
Assumptions are clearly stated.
It is not clear if the industry data is publically
available.
5. Generalizability of results Two of the three methods allow for
generalization of results to the entire country.
The sample for two of the three methods is the
entire country.
Sample size/sampling design of the trash
method does not allow for generalization of
results, but this is acknowledged.
The sample selection of the trash method is
biased, but this is acknowledged.
6. Measurements are defined correctly The manufactured and hand-rolled cigarettes
consumption is compared to the manufactured
and hand-rolled cigarettes sale.
Packs’ characteristics do not distinguish
between tax avoidance/evasion but this is
acknowledged.
7. Presentation of results Estimates are presented as a range. The size of illicit market is expressed as a share
of licit market.
8. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Uses multiple methods and/or corroborating
information to cross-verify the estimates.
9. Acknowledgement of methodological
weaknesses
Points to possible weaknesses of the applied
methodology/data and assesses the implication
of these shortcomings for the estimates.
Example 2
Another study that used dierent multiple methods
to cross-validate the results was Stoklosa and Ross
(2014).72 In 2011, they conducted a household survey
and collected discarded packs to estimate the size of the
illicit cigarette market in Warsaw, Poland. The goal was to
identify packs that were not destined for the Polish market
by inspecting excise tax stamps and health warnings.
The household survey used a quota sampling method
taking into account the size, the gender and age
composition of Warsaw’s population. During the survey
400 smokers were asked to show all open cigarette
packs in their possession. This was dierent from some
previous surveys that only asked for an open pack. The
interviewers found that in some cases, a smoker would
have two packs open at the same time: one low-tax
pack for their own use, and one full-tax pack if a visitor
stopped by. The pack data obtained during the survey
were weighted by the self-reported amount of monthly
cigarette consumption.
The collection of discarded packs followed the
methodology developed by Merriman73 by randomly
selecting 30 out of 783 voting districts in Warsaw to
systematically collect 754 discarded cigarette packs.
All observed tax stamps were compared with ocial tax
stamps provided by the Ministry of Finance. Packs with
a tax stamp issued by another country or without the
Polish health warning were classified as packs not taxed in
Poland. To account for the possibility that the tax stamp
was removed in the process of opening the pack, packs
with a missing or damaged tax stamp, but with a health
warning in Polish, were counted as full-tax cigarettes.
This could underestimate the share of illicit cigarettes if
some of the packs with Polish health warnings were not
taxed. The data from packs collected on streets could
not distinguish between tax avoidance and tax evasion.
The packs observed during the household survey could
have been supplemented by self-reporting data on the
packs’ provenance (e.g., brought to Poland legally by
-36-
travellers), which would allow estimation of the scope of
tax avoidance and tax evasion separately, but this analysis
was not performed.
The study found that 14.6% and 15.6% of cigarette
packs were not intended for the Polish market, using the
survey and the discarded pack method, respectively. The
test of independence comparing the results based on the
two methods determined that they are not statistically
dierent. Since the two dierent methods agreed on the
share of non-domestic cigarettes on the market, it seems
unlikely that the study’s disclosed weakness related to its
relatively small sample size would have biased the results.
Table 5 demonstrates how various criteria for assessing
the quality of the estimates were applied to Stoklosa
and Ross72 (2014). The main strength of the study is the
use of multiple methods to cross-validate the results.
The main weaknesses is the small sample size and the
failure to separate tax avoidance from tax evasion using
the self-reported survey data. The small sample size is
acknowledged as a study weakness. Overall, the study
can be categorized as well-executed and its results can be
trusted, given that two dierent methods resulted in the
same estimate.
Example 2. Table 5
Assessing Stoklosa and Ross (2014)72
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed Published in a peer-reviewed journal.
2. Funding Funding acknowledged.
Funding entity has no potential conflict of interest with
respect to the subject of the study.
3. Grounded in theory Study acknowledges that it measures a combination of
tax avoidance (cross-border shopping) and tax evasion
(illegal smuggling).
The impact of other factors is controlled by the limited
time span of the study period.
Study failed to separate tax avoidance
(cross-border shopping) from tax
evasion (illegal smuggling) using the
survey data.
4. Transparency and replicability Methods and data are adequately described so that the
results can be replicated; data is publically available or
can be made available upon request.
Assumptions are clearly stated.
5. Generalizability of results Sample size and sampling design is well-described.
The sample is selected objectively using a quota
sampling method. Quotas for each district reflected the
size of the district’s population, and the gender and age
distribution of the Polish population were taken into
account.
Sample size is too small to allow for
generalization of results, but this
shortcoming is acknowledged.
The limitation of the quota sampling
methods is not acknowledged.
6. Objective criteria preferred over
subjective criteria Low-tax purchases are identified based on a set of
objective criteria: the presence of a tax stamp or an
appropriate health warning.
7. Measurements are defined correctly Survey questionnaire distinguishes between dierent
tax avoidance/evasion categories. The information that would allow the
distinction between tax avoidance and
tax evasion is not used.
8. Identification of counterfeit products The identification of counterfeit
cigarettes is not performed due to
budget constraints.
9. Presentation of results Estimates are presented with confidence intervals.
The size of illicit market is expressed as a share
of total market.
10. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Uses multiple methods to cross-verify the estimates.
11. Acknowledgement of methodological
weaknesses Points to possible weaknesses of the applied
methodology/data and assesses the implication
of these shortcomings for the estimates.
-37-
Example 3
Blecher E (2010)85 employed gap analysis to estimate
the size of the illicit cigarette market in South Africa
between 1997 and 2007. First, he calculated the tax-
paid sales by dividing the total excise tax revenue by
the specific excise tax. Then, he estimated the cigarette
consumption using adult smoking prevalence (from
a national representative survey), ocial population
estimates and average smoking intensity. Since the
national survey did not collect data on smoking
intensity for the period of 1997 – 2001, the author
calculated it using non-linear decay function and the
assumption that the illicit market did not exist in 1997
(i.e., dividing the tax-paid sales by number of smokers
in 1997). The assumption of 0%, 5% and 10% smoking
intensity underreporting allowed the author to test the
sensitivity of the estimates. The study found that the
size of the illicit market grew from 1997 until 2000
when it reached 9.4% – 11.5% of the total market.
Between 2000 and 2007, the share of the illicit market
was stable. In 2007, the share of illicit cigarettes was
7.0% – 11.2% of the total market.
As suggested by Table 6, Blecher E (2010)85 meets
most of the criteria for high quality estimates. The
most valuable features of the study are the focus on the
changes in tax avoidance/evasion over time using the
same established methodology, and finding a solution
for missing smoking intensity data. The study did not
distinguish between various types of tax avoidance/
evasion, which is typical for a macro-level analysis,
and did not take into account the smaller size of the
market when calculating the revenue loss. Both of these
shortcomings are acknowledged.
Example 3. Table 6
Assessing Blecher E (2010)85
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed Published in a peer-reviewed journal.
2. Funding The author’s institution has no potential
conflict of interest with respect to the
subject of the study.
No funding acknowledged in the
Acknowledgement, but most likely funded by
the author’s institution.
3. Grounded in theory Discusses factors that could have
influenced the scope of tax avoidance/
evasion.
Study doesn’t distinguish between tax
avoidance and tax evasion due to data
limitations and due to the specific nature of the
illicit cigarette market supply in South Africa.
4. Transparency and replicability Methods and data are adequately described
so that the results can be replicated; data is
publically available or can be made available
upon request.
Assumptions are clearly stated.
5. Generalizability of results The use of macro-level data and nationally
representative surveys allow for generalization
of results to the entire country.
6. Measurements are defined correctly Measures of tax-paid sales and cigarette
consumption are clearly defined.
7. Presentation of results Estimates are presented as a range or with
confidence intervals that account for various
assumptions used in generating the estimate.
The size of the illicit market is expressed as a
share of total market.
8. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Estimates changes in tax avoidance/evasion over
time using the same method.
9. Acknowledgement of methodological
weaknesses Points to possible weaknesses of the applied
methodology/data and assesses the implication
of these shortcomings for the estimates.
-38-
Example 4
Data quality is a primary concern in low- and middle-
income countries. Some of this quality deficiency can
be overcome by applying appropriate and/or innovative
methods.
Pavananunt (2011)86 used two methods to estimate
the size of the illicit cigarette market in Thailand for
six selected years from 1991 to 2006. The first was the
gap method that compared estimates of consumption
with ocial sales. Cigarettes sales were estimated
using tax data and 1% of the estimated amount was
subtracted to account for damage or product loss.
There is no information regarding the source of the
tax data and no justification is provided for the 1%
adjustment. Consumption of cigarettes was estimated
by combining data from three dierent surveys (two
dierent surveys for prevalence and one additional
survey for average number of cigarettes consumed by
a smoker). Even though it was not discussed in the
paper, there is a high probability that each survey used
a dierent sampling frame and/or dierent questions
to collect data. This makes it dicult to compare
trends over time. The author subtracted an estimate
of consumption of hand-rolled cigarettes from total
cigarette consumption since hand–rolled cigarettes
were not included in the calculation of the sales data.
However, the estimation of hand-rolled cigarette
consumption required making numerous assumptions,
which introduced additional noise into the time series
data. The impact of these assumptions on the estimates
are not discussed. The author also did not clarify
whether the estimates of hand-rolled cigarettes shares
came from the same surveys that were used to estimate
cigarette consumption. The estimated consumption was
considerably lower than legal cigarette sales, fluctuating
from 29% to 54% of legal sales depending on the year
of the survey. Such results would suggest illegal export
of cigarettes taxed in Thailand to other countries. This
is highly unlikely given that cigarette prices in Thailand
are higher compared to its neighboring countries.
The author attributed the gap between the tax-paid
sales and consumption to possible under-reporting of
cigarette use in the surveys and to the surveys’ designs
that did not capture consumption of migrant workers
and visitors to Thailand. The author concluded that
comparing sales with consumption is not a feasible
approach to estimate the scope of tax avoidance/evasion
in Thailand due to the shortcomings of the survey data.
Therefore, the author focused on the second method
that was based on discrepancies between export volume
from countries exporting cigarettes to Thailand (UN
Comtrade) and ocial data on imports from the same
countries (Customs Department of Thailand ) during
1991-2006.
On the country level, this method is primarily useful
for identifying the source of illicit cigarettes, not
for estimating the scope of tax evasion. In addition,
using the trade discrepancy method is complicated
by many factors, including the dierences in trade
classification systems and their changes over time. In
this case, using trade data from two dierent sources
introduced additional noise into the data and increased
the likelihood that the discrepancy between recorded
exports and imports does not accurately reflect the level
of illicit trade. Nevertheless, the author adjusted these
discrepancies using a three-year moving average to
account for dierent time lags in data recording across
countries, but no explanation was given for why three
years was considered the most appropriate adjustment
period. The results revealed huge dierences between
Thailand and their trade partners’ records, ranging
from 83% to 15% of the partner’s export to Thailand.
The missing import volume was assumed to be all
consumed in Thailand even though there is no evidence
to support this assumption. Based on this calculation,
the author reported that the level of smuggling in
Thailand rose from 3% in the early 1990s to a peak of
17% in 1998, then declined to 7% by 2004 and rose to
10% in 2005. The study further attempted to compare
the trend in illicit cigarette trade obtained from trade
discrepancies with excise tax rates over time. Even
though it states that there was an increase in both the
excise tax rate and illicit trade during the study period,
the study concludes that there was no relationship
between the size of the illicit cigarette market and the
excise tax rate.
Table 7 summarizes the strengths and the weaknesses
of Pavananunt (2011)86 and shows that this study only
partially meets the criteria of a well-executed study.
The main contribution of the study was that it was the
first systematic attempt to measure the size of the illicit
market in Thailand. However, the study struggled with
data quality and selection of the appropriate method.
It acknowledged its shortcomings and called for more
research to quantify the scope of tax avoidance/evasion
in Thailand.
-39-
Example 4. Table 7
Assessing Pavananunt (2011)86
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed Published in a peer-reviewed journal.
2. Funding Funding acknowledged.
Funding entity has no potential conflict of interest
with respect to the subject of the study.
3. Grounded in theory Takes into account relevant factors that could
influence the scope of tax avoidance/evasion. Study doesn’t distinguish between tax
avoidance and tax evasion and confuses
legal tax avoidance with illegal tax evasion.
4. Transparency and replicability Data is publically available or can be made available
upon request.
Assumptions are clearly stated.
The methodology and data are not fully
described.
Some assumptions are not justified.
5. Generalizability of results The use of macro-level data and nationally
representative surveys allow for generalization of
results to the entire country.
6. Measurements are defined correctly Measures of tax-paid sales and cigarette
consumption are clearly defined.
7. Presentation of results The size of the illicit market is expressed as a share
of total market. Results are not presented as a range or
with confidence intervals. Results are not
robust with respect to assumptions made.
8. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Estimates changes in tax avoidance/evasion over time
using the same method. No cross-validation of results since one
method is deemed unreliable.
9. Acknowledgement of methodological
weaknesses Points to possible weaknesses of the applied
methodology/data. The impact of weaknesses on the
estimates is not assessed.
Example 5
In Australia, three major tobacco companies
commissioned a series of reports on the illicit tobacco
market in Australia from major global consulting
groups.87-92 All these reports suered from a lack of
transparency and academic rigor93-96 while relying
survey data collected by Roy Morgan Research
(RMR) and on empty pack surveys conducted by
Klynveld Peat Marwick Goerdeler (KPMG). The
survey methodologies employed by RMR and by
KPMG were not adequately described and their data
conflicted with the results of much larger nationally
representative surveys. The major problem with the
RMR survey was that it asked respondents if they
purchased any contraband or any counterfeit cigarettes.
People may not be always aware of the fact that they
have purchased an illegal product unless these products
can be clearly identified, which is not always the case.
It is likely that the RMR questionnaire recorded the
same purchase multiple times due to overlapping
questions (e.g., counterfeited cigarettes were reported
both as contraband cigarettes and also separately as
counterfeited cigarettes), which could have resulted
in double or triple counting of illicit purchases. Based
on the average amount of illicit tobacco purchased,
those who admitted purchasing illicit tobacco would
have to use it almost exclusively. This conflicted with
the results of a national survey that showed that the
vast majority of smokers who used illicit tobacco
products use them only occasionally. The Deloitte
studies assumed that an illicit cigarette stick weighed
1.0 gram, but the assumption applied for the licit
cigarettes was not disclosed. Most studies assume
that the weight of a cigarette stick is 0.7 gram, the
Australian government uses 0.8 gram for the weight
of a legal stick, and the follow up study produced by
KPMG91 assumed that between 0.6 and 0.75 grams
of RYO tobacco is used in each cigarette. If there was
a dierent weight applied for licit and illicit cigarettes,
and this dierence was not justified, the scope of the
illicit tobacco market could have been overestimated.
The 2011 report claimed that illicit tobacco products
represented 15.9% of the licit market in 2010, instead
of reporting the estimate as a share of the total market,
which would have been 13.7%. Reporting the estimate
as a share of the licit market generates a larger, more
dramatic figure. In that year, a tobacco control NGO,
Quit Victoria, provided an alternative estimate based
on publically available data: about 2–3% of the total
market was estimated to be illicit.93 The subsequent
reports89-91 provided estimates of a trend. The estimated
share of illicit cigarettes in the legal market dropped
from 15.9% in 2010 to 13.4% in 2011, to 10.5% in
2012, and increased to 14.2% in 2013. This would
have been a drop from 13.7% in 2010 to 9.5% in 2012,
and an increase to 12.4% in 2013 if the estimates were
expressed in term of the share of the total market.
Even though these estimates may have suered from
the same weaknesses, to the extent that the same
methodology was used, the estimate of a trend should
be valid. This means that the share of the illicit market
-40-
declined between 2010 and 2013, a result that was not
highlighted in any of these studies. Aware of the criticism
of the methodologically weak RMR survey, KPMG
(2013b)91 added a discarded pack collection survey.
However, the representativeness of their samples is
questionable, and the number of legal non-domestic
cigarettes was underestimated, thus overestimating the
size of illicit cigarette market.95 Even though the KPMG
(2013b)91 report tried to cross-verify the estimates of
unbranded illegal tobacco use by estimating the sale
of rolling paper, the result of this exercise was highly
sensitive to assumptions about the amount of RYO
tobacco used in each cigarette. The report assumed that
each RYO cigarette uses between 0.6 and 0.75 grams per
cigarette, while another study suggested that that amount
is close to 0.45 grams.95 Using higher amount of tobacco
per tube generated an upward bias in the estimates of
use of unbranded tobacco.
Table 8 assess the quality of the reports on the scope
of the illicit cigarette market in Australia. These studies
clearly do not meet the good quality criteria. The
only positive feature is the assessment of the trend,
but given the lack of transparency, the consistent
application of the same method cannot be guaranteed.
KPMG (2014)92 tried to address some weaknesses
of the earlier reports (e.g., it stop relying on the
methodologically weak RMR survey and accounting
for legal consumption of foreign packs), but the lack of
transparency about the methods and about contractual
agreements with the tobacco companies who have a
vested interest in the results put these estimates into the
unreliable category.
Example 5. Table 8
Assessing the Industry-Funded Estimates in Australia
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed No reference to a peer-review process; terms under
which study was prepared and reviewed are not
disclosed; disclaimer about using the results at your own
risk.
2. Funding Funding by the tobacco industry
acknowledged. Funding entity has a potential conflict of interest with
respect to the subject of the study.
3. Grounded in theory Study distinguishes between various types
of tax avoidance/evasion, and clarify which
types are subject of the study.
Takes into account some factors that
could influence the scope of tax avoidance/
evasion.
Fails to account for the presence of tourist and foreign
students that could influence the scope of tax avoidance/
evasion.
4. Transparency and
replicability Methods and data are not adequately described; the
results cannot be replicated using the information
provided in the study; data is not publically available.
Assumptions are not stated or stated clearly.
5. Generalizability of results Sampling design is not adequately described.
Survey sample suers from high non-response rate and
there is no attempt to correct for this or to establish the
representativeness of the sample.
6. Objective criteria preferred
over subjective criteria Low-tax purchases are identified by respondents’
self- report, and there is no attempt to cross-verify the
information using objective criteria.
7. Measurements are
defined correctly Survey questionnaire doesn’t clearly distinguish between
dierent tax avoidance/evasion categories; categories
may overlap and the same purchase might be counted
multiple times.
Conversion of cigarette sticks to/from weight measure is
not transparent or is not justified.
8. Identification of
counterfeit products Identification of counterfeit products is performed by a
party with a vested interest in in the results.
9. Presentation of results The size of the illicit market is expressed
as a share of the total market (KPMG and
PwC reports).
Results are not presented as a range or with confidence
intervals. Results are not robust with respect to
assumptions made.
The size of the illicit market is expressed as a share of
the licit market (Deloitte reports).
10. Measures change over time using
the same method or cross-validates
a point estimate using multiple
methods
Changes in tax avoidance/evasion over
time are presented in later reports. Earlier reports present a point estimate of the scope of
tax avoidance/evasion.
Corroborating evidence used to cross-verify results
cannot be trusted based on criteria presented in this
table.
11. Acknowledgement of
methodological weaknesses Weaknesses of the applied methodology/data are not
acknowledged/discussed.
-41-
Example 6
In Europe, Project Star conducted by KPMG has
provided an annual estimate of illicit tobacco market
volume and market share at both national and EU
levels since 2006, even though the first report that
became publically available is from 2011 and provided
estimates for 2010.77 The project is the result of an
agreement between Philip Morris International (PMI),
the European Commission, OLAF (the European
Anti-Fraud Oce) and the EU Member States. The
data comes from statistics on legal cigarette sales,
discarded packs surveys, and from consumer surveys.
The sale data are in most cases provided by PMI, and
corroborated by prevalence data, also obtained from
PMI despite the fact that ocial WHO prevalence
data are publically available. Since the scope of
tax avoidance/evasion depends to some extent on
prevalence estimates, and the PMI and WHO data
on prevalence dier, the results may be suspect on
this ground alone. The discarded pack collection at
a country level is conducted by various commercial
entities under contracts with the main tobacco
manufacturers. In each country, a sample of littered
cigarette packs is periodically collected in several
medium and large cities in order to determine the
prevalence of non-domestic and counterfeit products,
while also studying the non-domestic (i.e., both legal
via the import allowance and illegal via smuggling)
cigarette market shares of four main manufacturers.
However, the methodology of empty pack surveys and
the model used to generate estimates are not suciently
described to judge the quality of the estimates, but
there is evidence that the tobacco industry used selective
sampling in order to systematically misrepresent the size
of the illicit cigarette market.66 Specifically, the sampling
method in Germany overrepresented geographic regions
along the country’s eastern border and around U.S.
military bases, where evidence of more tax avoidance/
evasion can be expected. In some cases, the sample
sizes of the empty pack survey vary substantially from
its original targets without any explanation. For
example, the target sample size for Poland in 2011 was
34,000 packs, but the final sample consisted of 694,547
discarded cigarette packs.72 All data are collected in
cities, thus underrepresenting the rural population.
The estimates of non-domestic packs rely heavily on
expertise and data provided by the tobacco industry,
which has an obvious conflict of interest. For example,
the manufacturers determine if a discarded pack is
genuine or counterfeit while knowing that there are
penalties if the genuine pack share rises to a certain
level.18 A close examination of the 2010 results revealed
that almost a quarter of the EU illicit cigarette market
consisted of PMI’s own brands, while counterfeited
PMI brands represented just 5% of this market — a
finding obscured by PMI’s public presentation of the
data.18 The consumer survey data are used to estimate
the scope of legal non-domestic consumption, but this
method fails to account for legal cross-border purchases
by migrant workers, foreign students, and those living
in the border areas, thus overestimating the size of
illicit cigarette market. Project Star estimated that 9.9%
of the total EU market consisted of illicit cigarettes
in 2010.56 An independent study estimating the scope
of illicit cigarette consumption in 16 EU countries
found that only 6.5% of the total market consisted
of illicit cigarettes. This is 33% lower compared to
the 9.7% Project Star estimate for similar countries.
The 2012 Project Star estimates stated that that 65.5
bn cigarettes, or 11.1% of total EU consumption, is
illegal.98
In Asia, Philip Morris International (PMI) funded the
Asia-11 and Asia-14 illicit trade studies, which relied on
a methodological approach similar to the KPMG Project
Star and were executed by the International Tax and
Investment Center (ITIC) and Oxford Economics.99,100
The International Tax and Investment Center (ITIC)
claims to be an independent non-profit research and
education foundation that serves as a clearinghouse for
information on best practices in taxation and investment
policy. However, it is funded by the major transnational
tobacco companies (Philip Morris International, British
American Tobacco, and JT International), which calls
into question ITIC’s independence.101 Oxford Economics
is a commercial consulting group. The Asia-11 and
Asia-14 reports, like the KPMG reports for the EU,
lacks specifics about its methodology, sampling design,
and data sources.102-104 The data analysis often relies on
industry-provided data and on methodologically weak
estimates, while the description of the surveys makes
it dicult to assess the extent to which they provide
representative and meaningful data. Surveys conducted
via telephone interviews are particularly problematic
since many households in lower income countries may
not have telephones. The scope of the empty pack surveys
is not well described. If these surveys were conducted in
popular public places in urban areas and/or during peak
tourist times, or were conducted in a nonrandomized
unscientific manner (e.g., selecting known hotspots for
illicit consumption), such surveys will overestimate the
extent of illicit cigarette trade. In addition, the report does
not provide a clear rationale for selecting countries for
this report.102 Both the Asia-11 and the Asia-14 reports as
well as the Project Star reports ignore the possibility that
the tobacco industry could be a source of illicit products
and that the legal supply chain is not secure.
The Project Star and Asia-11 reports are assessed
against the quality criteria in Table 9. Each study
clearly failed to meet the majority of the criteria and,
as advised by the studies themselves, the public should
rely on these estimates at their own risk.
-42-
Example 16. Table 9
Assessing Project Star105,98 and Oxford Economics (2012)99
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed No reference to a peer-review process; specific
terms under which study was prepared are not
disclosed; disclaimer about using the results at
your own risk.
2. Funding Funding by tobacco company acknowledged. Funding entity has conflict of interest with
respect to the subject of the study.
3. Grounded in theory Distinguishes between various types of tax
avoidance/evasion.
Takes into account only some factors that could
influence the scope of tax avoidance/evasion.
Fails to account for some cross-border
shopping and purchases by migrant workers
and foreign students, which could overestimate
the scope of tax evasion.
4. Transparency and replicability Methods and data are not adequately
described; the results cannot be replicated
using the information provided in the study;
data is not publically available.
Assumptions are not stated or stated clearly.
5. Generalizability of results Sample size and sampling design is not
adequately described; sample size is too small
in some countries to allow for generalization of
results.
The sample selection for discarded packs is
biased – urban areas only; biased prevalence
estimates.
Sample suers from high non-response rate
and there is no attempt to correct for this or to
establish the representativeness of the sample.
6. Objective criteria preferred over
subjective criteria Low-tax products are identified based on a set
of objective criteria. The identification criteria are defined by a
stakeholder with a vested interest in the results.
7. Measurements are defined correctly The method cannot distinguish between
various types of tax avoidance and tax evasion
categories.
Questionnaire does not distinguish between
dierent types of tax avoidance/evasion (Asia
study).
8. Identification of counterfeit products Identification of counterfeit products is
performed by a party with a vested interest in in
the results.
9. Presentation of results The size of the illicit market is expressed as a
share of total market. Results are not presented as a range or with
confidence intervals.
10. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Estimates changes in tax avoidance/evasion
over time using the same method in Europe.
Multiple approaches are used, but they are
complementary and cannot be relied upon in
isolation.
Only point estimate presented in the Asia
report.
Corroborating evidence used to cross-verify
results cannot be trusted based on criteria
presented in this table.
11. Acknowledgement of methodological
weaknesses Weaknesses of the applied methodology/data
are not acknowledged/discussed.
Example 7
A study from Italy Calderoni (2014)97 claimed to
use a new method to estimate the volume of the illicit
cigarette market and the associated revenue loss for
a four year period (2009–2012) at the regional level.
The study relied on data collected for the industry-
sponsored Project Star (KPMG, 2013a)98 and on data
directly provided by Philip Morris International. Even
the author himself expressed concerns about data quality,
but he used them anyway to generate regional estimates
by simply weighting the national estimates by the regional
estimate of smokers’ share in the population and ‘non-
domestic’ packs in the empty pack surveys. The author
recognized that the non-domestic packs share was a
poor proxy for illicit trade share, because empty pack
collection cannot distinguish between tax avoidance and
tax evasion. The study reported that the revenue loss
due to tax avoidance/evasion increased from €0.5 bn in
-43-
2009 to €1.2 bn in 2012, and suggested that the share
of illicit cigarettes on the market is driven by proximity
to countries with cheaper cigarettes. The author failed
to mention that proximity can also motivate legal cross-
border shopping.
The author of the study, Professor Calderoni, is a
researcher at Transcrime, a research center associated
with the Università Cattolica del Sacro Cuore in Milan
and funded by Philip Morris International. Although
Transcrime researchers claim to have full control
over their research results, their reports closely reflect
tobacco industry views on public policies. In addition,
tobacco companies often present Transcrime’s work
in policy debates without mentioning their industry
funding, creating the impression of a broad independent
constituency in favor of the industry’s arguments against
tobacco control policies.63
Table 10 summarizes the critique of Calderoni (2014).
Despite the study being published in a peer-reviewed
journal and meeting some criteria of a well-executed
study, it suers from a major weakness of relying on
poor quality data. In addition, the author works for
an institution funded by the tobacco industry. These
two features call the results of Calderoni (2014) into
question.
Example 7. Table 10
Assessing Calderoni (2014)97 Estimates
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed Published in a peer-reviewed journal.
2. Funding Funding from tobacco industry not disclosed/
acknowledged.
Funding entity has a potential conflict of
interest with respect to the subject of the study.
3. Grounded in theory Study describes various types of tax
avoidance/evasion. Uses results of a study that cannot distinguish
between tax avoidance/evasion.
Takes into account various factors that could
influence the scope of tax avoidance/evasion.
4. Transparency and replicability Methods and data are adequately described so
that the results can be replicated. Data is not publically available.
Assumptions are clearly stated.
5. Generalizability of results Sample size and sampling design are not
adequately described; sample size might be too
small to permit regional level estimates.
The sample selection is biased toward urban
areas.
There is no attempt to establish the
representativeness of the sample.
6. Objective criteria preferred over
subjective criteria Low-tax products are identified based on a
set of objective criteria. The identification criteria are defined by a
stakeholder with a vested interest in the results.
7. Measurements are defined correctly The method cannot distinguish between
various types of tax avoidance and tax evasion
categories.
8. Identification of counterfeit products Identification of counterfeit products is
performed by a party with a vested interest
in the results.
9. Presentation of results Estimates are presented in a range that
accounts for various assumptions used in
generating the estimate.
It is not clear that the size of revenue loss is
expressed correctly, and it is not clear how it
was calculated.
There are some indications that the size of
revenue loss due to illicit market is expressed
as a share of potential revenue.
10. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Estimates changes in tax avoidance/evasion over
time using the same method. No cross-validation of results.
11. Acknowledgement of methodological
weaknesses Points to possible weaknesses of the data.
-44-
Given the problematic assumptions and definitions
of some variables as well as issues with the model
specification, the results should be interpreted with
caution. The tax dierential with a neighbouring state,
which was supposed to estimate the extent of cross-
border shopping, was statistically significant, but had
the wrong sign, meaning that higher tax in a state
would motivate people from a lower-tax state to cross
the border and buy cigarettes with higher taxes. The
authors dealt with this unexpected result by including
an interactive term between tax dierence and the
percentage of population living near the border.
However, this variable is problematic, as indicated
above. The state tax dierential with the state of
North Carolina had the correct sign, and the authors
concluded that this dierential causes commercial
smuggling. However, the model could not capture
large scale tax evasion by diversion, which involves the
manipulation of accounting records, reporting only
a portion of their sales, or importing illegal cigarettes
from overseas. The country-level Michigan model
evaluating the impact of tax increases in neighbouring
states also generated mixed results, with several
coecients having a wrong sign. The results that had
the wrong signs were ignored by the authors.
Table 11 outlines the shortcoming of LaFaive et
al,2008 The study meets some criteria of a well-
executed study, but suers from major limitations such
as incorrect model specification and ignoring the results
that do not fit the study’s hypothesis. These issues
might have been addressed if the study had been peer
reviewed. In addition, the Mackinac Center for Public
Policy has been criticized by leading academics for the
low quality of its research.28 Therefore, these results
should be treated with considerable skepticism.
Example 8
In the USA, LaFaive and colleagues from the Mackinac
Center for Public Policy generated estimates of tax
avoidance and tax evasion for 47 US states from
1990 to 2006 using econometric analysis and taking
advantage of cross-state and cross-time variation in
state excise taxes and smoking prevalence.106 Their
model consisted of two parts. The first part estimated
state per capita tax paid sales as a function of state-
level smoking prevalence and a time trend. The time
trend was supposed to represent smoking intensity
and consumption underreporting, assuming that
smoking intensity and consumption underreporting
are similar across states and exhibit similar trend, while
there was no underreporting of smoking prevalence.
However, this assumption is not realistic since there are
dierences in smoking intensity across states26, smoking
prevalence is being underreported69, and social norms
that influence underreporting can have dierent trends
across states given their diverse approaches to tobacco
control.22
The second part of the model estimated the gap
between actual sales and sales predicted by the first part
of the model as a function of average tax dierential
with neighbouring states weighted by the population
living near the border, the population living on both
sides of the border as the share of the state population,
the state tax dierential with the state of North
Carolina (North Carolina was not included in the
model), and a state tax for states with Native American
reservations and/or states with Mexican or Canadian
borders. The population variable is problematic since
only the border population in a particular state is
motivated to either cross-border shop or not, and this
motivation is unaected by the population size on the
other size of the border. The mean of the population
variable is 1.305, meaning that on average 130.5% of
state population would be potentially motivated by
the tax dierentials. The model treated the presence
of a Canadian and Mexican border the same way as it
treated the presence of Native American reservations,
assuming that the motivation to shop there was to avoid
state taxes without considering the prices of cigarettes
in those two countries. In addition, the model assumed
that North Carolina is the only source of all illegal
bootlegging, given its no-tax-stamps requirement and
relatively low tax. This is not a realistic assumption
given the empirical evidence that other states are also
the source of cigarette bootlegging.64
-45-
Example 8. Table 11
Assessing LaFaive et al (2008)106
criteria
characteristics of studies that
meet the criteria
for good quality
characteristics of studies that
do not meet the criteria
for good quality
1. Peer reviewed No reference to a peer-review process.
2. Funding Funding acknowledged. Funding entities may have a potential conflict
of interest with respect to the subject of the
study.
3. Grounded in theory Study distinguishes between various types of
tax avoidance/evasion, and clarifies which types
are subject of the study.
The model estimating tax avoidance/evasion is
not correctly specified.
Fails to account for some factors that could
influence the scope of tax avoidance/evasion
(e.g. regulation of cigarette Internet sales).
4. Transparency and replicability Methods and data are adequately described;
data is publically available or can be made
available upon request.
Assumptions are clearly stated.
5. Generalizability of results Sample size and sampling design is well
described and allow for generalization of results
to the entire country/region/population.
The sample is selected objectively.
6. Measurements are defined correctly The motivation for tax avoidance/evasion is not
correctly defined.
7. Presentation of results The size of illicit market is expressed as a share
of total market. Results are not presented as a range or with
confidence intervals. Results are not robust
with respect to assumptions made.
8. Measures change over time using
the same method or cross-validates a
point estimate using multiple methods
Estimates the scope of tax avoidance/evasion
during one period of time without using
multiple methods to cross-verify the results.
Another method is used to establish
relationship between tax and tax avoidance/
evasion, but the results are not consistent with
the study’s hypothesis.
9. Acknowledgement of methodological
weaknesses Points to possible weaknesses. The impact of the weaknesses on the estimates
is not discussed; unexpected results are ignored.
-46-
Table 12
Summary of Studies Subject to the Quality Criter ia
study
number of
criteria for
high quality
study met
number of
criteria for
low quality
study met
results
can be
trusted
main reason
for trusting/
not trusting
the results
Lakhdar (2008) 12 8 Ye s Estimates cross-validated using multiple methods;
weaknesses are acknowledged.
Stoklosa and
Ross (2014) 15 5 Ye s Estimates cross-validated using multiple methods;
weaknesses are acknowledged.
Blecher E
(2010) 12 3 Ye s Focuses on the changes in tax avoidance/evasion
over time; weaknesses are acknowledged.
Pavananunt
(2011) 12 7 No Struggles with data quality and appropriate
methods; most weaknesses are acknowledged.
Industry
Funded
Estimates in
Australia
5 18 No Lacks transparency; data potentially biased; funder
with a conflict of interest.
Project Star
(KPMG 2011;
KPMG 2013a)
and Oxford
Economics
(2012)
7 17 No Lacks transparency; data potentially biased; funder
with a conflict of interest.
Calderoni
(2014) 11 13 No Relies on poor quality data; institute’s funder has a
conflict of interest.
LaFaive et al
(2008) 9 10 No Model not specified correctly; selective
interpretation of the results.
The summary in Table 12 clearly demonstrates that
studies supported by the tobacco industry cannot be
trusted due to lack of transparency and the use of
potentially contaminated data. The estimates presented
in these studies are consistently and substantially
higher compared to those produced by independent
researchers.72,93
Summary of examples
Even though assessing the quality of studies of tax
avoidance/evasion may requires technical skills and
experience, this chapter provides a set of simple
indicators that allow even a lay person to form an initial
opinion about a study. A study that does not meet one
or more of the quality criteria should be subject to
scrutiny by experienced and independent researchers.
The studies subject to such scrutiny in this chapter are
summarized in Table 12.
-47-
This Methodological Guide updates the previous set of
recommendations on how to estimate the scope of tax
avoidance and tax evasion. We draw on the results of
numerous empirical studies that tested the applicability
of five methods described by the World Bank Toolkit
#7 in a variety of settings. Over time, some of those
methods have been refined, and some revealed new
weaknesses, while new methods evolved as a response
to the evolving nature of illicit tobacco trade, the policy
debates surrounding the issue, and the development of
new technologies.
Our Methodological Guide expanded the original five
approaches into eleven distinct methods by adding five
new methods and by separating one method (observing
smokers) into two (survey of tobacco users and
examination of cigarette packs obtained from smokers)
due to their distinct features. Unlike the previous
Toolkit, we summarize the principles of sound research
at the beginning of the core section “How to Measure
the Scope of Tax Avoidance and Evasion” and do not
repeat it with each method. This allows us to focus
on the distinct features of each methodology. Another
new feature of this Guide is a critical assessment of the
existing estimates of tax avoidance and evasion. The
goal is to provide guidance on how to assess the quality
of existing estimates and help various stakeholders,
including the research community and policymakers,
to navigate through studies that are presented to them.
Since the sole focus of the Guide was to provide an
estimate of the scope of tax avoidance and tax evasion,
it does not describe methods to eliminate the problem.
The WHO FCTC Protocol to Eliminate Illicit Trade in
Tobacco Products3 provides a comprehensive overview
on that subject.
Conclusions and Summary
Based on the assessment of the methods presented
in this Methodological Guide, we recommend to
use multiple methods that suer from the minimum
weaknesses, execute them according to the principles
of the rigorous research and triangulate the results
in order to cross-validate the estimates and minimize
methodological limitation of individual methods. Such
an approach will result in methodologically sound
and objective quantitative estimates of tobacco tax
avoidance and tax evasion.
It is hoped that this Methodological Guide will
inspire the research community to study the scope of
tax avoidance and tax evasion, apply the described
methods, and build upon them in order to develop new
ones. This will improve our understanding of the scope
of the problem and allow for the development of tailor-
made solutions to minimize it.
-48-
References
1. Merriman D. Tool 7. Smuggling. Understand, measure and combat
tobacco smuggling. Economics of tobacco toolkit. Washington DC, The
World Bank.2001.
2. International Agency for Research on Cancer . IARC Handbooks of
Cancer Prevention, Tobacco Control, Vol. 14: Eectiveness of Tax and
Price Policies for Tobacco Control. Lyon, France: IARC; 2011.
3. World Health Organization. Protocol to Eliminate Illicit Trade in
Tobacco Products. Geneva, Switzerland: World Health Organization;
2013.
4. World Health Organization. WHO Framework Convention on Tobacco
Control. Geneva, Switzerland: World Health Organization; 2005.
5. World Customs Organization. Illicit Trade Report. Brussels, Belgium:
2012.
6. DTZ Pieda Consulting. The black market in tobacco products. London:
Tobacco Manufacturer’s Association; 2000.
7. Hargreaves S. The lucrative business of cigarette smuggling. CNN
Money. 2013 April 16.
8. Joossens L. Vietnam: smuggling adds value. Tobacco Control.2003;
12:119–120. doi:10.1136/tc.12.2.119 PMID:12773713.
9. Merriman D, Yurekli A, Chaloupka F. How big is the worldwide cigarette
smuggling problem? In: Prabhat J, Chaloupka F, eds., Tobacco control in
developing countries. Oxford, UK: Oxford University Press. 2000; 1-464.
10. Joossens L, Raw M. Progress in combating cigarette smuggling:
controlling the supply chain. Tobacco Control. 2008; 17:399–404.
doi:10.1136/tc.2008.026567 PMID:18784154.
11. Joossens L . Smuggling and cross-border shopping of tobacco products
in the European Union. London: The Health Education Authority. 1999.
12. Joossens L, Merriman D, Ross H et al. How eliminating the global
illicit cigarette trade would increase tax revenue and save lives. Paris:
International Union against Tuberculosis and Lung Disease; 2009.
13. Council of the European Union. Report on cigarette smuggling in the
European Union. Brussels: Council of the European Union; 2005.
14. Duy M. Tobacco consumption and policy in the United
Kingdom. Applied Economics. 2006 ; 38:1235–1257
doi:10.1080/00036840500392599
15. Joossens L, Chaloupka F, Merriman D et al. Issues in the smuggling
of tobacco products. In: Jha P, Chaloupka F, eds., Tobacco Control in
Developing countries. London, Oxford University Press. 2000; 393-406.
16. West R, Townsend J, Joossens L, Arnott D, Lewis S. Why combating
tobacco smuggling is a priority. BMJ. 2008; 337:1028–1029.
17. Lovenheim M. How far to the border? The extent and impact of cross-
border casual cigarette smuggling. Discussion Paper N° 06-40. Stanford,
CA, Stanford Institute for Economic Policy Research (SIEPR). 2007.
18. Gilmore AB, Rowell A, Gallus S, Lugo A, Joossens L, Sims M. Towards
a greater understanding of the illicit tobacco trade in Europe: a review
of the PMI funded ‘Project Star’ report. Tobacco Control.2013; 0: 1-11.
doi:10.1136/tobaccocontrol-2013-051240.
19. Liber AC, Ross H, Omar M . The Impact of the Malaysian Minimum
Cigarette Price Law: Findings from the ITC Malaysia Survey Tobacco
Control. Print. 2015.
20. Merriman D . Cigarette smuggling does not reduce the public health
benefits of cigarette taxes. Applied Economics Lett. 2002; 9:493–496
doi:10.1080/13504850110095468.
21. Chernick H, Merriman D . Using Littered Pack Data to Estimate
Cigarette Tax Avoidance in NYC. National Tax Journal. 2013 June;
635–668.
22. National Center for Tobacco-Free Kids. FY2004 Rankings of State
Funding for Tobacco Prevention, National Center for Tobacco-Free
Kids. 2003 November 4. www.tobaccofreekids.org.
23. Canadian Cancer Society , Non-smokers’ Rights Association, Physicians
for a Smoke-free Canada, et al. Surveying the damage: cut-rate tobacco
products and public health in the 1990s. Ottawa, Ontario: 1999.
24. Physicians for a Smoke-free Canada, Tobacco Tax Revenues Fact Sheet.
2013 November. www.smoke-free.ca/factsheets/pdf/totaltax.pdf; accessed
12/4/14.
25. Baudier F . Le tabagisme en 1997: état des lieux. La santé de l’homme.
1997. 331(14).
26. VizualStatistix. Average Daily Cigarette Consumption per Adult Smoker
by State shared by VizualStatistix on Sep 28, 2013 in Lifestyle. Available
at: http://visual.ly/average-daily-cigarette-consumption-adult-smoker-
state.
27. Emer y S, White MM, Gilpin EA, Pierce JP . Was there significant
tax evasion after the 1999 50 cent per pack cigarette tax increase in
California? Tobacco Control. 2002; 11:130–134.doi:10.1136/tc.11.2.130
PMID:12035006.
28. Reuter P, Majmundar M, Committee on the Illicit Tobacco Market,
National Research Council. Understanding the U.S. Illicit Tobacco
Market: Characteristics, Policy Context, and Lessons from International
Experiences. National Academy of Sciences. 2015.
29. Baltagi B, Levin D. Estimating dynamic demand for cigarettes
using panel data: the eects of bootlegging, taxation and advertising
reconsidered. Rev Econ Stat. 68:148–155 doi:10.2307/1924938.
30. Licari MJ, Meier KJ. Regulatory policy when behavior is addictive:
smoking, cigarettes taxes and bootlegging. Polit Res Q. 1997; 50:5–24.
31. Goolsbee A, Lovenheim M, Slemrod J . Playing with fire: cigarettes, taxes
and competition from the Internet. Working Paper No. 2007-5. Ann
Arbor, Michigan, Oce of Tax Policy Research, Michigan Ross School of
Business.2007.
32. Townsend J, Roderick P, Cooper J . Cigarette smoking by socio-
economic group, sex, and age: eects of price, income, and health
publicity. BMJ. 1994; 309:923–927. PMID:7950662.
33. DeCicca P, Kenkel D, Liu F. Excise Tax Avoidance: the case state of
cigarette taxes. NBER Working Paper Series. Working Paper #15941.
Cambridge, MA, National Bureau of Economic Research. 2010.
34. Hyland A, Bauer JE, Li Q et al. Higher cigarette prices influence cigarette
purchase patterns. Tobacco Control. 2005; 14:86–92.doi:10.1136/
tc.2004.008730 PMID:15791017.
35. Fix BV, Hyland A, O’Connor RJ, Cummings KM, Fong GT, Chaloupka
FJ, Licht AS. A Novel Approach to Estimating the Prevalence of Untaxed
Cigarettes in the USA: Findings from the 2009 and 2010 International
Tobacco Control Surveys. Tobacco Control. 2014; 23:i61–i66.
doi:10.1136/tobaccocontrol-2013-051038.
36. Wiltshire S, Bancroft A, Amos A, Parry O . “They’re doing people a
service”-qualitative study of smoking, smuggling, and social deprivation.
BMJ.2001; 323:203–207.doi:10.1136/bmj.323.7306.203 PMID:
11473911.
37. Moodie C, Mackintosh A, West R . Adolescents’ awareness of,
and involvement with, illicit tobacco in the UK. Tobacco Control.
2010(19):521e522.
38. Lee JM, Chen SH, Chen HF, Jeng HY . Price sensitivity and smoking
smuggled cigarettes. Eur J Public Health, 19:23–27.doi:10.1093/eurpub/
ckn115 PMID: 19039020.
39. Gruber J, Sen A, Stabile M. Estimating price elasticities when there is
smuggling: the sensitivity of smoking to price in Canada. NBER Working
Paper Series. Working Paper #8962. Cambridge, MA: National Bureau
of Economic Research. 2002.
40. Callaghan RC, Veldhuizen S, Leatherdale S et al. Use of contraband
cigarettes among adolescent daily smokers in Canada. CMAJ. 2009;
181:384–386. PMID: 19737829.
41. Cantreill J, Hung D, Fahs MC, Shelley D . Purchasing patterns and
smoking behaviors after a large tobacco tax increase: a study of Chinese
Americans living in New York City. Public Health Rep. 2008; 123:135–
146. PMID: 18457066.
42. Market Tracking International Ltd . World Tobacco file 1996: emerging
markets in Central and Eastern Europe. London, Argus Business; 1997.
43. Joossens L, Raw M. Cigarette smuggling in Europe: who really benefits?
Tobacco Control. 1998; 7:66–71.doi:10.1136/tc.7.1.66 PMID: 9706757.
44. World Customs Organization . Customs and Tobacco Report 2006.
Brussels: World Customs Organization; 2007.
45. LeGresley E, Lee K, Muggli ME et al. British American Tobacco
and the “insidious impact of illicit trade” in cigarettes across Africa.
Tobacco Control.2008; 17:339–346. doi:10.1136/tc.2008.025999
PMID:18617598.
46. Collin J, Legresley E, MacKenzie R et al. Complicity in contraband:
British American Tobacco and cigarette smuggling in Asia. Tobacco
Control.2004; 13 Suppl 2;ii104–ii111. doi:10.1136/tc.2004.009357
PMID:15564212.
47. Joossens L. Tobacco smuggling: an optimal policy approach. In: Abedian
A, van der Merwe R, Wilkins N, et al., eds., The Economics of Tobacco
Control. Cape Town: Applied Fiscal Research Center; 1998.
48. Gilmore AB, McKee M. Tobacco and transition: an overview of industry
investments, impact and influence in the former Soviet Union. Tobacco
Control.2004a; 13:136–142.doi:10.1136/tc.2002.002667
49. Gilmore AB, McKee M. Moving East: How the transnational tobacco
industry gained entry to the emerging markets of the former Soviet
Union - Part I: Establishing cigarette imports. Tobacco Control. 2004b;
13:143-150. doi:10.1136/tc.2003.005108
50. Gilmore A, Collin J, Townsend J . Transnational Tobacco Company
Influence on Tax Policy during Privatization of a State Monopoly: British
American Tobacco and Uzbekistan American Journal of Public Health.
2007; 97(11), November 2007.
-49-
51. Nakkash R, Lee K . Smuggling as the “key to a combined market”:
British American Tobacco in Lebanon. Tobacco Control. 2008; 17:324–
331.doi:10.1136/tc.2008.025254 PMID:18818226.
52. Skafida V, Silver KE, Rechel BPD, Gilmore AB . Change in tobacco
excise policy in Bulgaria: the role of tobacco industry lobbying
and smuggling. Tobacco Control. 2012; 1-10. doi:10.1136/
tobaccocontrol-2012-050600
53. World Health Organization .The Cigarette <transit> road to the Islamic
Republic of Iran and Iraq: Illicit tobacco trade in Middle East. Report
no. WHO/EM/TFI/011/E/G. Cairo, WHO: Regional Oce for the
Mediterranean; 2003.
54. Walton I. Combating illicit and counterfeit cigarettes: the elements of the
problem. Brussels: ITIC; 2009 Nov.
55. House of Commons. Tobacco smuggling. First Report of Session 2014–
15. Home Aairs Committee, House of Commons, HC 200, UK, 2014
June 14.
56. Joossens L, Ross H, Stoklosa M. ‘EU Policy and Illicit Tobacco
Trade: Assessing the Impacts. Prepared for the European Parliament’s
Committee on Budgetary Control Workshop “Cigarette Smuggling”
Proceedings. Brussels, 2014 January. www.europarl.europa.eu/document/
activities/cont/201401/20140116ATT77675/20140116ATT77675EN.
pdf (accessed 07/01/2014)
57. Ribisl KM, Kim AE, Williams RS. Web sites selling cigarettes: how
many are there in the USA and what are their sales practices? Tobacco
Control.2001; 10:352–359. doi:10.1136/tc.10.4.352 PMID: 11740027.
58. Ribisl KM, Williams RS, Kim AE. Internet cigarette sales: Briefing to the
Congressional task force on tobacco and health. Washington DC: The
Robert Wood Johnson Foundation; 2006.
59. Stephens WE, Calder A, Newton J. Source and health implications of
high toxic metal concentrations in illicit tobacco products. Environ Sci
Technol. 2005; 39:479–488.doi:10.1021/es049038s PMID:15707047
60. Pappas RS, Polzin GM, Watson CH, Ashley DL. Cadmium, lead, and
thallium in smoke particulate from counterfeit cigarettes compared
to authentic US brands. Food Chem Toxicol. 2007; 45:202–209.
doi:10.1016/j.fct.2006.08.001 PMID:17011104
61. Fleenor P. Cigarette taxes, black markets, and crime: Lessons from New
York’s 50-year losing battle. Policy Analysis Report N°.468. Washington
DC: The Cato Institute; 2003.
62. FIA International Research Ltd. Organized crime and the smuggling
of cigarettes in the United States - The 1999 update. Washington: FIA
International Research Ltd; 1999a
63. Fooks G, Peeters S, Evans-Reeves K. Illicit trade, tobacco industry-
funded studies and policy influence in the EU and UK. Tobacco
Control.2014; 23:81–83. doi:10.1136/tobaccocontrol-2012-050788
64. Thursby JG, Thursby, MC. Interstate cigarette bootlegging; extent,
revenue losses, and eects of federal intervention. Natl Tax J.2000;
53:59–77.
65. Shepard E, Blackley P. The Impact of Marijuana Law Enforcement in an
Economic Model of Crime. Journal of Drug Issues. 2007; 37: 403.
66. Adams M, Eertz T. Tabaksteuern, Schmuggel und unversteuerte
Zigaretten – Zur Glaubwürdigkeit der “Entsorgungsstudie” der
Tabakindustrie [Excise Taxes on Tobacco and the Problem of Smuggling
– Concerning the Credibility of the Tobacco Industry’s “Discarded-
Cigarette Packages-Study”]. Gesundheitswesen. 2011;73:705–12.
67. Stoklosa M, Ross H. Contrasting Academic and Tobacco Industry
Estimates of Illicit Cigarette Trade: Evidence from Warsaw, Poland.
Tobacco Control.2014; 23:e30-e34 Published Online First: 2013 Aug
13, doi:10.1136/tobaccocontrol-2013-051099
68. Government Monitor. Singapore Sees 40% Fall in Supply of Illegal
Cigarettes. Government Monitor; Public Sector News & Information.
Government of Singapore. Posted on: 2010 January 2010 25.
69. Warner KE. . Possible increases in the underreporting of cigarette
consumption. Journal of the American Statistical Association.1978;
73:314-318.
70. Chiou L, Muehlegger E. Crossing the line: direct estimation of cross-
border cigarette sales and the eect on tax revenue. Regulatory Policy
Program Working Paper RPP-2008-02. Cambridge, MA, Mossavar-
Rahmani; 2008.
71. Kurti M, Lampe K, Thompkins D. The illegal cigarette market in a
socioeconomically deprived inner-city area: the case of the South Bronx.
Tobacco Control. 2013; 22: 138-140
72. Stoklosa M and Ross H. Tobacco control funding during the
global economic crisis. Tobacco Control. 2014; doi:10.1136/
tobaccocontrol-2013-051467
73. Merriman D. The Micro-Geography of Tax Avoidance: Evidence from
Littered Cigarette Packs in Chicago. Am Econ J - Econ Policy. 2010;
2:61–84 doi:10.1257/pol.2.2.61.
74. Williams E, Curnow R, Streker P. Understanding Littering Behavior
in Australia. Beverage Industry Environmental Council Community
Change Consultants Report. 1997; ACN 008 542 765.
75. Guindon GE, Driezen P, Chaloupka FJ, & Fong GT. Cigarette Tax
Avoidance and Evasion: Findings from the International Tobacco
Control Policy Evaluation Project. Tobacco Control. 2014; 23:i13–i22.
doi:10.1136/tobaccocontrol-2013-051074
76. Joossens L, Lugo A, Vecchia C, Gilmore A, Clancy L, Gallus S. Illicit
Cigarettes and Hand-Rolled Tobacco in 18 European Countries: a Cross-
sectional Survey. Tobacco Control. 2014; 23:e17–e23. doi:10.1136/
tobaccocontrol-2012-050644
77. KPMG. Project Star 2010 Results. Project Star. KPMG, (2010)
Accessed January 27, 2014.
78. Nagelhout G, Putte B, Allwright S, Ute Mons, McNeill A, Guignard
R, Beck F, Siahpush M, Joossens L, Fong G, Vries H, Willemsen
M. Socioeconomic and country variations in cross-border cigarette
purchasing as tobacco tax avoidance strategy. Findings from the ITC
Europe Surveys. Tobacco Control.2014; 23:i30–i38. doi:10.1136/
tobaccocontrol-2012-050838
79. Scollo M, Bayly M, Wakefield M. Availability of illicit tobacco in small
retail outlets before and after the implementation of Australian plain
packaging legislation. Tobacco Control 2014; 0:1–7.
80. HM Revenue & Customs (HMRC) Measuring tax gaps 2011. An ocial
statistics release. 2011 September21.
81. Yürekli A and Sayginsoy Ö. ‘Worldwide organized cigarette smuggling:
an empirical analysis’, Applied Economics, 2010; 42; 5:545 – 561.
82. Walbeek C. Measuring changes in the illicit cigarette market using
government revenue data: the example of South Africa. Tobacco Control.
2014;(23):e69-e74 Published Online First: 2014 Jan 15, doi:10.1136/
tobaccocontrol-2013-051178
83. Leverett M, Ashe M, Gerard S et al. Tobacco use: the impact of prices. J
Law Med Ethics. 2002; 30 Suppl;88–95. PMID:12508509
84. Lakhdar CB. Quantitative and qualitative estimates of cross-border
tobacco shopping and tobacco smuggling in France. Tobacco
Control.2008; 17:12–16. doi:10.1136/tc.2007.020891 PMID:18218801
85. Blecher E. A mountain on a molehill: is the illicit trade in cigarettes
undermining tobacco control policy in South Africa? Trends Organ
Crimes. 2010; 13:299–315. doi:10.1007/s12117-010-9092-y.
86. Pavananunt P. Illicit Cigarette Trade in Thailand. Southeast Asian
Journal of Tropical Medicine and Public Health. 2011; Vol 42 No. 6
November 2011.
87. PriceWaterhouseCoopers (PwC). Australia’s illegal tobacco market:
Counting the cost of Australia’s black market. Sydney: British American
Tobacco Australia, 2010.
88. Deloitte. Illicit trade of tobacco in Australia. Prepared for British
American Tobacco Australia Limited, Philip Morris Limited and
Imperial Tobacco Australia Limited, Sydney. 2011 Februar y.
89. Deloitte. Illicit trade of tobacco in Australia: Report for 2011. Prepared
for British American Tobacco Australia Limited, Philip Morris Limited
and Imperial Tobacco Australia Limited, Sydney, 2012 May.
90. Deloitte. Illicit trade of tobacco in Australia: Update for 2012. Prepared
for British American Tobacco Australia Limited, Philip Morris Limited
and Imperial Tobacco Australia Limited, Sydney, 2012 December.
91. KPMG. Illicit Tobacco in Australia. 2013 Half Year Report. 2013b
October.
92. KPMG. Illicit Tobacco in Australia. 2014 Half Year Report. 2014
October.
93. Cancer Council Victoria . Illicit trade of tobacco in Australia. A report
prepared by Deloitte for British American Tobacco Australia Limited,
Philip Morris Limited and Imperial Tobacco Australia Limited.
Quit Victoria, Cancer Council Victoria. 2011 March, updated 2011
November.
94. Cancer Council Victoria . Illicit trade of tobacco in Australia: Report for
2011: a report prepared for British American Tobacco Australia Limited,
Philip Morris Limited and Imperial Tobacco Australia Limited. A
critique by Quit Victoria, Cancer Council Victoria. 2012 May.
95. Cancer Council Victoria. Analysis of KPMG LLP report on use of illicit
tobacco in Australia. Prepared by Quit Victoria, Cancer Council Victoria;
2013November 11, updated 2014 March.
96. Cancer Council Victoria. Analysis of KPMG LLP report on use of illicit
tobacco in Australia 2013 Full year report. Prepared by Quit Victoria,
Cancer Council Victoria. 2014 April 11.
97. Calderoni F. A new method for estimating the illicit cigarette market at
the subnational level and its application to Italy, Global Crime, 2014
February 26; DOI:10.1080/17440572.2014.882777
98. KPMG. Project Star 2012 Results. Project Star. KPMG, (2013a).