Data and methods

We used two observational methods: data collection from packs in smokers’ personal possession and packs discarded on streets. The collection took place in Warsaw in September–October 2011. Two pack features determined whether the pack was intended for the Polish market: excise tax stamps and health warnings. Each legitimate cigarette pack must bear a Polish tax stamp and an approved health warning in Polish. The tax stamp is affixed under a cellophane cover.22 In order to determine whether a tax stamp was legitimate, we compared it with a sample of official tax stamps provided by the Ministry of Finance.23

A missing tax stamp is not always an indicator of tax avoidance/evasion. The stamps may be removed by a consumer, or may fall off a pack, especially in case of discarded packs. Therefore, only packs with a tax stamp issued by another country or without the Polish health warning were classified as packs non-taxed in Poland. Packs with missing or damaged tax stamp, but with a Polish health warning were counted as tax-paid cigarettes.

Data from packs in personal possession were collected by a market research agency, GfK Polonia, in Warsaw in October 2011 using a quota sampling method. The data collection protocol was approved by the Northeastern Illinois University Institutional Review Board. The population of smokers in Warsaw was segmented by districts and by gender and age. The quota of 400 observations was distributed among each of Warsaw's 18 districts proportional to the districts’ population. As no Warsaw-level data on gender and age of smokers were available, the gender and age sample proportions were based on characteristics of smokers living in Polish cities with more than 500 000 inhabitants (see online supplementary appendix 1 for the gender/age and the district sample proportions).

The interviews were conducted at various times of the day in an effort to include smokers who are not at home during working hours. Each interviewer had a number of randomly assigned street addresses as starting points in a district. This number depended on the quota for that district. From a given starting point, the surveyor had to conduct five interviews before moving on to their next starting point in another neighbourhood. If the starting point address was an apartment building, the interviewer would go to a randomly selected apartment. In each household, the surveyors would ask for a smoker, and if the person met the age and sex criteria, the interview was conducted. To choose the next household, the surveyor could ask if any of the neighbours was a smoker (snowball method) or go to the next household three doors away.

During each interview the smoker was prompted to show all open cigarette packs in their possession. This is different from the Global Adult Tobacco Survey that asked smokers to show only one open pack. The information on number of open packs and the presence/absence of Polish tax stamp and health warning on each pack was collected. If more than one pack was open, the tax avoidance/evasion was assessed based on the pack being smoked most often, as indicated by the smoker. Additionally, detailed information on the non-tax-paid packs was collected, including the information on the tax stamp (country issuing the stamp/stamp damaged/no stamp), the health warning (language of the health warning/no warning), the maximum price of the pack (printed on the bottom of the pack), the brand and the year the tax stamp was issued. The smoker was asked about the place of purchase of the pack. Sociodemographic characteristics of each smoker were also collected. The pack characteristics data were weighted by the self-reported amount of monthly cigarette consumption (eg, data collected from a person who smokes 60 packs a month would count twice as much as the data collected from a person who smokes 30 packs a month).

The discarded packs were collected in September 2011 in 30 of 783 voting districts in Warsaw. Our methodology was inspired by methods developed by David Merriman and applied in his studies in Chicago16 and New York.24 The sample voting districts were selected randomly using the probability proportional to size method, so that the probability of each district being selected was weighted by the number of adult permanent residents of that district. In each selected district, we designed a 3 km (1.86 miles) route on which two people picked up every littered cigarette pack. A typical route required data collectors to walk 1.5 km along several streets in the district and returning back to the starting point on the other side of the streets. Because voting districts are small geographical areas, a typical route would cover a substantial portion of each district (see online supplementary appendix 2 for more information on the routes design process). A total of 754 packs were collected, each placed in plastic bags, labelled with a unique serial number, and returned to the research office where pack data were entered to a database. That information included cigarette brand, package type (hard/soft), cigarette length, cigarette type (filtered/unfiltered), tax stamp details (country issuing the stamp/stamp damaged/no stamp) and health warning details (the language of the health warning/no warning).

We compared our results with results from an illicit cigarette study commissioned by four major tobacco companies (British American Tobacco, Imperial Tobacco Group, Japan Tobacco International and PMI) and conducted by the Almares market research company in August–September 2011.25 Despite the original sampling target of 34 000 packs, the study collected 694 547 discarded cigarette packs from streets and public bins in 70 cities in Poland. The reason this sample size changed so dramatically remains unclear. The sampling plan called for 15.6% of the sample packs to be collected in Warsaw; therefore we estimate that over 100 000 cigarette packs were collected in this city alone, a sample over 100 times larger than our sample.

Despite internet searches and multiple attempts to contact the tobacco manufacturers and the research company, we do not have all details of the method used by the tobacco industry. For example, we lack information on how the sample paths and bins for the discarded pack collection were selected, what pack features were taken into account when deciding whether the pack is tax-paid or non-tax-paid. We only know that the packs were examined by the four respective producers to find counterfeit cigarettes.

We used a Pearson χ² statistic to compare the results of our pack collection with the industry results. To account for the weights in the survey data when comparing the results of our two methods, we employed a second-order correction of Rao and Scott that uses the F statistic.26

Discussion and conclusions

Measuring tobacco tax avoidance/evasion is a complex task due to the illegal nature of some activities involved.27 Observational studies provide a direct way to assess tobacco tax avoidance/evasion through collection of reliable data from tobacco products’ packs/containers.28 We found that the proportion of packs classified as not intended for the Polish market is not statistically different between the packs in personal possession and those discarded on the streets.

Both methods, the survey of packs in smokers’ possession and the discarded pack collection, have limitations. First, it may be hard or even impossible to differentiate between packs intended and not intended for the domestic market, when there are no unique characteristics of the incountry tax-paid packs (eg, when no tax stamps are required, the stamps are not attached directly to a pack or the health warnings are not unique). Second, there might be mistakes in data coding. Third, in the household survey, smokers might be reluctant to present their non-taxed packs or participate in the survey, even though possessing a non-taxed pack is not illegal in Poland. Additionally, both methods cannot detect counterfeit packs that have all required characteristics, unless more sophisticated processes are included (eg, engaging experts from Ministries of Finance). Implementation of tracking and tracing measures, such as unique codes on every pack,29 would facilitate the estimations of tobacco tax avoidance/evasion. The Protocol to Eliminate Illicit Trade in Tobacco Products adopted by the parties to the WHO Framework Convention on Tobacco Control requires the parties to implement tracking and tracing systems within 5 years after the protocol enters into force, which is 5 years after the 90th day following the 40th party joining the protocol.30

Finally, among the packs classified as not intended for the domestic market, both methods fail to distinguish smuggled or illegally manufactured packs from legitimate packs that were purchased in duty-free shops or outside Poland in the amount allowed by the law. Nevertheless, Polish tobacco taxes were circumvented, which is a concern for tobacco tax revenue authorities and for those concerned with public health.

Our study suggests that the researchers studying tobacco tax avoidance/evasion using the discarded pack collection method should not base their estimates on missing stamps alone. We found statistically more packs without the Polish tax stamp in the discarded pack collection than in the survey of packs in smokers’ possession. This suggests that many of the discarded packs might have lost their stamps before the collection. The stamps might have been torn off by smokers before being littered, or washed off by rain while lying on streets. When only the tax stamp data are taken into account, collecting discarded packs will overestimate the tax avoidance/evasion. Redesigning tax stamps so that they are permanently affixed to cigarette packs would make the litter sampling method more reliable, and could assist with enforcing the law.

Unlike previous surveys of packs in personal possession,14 ,17 ,19 our study prompted smokers to show all open cigarette packs. We found that some smokers have taxed and non-taxed packs in their possession. This should be taken into account in further studies of packs in personal possession.

The estimate of cigarette tax avoidance/evasion from the industry-commissioned study is higher than our estimate based on discarded pack collection by nearly half. The reason for this difference is not known, since the industry is reluctant to disclose the details of their method. Some of the difference between our estimates and those produced by the industry-commissioned studies might result from the inability of our method to detect counterfeit cigarettes. However, the problem of cigarette counterfeiting in Poland is relatively minor (industry studies estimate that 1.9% of cigarettes in Poland are counterfeit 25). The industry-commissioned study has also collected data from packs found in public bins. However, Merriman found that the share of non-taxed packs among those that are legally discarded is not different from the share among the packs found on the streets.16 This suggests that our measures of prevalence of packs not intended for the domestic market based on the discarded packs method are comparable with the industry measure, and the observed discrepancy between our estimate and the estimate from the industry-commissioned study might result from the industry misuse of their method, as found in Germany.21

The tobacco industry has been very effective in disseminating its messages about the size of the illicit cigarette markets and its proposed solutions for their elimination. In Poland alone, there have been at least four corporate social responsibility campaigns focusing on illicit cigarette trade since 2010.31 These campaigns used the estimates from the industry-sponsored studies 32 ,33 and involved presentations on TV, in the press and on billboards, as well as the distribution of over a million leaflets to the public. The information was disseminated directly at schools and Catholic parishes, and the industry even organised workshops for customs officers.31 Additionally, since the industry estimates have been used by market intelligence firms,34 these inflated figures have been disseminated to other important stakeholders and repeated in many market analyses.

A government's reliance on the tobacco industry estimates of tobacco tax avoidance and evasion involves many risks. If governments are presented with such estimates, they should strictly require the industry to publically disclose all methodological details and data. Governments should also seek advice from independent experts on any matters introducing such conflict of interest.