Article Text
PDF
Abstract
Objective To estimate with a rigorous statistical methodology and independent from the tobacco industry the prevalence and consumption of illicit cigarettes in Metropolitan Santiago de Chile, in addition to identifying the variables statistically associated with choosing to smoke illicit cigarettes.
Methods Surveys of 851 smokers who reside in the Metropolitan Santiago were collected using a sampling design that combined a randomisation of high-traffic points and a quota sampling to approximate the smoking population. Photographs of packs along with questions on where they were bought were used to define whether cigarettes were licit or illicit. After this identification, the statistical association between the decision to smoke illicit cigarettes and sociodemographic variables and smoking habits was estimated using probit models.
Results The proportion of smokers smoking illicit cigarettes in Metropolitan Santiago was 10.9%. Adjusted by smoking intensity, 16.3% of cigarettes smoked in a month were illicit. Models show that the probability of smoking illicit cigarettes is inversely associated with employment status (ie, employed/inactive/unemployed), and smokers with lower levels of education are more likely to smoke illicit cigarettes. Though smokers’ incomes are not directly measured, both employment status and educational levels are indicative that illicit cigarette consumption is more prevalent among low-income groups.
Conclusions The proportion of smokers consuming illicit cigarettes estimated in this research is less than half of the widely publicised claims of the tobacco industry. Furthermore, past and present pricing strategies by the tobacco industry indicate that, contrary to public statements, the tobacco industry is not concerned by illicit trade.
- illegal tobacco products
- public policy
- taxation
- tobacco industry
Statistics from Altmetric.com
Introduction
Chile has one of the highest smoking prevalences in the Americas, with an adult 30-day prevalence of 33.4% in 2016.1 2 In the case of certain age groups, for instance, women under the age of 18, Chile has one of the highest 30-day prevalences in the world at 28.3%.3 4 Following the ratification of the WHO Framework Convention on Tobacco Control (FCTC) in 2006, excise taxes on tobacco products were increased, although modestly. In 2010, the ad valorem rate was raised from 50.4% to 62.5% and a specific tax of approximately CLP2.5 (US$0.005) per cigarette was introduced. In 2012, the ad valorem rate fell to 60%, and the specific tax doubled to CLP5.1 (US$0.01) per cigarette. Finally, the ad valorem fell to 30% in 2014, but the specific tax increased to CLP43.7 (US$0.08) per cigarette.
Mainly due to the tobacco industry’s (TI) pricing strategies, real cigarette prices increased sharply before and after those tax changes. Real cigarette prices rose by 40% between January 2000 and May 2010, despite the fact that there was no change in tobacco taxes during this period.5 The real price further rose 68% between June 2010 and December 2017. Figure 1 shows such an evolution and the times at which tobacco tax changed (vertical dashed lines).
Real price of cigarettes in Chile (yearly data, December of each year) 2000=100.
After the last tax change in 2014, British American Tobacco (BAT), which controls 96% of the Chilean market,6 started a persistent campaign in the press relating an alleged explosive rise in cigarette smuggling to the tax changes,7–9 omitting the fact that a significant part of the price increase was due to their own pricing strategies. In all cases, the origin of the numbers on smuggling was BAT itself and/or international market consultancies (eg, Kantar and KPMG), often consulting for the TI, which do not disclose the sampling framework or broader methodologies used (eg, gap analysis and smoker survey). Euromonitor International, for instance, reported an increase in illicit trade penetration from 5% in 2013 to 19.4% in 2017, stating that ‘there is a direct correlation between taxes, the average unit prices of cigarettes and illicit trade. Increases in taxes are usually passed on to the average unit price of cigarettes, boosting the growth of illicit trade in Chile’.10 In this case, there is no explanation on how these estimates where produced, who was interviewed, how the different estimates were weighted and so on. Even the official bodies in charge of illicit trade control replicate BAT figures. In its most up-to-date study, the Servicio de Aduanas (Chilean Customs Service) acknowledge that they do not know the extent of the illicit cigarette market and, because of that, they have relied on BAT figures.11
In 2017, BAT stated that the illicit cigarette market had multiplied 6.2 times during the previous 5 years, accounting for 22% of the national market and 24% of the Metropolitan Santiago market, where 40% of the country’s total population lives.12 13 It is not clear from BAT statements if they are referring to prevalence of smokers of illicit cigarettes (which may be the case, given they use a pack-swap method to obtain such estimates) or the market penetration of illicit cigarettes, as methodological aspects of these estimates are largely not reported by the industry.
The objective of this work is twofold: first, to estimate the proportion of smokers of illicit cigarettes and the market penetration of such cigarettes in Metropolitan Santiago de Chile; and second, to identify the variables that are statistically associated with choosing illicit cigarettes. Though typical illicit trade control measures are supply side oriented (eg, increasing penalties for contraband-related crimes, reducing or eliminating duty-free allowances, increasing controls at sensitive borders and implementing track and trace systems), identifying the socioeconomic variables associated with the purchase of illicit cigarettes may help authorities to understand relevant aspects of this phenomenon and help them to design more effective policies to control the illicit cigarette trade from the demand side. This follows similar studies conducted in the region, which also inquire about socioeconomic variables linked to illicit cigarette consumption.14–16 In this regard, the results obtained here can be compared with those obtained in similar countries.
For the first time in Chile, a direct method (survey of smokers) is used, and one that is independent of the TI. Because the use of direct methods to estimate illicit cigarette trade is relatively new in the Latin American region,15 16 this research makes an important new contribution to this growing literature and helps to establish patterns of both industry behaviour and that of smokers.
Methodology
Surveys of 851 smokers who reside in Metropolitan Santiago were implemented in May and June 2017. The sampling framework sought to represent the smoking population living in Metropolitan Santiago. The sample, with a sampling error estimated at 3.3%, combined a randomisation of high-traffic points and a quota sampling to approximate the smoking population. Thus, the first step of the sampling selection was to determine the geographic locations at which smokers would be interviewed. Data form the 2012 Encuesta de Origen y Destino del Gran Santiago (Survey of Origin-Destination in the Great Santiago Area, or EOD in its Spanish acronym)17 and 2016 data from the registry of built surfaces provided by the Servicio de Impuestos Internos (Inland Revenue Service, or SII in its Spanish acronym) were used. The EOD data were used to identify the number of travels made to 767 high-traffic areas of Metropolitan Santiago (without considering back-to-home travels). Data from the SII permitted the identification of 66 467 blocks (manzanas) for those 767 high-traffic areas, and for each block, it was considered the built surface (in square metres) for each type of building of the 22 types defined by the SII. These types include commercial buildings, sport facilities, schools, universities, office buildings, healthcare facilities and parks, among others. With this information, the predicted foot traffic was estimated for each point, using an Ordinary Least Squares (OLS) model without intercept:
where 
is the number of visits to the area i (predicted foot traffic), while the independent variables are the square meter areas of commercial buildings (
), sport facilities (
), educational and cultural facilities (
), office buildings (
), churches and cultural places (
), healthcare facilities (
), transportation and telecommunications (
) and parks (
).
Using the coefficients obtained from the regression (not shown but available from the authors), the predicted foot traffic was estimated for each of the 66 467 blocks of the SII registry. This exercise was conducted by the Centro de Inteligencia Territorial (Centre for Territorial Intelligence, or CIT in it Spanish acronym) of Universidad Adolfo Ibáñez.
Once this variable was estimated for each block, to have a wide coverage of Metropolitan Santiago, the 424 high-traffic areas with the highest potential of attraction among their neighbours were selected.
Of those 424 areas, 40 were randomly selected, with a probability that was proportional to the potential to attract travels estimated from the OLS regression. The distribution of the 424 high-traffic areas selected (red dots) and the 40 that were finally used to contact smokers (blue stars) are shown in figure 2.
Selected high-traffic areas used to interview smokers.
Based on the age and sex stratification of the sample of the 2014 national survey on drug use,18 a sample of 800 smokers (resulting in an estimated 3.3% sampling error) was distributed among the 40 high-traffic areas selected (the distribution of 20 smokers by point is included as online supplemental material Table A). Smokers were interviewed following a quota sampling, which consisted of defining the number of smokers for each age/sex group that had to be interviewed at each area and conducting interviews until predetermined quotas for each of those groups were filled. In case a specific quota was not met on a specific day, pollster returned to the same point the next day until such a quota was met. Pollster collected surveys mainly at rush hours, either early in the morning or late in the afternoon.
Supplemental material
In terms of age, four groups were defined: 18 years and less; 19–40 years; 41–60 years; and 61 years and more. At the end, 851 interviews were conducted (more than the 800 predetermined to achieve the 3.3% sampling error fully). Due to the fact that after the completion of the field work, a new data for the Metropolitan Santiago from the 2016 national survey on drug use were released2; there was an ex-post weighing of observations to make the completed sample comparable to such a survey.
The field work was conducted by the Centro de Microdatos (Centre for Microdata) of the Universidad de Chile, a nationally respected group specialising in surveys. The questionnaire19 collected information on sociodemographic characteristics (sex, age, municipality of residence, years of education and employment status), consumption habits (age of initiation, consumption intensity, preferred brand, last-purchased brand and so on) and other characteristics, such as the place of last purchase (the purchase for which the interviewee showed the pack), the price paid, the attribute for which the last purchase was chosen (the most important one, between ‘price’, ‘flavor’ or ‘harmfulness’) and so on. Smokers over the age of 13 years were interviewed, as was done in the 2016 national survey on drug use.2
Pollsters approached people that were smoking and declared to be willing to show their pack of cigarettes. Only residents of Metropolitan Santiago were interviewed. Photographs of the front and side of the smoker’s pack of cigarettes were taken, in order to help determine whether they were licit or illicit. A number of photographs (related to 41 individuals) were not clear enough to determine the country of origin of the pack and, thus, those observations were discarded. The final sample consisted of 810 smokers/observations, reweighted to make them representative of the smoking population of Metropolitan Santiago.
Cigarette packs without the national health warning and that were not reported to have been bought overseas or in a duty-free zone/shop for own consumption were determined to be illicit. Duty-free allowances for cigarettes are relatively generous in Chile; travellers 18 years and older can legally enter up to 400 cigarettes per trip, with no limits on the number of trips per day. In some areas of the country (northern border with Peru and sectors of the border with Argentina) authorities have detected that individuals conduct so-called ant-smuggling by crossing the border several times per day to bring in cigarettes.11 If cigarettes are purchased by the interviewed smoker in a duty-free shop for own consumption, they are considered licit. In contrast, cigarettes brought in using allowances but then sold to third persons are considered illicit. The proportion of smokers that reported buying their cigarettes in duty-free shops/zones or abroad was less than 1% of the final sample.
Once the smokers of illicit cigarettes were identified, the statistical associations between the decision to smoke illicit cigarettes and sociodemographic variables and smoking habits were estimated with two probit models with robust SEs (table 1). The models include as covariates sex, employment status (employed or not), preferred attribute in the choice of brand (with ‘less harmful’ as the reference category) and educational level achieved (with complete or incomplete university education as the reference category). These variables have been considered in similar studies in the region.15
Descriptive variables of the collected sample
Model 1 considers the age, age squared and onset age among the covariates, while model 2 considers the number of years as a smoker but not the age (due to the high correlation between the two variables).
Results
Table 1 shows descriptive statistics (means and SD) for the variables considered in the models. Women accounted for 50.2% of the sample. In terms of age, the average age was 39.3; 3.2% were smokers under the age of 18 years; 49% were between 19 years and 40 years; and 42.4% were between 41 years and 60 years. The average age of initiation was 16.9 and the average smoking intensity was 9.2 cigarettes per day. These results are similar to the 2016 national drug use survey in which the average age of smokers in Metropolitan Santiago was 39.7; the age of initiation was 16.3; and about 46% were men2 (see online supplemental material table B).
The proportion of smokers smoking illicit cigarettes in Metropolitan Santiago was 10.9% (95% CI 8.7% to 13%). Incorporating smoking intensity, it is calculated that 16.3% (95% CI 12.8% to 19.8%) of cigarettes smoked are illicit.
Table 2 present the average marginal effects for the probit models (coefficients of the regressions are shown as online supplemental material table C), estimated with Stata V.15, using analytical weights to consider the statistical representativeness of each observation. None of the models shows a significant statistical association between the decision to smoke illicit trade cigarettes and age-related variables (eg, age, squared age, initiation age and/or squared initiation age). Both models show that employed people are less likely to consume illicit cigarettes, while people who prefer price, who smoke with greater intensity and/or have less education show a significantly greater probability of doing so.
Marginal effects of probit models of the decision to smoke illicit cigarettes (reference: smoking licit cigarettes)
For instance, those who express preference for price when choosing cigarettes have a 23% higher probability to choose illicit cigarettes, over those who express preference for ‘less harmful’ cigarettes. The fact that the preference for price has a strong statistical relationship with the decision to smoke illicit cigarettes may signal that illicit cigarettes are considerably less expensive than licit ones. In table 3, the average reported prices per stick of licit and illicit cigarettes are shown, along with a test of the difference of means. Illicit cigarettes are, on average, 61% cheaper than licit ones. The test of the difference of means shows that this difference is significant (at 1%). It also shows that median prices for illicit cigarettes are one-third of the licit ones and that the most-sold brand of illicit cigarettes is Fox (manufactured in Paraguay).
Characteristics of licit versus illicit cigarette prices and brands
Discussion
The proportion of illicit cigarettes identified in this study is higher than the one recently found for Colombia,15 though far below the level reported for Panama16 and Brazil.14 20 We do not have estimates for illicit cigarettes penetration in other areas of the country, but it could be expected that in border areas or those near duty free zones, the penetration could be higher, based on what was found in other Latin American countries, such as Colombia15 and Panama.16 Given that 40.5% of the national population lives in Metropolitan Santiago,21 it is very likely that the national total is close to what was reported here. In any case, estimates reported here are significantly lower than those reported by the TI, even when the TI does not make clear in its narrative if it is reporting prevalence of smokers or market share. If the TI estimates are prevalence of smokers consuming illicit cigarettes, those presented here are 55% lower. If the TI estimates refer to market share of illicit cigarettes, those presented here are 33% lower.
Given that, for Chile, this is the only independent study (ie, non-TI) to measure illicit trade rigorously, we cannot effectively link changes in cigarette real prices to changes in illicit trade, as there are no comparable surveys collected before tax changes (often the main exogenous cause for change in prices). However, a recent study using gap analysis for Chile— identifying the difference between tax-paid sales and overall consumption— found that between 2008 and 2014, a period when cigarette real prices increased by more than 50%, there were no statistically significant changes in the penetration of illicit cigarettes.22 This evidence casts doubts on the constant TI claims that cigarette price increases are directly related to contraband.
According to the econometric analysis presented, the decision to consume illicit cigarettes is positively associated to having lower income (inactive/unemployed and those with lower levels of education are more likely to smoke illicit cigarettes). Though no inference can be made between the association of price and illicit trade, considering the TI’s past and present pricing behaviour is useful for understanding how the TI links tax increases to illicit trade particularly when there is an industry pricing strategy that increases prices. First, notably, the TI increased the real price of cigarettes even before taxes were increased, as shown in the introduction. Second, because the TI pass-through of tax increases has been higher than 1, it casts serious doubt that illicit trade was the concern and instead suggests a strategy of raising prices to increase profits while rhetorically linking the higher prices to the tax increases. According to our own estimates (based on public data on cigarette prices), the pass-through rate for accumulated tax increases from early 2010 to the end of 2017 was 1.12 for the most widely sold cigarette, suggesting that BAT passed on 112% of the tax increases on to the price of its most-sold product. This pass-through was obtained by considering the price and tax structure of the most-sold brand in early 2010 and the same at the end of 2017, assuming the producer gross real revenue per stick is the same at both points in time.
Lastly, it should be mentioned that, while illicit trade is a serious problem that must be addressed by the authorities, such trade has not prevented a sustained decline in the 30-day prevalence of tobacco smoking, which fell from 43.6% in 2004 to 33.4% in 2016.2 The authorities must continue to build vigorously on recent progress in the implementation of FCTC tools such as taxes, in addition to ratifying the FCTC Protocol to Eliminate Illicit Trade in Tobacco Products, which has excellent tools to tackle these challenges, but for which the government has not yet indicated willingness to sign. In spite of this lack of initiative, there has been slow progress to implement a track and trace system (it should be implemented by the end of 2018), which will allow the SII better control of domestic cigarette production for tax collection purposes. The impact of such a policy on illicit trade is not clear but will at least provide transparent information on domestic production by type of product. Far less progress has been achieved on internal coordination of the different domestic official agencies in charge of the control of illicit trade and the coordination with agencies from other countries. The permanently tense diplomatic relationship with, for instance, neighbouring Bolivia (with which Chile has no diplomatic relations) is a barrier for achieving such cooperation. The lack of interest by Paraguay, a major source of cigarette illicit trade in the region, in discussing measures to curb illicit trade is also a significant barrier to advance international cooperation for Chile and for other countries.14
These results have several limitations. First, as explained in the Methodology section, individuals that reported buying individual sticks were not interviewed. In Chile, selling packs with less than 10 cigarettes or individual sticks is forbidden, and the prevalence of this phenomenon is not considered a problem by health authorities (though there is no formal study analysing this issue). The reason for not considering this group is that there would no clear way of determining whether such sticks were licit or illicit, as no inspection of the original pack could be completed. In such a case, the classification of those sticks would rely only on the information reported by the smokers, who would not necessarily know if cigarettes were licit or not (as both types of cigarettes could be sold by single units). It is not clear the bias that ignoring single-cigarette buyers would have on the estimates presented here, as no information exists on this group. It is likely that single sticks are bought by low-income individuals, but it could be possible that these individuals would choose to buy single sticks of a licit brand. In any case, even the TI estimates do not consider this group, as they report using the pack-swap method, which also only consider those buying packs.23
Second, we believe the method used here is the soundest to find smokers (at an affordable cost of data collection) as it has a wide geographical coverage and is representative of the age/sex stratification of smokers. However, because of its characteristic of interviewing smokers on the streets, those that tend to stay at home are under-represented. There is no way of knowing the impact of this bias and, much less, on how to correct it. We do not have any prior information on the smoking patterns of those staying at home and whether they have a higher or lower probability of choosing illicit cigarettes. The alternative sampling strategy, going to homes to find smokers, is far more expensive, difficult and does not assure an unbiased result (eg, minors smoking without the knowledge of their parents would tend not to report their smoking habits).
Third, smoker surveys may be affected by the occasional unwillingness of smokers to show their packs. One way to address this is to instead (or also) pick up discarded packs from the streets. Previous research, however, has demonstrated little statistical difference in the results when both methodologies are used at the same time and in the same place.24
Fourth, given that this study is the first of its kind in Chile, we are not able to link tax increases to illicit trade changes. We do not know if illicit trade has increased or not as a result of tax changes. We emphasise that there is no direct evidence of this happening and we doubt this claim, particularly given that the TI was responsible for significant price increases well before the tax changes. This study will provide an independent baseline to assess the future evolution of the cigarette illicit market, at least in Metropolitan Santiago with the possibility of linking those changes to price/tax movements.
Finally, because of lack of resources, the study was conducted only in the capital city, where 40% of the country population lives. As mentioned above, it is possible that areas closer to the border have higher penetration of illicit cigarettes. However, it is perhaps just as likely that such a penetration at the national level is close to the one reported here because of the similarities in the economic variables that typically condition illicit cigarette smoking, such as per capita gross domestic product, employment, poverty, labour participation rate and educational attainments.
What this paper adds
What is already known on this subject
Tobacco industry estimates of illicit trade are typically overestimated, and they almost always automatically link tobacco tax increases to illicit trade. There are no independent estimates of illicit trade for Chile and few in Latin America to contrast with the highly publicised estimates of the tobacco industry. Furthermore, little is known about what variables are associated with the decision to buy illicit cigarettes in this region.
What this paper adds
This research generates the first independent estimate of illicit cigarette consumption for Chile (for its capital city). Statistical analyses that link the decision to smoke illicit cigarettes and socioeconomic and smoking habit variables are conducted.
Footnotes
Contributors GP designed the questionnaire and the study in general. GP and DA run the models and interpreted and analysed results. All authors discussed findings, interpreted results in terms of policies and wrote the manuscript.
Funding This work was carried out with the aid of a grant from the American Cancer Society (ACS) and a grant from the Global Tobacco Control Leadership Program of the Bloomberg School of Public Health of the Johns Hopkins University. ACS, as an independent research institution publishes widely, funded primarily by small donors.
Competing interests None declared.
Ethics approval The Institutional Review Board clearance was obtained from the Ethics Committee of the Facultad the Economía y Negocios (School of Economics and Business) of the Universidad de Chile, which oversees Centre for Microdata surveys.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Dataset used to conduct the analyses is available from corresponding author.
Collaborators Campos, Natalia; Ruiz-Tagle, Jaime; Quijada, Sandra.
Patient consent for publication Not required.