Introduction

The European Union’s (EU) Tobacco Tax Directive1 through its minimum tax requirements is the major driving force behind cigarette tax increases in the region.2 3 Those tax increases have resulted in significant drops in smoking prevalence in Europe4 and serve to safeguard Member States’ budgetary revenue.5 Nevertheless, large differences in cigarette prices between Member States still exist.6 A revision of the minimum rates is back on the Commission’s agenda and is now being reconsidered,7 making the analysis of the impact of cigarette prices on tax avoidance and evasion timely and germane.

The tax harmonisation mechanisms might be the only way to curb cross-border cigarette purchasing. Since the limits on the number of cigarette packs that can be legally transported across EU borders for personal use are high (40 packs),8 those purchases are most often legal. Therefore, no measures designed to curb illicit cigarette trade, such as law enforcement or the measures listed in the Protocol to Eliminate Illicit Trade in Tobacco Products,9 would apply to cross-border purchases that are done within the legal limits for smokers’ own consumption.

There is some evidence suggesting that cross-border cigarette purchasing is affected by between-country differences in cigarette prices, with lower cigarette prices in the neighbouring jurisdiction10–12 and larger price difference between jurisdictions13–17 being significantly associated with more cross-border cigarette purchasing. Proximity provides opportunity for legal cross-border purchases and facilitates illegal activities such as reselling legally purchased cross-border cigarettes, while between-country price difference is an incentive for such activities. However, there is a dearth of studies quantifying both the magnitude and the nature of this effect.

There are ever more studies measuring tobacco tax avoidance and evasion that are independent from the tobacco industry.18 In the EU, such studies involve analysing the gap between survey-reported cigarette consumption and tax-paid sales,19 examining packs presented by smokers,20 21 and collecting littered packs.22 These studies, however, do not distinguish between the large-scale illicit trade and the cross-border purchasing. Only a few studies use European data to focus on cross-border shopping specifically. A 2014 study by Nagelhout and colleagues10 finds that cross-border purchasing might be more frequent in regions bordering countries with lower cigarette prices, but that the overall magnitude of the problem is not large. A more recent study by Agaku et al 13 finds that, although the effects of between-country price differences on cross-border cigarette purchasing are significant, only 4% of all EU smokers purchased most of their products via cross-border transitions. However, both studies use survey-reported data, which might be affected by under-reporting of tax avoidance activities. Moreover, because these studies asked about cross-border purchasing by smokers themselves, they do not capture instances when the cigarettes were bought abroad by other individuals, such as a smoker’s family member or a friend. Finally, both studies use a single cross-section of data, and therefore a causal relationship cannot be established.

This study aims to shed light on the effects of cigarette price differences between EU Member States on cross-border cigarette purchases. Additional evidence on this relationship is needed to guide tobacco tax policies in Europe. This study verifies findings by Agaku and colleagues13 using methods that are free from potential reporting bias attached to analysis of survey data.

Data and methods

Primary independent variables

The primary independent variables are the population-weighted between-country differences in cigarette prices. Those variables are constructed following the methods developed by Becker et al.14 The import incentive variable for cross-border cigarette purchasing is defined as follows:

where K_ij is the fraction of the population of the higher price, importing country i living in the border regions neighbouring the lower price country j (K_ij=Border_POP_ij/POP_i). Weight K is included in the formula to reflect the fact that the greater the proportion of the people living near the border with the lower price country, the more likely cross-border cigarette purchases from that country are. The sum is taken over all lower price neighbouring countries. The export incentive variable is defined as follows:

where K_ji is, again, the fraction of the population of the higher price country j living in the border regions neighbouring the lower price, exporting country i. POP_j and POP_i represent the total populations of the high-price and low-price countries, respectively. This relative-population weight is included in the formula to reflect the fact that exporting cigarettes to a larger neighbour is more likely to affect per capita sales in the lower price country, than exporting to a smaller neighbour is, simply because more cigarettes are expected to be exported to a larger neighbour than to a smaller neighbour. The sum is taken over all higher price neighbouring countries.

Cigarette prices used in this study are the weighted average prices (WAP) expressed in euro. Before 2010, the EU Member States did not report WAP to the European Commission and reported the most popular price category (MPPC) instead. However, the difference between MPPC and WAP is very small (4% on average among those countries which continued reporting both WAP and MPPC).6 Therefore, MPPC was used as a proxy for WAP for the EU Member States for years before 2010. For non-EU countries, WAP was calculated using the formula from the 2010 EU Tax Directive,29 that is by dividing the total cigarette value by the total quantity of cigarettes reported by Euromonitor.26

The data set of the EU border regions, obtained from the European Commission, consisted of 973 level 3 geographical regions from the Nomenclature of Territorial Units for Statistics classification (NUTS3 regions).30 The data set covers both regions along terrestrial borders as well as maritime borders. Maritime borders are defined as coastlines located at less than 150 km from overseas coastlines (see figure 1). Additional to the border regions in the EU Member States, the data set includes border regions in several non-EU countries that border EU countries (Liechtenstein, Macedonia, Norway, Switzerland and Turkey). In the data set, there are also 2011 estimates of each region’s border population, defined as the number of people living within 25 km from the border.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Border NUTS3 regions in the European Union. Although regions vary in size, the weights used in this study are based only on the estimated number of individuals living within 25 km from each border. Source: Author’s own map created based on the European Commission data. NUTS3, Nomenclature of Territorial Units for Statistics classification.

To construct the weights for the import and export incentive variables, I first dropped overseas regions of Spain (Fuerteventura and Lanzarote) and France (Guadeloupe, Martinique, Guyane, La Réunion and Mayotte) that were included in the data set, but that are not relevant for the analysis. The 2011 estimates of border populations for the remaining European regions were then added up along each border and, together with 2011 estimates of the total countries’ populations, were used to construct the weights. These computed weights and 2004–2017 cigarette prices were then used to compute the import and export incentive variables for each country and each year, according to the formulas described above. All import and export variables were adjusted for eurozone inflation.

The data set of the EU border regions includes border population estimates in several EU neighbouring countries with relatively high cigarette prices (Lichtenstein, Norway and Switzerland). This permits the computation of import and export variables covering internal and external EU borders. For three cases when the border population data for a non-EU member, necessary to construct the variables, were not available (all three for exports to Serbia), it was assumed that there was no between-country price difference in those cases.

Subsequently, I constructed import and export incentive variables covering internal and external EU borders, both terrestrial and maritime. I also constructed separate variables for EU internal versus EU external borders and for terrestrial versus maritime borders. These additional variables permit analyses that are more granular and advanced than any of the previous studies employing a similar methodology.14–17

Another way of looking at the import and export incentives is through the lenses of transaction costs. With this approach, the decision to purchase cigarettes across the border is influenced by the cost of the trip. Therefore, in another approach, I used gasoline prices as a proxy of the transaction costs of cross-border cigarette purchasing. Specifically, I weighted the cross-border price differences by gasoline prices rather than by border population. The methods to construct the gasoline price-weighted incentive variables as well as the results from the models using those incentive variables are presented in online supplementary appendix table S1.

Supplementary file 1

[tobaccocontrol-2018-054678supp001.pdf]

Results

Table 1 summarises the key variables. All extremely high values of per capita cigarette sales (above 4000 cigarettes per adult annually) are for Luxemburg. The results from the models capturing the effects of between-country price difference on cross-border cigarette purchases are presented in table 2. Cigarette price significantly impacts cigarette sales in the EU. The estimated price elasticity varies, depending on the model specification (from −0.47 to −0.35). The estimated income elasticity is positive and significant in all models (from 0.66 to 0.70). The estimated effects of non-price tobacco control policies have all the expected sign and reach significance at the 0.05 level in model 1 and at the 0.10 level in models 2 and 3. The coefficients for both import and export incentive variables have the expected sign in all models, except for the export incentive across maritime borders (2b) in model 2. In model 1, the overall effect of price differences on cross-border shopping is not significant. When the effects of between-country price differences are estimated separately for maritime and terrestrial borders, the import and export incentive variables are significant for terrestrial borders (joint test: F=9.25, p<0.01), but not for maritime borders (joint test: F=0.25, p=0.78). After further disentangling the effects of those incentives into separate effects for EU internal terrestrial borders and EU external terrestrial borders, the incentives for EU countries to trade within the EU (across internal borders) are jointly significant (joint test: F=8.19, p<0.01), while the incentives for EU countries to trade with non-EU countries (across external borders) are not (joint test: F=1.56, p=0.21). When the model is only estimated on countries with external EU borders (n=251), the incentive variables for external borders are still not significant (F<0.01, p=0.99).

Models where, to reflect the transaction costs, the EU cross-border price differences were weighted by gasoline prices instead of by border population yielded similar results. The import and export incentive variables had the expected sign, although the import incentives were no longer significant in the models (online supplementary appendix table S1).

Discussion

This study shows that in an average EU Member State only a small portion of cigarette sales can be explained by cross-border purchasing. When taking the value of the estimated coefficient for imports across EU internal terrestrial borders (coefficient: −0.38 from model 3), the value of that coefficient indicates that in a theoretical scenario when incentives for cross-border imports fall from its mean level to zero, ceteris paribus, sales would increase by about 1.5% (0.38×0.04). At the same time, reducing incentives from cross-border exports through EU internal terrestrial borders down to zero, ceteris paribus, would reduce sales by about 6% (0.27×0.23). These results were similar (from 4% to 6%) in the models where the EU cross-border price differences were weighted by gasoline prices instead of by border population (see online supplementary appendix table S1). In reality, reducing the between-country price differences would require an adjustment in domestic price as well. Therefore, the equilibrium for tax-paid sales given no cross-border price differences is harder to predict. However, some useful inferences can still be made based on the value and ratio of those effects. The small value of these two effects suggests that the part of variation in tax-paid sales that can be explained by cross-border purchasing incentives is not large. This is consistent with findings from studies by Nagelhout et al 10 and Agaku et al,13 which both find that in most EU countries only a small portion of all smokers reported frequent cross-border cigarette purchasing.

The fact that the effects of the incentive variables are larger for the export (0.27×0.23) than for the import (0.38×0.04) incentive suggests that cross-border cigarette purchasing has a larger effect on sales in the exporting country than in the importing country. A similar conclusion can be made from the fact that, in different model specifications, export incentive variables are more often significant than import incentive variables. This could mean that some of the cigarettes purchased from across the border are smoked in addition to domestic duty-paid cigarettes, as opposed to being substitutes for domestic cigarettes. The overall cigarette consumption would be lower, if not for the availability of the lower-priced cigarettes from across the border. Thus, increasing cigarette taxes to reduce between-country price differences would curb tobacco use in two ways: first, through the effects of price increases on cigarette consumption alone, and second through reduced cross-border purchasing.

Another finding of this study is that patterns in cross-border cigarette purchasing differ by border type. Even though incentives to purchase cigarettes from across maritime borders were higher than incentives for terrestrial borders (table 1: 2a vs 2b), between-country price differences were not significantly associated with purchases across maritime borders. Perhaps more interestingly, between-country difference in cigarette prices did not affect cigarette purchasing across EU external, terrestrial borders. This is likely because of stricter restrictions associated with crossing EU external borders, both in terms of required travel documents as well as the number of cigarette packs that can be legally transported across those borders. This finding suggests that further increases in EU cigarette prices, which increase the price differences with non-EU neighbours, are not expected to significantly impact cross-border purchases from outside of the EU. Thus, the larger price differences with non-EU countries should not distort the functioning of the EU internal market.

The findings from this study consistently indicate that cigarette prices affect cigarette sales. The estimated price elasticity varies from −0.47 to −0.35, depending on the model specification. These findings are consistent with other studies from the EU.4 34 The estimated income elasticity ranging from 0.66 to 0.70 is also within the expected range.35 These findings confirm that higher tobacco taxes that lead to higher cigarette prices in the EU will significantly reduce demand for cigarettes in the EU. Because higher income is associated with higher cigarette sales, future tax increases need to account for the effects of growing income, to make cigarettes less affordable over time.36

This study also contributes to research aiming to advance methods to estimate tobacco tax avoidance and evasion. First, creating separate import and export incentive variables by border type (internal vs external borders and terrestrial vs maritime borders) was novel. It allowed the distinction between the effects of between-country price differences on cross-border cigarette purchasing within the EU and the effect on cross-border trading with the non-EU, neighbouring countries. This information can guide policymakers designing policies to curb cross-border purchasing. Additionally, the fact that findings from this study are consistent with findings from studies based on survey data10 13 suggests that surveying smokers about their cross-border purchasing is not affected by the under-reporting bias. Therefore, methods relying on official data of tax-paid sales and prices, such as methods employed in this study, as well as the methods relying on surveys of smokers, might both be adequate for measuring the impact of between-country price differences on cross-border purchases. These two methods can be used independently, depending on the availability of data. Finally, the longitudinal set-up of this study permits stronger conclusions about the causal relationships between cigarette price differences and cross-border cigarette purchases.

Some limitations exist in this study. First, although the number of cigarettes released for consumption should closely reflect the tax-paid sales in the EU Members, it is not an exact measure of sales, especially in countries without antistockpiling rules in place. Second, the fact that the EU switched from reporting MPPC to reporting WAP affected the quality of the price measure used in this study. However, this switch should be picked up by the models’ year fixed effects. Third, the fact that the price data used for the non-EU, neighbouring countries come from an unofficial commercial source (Euromonitor) might have affected the quality of the price measure used for those non-EU countries. Fourth, the estimates of border populations for years other than 2011 were not available. Therefore, any change in border population throughout the period of this analysis might have affected the accuracy of the incentive variables. However, weighting the cross-border price differences by yearly gasoline prices instead by the 2011 border population did not change the results substantially. This suggests that the results are robust to the type of weight used for the cross-border price difference. Fifth, there is no EU-wide, longitudinal database for annual data on non-price tobacco control measures, such as smoking bans or advertising bans. Consequently, values for these other tobacco measures were imputed for every other year from 2008 to 2017. The small potential inaccuracy resulting from this imputation might be contributing to the non-price tobacco control variable reaching significance only at the 0.10 level in models 2 and 3. Finally, uneven enforcement of the limits on the number of cigarette packs that can be legally transported across EU borders could affect the findings. However, despite the shortcomings, the model produces results that are robust, with the estimated effects having the expected signs.

Conclusion

This study finds that differences in cigarette prices across the Member States significantly affect cigarette purchasing across terrestrial, internal EU borders. No such effect was found for cigarette purchasing across maritime borders and for cross-border cigarette trading with countries from outside the EU. These findings underscore the need for further efforts to harmonise taxation within the EU, such as through increasing minimum tobacco tax rates required in the Tobacco Tax Directive. Increases in minimum tobacco tax rates resulting in higher cigarette prices and in convergence of those prices across the Member States would reduce cross-border cigarette purchasing, ensuring proper functioning of the internal EU cigarette market and improving public health by contributing to decreases in cigarette consumption.