To convert euros to US dollars, multiply by 1.1787.
eTable 1. Median GDP per Capita, Score of the Smokefree Component of the Tobacco Control Scale, Unemployment, Education, and Births by Mothers at High Risk in NUTS-2 Regions of 23 EU Countries, 2004-2014
eTable 2. Infant Mortality Rate (per 1,000 Live Births) in 23 EU Countries, 2004-2014
eTable 3. Minimum Cigarette Prices per Pack In 23 EU Countries, 2004-2014
eTable 4. Median Cigarette Prices per Pack in 23 EU Countries, 2004-2014
eTable 5. Price Differential Between Median and Minimum Cigarette Prices (% of Median Price) in 23 EU Countries, 2004-2014
eTable 6. Sensitivity Analysis: Association of Infant Mortality (Infant Deaths per 1,000 Live Births) With Median Cigarette Price and Price Differential Between Median and Minimum Cigarette Price, in 23 EU Countries (950 Region-Years)
eFigure 1. Tax Increases and Minimum and Median Cigarette Prices in the United Kingdom, 2004-2014
eFigure 2. Potential Causal Pathways Between Tobacco Prices and Infant Mortality
Customize your JAMA Network experience by selecting one or more topics from the list below.
Filippidis FT, Laverty AA, Hone T, Been JV, Millett C. Association of Cigarette Price Differentials With Infant Mortality in 23 European Union Countries. JAMA Pediatr. 2017;171(11):1100–1106. doi:10.1001/jamapediatrics.2017.2536
Are differentials in cigarette prices associated with increased infant mortality?
In this longitudinal, ecological study, increases in the median price of cigarettes were associated with reductions in infant mortality across Europe between 2004 and 2014. However, pricing differentials between median and minimum cigarette prices were associated with increases in infant mortality.
Pricing differentials between cigarette products with widespread availability of budget cigarettes are likely to produce adverse child health outcomes; legislators should implement measures to eliminate budget cigarettes globally.
Raising the price of cigarettes by increasing taxation has been associated with improved perinatal and child health outcomes. Transnational tobacco companies have sought to undermine tobacco tax policy by adopting pricing strategies that maintain the availability of budget cigarettes.
To assess associations between median cigarette prices, cigarette price differentials, and infant mortality across the European Union.
Design, Setting, and Participants
A longitudinal, ecological study was conducted from January 1, 2004, to December 31, 2014, of infant populations in 23 countries (comprising 276 subnational regions) within the European Union.
Median cigarette prices and the differential between these and minimum cigarette prices were obtained from Euromonitor International. Pricing differentials were calculated as the proportions (%) obtained by dividing the difference between median and minimum cigarette price by median price. Prices were adjusted for inflation.
Main Outcomes and Measures
Annual infant mortality rates. Associations were assessed using linear fixed-effect panel regression models adjusted for smoke-free policies, gross domestic product, unemployment rate, education, maternal age, and underlining temporal trends.
Among the 53 704 641 live births during the study period, an increase of €1 (US $1.18) per pack in the median cigarette price was associated with a decline of 0.23 deaths per 1000 live births in the same year (95% CI, –0.37 to –0.09) and a decline of 0.16 deaths per 1000 live births the following year (95% CI, –0.30 to –0.03). An increase of 10% in the price differential between median-priced and minimum-priced cigarettes was associated with an increase of 0.07 deaths per 1000 live births (95% CI, 0.01-0.13) the following year. Cigarette price increases across 23 European countries between 2004 and 2014 were associated with 9208 (95% CI, 8601-9814) fewer infant deaths; 3195 (95% CI, 3017-3372) infant deaths could have been avoided had there been no cost differential between the median-priced and minimum-priced cigarettes during this period.
Conclusions and Relevance
Higher cigarette prices were associated with reduced infant mortality, while increased cigarette price differentials were associated with higher infant mortality in the European Union. Combined with other evidence, this research suggests that legislators should implement tobacco tax and price control measures that eliminate budget cigarettes.
Increasing cigarette prices through increased taxation has proven to be the single most effective tobacco control measure.1 Previous studies have documented substantial health benefits of tobacco taxation derived from associated reductions in smoking prevalence, cigarette consumption among smokers, and exposure to second-hand smoke (SHS),2 including reductions in infant mortality3,4 most likely mediated by decreases in maternal smoking and fetal and infant exposure to SHS.5,6
Transnational tobacco companies have responded to increased tobacco taxation with differential pricing strategies, whereby tax increases are loaded onto premium brands (eFigure 1 in the Supplement).7 The resulting price gap between premium and budget cigarettes attenuates the effectiveness of tax increases in reducing smoking prevalence, as smokers can switch to less-expensive products. The European Union (EU) has adopted a mixed tax system for tobacco products, including a minimum excise tax burden and an excise tax floor.8 As a result, overall prices have increased in both high-income and middle-income countries.9 Cigarettes have become less affordable in the EU since 2003, although this is not uniform across countries,10,11 as member states can impose excise taxes exceeding minimum EU requirements and may have different value added tax rates. There is still a considerable gap between prices for premium and budget brand cigarettes across the EU.
Research on the effect of price differentials in cigarette markets on tobacco-related health outcomes is sparse, which constitutes an important knowledge gap given transnational tobacco company strategies to load tax increases onto premium cigarettes.7 Tobacco pricing strategies are also important for tobacco control in low- and middle-income countries (LMICs) where cigarette price differentials are generally larger than in high-income countries.1 We used data from 23 EU countries to examine associations between median cigarette prices, price differentials between cigarette brands, and infant mortality. We hypothesized that increased price differentials in EU cigarette markets would be associated with higher rates of infant mortality.
We conducted a longitudinal ecological study using region-level data from 23 EU countries to assess associations between median cigarette prices, cigarette price differentials, and infant mortality. Because this study used routinely collected anonymized data aggregated to regional and national levels, ethical approval was not required.
We extracted region-level data from Eurostat,12 the official EU statistical authority. The EU, which has 28 member countries, is further divided into 3 levels of Nomenclature of Territorial Units for Statistics (NUTS) regions, including 276 subnational (NUTS-2) regions. We extracted annual data on rates of infant mortality (infant deaths under 1 year of age per 1000 live births), educational level (percentage of the population aged 25-64 years with tertiary education), current market gross domestic product per inhabitant, and unemployment rate (among the population aged >15 years) at a NUTS-2 level.12 We also extracted annual data on maternal age at delivery and calculated the proportion of mothers younger than 18 years or 40 years of age or older for each NUTS-2 region, as these age groups are known to be associated with higher infant mortality.13 Data were available for all countries from January 1, 2004, to December 31, 2014.
Euromonitor International14 collects the prices of tobacco products in EU countries (except for Croatia, Malta, Cyprus, Luxembourg, and Austria). They record the prices of multiple cigarette products in retail outlets every year, covering at least the top 10 brands with the highest market share in each country. We obtained annual price data for 23 EU countries for 2004 and 2006-2014. Data for 2005 were not available, as Euromonitor International was not collecting data annually before 2006. A median of 93 (interquartile range, 58-184) cigarette products were sampled in each country each year. Prices were recorded in the national currency; for countries not using the euro, prices were transformed into Euros using the exchange rate on June 30 of the respective year. We calculated minimum and median prices per 20 cigarettes (1 pack) for each country and year. Data for 2005 were imputed with linear interpolation using data from 2004 and 2006. To assess the validity of this approach, we used existing data points (ie, for other years) and found that interpolation accurately estimated price data in years for which price data were available (median difference between actual and interpolated prices was 1.1% for median and 0.2% for minimum cigarette prices). To capture differential pricing of cigarette products that may lead smokers to switch from expensive to cheaper products, we calculated the price differential between the minimum and the median cigarette price. We expressed this as a percentage of the median cigarette price for every year and country.
Smoke-free legislation has been associated with reduced infant mortality.15 We used scores from the Smokefree Work and Other Public Places subscale of the Tobacco Control Scale (TCS)16-18 to adjust for smoke-free policies within each country. The highest possible score in this subscale is 22. Country-level TCS scores have been published for 2005, 2007, 2010, and 2013.16-18 We also collected data from official reports of the European Commission on smoke-free legislation implemented in EU countries during the study period.19,20 Based on legislation and scoring criteria of the TCS,18 we estimated TCS smoke-free scores for years when official TCS scores were not available. We created a binary variable (comprehensive bans vs noncomprehensive bans) from the TCS smoke-free score using a cutoff of 18 points, as substantial health benefits and reductions in SHS exposure are generally achievable through comprehensive smoke-free legislation,21 but we also conducted a sensitivity analysis using TCS smoke-free score as a continuous variable.
The main outcome of our analysis was the rate of infant mortality (infant deaths per 1000 live births). Consistent with previous research on this topic,3,4 we fitted a linear fixed-effects panel regression model with robust standard errors. Panel regression models are widely used for exploiting data routinely collected over time across multiple units. In our study, the unit of analysis was NUTS-2 region with annual observations for the period 2004-2014. Panel regressions account for the clustered nature of the observations (ie, annual observations within region).22 We used a fixed-effects specification to control for time-invariant region-level unobserved factors that could be associated with independent variables. Fixed-effects panel regressions estimate associated changes between independent and outcome variables within regions over time, but do not estimate between-region associations. The choice of a fixed-effects specification was supported by the Hausman specification test.22
We have drawn a detailed directed acyclic graph with DAGitty23 to illustrate potential causal pathways from cigarette prices to infant mortality, identify potential sources of confounding, and guide covariate selection (eFigure 2 in the Supplement). Model specification was determined by comparing Bayesian information criterion and Akaike information criterion and considering availability and completeness of data at the NUTS-2 level. Some health outcomes of reduced cigarette consumption and exposure to SHS during pregnancy may only be detected the following year (owing to 9 months’ gestation). Therefore, our final model included within-year and 1-year lagged terms for the median cigarette price and the price differential, adjusting for potential confounders (gross domestic product per capita, unemployment, TCS smoke-free score, educational level, and maternal age), and time (quadratic term) to account for underlying nonlinear time trends. Mediators of the association between cigarette prices and infant mortality, such as population-level cigarette consumption, smoking in pregnancy, exposure to SHS in pregnancy and infancy, congenital anomalies, and preterm birth were not included in the model, consistent with our directed acyclic graph.24,25
Gross domestic product per capita and tobacco prices were adjusted for inflation using the Harmonised Index of Consumer Prices.12 A total of 84 of 2527 potential data points (3.3%) with missing values of infant mortality (60 [2.4%]), unemployment (75 [3.0%]), education (56 [2.2%]) or maternal age (75 [3.0%]) were excluded from the analysis. We performed a sensitivity analysis using values of the smoke-free component of TCS only for years when the TCS was published (2005, 2007, 2010, and 2013). Results are presented as β coefficients with robust 95% CIs and can be interpreted as changes in rates of infant mortality (infant deaths per 1000 live births) in the region per 1-unit increase of the independent variable. The variable for price differential was scaled to show β coefficients for a 10% increase in price differential.
We used the “predict” command in Stata (StataCorp LLC) and the coefficients from our main model to estimate the expected number of infant deaths by region and year in the following counterfactual scenarios: if median cigarette prices had remained at 2004 levels throughout the study period, and if the price differential between median and budget cigarettes was fixed at zero for the entire study period. We used the same approach to calculate the number of deaths expected with our model using actual observed values of prices and price differentials. We subsequently compared the 2 scenarios with our model’s expected number of deaths made on observed values to estimate the averted number of infant deaths (with 95% CI) associated with increases in median cigarette price during the study period and the increase in the number of infant deaths associated with cigarette price differentials. Both estimates accounted for both within-year and lagged associations and refer to the period 2005-2014.
There were a total of 53 704 641 live births recorded in the 276 NUTS-2 regions with complete data between 2004 and 2014, representing 96% of all live births across the 23 EU countries. Descriptive characteristics are shown in eTable 1 in the Supplement.12 Infant mortality declined in all countries during the study period, with reductions up to 50% in countries that joined the EU after 2003 (Figure 1). The median infant mortality rate was 4.4 deaths per 1000 live births in 2004 (range, 3.1 in Sweden to 16.8 in Romania) and 3.5 per 1000 live births in 2014 (range, 1.8 in Slovenia to 8.4 in Romania) (eTable 2 in the Supplement).
Both median and minimum cigarette prices (adjusted for inflation) increased in all 23 countries during the study period (Figure 2). Ireland and the United Kingdom consistently had the highest minimum and median prices. The lowest cigarette prices were in countries that became EU members after 2003. Overall, minimum prices for a pack of 20 cigarettes in 2004 ranged from €0.42 (US $0.49) in Bulgaria and Romania to €6.32 (US $7.45) in Ireland (median minimum price across the 23 countries, €2.16 [US $2.55]) and in 2014 ranged from €2.09 (US $2.46) in Bulgaria to €8.32 (US $9.81) in Ireland (median minimum price, €3.60 [US $4.24]) (eTable 3 in the Supplement). Similarly, median prices per pack in 2004 ranged from €0.93 (US $1.10) in Bulgaria to €8.14 (US $9.59) per pack in the UK (median, €2.68 [US $3.16]) and in 2014 ranged from €2.40 (US $2.83) in Bulgaria to €10.00 (US $11.79) in Ireland (median, €4.10 [US $4.83]) (eTable 4 in the Supplement). The relative price differential decreased over time, except for Ireland and Portugal, where it was higher in 2014 compared with 2004. The median price differential was 24.6% in 2004 (range, 5.0% in France to 65.5% in Romania), meaning that the least-expensive cigarettes were 24.6% cheaper than the median-priced cigarettes. In 2014, the median differential was 12.8% (range, 0.5% in France to 25.4% in Poland) (eTable 5 in the Supplement).
In the multivariable linear fixed-effects panel regression model, an increase of €1 per pack in the median cigarette price was associated with 0.23 fewer deaths per 1000 live births (95% CI, –0.37 to –0.09) in the same year and an additional 0.16 fewer deaths per 1000 live births in the following year (95% CI, –0.30 to –0.03) (Table). An increase of 10% in the price differential between median and minimum cigarette price was associated with 0.07 more deaths per 1000 live births (95% CI, 0.01-0.13) the following year, while the association was nonsignificant in the same year (β, –0.04; 95% CI, –0.08 to 0.01). Within-region changes in educational level, smoke-free legislation, and macroeconomic factors were not associated with the rate of infant mortality rate in the same year. Maternal age younger than 18 years or 40 years or older was associated with a higher infant mortality rate (Table). Results from the sensitivity analyses were largely consistent with our main findings (eTable 6 in the Supplement).
In comparison with a counterfactual scenario in which cigarette prices remained unchanged since 2004, actual median price increases for cigarettes were associated with 9208 (95% CI, 8601-9814) fewer infant deaths between 2005 and 2014. In a counterfactual scenario in which there was no price differential between median and minimum cigarette price, we estimated that a further 3195 (95% CI, 3017-3372) infant deaths would have been averted.
We analyzed regional data from 23 EU countries between 2004 and 2014 and found that increases in cigarette prices were associated with lower rates of infant mortality. Larger differences between median and minimum cigarette prices were associated with increased rates of infant mortality. Median price increases during the study period were associated with 9208 fewer infant deaths, but a further 3195 infant deaths could have been averted if no price difference between minimum-priced and median-priced cigarettes existed.
To our knowledge, no previous studies have examined the association of price differentials in tobacco markets with health outcomes. A previous study in the United States found that a cigarette tax increase of $1 per pack was associated with 0.19 fewer deaths per 1000 live births.3 Canadian data showed that a 10% increase in tobacco taxes was associated with a 1.8% decline in infant mortality.4 Both estimates are consistent with our results. The association between tobacco prices and lower infant mortality is likely mediated by decreases in maternal smoking and fetal and infant exposure to SHS26 (eFigure 2 in the Supplement). Maternal smoking,5,27,28 especially among socially disadvantaged groups,29 as well as exposure to SHS decline significantly following increases in cigarette prices.3-5,30
We identified a lagged association of cigarette price increases and price differentials and infant mortality. This finding is plausible considering the delay between maternal smoking during gestation—when harm may be inflicted—and delivery dates.15,28,29 It may also reflect lagged associations of price changes with smoking behavior. Although previous work has linked increases in tobacco price to smoking cessation31 and reduced smoking rates overall,32 changes in smoking behaviors may not be immediate. Comprehensive smoke-free legislation has previously been associated with lower infant mortality in England.15 However, our analysis showed no association at the EU level. The TCS smoke-free score—as a continuous variable—was also not associated with infant mortality. We used annual data, which may not have fully captured outcomes of changes in legislation implemented during the calendar year, such as the introduction of comprehensive smoke-free legislation in England banning smoking in virtually all indoor public places.
We analyzed EU data from a period during which tobacco taxation and prices increased in all EU countries. Countries that were members of the EU before 2004 implemented smaller tax increases, but still had higher taxes compared with the newest members by 2010.10 Prices33 and affordability of cigarettes continue to vary greatly among EU countries11 and free movement across borders within the EU allows some smokers to exploit price differences between countries to obtain cheaper cigarettes.33,34 Although our study could not measure this behavior, it highlights the need for further harmonization of taxation strategies to strengthen individual countries’ tobacco control measures.
Reliance on heavy excise taxes (which the World Health Organization recommends should account for >70% of the retail price),1 annual price increases above inflation, and strengthening mechanisms of tobacco tax administration maximize the health benefits of tobacco taxation.2 This robust body of evidence can motivate governments to increase taxes. We could not explore the effect of price differences between cigarettes and other tobacco products, such as loose tobacco, which has been subject to lower taxation in the EU.35 However, our findings underscore the additional importance of addressing cigarette price differentials through taxation for population health. This is an important and topical issue within Europe, as a revision of the EU tobacco taxation directive is currently under way.36 Tax measures introduced in the UK that explicitly target budget cigarettes appear to have been effective at reducing the differential in cigarette prices (eFigure 1 in the Supplement).37 In 2016, the UK government additionally signaled an intention to introduce a minimum excise tax,38 which will help establish an effective minimum cigarette price, recognizing that smokers commonly switch to less expensive types of tobacco.39
Our findings highlight the potential benefits of increasing taxation and reducing cigarette price differentials on child health outcomes in Europe. Although high-income countries can further improve child health outcomes through strengthening tobacco control, our findings are particularly important for LMICs, where rates of infant mortality remain unacceptably high and cigarette price differentials are generally greater than in high-income countries. Although maternal smoking is often low in LMICs and the principal harm is derived from exposure to SHS during pregnancy, this scenario is likely to change as the tobacco industry is increasingly targeting young women in these settings.40-42 Reduced reliance on ad valorem taxes in the EU means that tobacco prices are generally much higher than in LMICs, where transnational tobacco companies may have greater flexibility in adopting differential pricing strategies.1,2,43 A recent modeling analysis in China, the largest tobacco consumer in the world, found that an increase of tobacco taxes by 50% could achieve substantial health and financial gains, particularly among the poor.44 Research and policy recommendations specifically targeting price differentials are scarce. Bespoke taxation structures that minimize the availability of budget cigarettes and other tobacco products may be required to effectively counteract potential transnational tobacco company pricing strategies. Further research on the effects of price differentials between cigarettes and other tobacco products, such as loose tobacco, and the effect of cross-country price differences is also warranted.
To our knowledge, this is the first study to explore the association between cigarette price differentials and infant mortality. We conducted an ecological analysis precluding individual level inference. Although we used fixed-effects models and adjusted for key covariates, the use of NUTS-2 regions as our unit of analysis means that we cannot rule out the possibility of residual confounding and that ecological fallacy may be an issue. Including the price differential, rather than minimum price, in our model makes our findings applicable and easier to interpret in settings with different levels of cigarette prices; it is a better marker of opportunities for smokers to switch to less expensive cigarettes, which is especially relevant in the context of rising prices. Sociodemographic or infant mortality data were missing for some regions and years, and price data were unavailable for 5 EU countries, but we included 96% of total live births in the 23 countries assessed in our analysis. No data on the prevalence of maternal smoking or exposure to SHS were available at the NUTS-2 level, which would have enabled a fuller understanding of the likely causal pathways.
Higher median cigarette prices were associated with reduced infant mortality, while cigarette price differentials were associated with higher infant mortality in the EU. Combined with other evidence, this research suggests that legislators should implement tobacco tax and price control measures that eliminate budget cigarettes.
Corresponding Author: Filippos T. Filippidis, PhD, Public Health Policy Evaluation Unit, School of Public Health, Imperial College London, St Dunstan’s Road, 310 Reynolds Bldg, London W6 8RP, England (email@example.com).
Accepted for Publication: June 20, 2017.
Published Online: September 18, 2017. doi:10.1001/jamapediatrics.2017.2536
Author Contributions: Dr Filippidis had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Filippidis, Laverty, Been, Millett.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Filippidis, Laverty, Millett.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Filippidis, Laverty, Hone, Millett.
Obtained funding: Been, Millett.
Administrative, technical, or material support: Hone.
Study supervision: Laverty, Been, Millett.
Conflict of Interest Disclosures: None reported.
Funding/Support: Drs Millett and Laverty are funded through Research Professorship award NIHR RP_2014-04-032 from the National Institute for Health Research. Dr Been is supported by fellowship grants from the Erasmus University Medical Centre and grant 4.2.14.063JO from the Netherlands Lung Foundation.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: