ESTIMATING THE DISTRIBUTIONAL IMPACT OF INCREASING TAXES ON TOBACCO PRODUCTS IN ARMENIA ODELING THE LON Results from an extended cost-effectiveness analysis ERM HEALTH AND OST IMPACTS OF EDUCING SMOKING REVALENCE Iryna Postolovska Rouselle Lavado HROUGH TOBACCO Gillian Tarr Stéphane Verguet AXATION IN 4 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia Contents Acknowledgments 6 Abstract 7 Introduction 8 Methods 11 • Modeling approach 11 • Stakeholder interviews 18 Results 19 • Distributional impact 19 • Stakeholder interviews 20 Discussion 23 References 26 Supplementary appendix 28 Figures Figure 1 12 Conceptual framework for modeling the health and financial impact of higher excise taxes on tobacco products among male smokers in Armenia 5 Acknowledgments Iryna Postolovska was a consultant to the World Bank for this study. We appreciate the assistance of Samvel Kharazyan and Arpine Azaryan in arranging the stakeholder interviews, and are grateful to all interview respondents for their participation in this study. We thank Volkan Çetinkaya, Alan Fuchs, Moritz Meyer, and Patricio Marquez for their helpful comments and suggestions on an earlier version of this manuscript. Support for the preparation of this report was provided by the World Bank’s Global Tobacco Program, co-financed by the Bill and Melinda Gates Foundation and the Bloomberg Foundation. The findings, interpretations and conclusions in this research note are entirely those of the authors. They do not necessarily represent the view of the World Bank Group, its Executive Directors, or the countries they represent. Yerevan, Armenia, Washington, D.C., and Boston, United States April 2017 Authors’ affiliations: Iryna Postolovska and Stéphane Verguet, Harvard T.H. Chan School of Public Health, Department of Global Health and Population, Boston, MA, USA; Rouselle Lavado, The World Bank, Washington, DC, USA; Gillian Tarr, Department of Epidemiology, University of Washington, Seattle, WA, USA Correspondence to: Stéphane Verguet Department of Global Health and Population Harvard T.H. Chan School of Public Health 665 Huntington Avenue Boston, MA 02115, USA Email: verguet@hsph.harvard.edu Recommended citation: Postolovska I., Lavado, R., Tarr, G., Verguet, S. 2017. Estimating the distributional impact of increasing tobacco taxes in Armenia: Results from an extended cost-effectiveness analysis. Washington, DC: The World Bank 6 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia Abstract Background At present, tobacco taxes in Armenia are among the lowest in Europe and Central Asia. Global experience has shown that increasing taxes on tobacco is one of the most cost- effective public health interventions. This is particularly relevant for Armenia, where smoking is among the leading risk factors of mortality among the population. Methods We conducted an extended cost-effectiveness analysis (ECEA) to assess the health, financial, and distributional consequences of increases in the excise tax on cigarettes in Armenia. Specifically, we estimated (i) the number of premature tobacco-related deaths averted, (ii) out-of-pocket (OOP) expenditures related to treatment of tobacco-related disease averted, (iii) the number of averted poverty cases (number of individuals falling below the national poverty line as a result of incurred OOP medical expenditures for tobacco-related disease treatment), (iv) the number of averted catastrophic health expenditures (individuals spending more than 10 percent of their consumption expenditure on tobacco-related treatment), and (v) government savings resulting from averted treatment costs for those eligible for the government- funded basic benefits package. We simulated a hypothetical price hike leading to an excise tax rate of 75 percent of the retail price of cigarettes, as recommended by the World Health Organization, which would be fully passed onto the consumers (current Armenian male smoking population). In addition, we conducted a series of stakeholder interviews to gain a better understanding of how tobacco tax increases were placed on the political agenda in Armenia Results Increased excise taxes on tobacco would bring large health and financial benefits to Armenian households and be pro-poor: about 88,000 premature deaths, US$ 63 million of OOP medical expenditures, 22,000 poverty cases, and 33,000 cases of catastrophic health expenditures would be averted. Government savings on tobacco-related treatment costs would amount to US$ 26 million. Half of the premature deaths and 27 percent of poverty cases averted would be concentrated among the bottom 40 percent of the population. The findings from the qualitative analysis suggest that the accession to the Eurasian Economic Union in 2015 and the fiscal constraints faced by the government created a window of opportunity for tobacco taxation to be placed on the policy agenda. Conclusions ECEA can be an important tool and input for policy decisions. In the case of Armenia, the ECEA findings point to the potentially largely pro-poor aspect of increased tobacco taxation. 7 Introduction Tobacco use is a leading risk factor for premature mortality. Today, almost 7 million premature deaths per year are attributed to tobacco use globally, with almost 20 percent of those deaths occurring in Europe and Central Asia. (1) This can result in substantial societal costs, as half of those who die of tobacco-related non-communicable diseases (NCDs) are in the prime of their productive years.(2) Recent estimates indicate that the total economic cost of smoking, including health expenditures and productivity losses, amounted to US$ 1436 billion in 2012 (approximately 1.8 percent of the world’s annual GDP).(3) Acknowledging the dire consequences of the tobacco epidemic, in 2003 the World Health Assembly adopted the World Health Organization’s (WHO) Framework Convention on Tobacco Control (FCTC),(4) which has since been ratified by 180 countries. The FCTC recommends a multidimensional approach to reducing smoking, including demand-side interventions, such as tax measures on tobacco products.(4) In 2015, countries renewed their commitment for the fight against tobacco by pledging to strengthen the implementation of the FCTC under the Sustainable Development Goals (SDGs).(5) Price measures have been at the forefront of the fight against tobacco.(6–10) Evidence has shown that price is the key determinant of smoking uptake and cessation, with numerous studies having found that price increases on cigarettes are highly effective in reducing demand by inducing smokers to quit and deterring non-smokers from initiating.(8) In addition, higher prices also result in current smokers reducing the number of cigarettes smoked daily and prevent ex-smokers from returning to smoking.(8,9) While tax hikes can generate additional revenue for development financing, as stated in the 2015 Addis Ababa Action Agenda and endorsed by the United Nations as part of the SDGs,(11) the main objective of tobacco taxes is to discourage product use and, as a result, avert the adverse health consequences of smoking. The WHO recommends that countries increase the excise tax rate on tobacco products to 75 percent.(7,12) There are two types of excise taxes: specific and ad valorem. A specific excise tax is a fixed monetary value per quantity (e.g. per pack or kilogram of tobacco), while an ad valorem excise tax is levied as a 8 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia percentage of the value of tobacco products (e.g. per retail price).(8) Unlike other types of taxes, such as value-added tax (VAT), high specific excise taxes can narrow the price gap between the types of cigarette brands and encourage cessation rather than substitution to lower-priced cigarettes as a result of tax increases.(9,13) However, only 33 countries so far have raised tobacco excise tax rates to the WHO-recommended rate of 75 percent. (7) Opponents, particularly the tobacco industry, have used the potential “regressivity” of excise taxes as an argument against further tax increases to build coalitions in opposition to cigarette price increases.(14–16) According to the argument, since the poor spend a larger proportion of their disposable income on smoking than the rich, increases in cigarette taxes and prices could disproportionately hurt the poor.(17) Other commonly cited arguments against increasing tobacco taxes include lost government revenue, job losses, smuggling, and political unpopularity with the voters.(6,9) With the SDGs, many international agencies such as the World Bank, are encouraging governments to adopt policies that would reduce poverty and boost shared prosperity by improving the living standards of the bottom 40 percent of the population.(18) Achieving these goals implies that policy recommendations pay special attention to the distributional impact of any reform to ensure that the poorer populations are benefiting the most. Armenia represents a fitting country to examine the distributional impact of increased tobacco taxes. It is a lower middle-income country with almost 30 percent of its population living today below the national poverty line of around 41,700 AMD per month (approximately US$100).(19) In recent years, Armenia’s economy has been hard hit due to regional and global economic conditions. With a public debt approaching almost 55 percent of its gross domestic product (GDP) and fiscal revenues representing only 22 percent of its GDP, Armenia is currently facing significant fiscal pressures. (20) Furthermore, smoking prevalence is high and tobacco use is one of the leading risk factors for premature mortality.(1) Almost 26 percent of Armenian adults smoke, largely the men (53 percent smoking prevalence among males as opposed to 2 percent among females).(21) Prevalence of smoking is particularly high among men in the second and third wealth quintiles of whom almost 60 percent smoke compared to 49 percent in the poorest quintile and 42 percent in the highest quintile.(22) In spite of an explicit publicly funded health benefits package, out-of-pocket (OOP) 9 healthcare expenditures represent almost 54 percent of the country’s total health spending,(23) and 9 percent of households incur catastrophic health expenditures (spending more than 25 percent of their nonfood expenditures on health).(24) Armenia was the first among the former Soviet Union countries to adopt and ratify the FCTC, which was shortly followed by the adoption, in March 2005, of a national law on “Restrictions on the sale, consumption, and use of tobacco” and a state program to control tobacco use.(25) Despite these initial moves, the government subsequently failed to act,(26) and Armenia now ranks behind many other countries in the region, such as Ukraine, Russia, and Georgia, on tobacco control efforts.(7) In particular, tax measures have remained inadequate to reduce demand for tobacco: tax as a share of the price of the most-sold cigarette brand constituted 34 percent in 2014 (17 percent excise tax and 17 percent VAT), with a mean price per cigarette pack of around US$ 1.25.(27) In 2015, the Armenian government approved a package of draft laws to revise its tax code, including a proposal to increase tobacco taxes, in order to raise revenues.(28) In this paper, we explored the potential distributional impact of increasing tobacco taxes in Armenia. We applied extended cost-effectiveness analysis (ECEA) methods (29–31) to assess the health, financial, and distributional consequences among smokers (males only, and by individual consumption quintile) of increases in cigarette taxes. We also conducted a series of qualitative interviews with key stakeholders to examine the agenda-setting discourse surrounding the recently proposed tobacco tax increases in Armenia. 10 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia Methods Modeling approach ECEA has been developed for health policy assessment (29,30) and applied to a wide array of policies including tobacco taxation.(31,32) ECEA intends to explicitly examine the outcomes of policy in multiple domains: the health benefits (e.g., premature deaths averted), the financial consequences for individuals and households (e.g., OOP expenditures averted due to disease treatment averted), the corresponding financial risk protection (e.g., cases of medical impoverishment or catastrophic health expenditures averted), and the distributional consequences among the population (e.g., per socioeconomic group, geographical setting). In doing so, it goes beyond traditional cost-effectiveness analysis in enabling quantification of the financial risk protection and equity (distributional) benefits of policy.(30) Previous studies modeling the impact of tobacco tax increases have investigated their aggregate impact, but few have explored their distributional consequences. In addition, such studies focused their analysis primarily on health gains and often did not explore the smokers’ financial consequences related to treatment of tobacco-related diseases. Here we built on a previously developed ECEA model for examining increases in tobacco taxes (31,41) to examine the premature deaths averted, the OOP expenditures averted and financial risk protection provided, and their distributions across socioeconomic groups among male smokers, by an increase in the excise taxes on tobacco products in Armenia. Using the baseline excise tax rate of 17 percent and an average price per pack of 525 AMD (approximately US$ 1.25),(27) we applied ECEA to simulate a price increase leading to the WHO-recommended excise tax rate of 75 percent.(7) We estimated that correspondingly the average price per cigarette pack would increase by 45 percent (to 756 AMD or US$ 1.80). We assessed distributional impact in terms of: (i) averted premature tobacco- related deaths; (ii) averted OOP expenditures on tobacco-related disease treatment; (iii) government savings resulting from averted tobacco-related disease treatment costs for those eligible for the government-funded basic benefits package; (iv) averted cases of medical impoverishment (number of individuals falling below the national poverty line as a result of OOP 11 tobacco-related treatment costs); and (iv) averted cases of catastrophic health expenditures (number of individuals spending more than 10 percent of their individual consumption on tobacco-related treatment costs) (Figure 1). Figure 1. Conceptual framework for modeling the health and financial impact of higher excise taxes on tobacco products among male smokers in Armenia. Averted premature mortality Fewer cases of impoverishment Reduced Lower out-of- as a result of consumption of pocket health out-of-pocket tobacco products Lower incidence spending on expenditures [higher reduction of tobacco- tobacco-related among the poorer related diseases diseases wealth quintiles] Fewer cases of catastrophic Lower health Higher Higher government expenditures excise health spending taxes price on tobacco- related diseases Additional Incremental financing for government health and revenues other sectors In Armenia, smoking is largely concentrated among men: 53 percent of males smoke compared to 2 percent of females.(22) Hence, we restricted our analysis to the current male population only, which we divided into five-year age groups (age 0 to 84 and a single group for all men above 84) using population estimates from the World Bank’s Health, Nutrition, and Population Statistics database.(33) The population was subsequently divided into individual consumption quintiles, and the maximum consumption for each quintile was estimated using the 2014 Integrated Living Conditions Survey (ILCS).(24) We applied these consumption quintile cutoffs and the Gini index (estimated from the 2014 ILCS) to simulate an individual consumption distribution using a Gamma distribution.(34,35) Age- and quintile-specific smoking prevalence rates were used to calculate the total number of smokers per each age group and consumption quintile. 12 following the approach of Hu et al.(37) and Adioetomo et al.(38) This elasticity falls within the - 0.54, which was estimated from the 2015 Kyr 0.54, which 0.40 to -0.80 estimated range of price elasticity of demand for tobacco in estimated was developing from the 2015 Kyrgyz Integrated following the approach of Hu et al.(37) and Adioe following the approach of Hu et al.(37) and Adioetomo et al.(38) T countries.(8,10) In line with findings from other studies,(8,9,32) the 2015 KIHS estimates also Estimating the distributional impact of increasing taxes on 0.40 tobaccoto In Armenia range of price elasti -0.80 estimated products: 0.40 to -0.80 estimated range of price elasticity of demand indicate that the poor are more responsive to price changes, with the elasticity ranging from - countries.(8,10) In line with findings from other countries.(8,10) In line with findings from other studies,(8,9,32) 0.74 in the poorest quintile to -0.28 in the richest quintile (see supplementary webappendix). Due to data restrictions, we were not able to calculate indicate that the poor are more responsive to pri the price elasticity indicate that the poor are more responsive to price changes, wit of demand for tobacco products for Armenia. Rather, we assumed a price Although the Kyrgyz Republic exhibits a lower smoking prevalence among men (26 percent) and 0.74 in the poorest quintile to -0.28 in the riches elasticity of demand for tobacco of -0.54, which was estimated from the 2015 0.74 in the poorest quintile to -0.28 in the richest quintile (see s Kyrgyz Integrated Household Survey (KIHS)(36) following the approach of Although the Kyrgyz Republic exhibits a lower smo a lower average price of cigarette pack (0.60 US$) compared to Armenia, excise tax rates are Although the Kyrgyz Republic exhibits a lower smoking prevalence Hu et al.(37) and Adioetomo et al.(38) This elasticity falls within the -0.40 to -0.80 estimated range of price elasticity of demand for a lower average price of cigarette pack (0.60 US$ tobacco in developing similarly low (16 percent specific and 8% ad valorem excise tax rates).(7) While we were able to a lower average price of cigarette pack (0.60 US$) compared to countries.(8,10) In line with findings from other studies,(8,9,32) the 2015 KIHS similarly low (16 percent specific and 8% ad valore estimate quintile-specific elasticities from the estimates KIHS, we also indicate that did thenot poorhave the responsive are more data necessary to priceto changes, similarly low (16 percent specific and 8% ad valorem excise tax ra with the elasticity ranging from -0.74 in the poorest quintile estimate -0.28 in the to quintile-specific elasticities from the K estimate estimate price elasticity by age-group. Based on evidence from reviews (8,39) we assumed that quintile-specific richest quintile (see supplementary webappendix). Although the Kyrgyz elasticities from the KIHS, we did not Republic exhibits a lower smoking prevalence among estimate price elasticity by age-group. Based on e men (26 percent) and estimate price elasticity by age-group. Based on evidence from re those under the age of 25 were twice as responsive to price changes as those above the age of a lower average price of cigarette pack (0.60 US$) compared to Armenia, those under the age of 25 were twice as responsi excise tax rates are similarly low (16 percent specific and 8 percent ad those under the age of 25 were twice as responsive to price chan 24. valorem excise tax rates).(7) While we were able to estimate 24. quintile-specific 24. elasticities from the KIHS, we did not have the data necessary to estimate price elasticity by age-group. Based on evidence from reviews (8,39) we We updated a simple static model (41) following a single cohort of all men alive in 2015. We We updated a simple static model (41) following assumed that those under the age of 25 were twice as responsive to price We updated a simple static model (41) following a single cohort assumed that the excise tax increase would be fully passed onto the consumers through a 45 changes as those above the age of 24. assumed that the excise tax increase would be fu assumed that the excise tax increase would be fully passed onto Weof percent increase in the retail price tobacco updated a simple half of and static that model price (41) elasticity following percent a singlewas increase due cohort to of in the retail price of tobacco all men percent increase in the retail price of tobacco and that half o alive in 2015. We assumed that the excise tax increase would be fully passed participation elasticity.(8) We calculated the number of individuals by age group and individual participation elasticity.(8) We calculated the num onto the consumers through a 45 percent increase in the retail price of participation elasticity.(8) We calculated the number of individua tobacco and that half of price elasticity was due to participation consumption quintile who would quit (from the c elasticity. consumption quintile who would quit (from the current adult male smoking population) or not consumption quintile who would quit (from the current adult ma (8) We calculated the number of individuals by age group and individual consumption quintile who would quit (from the initiate current adult smoking male (among those < 15 years) as a r smoking initiate smoking (among those < 15 years) as a result of higher tobacco prices. initiate smoking For each (among age those < 15 years) as a result of higher population) or not initiate smoking (among those < 15 years) group as a result of ! and consumption quintile " , the number group ! and consumption quintile higher tobacco prices. For each age group and consumption quintile " , the number of individuals w group ! and consumption quintile " , the number of individuals who would quit or not initiate the number of individuals who would quit or not initiatesmoking smoking ( ∆$%,' ) was was calculated depending on smoking ( ∆$%,' ) was calculated depending on the initial num smoking ( ∆$%,' ) was calculated depending on the on calculated depending initial number the initial number of of smokers ( $%,' ), the smokers the ( ( ∆+ ∆+ participation elasticity ( participation elasticity ()), price elasticity * , and relative change in price ( participation elasticity ( %,' elasticity (1/2), participation ) ( ), price elasticity ), price elasticity ** price elasticity %,'%,' , and relative change in price ( : and relative change in price ++ ) , and relative change in price ( ) ∆+ : ) : ) + (( ∆+ ∆+ ∆$%,' = * $ . . (1) (1) ∆$%,' = * $ ( ∆+ ∆$%,' = * $%,' . )) + ',% ',% + (1) %,' %,' ) ',% + 12 To calculate the premature deaths averted ( To calculate the premature deaths averted ( ∆- ∆- ), we used estimates from Doll et al.(40,41) to %,' ), we used estimates from Doll et al.(40,41) to %,' To calculate the premature deaths averted (∆-%,' ), we used estimates from Doll et al.(40,41) to model the changes in expected mortality based on the age at smoking cessation ( model the changes in expected mortality based on the age at smoking cessation ( .. % ), assuming % ), assuming model the changes in expected mortality based on the age at smoking cessation (.% ), assuming that half of smokers would die from their addiction.(40,42,43) Hence, the number of premature that half of smokers would die from their addiction.(40,42,43) Hence, the number of premature that half of smokers would die from their addiction.(40,42,43) Hence, the number of premature deaths averted would be: deaths averted would be: deaths averted would be: (( ∆+ ∆+ ∆- ∆- = %,' = %,' ** ',% . .% -%,' %- . %,' . (2) (2) )) ',% + + ( ∆+ ∆-%,' = * .% -%,' . (2) ) ',% + 13 While higher prices are also likely to lower the intensity of smoking among continuing smokers, While higher prices are also likely to lower the intensity of smoking among continuing smokers, we only we calculated the only calculated the health health benefits benefits associated with quitting associated with While higher prices are also likely to lower the intensity of smoking among continuing smokers, quitting and and did not model did not model any any substitution effects of individuals switching to lower price cigarettes. substitution effects of individuals switching to lower price cigarettes. we only calculated the health benefits associated with quitting and did not model any ( ∆+ on elasticity ()), price elasticity *%,' , and relative change in price ( + ): ( ∆+ participation elasticity ( e elasticity *%,' , and relative change in price ( ∆+ ): ) ), price elasticity *%,' , and relative change in price ( + ): ( + ∆+ ∆$%,' = *',% $%,' . (1) ) + ( ∆+ (1) ∆$%,' = ) *',% + $%,' . (1) ( ∆+ ∆$%,' = *',% $%,' . te the premature deaths averted (∆-%,' ), we used estimates from Doll et al.(40,41) to ) + changes in expected mortality based on the age at smoking cessation ( aths averted ( To calculate the premature deaths averted ( To calculate the premature deaths averted ∆-%,' ∆-%,' ), we used estimates from Doll et al.(40,41) to .%)), assuming , we used estimates from Doll et al.(40,41) to we used estimates from Doll et al.(40,41) to model the changes in expected mortality based on the f smokers would die from their addiction.(40,42,43) Hence, the number of premature model the changes in expected mortality based on the age at smoking cessation ( .% ), assuming age at smoking cessation .% ), assuming d mortality based on the age at smoking cessation ( assuming that half of smokers would die from erted would be: their addiction.(40,42,43) Hence, the number of premature deaths averted that half of smokers would die from their addiction.(40,42,43) Hence, the number of premature from their addiction.(40,42,43) Hence, the number of premature would be: deaths averted would be: ( ∆+ ∆-%,' = *',% .% -%,' . (2) ) + ( ∆+ (2) ∆- %,' = ) *',% + . % -%,' . (2) ( ∆+ ∆-%,' = *',% While .% -%,' . higher prices are also likely to lower the intensity her prices are also likely to lower the intensity of smoking among continuing smokers, of smoking among ) + continuing smokers, we only calculated the health benefits associated with quitting associated calculated the health benefits and did notwith quitting not model and did effects any While higher prices are also likely to lower the intensity of smoking among continuing smokers, 14 model any substitution ely to lower the intensity of smoking among continuing smokers, of individuals switching to lower price cigarettes. on effects of individuals switching to lower price cigarettes. we only calculated the health benefits th benefits associated with quitting and did not model any associated with quitting and did not model any For OOP and government medical expenditures averted, we allocated the substitution effects of individuals switching to lower price cigarettes. als switching to lower price cigarettes. and expenditures government medical averted averted, we allocated the averted premature premature deaths above (2) to four main causes of deaths: heart disease, neoplasms (lung cancer), stroke, and chronic obstructive pulmonary ove (2) to four main causes of deaths: heart disease, neoplasms (lung cancer), stroke, OOP and government For we medical expenditures averted, we allocated the averted premature dical expenditures averted, diseaseallocated (COPD).(1) the averted Healthcare premature utilization rates for each cause were estimated nic obstructive pulmonary disease using the total(COPD).(1) Healthcare annual number utilization rates of hospitalizations byfor the each International deaths above (2) to four main causes of deaths: heart disease, neoplasms (lung cancer), stroke, es.(47–49) causes of deaths: heart disease, neoplasms (lung cancer), stroke, Statistical Classification of Diseases and Related Health Problems (ICD-10) re consumption estimated using quintile as annual the total a reference and applied to the disease-specific hospitalization rates. onary disease (COPD).(1) and group in thenumber chronic Healthcare Ministry of hospitalizations of obstructive utilization pulmonary Health rates 2015 disease Statistical for each by the International (COPD).(1) Yearbook Healthcare for Armenia (44)utilization rates for each owever, have suggested that quitting is and the prevalence rates of the four diseases as estimated by the Institute Classification of Diseases and Related Health Problems (ICD-10) group in the Ministry The average cost of treatment per disease was obtained from Armenia’s basic benefits package he total annual number cause were estimated of hospitalizations using the the total annual number of hospitalizations by the International International by Evaluation.(1) sts that would be incurred as a result of for Health Metrics and To estimate hospitalizations by quintile, 7 07 87,' 97 % Δ-%,' 2015 (BBP) Statistical Yearbook price list.(45) we for Statistical Classification of Diseases and Related Health Problems (ICD-10) group in the Ministry The used BBP Armenia data fully on (44) and the prevalence quintile-specific funds services utilization for rates socially rates of for the vulnerable four inpatient services groups, from the including ases and Related Health Problems (ICD-10) group in the Ministry the 2014 ILCS.(24) The utilization rates were normalized using the middle (4) . as estimated by the Institute of Health for 2015 Health Metrics Yearbook Statistical and Evaluation.(1) for applied Armenia estimate To (44) and the prevalence rates of the four arbook for Armenia (44) and the prevalence rates of the four poor and those with disabilities (46). According to the 2014 ILCS data (24), almost 28 percent of consumption quintile as a reference and to the disease-specific ations by quintile, we used hospitalization data as diseases on rates. The average quintile-specific estimated by the cost of treatment Institute rates utilization for per for Health disease was obtained inpatient Metrics and Evaluation.(1) To estimate Institute for Health Metrics and Evaluation.(1) To estimate e the population is eligible for the BBP. We assumed that the government would pay the full cost from Armenia’s basic benefits package (BBP) price list.(45) The BBP fully funds rom the 2014 ILCS.(24) The utilization hospitalizations services for rates by socially were we quintile, vulnerable normalized using used including groups, data the on quintile-specific the middle poor and thoseutilization with rates for inpatient we used data on quintile-specific utilization rates for inpatient of tobacco-related disease treatment for those covered by the BBP and that these individuals disabilities.(46) According to the 2014 ILCS data,(24) almost 28 percent of the d deaths among those covered by BBP ealth services for disease : per quintile services .(24) The utilization rates were from the 2014 normalized using ILCS.(24) the We The utilization rates were normalized using the middle middle population is eligible for the BBP. assumed that the government would would not incur any additional expenses. Individuals who are not eligible for the BBP would pay pay the full cost of tobacco-related disease treatment for those covered by premature deaths, 97 is the cost of 13 the BBP and that these individuals would not incur any additional expenses. the full BBP price out of pocket. The change in OOP spending per quintile as a result of tax hike d by the BBP in quintile " , 07 is the Individuals who are not eligible for the BBP would pay the full BBP price out 13 13 would be: of pocket. The change in OOP spending per quintile as a result of tax hike would be: 7 07 87,' 97 ' ∆//0' = (1 − 4' ) % Δ-%,' 7 07 87,' 97 , 3) 3) , where 4' represents the share of population covered by the BBP in quintile " , 07 is the contribution (in %) of disease : to tobacco-related premature deaths, 97 is the cost of nding per quintile as a result of tax hike treatment for disease : , and 87,' is the utilization of health services for disease : per quintile o are not eligible for the BBP would pay 14 " . Likewise, government savings as a result of averted deaths among those covered by BBP d by the BBP and that these individuals would be: the government would pay the full cost 014 ILCS data (24), almost 28 percent of ease treatment for those covered by the BBP and that these individuals would not incur any additional expenses. Individuals who are not eligible for the of tobacco-related disease treatment for those covered by the BBP and that these indivi would be: nt for those covered by the BBP and that these individuals would not incur any additional expenses. Individuals who are not eligible for the BBP would pa the full BBP price out of pocket. The change in OOP spending of tobacco-related disease treatment for those covered by the BBP and that these individuals would not incur any additional expenses. Individuals who are not eligible for the BBP would pay dditional expenses. Individuals who are not eligible for the BBP would pay the full BBP price out of pocket. The change in OOP spending per quintile as a r would not incur any additional expenses. Individuals who are not eligible for the BBP woul nses. Individuals who are not eligible for the BBP would pay the full BBP price out of pocket. The change in OOP spending per quintile as a result of tax hik would be: would not incur any additional expenses. Individuals who are not eligible for the BBP would pay the full BBP price out of pocket. The change in OOP spending per quintile as a result of tax hike of pocket. The change in OOP spending per quintile as a result of tax hike (1 − 4' ) would be: % Δ-%,' 7 07 87,' 97 , ∆//0' = the full BBP price out of pocket. The change in OOP spending per quintile as a result of tax 3) e change in OOP spending per quintile as a result of tax hike would be: the full BBP price out of pocket. The change in OOP spending per quintile as a result of tax hike would be: Estimating the distributional impact of increasing taxes on tobacco products: ∆//0 In Armenia would be: ' = (1 − 4' ) % Δ-%,' 7 0 where 4' represents the share of population covered by the BBP ∆//0 in quintile = (1 " − , 4 0 7) is the Δ- %,' 7 7 7,' 7 , 0 8 9 would be: ' ' % ∆//0' = (1 − 4' ) % Δ-%,' 7 07 87,' 97 , 3 ∆//0' = (1contribution (in 7%) 07 8of disease : to tobacco-related 3) ∆//0 ' = (1 premature − 4' ) 4'% represents where Δ- deaths, 7 07 87,' 9 the 7 , share of population covered 3) by − 4' ) % Δ-%,' 7,' 9 7 , ∆//0 ' = (1 − %,' 4'9 )7 is the % Δ- %,'cost 7 07of 87,' 97 , = (1 − 4' ) % Δ-%,' 7 07 87,' 97 , 3) where 4' represents the share of population covered by the BBP in quin where 4' represents ∆//0' = the (1 − 4' ) of share % Δ- %,' 7 07 8covered population 7,' 97 , by the BBP in quintile 3) " , 07 is th contribution (in %) of disease 07 is the prem to tobacco-related :quintile treatment for disease : , and where where the share of population covered by the BBP in 7,' 8 4 is the utilization of health services for disease represents represents quintile " , 07 where the the is the share share of of population population covered covered by the by : BBP per quintile the in BBP quintilein " , ' represents contribution the share (in %) of population of disease : to covered by the BBP tobacco-related in quintile premature " , 079 deaths, is ' 4 7 population covered by the BBP in quintile " , 07 is is the contribution the contribution (in %) (in %)of of disease disease : to to tobacco-related tobacco-related premature premature deaths, 9 is the cost o where 4' represents the share of population treatment for disease covered by the BBP : , and 87,' is the utilization of health in quintile " , 07 is the 7 ". Likewise, government of disease : to tobacco-related contribution premature savings deaths, deaths, 9 as a is (in result the %) cost of of disease averted of : to deaths treatment for disease tobacco-related among : , and those premature is covered deaths, by BBP 9 is the 87,' is the utilization of health services for disea cost of 7 iscontribution the cost of (in %) of treatment for disease disease , to and tobacco-related the utilizationpremature of 7 deaths, 97 is the co to tobacco-related premature deaths, 97 is the cost of treatment for disease : , and 8 is the utilization of health services for disease : per quinti contribution health (in %) of services treatment for disease for disease disease : to : , and per 87,'tobacco-related quintile Likewise, ".. Likewise, 7,' premature government government deaths, savings is the utilization of health services for disease 97 is savings the asas a cost a result of of averted : per quintile dea : , and 87,' is the utilization of health services for disease would be: : per quintile treatment for disease " . Likewise, , and 87,' is the utilization of health services for disease :government savings as a result of averted deaths among : per qu those result of averted deaths among those covered by BBP would be: the utilization of health services for disease : per quintile treatment for disease " . Likewise, : , and government savings as a result of averted deaths among 87,' is the utilization of health services for disease would be: those covered by BB : per quintile " . Likewise, ent savings as a result of averted deaths among those covered government by BBP savings would be: as a result of averted deaths among those covered by BBP " . Likewise, government savings as a result of averted deaths among those covered by s a result of averted deaths among those covered by would be: BBP " . Likewise, government would be: ;<=>?%@ABC?,' savings =4 ' a result as % Δ-%,' of averted 7 07 87,' 97 . deaths among those covered (4) by BBP would be: ;<=>?%@ABC?,' = 4' % Δ-% would be: ;<=>?%@ABC?,' = 4' % Δ-%,' 7 07 87,' 97 . ;<=>?%@ABC?,' = 4' % Δ-%,' 7 07 87,' 97 . (4 Note that we did not estimate potential health care costs that would be incurred as a result of ;<=>?%@ABC?,' = 4' % Δ-%,' 7 07 8 Note 9 that . we did not estimate (4) potential health care ;<=>?%@ABC?,' =4 costs that Δ-%,' would 07 87,' 97 . be (4) 7,' 7 ;<=>?%@ABC?,' = 4' % Δ-%,' 7 07 87,' 97 . Note that we did not estimate potential health care costs tha ' % 7 <=>?%@ABC?,' = 4' % Δ-%,' 7 07 87,' 97 . incurred (4) as a result of years of life gained = among quitters. Previous studies, Note that we did not estimate potential health care costs that would be incurr ;<=> ?%@ABC?,' 4 ' % Δ- 0 %,' 7 7 7,' 7 8 9 . (4) years of life gained among quitters. Previous studies, however, have suggested that quitting is however, Note that we did not estimate potential health care costs that would be incurred as a result o have suggested that quitting isyears of life gained among quitters. Previous studies, howev associated with a reduction in Note that we did not estimate potential health care costs that would be incurred as a result of estimate potential health care costs that would be incurred as a result of years of life gained among quitters. Previous studies, however, have suggested overall health Note that we did not estimate potential health care costs that would be incurred as a res expenditures.(47–49) ntial health care costs that would be incurred as a result of years of life gained among quitters. Previous studies, however, have suggested that quitting associated with a reduction in overall health expenditures.(47–49) associated with a reduction in overall health expenditures.(47 Note that we did not estimate potential health care costs that would be incurred as a result of years of life gained among quitters. Previous studies, however, have suggested that quitting is mong quitters. Previous studies, however, have suggested that quitting is associated with a reduction in overall health expenditures.(47–49) years of life gained among quitters. Previous studies, however, have suggested that quitt Previous studies, however, have suggested that quitting is For the cases of medical impoverishment (poverty cases) averted, we associated with a reduction in overall health expenditures.(47–49) years of life gained among quitters. Previous studies, however, have suggested that quitting is associated with a reduction in overall health expenditures.(47–49) uction in overall health expenditures.(47–49) calculated the number of individuals that would have fallen below associated with a reduction in overall health expenditures.(47–49) ll health expenditures.(47–49) associated with a reduction in overall health expenditures.(47–49) the poverty line as a result of OOP tobacco-related disease treatment expenditures. Given that the national poverty line was estimated in per adult equivalent terms,(50) we identified an annual individual consumption 14 cutoff in the simulated consumption distribution corresponding to the 1 30th percentile (30 percent14 of the population lived below the poverty line 14 in 2015),(50) 14 which corresponded to approximately 1220 US$ per year. 14 Hence, we calculated the number of individuals for whom the simulated annual consumption was above this poverty line, but whose annual net consumption would have decreased to < 1220 US$ after paying for tobacco- related disease treatment. Likewise, for averted cases of catastrophic health expenditures, we calculated the number of individuals for whom OOP expenditures on tobacco-related disease treatment would be greater than 10 percent of annual individual consumption. In addition to the scenario of moving up to a 75 percent excise tax rate, we studied two additional scenarios i.e. shifts to a 50 percent excise tax rate (i.e., a 25 percent price increase) and to a 100 percent excise tax rate (i.e., 65 percent price increase). We also conducted a few sensitivity analyses. First, we tested the price elasticity of demand for tobacco, by applying a flat price elasticity of -0.54 to all quintiles. Second, we used two alternative poverty thresholds: a lower poverty line of US$ 79 per month (or US$ 948 per annum), and a food 15 poverty line of US$ 56 per month (or US$ 672 per annum).(50) Approximately 10 percent of the population was classified as poor using the lower poverty line and about 2 percent of the population lived below the food poverty line. (50) Third, we used two alternative thresholds in the estimation of cases of catastrophic health expenditures: 20 percent and 40 percent of individual consumption. Table 1 gathers all the input parameters used in the model. All analyses were conducted using R software (R 3.3.2). Table 1. Data inputs for the modeling of the increase in the tobacco excise tax in Armenia. INPUT VALUE SOURCE Male population 1,419,370 (33) < 15 21% 15-24 16% Male population distribution, 25-44 30% (33) age group (years) 45-64 25% ≥ 65 9% Q1 (poorest) < 1091 Q2 1092-1458 Individual annual consumption Q3 1459-1744 (24) (2014 US$) Q4 1745-2191 Q5 (richest) > 2191 15-24 38% Male smoking prevalence, per age 25-44 67% (22) group (years) 45-64 58% ≥ 65 31% Q1 (poorest) 49% Q2 61% Male smoking prevalence, Q3 59% (22) by wealth quintile Q4 49% Q5 (richest) 42% Daily cigarette consumption 24 Cigarettes (22) Price per pack of cigarettes $1.25 (7) (2014 US$) COPD $424 Tobacco-related disease treatment Stroke 350 (45) costs (2014 US$) Heart disease 1724 Neoplasm (lung cancer) 4781 16 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia INPUT VALUE SOURCE Q1 (poorest) 40% Share of population eligible for the Q2 30% publically financed basic benefits Q3 27% (24,64) package (%), by consumption quintile Q4 23% Q5 (richest) 19% Neoplasms 40% Authors’ Circulatory systems Utilization rates of healthcare 75% calculations diseases services per tobacco-related disease based on Respiratory systems (44) 27% disease Q1 (poorest) 0.72 Relative use of healthcare Q2 0.73 Authors’ services by consumption quintile calculations based Q3 1 (standardized to use Q3 as a on reference) Q4 1.06 (24) Q5 (richest) 1.17 15-24 97% Reduction in mortality risk by age 25-44 85% (age group in years) at quitting (40) smoking 45-64 75% ≥ 65 25% Q1 (poorest) -0.74 Authors’ Price elasticity of demand for Q2 -0.65 assumption based tobacco products, by consumption Q3 -0.65 quintile on estimates from Q4 -0.46 Kyrgyzstan (36) Q5 (richest) -0.28 National monthly poverty line 41,698 AMD (50) ($100) National poverty rate (percent of 30% (50) population) 17 Stakeholder interviews Following Bump and Reich (51), to gain a better understanding of the circumstances in which the current tax hike was proposed, we also conducted a series of interviews with Armenian stakeholders. We focused on the two stages of the policy cycle: the initial placement on the policy agenda or “agenda setting” and the technical design of the reform proposal.(52,53) Qualitative data for this analysis were collected through semi-structured interviews, as well as published and grey literature on Armenia’s tobacco control efforts. We used a purposeful sampling approach to identify interviewees by constructing a preliminary list of stakeholders prior to arriving in Armenia based on a literature review of tobacco control efforts in Armenia. The interviews were conducted in Yerevan in June 2016. Interviewees included representatives from the Ministry of Health (n=3), international organizations (n=3), health professionals (n=2), local non-governmental organizations (NGOs) (n=2), and universities (n=1). In total, we interviewed eleven individuals (Table 2), using a semi-structured interview guide, although stakeholders were encouraged to talk generally about tobacco control efforts in Armenia. Contemporaneous notes were taken during the interviews, and content analysis was performed once all interviews were completed to identify relevant themes to the research question. This work was supplemented by information and data extracted from national surveys, news releases, and published research relating to tobacco control in Armenia. The Harvard Human Research Protection Program granted an exemption for this study. Table 2. Number of interviews conducted with key stakeholders. Number of Stakeholder group interviews Ministry of Health 3 Health professionals 2 International organizations 3 Local non-governmental organizations 2 Universities 1 Total 11 18 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia Results Distributional impact Increasing the price of cigarettes by 45 percent would avert approximately 88,000 premature deaths among current estimated quitters and non-initiators (Table 3). Half of the premature deaths averted would be concentrated among the bottom two quintiles, with only 10 percent of deaths averted from the richest quintile. This is largely driven by the higher price elasticity among the poor (almost 2.6 times higher among the poorest than the richest quintile). In the sensitivity analyses, when assuming a flat price elasticity of -0.54 across all quintiles, unsurprisingly, while the total number of premature deaths averted would remain similar at about 86,000, its distribution would be more uniform across quintiles and in line with the quintile-specific smoking rates. Almost 20 percent of deaths would be averted in the richest quintile compared to 17 percent in the poorest quintile (Supplementary webappendix, Table A.1). Table 3. Extended cost-effectiveness analysis results by individual consumption quintile for a shift to a 75% tobacco excise tax rate (equivalent to a 45% price increase). Q1 Q5 TOTAL Q2 Q3 Q4 (poorest) (richest) Premature deaths averted 88 21 23 22 13 9 (in 1000s) (71, 106) (17, 25) (18, 28) (18, 27) (11, 16) (7, 10) Out-of-pocket expenditures related to 63 10 13 19 12 9 tobacco-related disease treatment averted (million (51, 77) (8, 12) (11, 16) (15, 22) (10, 15) (7, 11) US$) Government savings related to tobacco-related 26 7 6 7 4 2 disease treatment averted (20, 30) (6, 8) (5, 7) (5, 8) (3, 4) (2, 3) (million US$) Poverty cases averted 22 0 6 8 5 3 (in 1000s) (18, 27) 0 (5, 7) (7, 10) (4, 6) (2, 3) Cases of catastrophic health expenditures 33 5 7 8 6 5 (>10% of consumption) (28, 40) (4, 6) (6, 8) (8, 12) (5, 8) (4, 6) averted (in 1000s) Note: No poverty cases are averted in the poorest consumption quintile given that 30% of the population is already below the poverty line. Lower and upper bounds are indicated in parentheses. 19 There would also be substantial savings in OOP and government medical spending. As a result of averted tobacco-related disease treatment costs among those eligible for the BBP, the government would save a total of approximately US$ 26 million. In addition, almost US$ 63 million of OOP expenditures related to tobacco-related disease treatment would be averted among those not covered by the BBP. Almost 37 percent of these OOP savings would accrue to the bottom two quintiles, with an additional 30 percent accruing to the middle quintile in which fewer individuals were eligible for the BBP. When we assumed a flat price elasticity by quintile, OOP savings were slightly larger in magnitude (US$ 67 million), and almost 27 percent of those would accrue to the richest quintile compared to 28 percent in the bottom two quintiles (Supplementary webappendix, Table A.1). With a 45 percent tobacco price increase, almost 22,000 poverty cases would be averted. Given that 30 percent of the population already lived below the poverty line, no poverty cases would be averted among this bottom 30 percent of the population. Almost 27 percent of the averted poverty cases would accrue to the second poorest quintile and 14 percent to the richest quintile. Testing the sensitivity to the poverty threshold retained in the estimation, our results indicate that under a lower poverty line of about US$ 948 per year, the number of poverty cases averted would slightly rise to 23,000; under the food poverty line of US$ 672 per year, 24,000 poverty cases would be averted (Supplementary webappendix, Table A.2). Similarly, almost 33,000 cases of catastrophic health expenditures (defined as health spending representing more than 10 percent of individual consumption) would be averted. Stakeholder interviews Tobacco control efforts in Armenia have diminished after FCTC ratification. (26) Despite the existence of FCTC-recommended policies, the government has not been able to strengthen tobacco control measures, particularly in relation to raising excise taxes on tobacco products. In our interviews, all stakeholders emphasized that Armenia was the first among the former Soviet Union countries to ratify the FCTC. The early push for tobacco control measures in Armenia was largely attributed to former President Robert Kocharyan, himself a non-smoker, who strongly advocated for the FCTC implementation and encouraged other government members to quit smoking. In the absence of a strong public supporter, the importance of tobacco control measures subsided after Kocharyan left office in 2008. While 20 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia the Ministry of Health was a proponent of stronger tobacco control measures, particularly increased taxation, most interviewees suggested that it was not a powerful player in discussions on fiscal policy. In addition, the tobacco industry previously held a strong lobby in the Parliament, with several former tobacco industry executives having served on the Parliamentary Standing Committee on Financial Credit and Budgetary Affairs and Standing Committee on Economic Affairs. This resulted in several draft laws on tobacco control measures being recalled from the Parliament. Economic pressures, however, presented a window of opportunity for an overhaul of the existing tax system. In 2015, Armenia was facing continuing fiscal constraints, and the World Bank and the International Monetary Fund (IMF) supported measures to raise additional revenues.(20) In addition, Armenia’s accession to the Eurasian Economic Union in 2015 resulted in its own set of tax measures and regulations, including the mandated harmonization of rates of excise duties on alcohol and tobacco products over the next five years.(54,55) As a result, in October 2015, the government approved a package of draft laws on the tax code, which placed tobacco excise taxes on the government agenda. More specifically, the recently amended and approved tax code mandates that excise rates for alcohol and tobacco are to increase by 15 percent per year over 2017-2021, resulting in a tobacco excise tax of 44 percent by 2021.(20) Based on the discussions with key stakeholders and a literature review, it became evident that unlike previous unsuccessful attempts, two important contextual factors helped to garner support for the inclusion of higher excise tobacco taxes in the new tax code: tobacco tax increases were included alongside tax increases on other goods and services, including labor income tax; and it was seen as an inevitable step for the harmonization of taxes in the Eurasian Economic Union. The design of the tax reforms in Armenia was based on two key principles: the new tax system should enhance growth and equity; and it should generate revenue to support fiscal consolidation and allow higher social and capital expenditures.(20,56) In relation to tobacco excise taxes, the focus on equity was particularly important. Based on the discussions with stakeholders, regressivity appeared to be an important factor in delaying increased tobacco taxes since FCTC ratification. To address the equity concerns, the World Bank and the IMF provided technical assistance to simulate various scenarios 21 of proposed tax increases on various products, including tobacco.(20,56) Experience from other countries, such as the Philippines, played an important role in assuaging the regressivity concerns and allowing the Ministry of Finance to move forward with the proposed changes. Regressivity more so than any other commonly used argument against tobacco taxation – such as loss of revenues or smuggling – was at the center of the tobacco tax discussion. In several interviews, stakeholders stated that Armenia had strong tax and customs administration systems. Tobacco products, as other goods and imports, have holographic labels and unique identification codes; and tax officers commonly make sample purchases to scan and test the information provided on the products. This was argued to be a strong deterrent to smuggling. In addition, two individuals interviewed cited the ease of tobacco tax increases. Unlike other proposed tax changes, the interviewees noted that tobacco and alcohol taxes were easier to enforce and did not require any additional regulation. As a result, the higher tobacco and alcohol taxes entered into force on January 1, 2017, while the remaining changes to the budget code will be implemented in the following year.(20) While several interviewees cited examples from the Philippines and Thailand, where tobacco and alcohol taxes are earmarked for health, the possibility of earmarking was not discussed at length, and few interviewees supported this idea. Stakeholders cited the danger of setting a precedent, which would result in other ministries and government agencies requesting their own earmarked sources. In addition, one stakeholder cited the unsuccessful attempt to earmark proceeds from a VAT on medicines for health in 2001 as a reason why earmarking tobacco taxes would not be a viable policy in Armenia. Therefore, the discussion of earmarking was not pursued extensively for tobacco taxes. Tobacco tax was seen as an important measure to reduce consumption, but all stakeholders emphasized that other FCTC measures should be enforced. They stressed in particular the importance of raising public awareness and enforcing smoke-free zones. Moreover, they indicated that while they supported further tobacco tax increases, they believed that national cessation support services (currently not available in Armenia) should follow in order to realize the full benefits of higher prices for cigarettes. 22 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia Discussion The ECEA results indicate that higher excise taxes on tobacco in Armenia would avert large numbers of premature deaths and poverty cases. With a hike to a 75 percent excise tax rate, 88,000 premature deaths, US$ 63 million of OOP medical expenditures, 22,000 poverty cases (or 33,000 cases of catastrophic health expenditures) would be averted. Because the poor are more sensitive to price changes, the health benefits would be concentrated among the bottom two consumption quintiles of the population. Given that a larger share of the poor are eligible for the BBP and thus exempt from OOP payments, the benefits of averted tobacco-related disease treatment costs would accrue to the middle quintiles 3 and 4, less than 30 percent of whom are BBP-eligible. Government savings on tobacco-related treatment costs for those BBP-eligible would amount to US$ 26 million, which represents almost 12 percent of the annual health budget (estimated at US$ 220 million in 2014).(23) The fiscal constraints faced by the government and the accession to the Eurasian Economic Union in 2015 mandated a comprehensive overhaul of the existing tax policy and created a window of opportunity for tobacco tax increases. While previous attempts to increase tobacco taxes were unsuccessful in part due to veto power in the Parliament, the comprehensive nature of the tax reform allowed tobacco measures to be included in the proposal. Despite initial concerns about the regressivity of tobacco taxes, the findings from stakeholder interviews suggested that experience from other countries and simulations of the potential impact of such taxes on the poor were strong arguments for raising tobacco taxes as part of the overall fiscal reform. Our case study of Armenia presents evidence of what would be a successful attempt to increase tobacco excise taxes as part of a broader reform of a governmental tax system, yet the proposed excise tax of 44 percent (to be achieved by 2021) remains well below WHO’s recommendations. Nevertheless our analysis presents a number of limitations. First, we were not able to calculate the price elasticity of demand for tobacco products in Armenia, and our model was based on price elasticity estimates from the Kyrgyz Republic. Yet, the Kyrgyz price elasticities fell within the range of elasticities estimated in developing countries.(8) In addition, to test the sensitivity of our findings to price elasticity assumptions, we also simulated impact using a flat price elasticity across all quintiles. Second, we did not model substitution effects of individuals switching to lower-priced cigarettes as a result of price increases. However, unlike other types of taxes, high specific excise taxes would narrow the price gap between the most and least expensive cigarettes and encourage cessation rather than substitution to lower-priced 23 cigarettes as a result of tax increases.(9,13). Third, we assumed that a decline in the intensity of smoking would not yield any health benefits: individuals who would reduce their tobacco consumption and smoke fewer cigarettes per day as a result of tax hike would not improve their health outcomes in our model; nor did we model second-hand smoking. As a result, we are likely to underestimate the full impact of higher tobacco taxes in our premature deaths and financial risk protection findings. Fourth, in the absence of data on OOP expenditures per disease, we used the BBP price list as a proxy for the incurred OOP expenditures. Although this is the official government price for services in all government facilities, there is some evidence of informal payments.(57) In addition, data on pharmaceutical expenditures on medicines not covered by the BBP were not available and hence could not be included. Our results thus are likely to underestimate the expenditures related to tobacco-related disease treatment and the number of poverty cases averted, since OOP medical expenditures are likely to be higher than the established government fees for the BBP. Fifth, we only included the cost of inpatient care, as we were not able to obtain detailed data on utilization for each disease and associated costs at the primary health care level. Primary care, however, is free for all citizens in Armenia. Therefore individuals should not incur any OOP at the primary care level. Sixth, the health and health-related financial benefits are modeled into the future (for the current Armenian male population), when individuals are expected to face tobacco- related diseases. Hence, there is wide uncertainty in our assumptions, as we assume that key inputs (e.g., consumption, cost of medical services, utilization, BBP coverage) remain the same. Seventh, we assumed that the excise tax would be passed fully onto the consumer. Although this is a standard assumption in tobacco tax modeling studies,(6,8) the empirical evidence is mixed,(8,58–61) hence we may overestimate here the effect of increased excise taxes. Our study contributes to the literature on tobacco taxation and the distributional impact of higher cigarette prices and taxes. While the regressivity argument has been commonly used against price increases and was perceived to be a constraint to increase tobacco taxes in Armenia, similarly to other recent studies we do not find evidence of higher tobacco prices necessarily disproportionately burdening the poor. As recent studies have found, the higher responsiveness to prices among the poor may shift the burden of incremental taxes to the rich, thus making tobacco taxes more progressive.(31,32,62) Not only can higher excise taxes reduce the number of deaths through smoking cessation,(63) but they can also decrease potential OOP expenditures on treatment for tobacco-related disease. Given the large costs associated with such treatment, by encouraging smokers to quit or averting initiation, tobacco taxes can bring substantial financial risk protection to individuals by preventing such OOP medical expenditures altogether.(31) 24 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia While the health benefits associated with smoking cessation have been well established, this has not been necessarily enough to encourage countries to raise tobacco taxes. Identifying the potential windows of opportunities (e.g. fiscal constraints) has important policy implications and could enable a push for higher tobacco taxes. As the global health community encourages the use of fiscal policies to change behavior (e.g. tobacco taxes, sugar-sweetened beverage taxes) in order to achieve public health gains, the sole public health argument might be insufficient. The case of Armenia suggests that governments could successfully increase tobacco taxes by including them as part of broader fiscal reforms. While the proposal to raise excise taxes in Armenia marks an important step, the proposed rate of 44 percent remains well below the WHO recommendation. Concerted efforts need to be taken in order to ensure that further tax increases are implemented and other tobacco control measures are enacted according to the FCTC and best practices. 25 References 1. Institute for Health Metrics and Evaluation. GBD Compare | IHME Viz 20. International Monetary Fund. Republic of Armenia Third Review Hub [Internet]. 2017 [cited 2016 Nov 18]. Available from: http://vizhub. Under the Extended Arrangement: Staff Report. 2016 Jul. Report No.: healthdata.org/gbd-compare IMF Country Report No. 16/246. 2. Bloom DE, Cafiero E, Jané-Llopis E, Abrahams-Gessel S, Bloom LR, 21. National Institute of Health. Assessment of the health system in Fathima S, et al. The global economic burden of noncommunicable 2016 [Internet]. Yerevan, Armenia: Ministry of Health Armenia; 2016. diseases. Program on the Global Demography of Aging; 2012. Available from: http://moh.am/uploads/HSPA_2016_ARM_2016.pdf 3. Goodchild M, Nargis N, d’Espaignet ET. 2017. Global economic cost 22. Ministry of Health. Armenian National Health Sector Performance of smoking-attributable diseases. Tob Control. 2017. Assessment [dataset]. 2015. 4. World Health Organization. WHO Framework Convention on 23. World Bank. World Development Indicators | Data [Internet]. 2016 Tobacco Control [Internet]. 2003 [cited 2016 Nov 21]. Available from: [cited 2016 Nov 20]. Available from: http://data.worldbank.org/data- http://apps.who.int/iris/bitstream/10665/42811/1/9241591013.pdf catalog/world-development-indicators 5. United Nations. Sustainable Development Goals [Internet]. 24. National Statistical Service of the Republic of Armenia. Armenia 2015 [cited 2016 Nov 21]. Available from: http://www.un.org/ Household’s Integrated Living Conditions Survey [dataset]. 2014. sustainabledevelopment/health/ 25. Law of the Republic of Armenia on the sale, consumption 6. U.S. National Cancer Institute, World Health Organization. The and limiting the use of tobacco [Internet]. HO-72-N Dec 24, 2004. Economics of Tobacco and Tobacco Control. Bethesda, MD: U.S. Available from: http://www.tobaccocontrollaws.org/files/live/ Department of Health and Human Services, National Institutes of Armenia/Armenia%20-%20Law%20on%20Restrictions%20of%20 Health, National Cancer Institute; and Geneva, CH: World Health Sale%2C%20Consumption%2C%20and%20Use.pdf Organization; 2016. Report No.: NIH Publication No. 16-CA-8029A. 26. Movsisyan NK, Connolly GN. Measuring Armenia’s progress on 7. World Health Organization. WHO Report on the Global Tobacco the Tobacco Control Scale: an evaluation of tobacco control in an Epidemic, 2015. Geneva: World Health Organization; 2015. economy in transition, 2005–2009. BMJ Open. 2014;4(2):e004410. 8. International Agency for Research on Cancer. IARC Handbooks 27. Ministry of Health of the Republic of Armenia, World Health of Cancer Prevention, Tobacco Control, Volume 14: Effectiveness Organization. Armenia: Reporting Instrument of the WHO Framework of Tax and Price Policies in Tobacco Control. Vol. 14. Lyon, France: Convention on Tobacco Control. 2014. International Agency for Research on Cancer; 2011. 28. Package of Draft Laws on Tax Code Approved [Internet]. The 9. Chaloupka FJ, Hu T, Warner KE, Jacobs R, Yurekli A. The taxation of Government of the Republic of Armenia: Official News. 2015 tobacco products. In: Jha P, Chaloupka FJ, editors. Tobacco Control in [cited 2016 Dec 19]. Available from: http://www.gov.am/en/news/ Developing Countries. New York: Oxford University Press; 2000. item/8122/ 10. Jha P, Chaloupka FJ. Curbing the epidemic: governments and 29. Verguet S, Laxminarayan R, Jamison DT. Universal Public Finance the economics of tobacco control. Washington, D.C.: World Bank of Tuberculosis Treatment in India: An Extended Cost-Effectiveness Publications; 1999. Analysis. Health Econ. 2015;24(3):318–332. 11. Addis Ababa Action Agenda of the Third International Conference 30. Verguet S, Kim JJ, Jamison DT. Extended cost-effectiveness on Financing for Development 13-16 July, 2015 [Internet]. Addis analysis for health policy assessment: a tutorial. PharmacoEconomics. Ababa, Ethiopia; [cited 2016 Nov 16]. Available from: http://www. 2016;34(9):913–923. un.org/esa/ffd/wp-content/uploads/2015/08/AAAA_Outcome.pdf 31. Verguet S, Gauvreau CL, Mishra S, MacLennan M, Murphy SM, 12. World Health Organization. WHO technical manual on tobacco tax Brouwer ED, et al. The consequences of tobacco tax on household administration. Geneva: World Health Organization; 2010. health and finances in rich and poor smokers in China: an extended cost-effectiveness analysis. Lancet Glob Health. 2015;3(4):e206–e216. 13. Jha P, Peto R. Global effects of smoking, of quitting, and of taxing tobacco. N Engl J Med. 2014;370(1):60–68. 32. Salti N, Brouwer E, Verguet S. The health, financial and distributional consequences of increases in the tobacco excise tax 14. Koh HK. An analysis of the successful 1992 Massachusetts tobacco among smokers in Lebanon. Soc Sci Med. 2016;170:161–169. tax initiative. Tob Control. 1996;5(3):220–225. 33. World Bank. Health, Nutrition and Population Data and Statistics 15. Balbach ED, Herzberg A, Barbeau EM. Political coalitions and [Internet]. 2016 [cited 2016 Nov 21]. Available from: http://datatopics. working women: how the tobacco industry built a relationship with worldbank.org/hnp/ the Coalition of Labor Union Women. J Epidemiol Community Health. 2006;60(suppl 2):ii27–ii32. 34. Kemp-Benedict E. Income distribution and poverty: methods for using available data in global analysis. PoleStar Technical Note 4, 2001. 16. Smith KE, Savell E, Gilmore AB. What is known about tobacco industry efforts to influence tobacco tax? A systematic review of 35. Salem A, Mount T. A convenient descriptive model of income empirical studies. Tob Control. 2013;22(2):144–153. distribution: the gamma density. Econometrica 1974;1115–1127. 17. Remler DK. Poor smokers, poor quitters, and cigarette tax 36. National Statistical Committee of the Kyrgyz Republic. Kyrgyz regressivity. Am J Public Health. 2004;94(2):225–229. Integrated Household Survey [dataset]. 2015. 18. World Bank. The World Bank Group Goals: End Extreme Poverty and 37. Hu T-W, Ren Q-F, Keeler TE, Bartlett J. The demand for cigarettes in Promote Shared Prosperity. Washington, D.C.: The World Bank; 2013. California and behavioural risk factors. Health Econ. 1995;4(1):7–14. 19. World Bank, National Statistical Service of the Republic of Armenia. 38. Adioetomo SM, Djutaharta T, Hendratno. Cigarette consumption, Social Snapshot and Poverty in Armenia: Based on the results of the taxation, and household income: Indonesia case study. Washington, 2014 Integrated Living Conditions Survey of Households. Yerevan, D.C.: World Bank; 2005. (Health Nutrition and Population (HNP) Armenia; 2015. discussion paper). 26 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia 39. World Health Organization. WHO technical manual on tobacco tax 56. Gohar Gyulumyan. Is it Fair that 80-85 Percent of Personal Income administration. Geneva: World Health Organization; 2010. Tax Collections is Formed through Those Earning Less than AMD 120 Thousand Income per Month? [Internet]. 2016 [cited 2016 40. Doll R, Peto R, Boreham J, Sutherland I. Mortality in relation Dec 19]. Available from: http://www.worldbank.org/en/news/ to smoking: 50 years’ observations on male British doctors. Bmj. opinion/2016/07/01/interview-with-ms-gohar-gyulumyan-on-the- 2004;328(7455):1519. tax-code 41. Verguet S, Tarr G, Gauvreau C, Mishra S, Jha P, Liu L, et al. 57. Lewis M. Informal payments and the financing of health care Distributional benefits of tobacco tax and smoke-free workplaces in in developing and transition countries. Health Aff (Millwood). China: a modeling study [forthcoming]. 2007;26(4):984–997. 42. Jha P, Jacob B, Gajalakshmi V, Gupta PC, Dhingra N, Kumar R, et al. 58. Linegar DJ, van Walbeek C. The effect of excise tax increases on A nationally representative case–control study of smoking and death cigarette prices in South Africa. Tob Control. 2017. in India. N Engl J Med. 2008;358(11):1137–1147. 59. Hanson A, Sullivan R. The incidence of tobacco taxation: evidence 43. Jha P, Ramasundarahettige C, Landsman V, Rostron B, Thun M, from geographic micro-level data. Natl Tax J. 2009;677–698. Anderson RN, et al. 21st-century hazards of smoking and benefits of cessation in the United States. N Engl J Med. 2013;368(4):341–350. 60. Sullivan RS, Dutkowsky DH. The effect of cigarette taxation on prices: an empirical analysis using local-level data. Public Finance Rev. 44. National Institute of Health of Armenia. Health and Wellness 2012;40(6):687–711. Statistical Yearbook 2015 [Internet]. Yerevan, Armenia: Ministry of Health Armenia; 2016. Available from: http://moh.am/uploads/ 61. Delipalla S, O’Donnell O. Estimating tax incidence, market power Health_Health%20Care%20Year%20Book_2015_Arm.pdf and market conduct: The European cigarette industry. Int J Ind Organ. 2001;19(6):885–908. 45. Ministry of Health of the Republic of Armenia. Armenia Basic Benefits Package Price List. 2016. 62. Fuchs A, Meneses F. Are Tobacco Taxes Really Regressive? Washington, D.C.: World Bank; 2017. 46. Richardson E. Armenia: Health system review. Health Syst Transit. 2013;15(4):1–99. 63. Jha P, Joseph R, Li D, Gauvreau C, Anderson I, Moster P, et al. Tobacco Taxes: A Win-win Measure for Fiscal Space and Health. 47. Hodgson TA. Cigarette smoking and lifetime medical expenditures. November 2012. 2012. Milbank Q. 1992;81–125. 64. Jamal S. Improvements in Targeting of Basic Benefits Package in 48. Fishman PA, Khan ZM, Thompson EE, Curry SJ. Health care costs Armenia. 2016. among smokers, former smokers, and never smokers in an HMO. Health Serv Res. 2003;38(2):733–749. 65. Adioetomo S, Djutaharta T, Hendratno. Cigarette consumption, taxation and household income: Indonesia case study. World Bank 49. Rasmussen SR, Prescott E, Sørensen TI, Søgaard J. The total lifetime Health Nutr Popul HNP Discuss Pap Econ Tob Control Pap. 2005;(26). health cost savings of smoking cessation to society. Eur J Public Health. 2005;15(6):601–606. 50. World Bank, National Statistical Service of the Republic of Armenia. Social Snapshot and Poverty in Armenia: Based on the results of the 2015 Integrated Living Conditions Survey of Households. [Internet]. Yerevan, Armenia; 2016. Available from: http://www.armstat.am/file/ article/poverty_2016_eng_2.pdf 51. Bump JB, Reich MR. Political economy analysis for tobacco control in low-and middle-income countries. Health Policy Plan. 2013;28(2):123–133. 52. Kaufman RR, Nelson JM, others. Crucial needs, weak incentives: social sector reform, democratization, and globalization in Latin America. Washington, D.C.: Woodrow Wilson Center Press; 2004. 53. Roberts MJ. Getting Health Reform Right : A Guide to Improving Performance and Equity [Internet]. Oxford: Oxford University Press; 2004 [cited 2012 Sep 14]. Available from: http://web.ebscohost.com. ezp-prod1.hul.harvard.edu/ehost/detail?sid=d438bcc6-8315-4a84- 8016-0e18aabdf879%40sessionmgr15&vid=1&hid=21&bdata=JnNpd GU9ZWhvc3QtbGl2ZSZzY29wZT1zaXRl#db=nlebk&AN=176907 54. Eurasian Economic Commission. On the Draft Agreement on the tax policy principles in respect of excise duties on tobacco products in the Member States of the Eurasian Economic Union [Internet]. Order № 126 Nov 10, 2015. Available from: http://www. eurasiancommission.org/ru/Lists/Decisions/DispForm.aspx?ID=531 55. Draft Agreements on the tax policy principles in respect of excise duties on alcohol and tobacco products in the EAEU Member States are approved [Internet]. Eurasian Economic Commission. 2015 [cited 2016 Dec 19]. Available from: http://www.eurasiancommission.org/ en/nae/news/Pages/29-10-2015-5.aspx 27 Supplementary appendix Table A.1. Extended cost effectiveness analysis results by individual consumption quintile for a shift to a 75% excise tobacco tax rate (equivalent to a 45% price increase) assuming a flat price elasticity (same price elasticity across all quintiles). Q1 Q5 TOTAL Q2 Q3 Q4 (poorest) (richest) Premature deaths 86 15 19 19 16 17 averted (in 1000s) (68, 102) (12, 18) (15, 23) (15, 22) (12, 19) (13, 20) Out-of-pocket expenditures related to 67 8 11 15 15 18 tobacco-related disease treatment averted (53, 80) (6, 9) (9, 13) (12, 18) (12, 17) (14, 22) (million US$) Government savings related to tobacco- 24 5 5 6 4 4 related disease treatment (19, 28) (4, 6) (4, 6) (4, 7) (3, 5) (3, 5) averted (million US$) Poverty cases averted 23 0 5 7 6 5 (in 1000s) (18, 28) 0 (4, 6) (6, 8) (5, 7) (4, 6) Cases of catastrophic health expenditures 35 4 6 8 8 9 (>10% of consumption) (28, 42) (3, 5) (5, 7) (6, 10) (6, 9) (7, 11) averted (in 1000s) Note: No poverty cases are averted in the poorest consumption quintile given that 30% of the population is already below the poverty line. Lower and upper bounds are indicated in parentheses. 28 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia Table A.2. Extended cost effectiveness analysis results by individual consumption quintile for a shift to a 75% excise tobacco tax rate (equivalent to a 45% price increase) using: lower and food poverty lines; 20% and 40% thresholds for catastrophic health expenditures. Q1 Q5 TOTAL Q2 Q3 Q4 (poorest) (richest) Poverty cases averted Using lower poverty line 23 1 7 8 5 2 of 79 US$ per month (in 1000s) (19, 28) (1, 2) (5, 8) (6, 9) (4, 6) (2, 3) Using food poverty line 24 4 6 8 5 1 of 56 US$ per month (in 1000s) (19, 28) (3, 5) (5, 7) (6, 9) (4, 6) (1, 2) Cases of catastrophic health expenditures averted Using a threshold of >20% 32 5 7 10 6 4 of individual consumption (in 1000s) (25, 38) (4, 6) (6, 8) (8, 12) (5, 7) (3, 5) Using a threshold of >40% 27 5 6 8 5 4 of individual consumption (in 1000s) (22, 33) (4, 6) (5, 7) (6, 9) (4, 6) (3, 5) Note: The lower and food poverty lines are two alternative measures used for poverty calculations in Armenia. In 2014, approximately 10% and 2% of the population fell below the lower and food poverty lines, respectively.(50) Lower and upper bounds are indicated in parentheses. 29 Table A.3. Extended cost effectiveness analysis results by individual consumption quintile for shifts to 50% and 100% excise tobacco tax rates (equivalent to 25% and 65% price increases, respectively). Q1 Q5 TOTAL Q2 Q3 Q4 (poorest) (richest) 50% excise tax rate or 25% price increase Premature deaths averted 75 18 20 19 11 7 (in 1000s) (61, 91) (15, 22) (16, 24) (15, 23) (9, 14) (6, 9) Out-of-pocket expenditures related to 24 4 6 8 5 1 tobacco-related disease treatment averted (million (44, 66) (7, 11) (9, 14) (13, 19) (9, 13) (7, 10) US$) Government savings related to tobacco-related 22 6 5 6 3 2 disease treatment averted (17, 26) (5, 7) (4, 6) (5, 7) (3, 4) (2, 2) (million US$) Poverty cases averted (in 18 0 5 7 4 2 1000s) (15, 23) (0) (4, 6) (6, 9) (4, 5) (2, 3) Cases of catastrophic health expenditures 29 5 6 8 6 4 (>10% of consumption) (23, 35) (4, 6) (5, 7) (7, 10) (5, 7) (3, 5) averted (in 1000s) 100% excise tax rate or 65% price increase Premature deaths averted 100 24 26 25 15 10 (in 1000s) (80, 121) (19, 29) (21, 31) (23, 30) (12, 18) (8, 12) Out-of-pocket expenditures related to 73 12 15 21 14 11 tobacco-related disease treatment averted (million (58, 87) (9, 12) (12, 18) (17, 25) (11, 17) (9, 13) US$) Government savings related to tobacco-related 28 8 6 8 4 2 disease treatment averted (23, 34) (6, 9) (5, 8) (6, 9) (3, 5) (2, 3) (million US$) Poverty cases averted (in 26 0 7 10 6 3 1000s) (20, 30) (0) (5, 8) (8, 11) (5, 7) (3, 4) Cases of catastrophic health expenditures 37 6 8 11 7 5 (>10% of consumption) (30, 46) (5, 7) (6, 10) (9, 13) (6, 9) (4, 7) cases averted (in 1000s) Note: No poverty cases are averted in the poorest consumption quintile given that 30% of the population is already below the poverty line. Lower and upper bounds are indicated in parentheses. 30 Estimating price elasticity of demand for tobacco in the Kyrgyz Republic Estimating price elasticity of demand for tobacco in the Kyrgyz Republic Data from the 2015 Kyrgyz Integrated Household Survey (KIHS) were used to estim Data from the 2015 Kyrgyz Integrated Household Survey (KIHS) were used to estim Estimating price elasticity of demand for tobacco in the Kyrgyz Repu Estimating price elasticity of demand for tobacco in the Kyrgyz Republic elasticities for tobacco. First, in order to estimate the price of cigarettes faced by no from the 2015 Data Household Kyrgyz Integrated Household Survey (KIHS) were use Data from the 2015 Kyrgyz Integrated Survey (KIHS) were used to estimate price elasticities for tobacco. First, in order to estimate the price of cigarettes faced by no Estimating price elasticity of demand for tobacco in the Kyrgyz Republic Estimating price elasticity of demand for tobacco in the Kyrgyz Republic using Estimating an the OLS regression distributional impact ofwe predicted increasing taxes onthe price tobacco of cigarettes In Armenia for nonsmokers bas products: Data from the 2015 Kyrgyz Integrated Household Survey (KIHS) were used to estimate p elasticities for tobacco. First, in order to estimate the price of cigarettes fa elasticities for tobacco. First, in order to estimate the price of cigarettes faced by nonsmokers, using from Data an OLS the 2015 Kyrgyz regression we predicted Household Integrated the price of Survey (KIHS) cigarettes were for used to estim nonsmokers bas Estimating price elasticity of demand for tobacco in the Kyrgyz Republic individual’s consumption quintile, oblast, and whether the individual elasticities for tobacco. First, in order to estimate the price of cigarettes faced by nonsmo using OLS regression an price we for predicted the price cigarettes of on resided for in nonsm an using an OLS regression predicted we the Data from 2015 the Kyrgyz of cigarettes Integrated Household nonsmokers Survey (KIHS) based were used elasticities for tobacco. First, in order to estimate the price of cigarettes faced by no the to estimate price individual’s consumption quintile, oblast, and whether the individual resided in an Estimating using price Estimating price elasticity of demand for tobacco in the Kyrgyz Republic an OLS elasticity regression we of demand predicted the for tobacco price of cigarettes rural area. We assumed that non-smokers faced the predicted price. Following Hu et in for nonsmokers based on individual’s elasticities for tobacco. First, in order to estimate the price of cigarettes faced by nonsmokers, consumption quintile, individual’s oblast, and consumption whether the quintile, individual oblast, resided and in whether an urban the individual res the Kyrgyz Data from the 2015 Kyrgyz Integrated Household using Survey Republic an OLS (KIHS) regression were we predicted used to estimate price the price of cigarettes rural area. We assumed that non-smokers faced the predicted price. Following Hu et for or nonsmokers bas individual’s consumption quintile, oblast, and whether the individual resided in an urba Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the fir using an OLS regression we predicted the price of cigarettes for nonsmokers based on the rural area. We assumed that non-smokers faced the predicted price. Follow rural area. We assumed that non-smokers faced the predicted price. Following Hu et al.(37) and elasticities for tobacco. First, in order to estimate the price of cigarettes faced by nonsmokers, individual’s consumption quintile, oblast, and whether the Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the fir Data from the 2015 Kyrgyz Integrated Household Survey (KIHS) were used toindividual resided in an rural area. We assumed that non-smokers faced the predicted price. Following Hu et al.(37) individual’s estimate consumption estimated price elasticities quintile, the probability for tobacco.oblast, of an First, and in whether individual order the individual being to estimate a smoker the price ofresided !"#$ in % an urban or & = 1 using the using an OLS regression we predicted the price of cigarettes Adioetomo et al.(65), we used a two-part model to estimate the elasticities Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the first part, we for nonsmokers based on the rural area. We assumed that non-smokers faced the predicted price. Following Hu et estimated cigarettes faced the probability by nonsmokers, anan usingof OLS regression being individual a smoker we predicted the !"#$ %& = 1 using the Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the first part rural area. We assumed that non-smokers faced the predicted price. Following Hu et al.(37) and individual’s consumption quintile, oblast, estimated price and oflogit equation: cigarettes whether the probability of an individual the for nonsmokers estimated individual the based resided being a an in on urban probability smoker individual’s the!"#$ or of an = consumption %&individual 1 using being the a smoker !"#$ %& = 1 following Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the fir logit equation: estimated the probability of an quintile, oblast, and whether the individual resided in an urbana individual being smoker or !"#$ %& = 1 using the follo rural area. Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the first part, we rural area. We assumed that non-smokers faced the predicted price. Following Hu et al.(37) and logit equation: We logit equation: assumed that non-smokers faced ) the predicted price. Following Hu et !"#$ estimated %& = 1 = the logit equation: probability of an ,(./ 0123 456 3 47/ ) individual , being a smoker (1) !"#$ %& = 1 using the estimated al.(37) and the probability Adioetomo et of an )*+ al.(65), we individual ) used a being two-part a smoker model to !"#$ %& = estimate the 1 using the following !"#$ % Adioetomo et al.(65), we used a two-part model to estimate the elasticities. In the first part, we & = 1 = , (1) )*+ ,(./ 0123 4563 47/ ) ) In the first part, we estimated elasticities. logit equation: the probability ) of an individual !"#$ %& = 1 = logit equation: !"#$ % !"#$ = 1 % = , = 1 = ) (1) , , (1) by the (1) being estimated the probability of an individual )*+ being ,(. awhere /a 012 smoker smoker 3 456 3 9:! 47 / & is !"#$ ) & %the & =1 & log )*+ price using using ,(. the the / 012 )*+of ,(. cigarettes following following 3 456 3 47 / /)012 456 3 logit47faced /) 3 equation: individual, Xi is the vector where 9:!& is the log price of cigarettes faced by the individual, Xi is the vector ) ) logit equation: demographic characteristics, including age, sex, consumption quintile, and oblast, an !"#$ ) 456 , of , (1) faced vector (1) of % = 1 = !"#$ % & = 1 = where & is the log price of cigarettes by the individual, is the vector Xi individual, Xi of where 9:!& is the log price 9:! where cigarettes & )*+ faced 9:! is the by log 3 the price individual, cigarettes i is the faced X by the of socio- is so t ,(. / 012 456 47 ,(. 012 3 )*+ &3 // 3 47 /) demographic characteristics, including age, sex, consumption quintile, and oblast, an ) random error term. 9:!&demographic characteristics, including age, sex, consumption quintile, and oblast, and U demographic characteristics, including age, sex, consumption quintile, and 1 is !"#$ %& = 1 = demographic characteristics, including age, sex, consumption quintile, and oblast, and U , where where is where is (1) the the log log price price of of cigarettes cigarettes facedfaced by the by individual, the individual, 1 is the Xi is the Xi is the vector of socio- )*+ ,(./ 0123 4563 47/ ) 9:! & is the log price of cigarettes faced by the individual, Xi is the vector random error term. vector of socio-demographic characteristics, including age, sex, consumption random error term. random error term. random error term. demographic characteristics, including age, sex, consumption quintile, and oblast, and U 1 is the the a where 9:!& is the log price of cigarettes faced by In quintile, and the the second oblast, andXU individual, part, we is the used error random ordinary term. least squares regression to estimate demographic characteristics, including age, sex, consumption quintile, and oblast, an i 1is the vector of socio- In the second part, we used ordinary least squares regression to estimate the a random error term. In the second part, we used ordinary least squares regression to estimate the amoun In the cigarettes smoked per day by current smokers ( demographic characteristics, including age, sex, consumption quintile, and oblast, and U random error term. second part,In we the used second ordinary part, least 1 is the we used squares to %=> ln ordinary regression least & = 1 ): regression to estim %squares estimate the In the second part, we used ordinary least squares regression to estimate & the amount of cigarettes smoked per day by current smokers (ln %=>& %& = 1 ): cigarettes smoked per day by current smokers ln %=>& %& = 1 ): amount of cigarettes smoked per day by current smokers ( random error term. In the second part, we used ordinary least squares regression to estimate the amount of ln In the %&cigarettes smoked per day by current smokers ( cigarettes smoked per day by current smokers ( %=>& second ln %=> = ?@ 9:! = 1 part, we +& %& BC & used &=+D1 @) : least ordinary squares %=>& %& = 1 ): ln(2) regression to estimate the a ln %=> ln %=>% = 1 = ? & && %& = 1 = ? 9:! @@ 9:! + BC && + BC& + cigarettes smoked per day by current smokers ( D %=>& %& = 1 ): & + Dln @@ (2) (2) In the second part, we used ordinary least squares regression to estimate the amount of cigarettes smoked per day by current smokers ( + BC& + ln D@ %=>& % & = 1 = ?@ 9:! (2) ln %=>& %& = 1 = ?@ 9:!&The total price elasticity ε was calculated as: & + BC& + D@ln ): %=> & %& = 1 (2) The cigarettes smoked per day by current smokers ( lnln total %=> &The total price elasticity ε was calculated as: %=> %price = elasticity 1 1= ): ε was calculated ?@ 9:!& + BC& + D@ as: The total price elasticity ε was calculated as: (2) & &%& = ln=%=> %&!"#$ =1 % The total price elasticity ε was calculated as: The total price elasticity ε was calculated as: == ?@1 & + 9:!? BC& + D @ (2) E 1&− & ) + ?@ (3) ln %=>& %& = 1 = ?@ 9:!& + BC& + D@ The total price elasticity ε was calculated as: E= (2) 1− !"#$ ?@ (3) (3) E= 1− !"#$ %&%&==11 ??) + ) +? @ E = 1 − !"#$ %& where =1 ? The total price elasticity ε was calculated as: +the ) is ?@ E = 1 for − !"#$ coefficient = % log(price) in 1 (3) eq. ?and 1 + ? coefficient for is the (3) The total price elasticity ε was calculated as: E = 1− !"#$ %& = 1 ?) + ?@ & ) @ (3) log(price) in eq. 2. The elasticity was calculated for the whole sample, as well 1 − for as independently E = (3) each !"#$ %consumption & = 1 ?) + ?@ quintile. (3) E = 1 − !"#$ %& = 1 ?) + ?@ 39 39 39 31 Table 1 presents the estimated elasticities. The overall price elasticity is -0.54. This price elasticity is consistent with findings from other studies, which have found elasticities ranging from -0.4 to -0.8.(8) Similarly to findings from other studies,(8,32) we also found that price elasticities varied across quintiles, with the poor more responsive to prices than the rich (elasticities of -0.74 and -0.28 in quintiles 1 and 5, respectively). Table A.4: Estimated price elasticities of demand in the Kyrgyz Republic, 2015 Consumption Quintile Elasticity Quintile 1 (poorest) -0.74 Quintile 2 -0.65 Quintile 3 -0.65 Quintile 4 -0.46 Quintile 5 (richest) -0.28 Total -0.54 Source: Authors’ calculations using Kyrgyz Integrated Household Survey 2015 32 Estimating the distributional impact of increasing taxes on tobacco products: In Armenia 33