An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic Cover photos (clockwise from left): iStock Village Ushguli in Georgia, Boys fill up on fresh water. Photo: Nick van Praag/ The World Bank Group, Worker in a helmet. Kyrgyz Republic. Photo: Nick van Pragg / World Bank Group. AN EXTENDED COST- EFFECTIVENESS ANALYSIS OF TOBACCO PRICE INCREASES IN THE KYRGYZ REPUBLIC THIS STUDY SIMULATES THE IMPACT OF HIGH TOBACCO PRICE RESULTING FROM INCREASES IN TOBACCO EXCISE TAX IN THE KYRG REPUBLIC. TOBAC TAXES ARE AMON 2 // Introduction CONTENTS Acknowledgments 4 Executive summary 5 Introduction 7 Methods 13 Results 21 Conclusion 25 References 27 Annexes 30 3 ACKNOWLEDGMENTS This report was written by Iryna Postolovska, Young Professional (Economist), the World Bank, under the overall guidance and coordination of the World Bank’s Kyrgyz health team, Ha Thi Hong Nguyen, Asel Sargaldakova, and Rouselle Lavado (formerly Economist, the World Bank, and now at the Asian Development Bank). We acknowledge the funding from several sources for the study, including the Swiss Development Cooperation through an Externally Financed Output ar- rangement with the World Bank to support the Health Sector-Wide Approach-2 (SWAp2), the World Bank - Japan Trust Fund for Policy and Human Resource Devel- opment, and the World Bank Global Tobacco Control Program, supported by the Bill & Melinda Gates Foundation and the Bloomberg Foundation. We are grateful to Patricio V. Marquez, Konstantin Krasovsky, Alan Fuchs, Volkan Cetinkaya, Stéphane Verguet, and Gillian Tarr, for providing inputs, comments, and support. We thank the Kyrgyz Mandatory Health Insurance Fund for sharing data. The findings, interpretations, and conclusions in this research are entirely those of the authors. They do not necessarily represent the views of the World Bank Group, its executive directors, or the countries they represent. Washington, D.C. April 2018 4 // Introduction EXECUTIVE SUMMARY Tobacco taxes are among the most cost-effective tobacco control measures in the world. Yet often countries are reluctant to raise tobacco taxes due to their perceived regressivity. This study simulates the impact of higher tobacco prices resulting from increases in tobacco excise tax in the Kyrgyz Republic. The study uses extended cost-effectiveness analysis to measure the distributional consequences of proposed excise tax increases on: (a) averted premature tobacco-related deaths; (b) averted out-of-pocket (OOP) expenditures on treating tobacco-related disease; (c) government savings resulting from averted treatment costs for those covered under the State Guaranteed Benefit Package; and (d) averted poverty cases as a result of OOP spending. The estimates are made for the current male population only, as men engage disproportionately in smoking compared to women. Where possible, the model uses Kyrgyz Republic data, including the cost of treatment and smoking prevalence (by quintile). In addition, the quintile-specific price elasticity of cigarettes is calculated using the 2015 Kyrgyz Integrated Household Survey. Several scenarios are modelled to show the difference between various price increases. Under the base scenario of a 40 percent price increase, results suggest that the Kyrgyz Republic could avert approximately 55,000 premature deaths, US$1.3 million in OOP expenditures, US$3.8 million in government expenditure on treatment for tobacco-related disease, and 6,400 cases of poverty (using the national poverty line of US$610 per year). The benefits of higher cigarette prices would be concentrated among the poorer quintiles, with almost 49 percent of averted deaths and 45 percent of OOP expenditures averted accruing to the two poorest quintiles. However, given the higher smoking prevalence among richer quintiles, the price increase would also bring about benefits for the highest quintile in terms of reductions in OOP expenditures resulting from tobacco-related diseases. The Kyrgyz Republic has already introduced gradual tobacco tax increases that will take place up to 2022, but steps should be taken to ensure that these increases result in real price increases and to strengthen other tobacco control measures such as ensuring access to cessation services. 5 THIS STUDY SIMULATES THE IMPACT OF HIGH TOBACCO PRICE RESULTING FROM INCREASES IN TOBACCO EXCISE TAX IN THE KYRG REPUBLIC. TOBAC TAXES ARE AMON 6 // Introduction 1 INTRODUCTION Globally more than 7 million deaths a year are attributed to tobacco use, approximately 10 percent of which are among nonsmokers exposed to secondhand smoke. Most of these deaths occur in low- and middle-income countries (LMICs), and among a relatively young population (below the age of 70) (Institute of Health Metrics and Evaluation 2017). If current smoking patterns continue, tobacco will kill about 1 billion people this century. Tobacco use is a leading global disease risk factor for noncommunicable diseases (NCDs) and an underlying cause of morbidity and premature mortality. Smokers on average lose 10 years of their life compared to nonsmokers, but those who quit smoking by the age of 40 can recover nearly the full decade of life that they would have lost as a result of continued smoking (Jha and Peto 2014). Smoking-related illnesses exert a heavy economic toll on countries, due to both direct medical care costs and lost productivity among affected workers. Recent estimates suggest that the total economic burden of smoking amounts to US$1.4 trillion per year, of which approximately US$422 billion is due to health expenditures for tobacco-related illnesses (Goodchild, Nargis, and d'Espaignet 2017). Under the Sustainable Development Goals (SDGs), countries have pledged to eradicate extreme poverty, reduce premature mortality related to NCDs by one third, and achieve universal health coverage (UHC) – including financial risk protection against health-related impoverishment. These three targets are closely linked, and achieving them will not be possible without progress to increase tobacco cessation rates in most LMICs. Globally, the prevalence of smoking among the poor is higher than among the rich (Lavado and others 2017), thus placing the poor at a higher risk of NCDs and impoverishment as a result of tobacco-related illnesses. The World Health Organization Framework Convention on Tobacco Control (WHO FCTC) is a key instrument in the fight against tobacco use. It advocates for demand-side measures (e.g. price and tax increases, bans on tobacco advertising, and offering smokers to help quit); and supply-side interventions to implement bans on sales to minors and eliminate illicit trade in tobacco products. Tobacco taxes are a central component of the WHO FCTC, as price is a key determinant of smoking uptake and cessation. Evidence suggests that price increases are highly effective in reducing demand for cigarettes by inducing smokers to quit, deterring nonsmokers from initiating, reducing the number of cigarettes smoked daily, and preventing former smokers from returning to smoking (Bask and Melkerson 7 An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic 2004, Duffy 2006, Huang and others 2004, International Agency for Research on Cancer 2011, Jha and Chaloupka 1999, Önder 2002, Townsend and others 1994, Tremblay and others 1995, Wilson and others 2012). WHO currently recommends that countries increase the excise tax rate for tobacco to 75 percent (WHO 2015).1 High excise taxes on tobacco products, however, continue to be underused in LMICs, despite the high – and in many cases rising – prevalence of smoking. Introducing or increasing taxes on tobacco items has raised concerns about the impact of such increases on low-income consumers (i.e. the distributional impact). Indeed, perceived regressivity of tobacco taxation has been used as one of the main arguments against tax increases. However, looking beyond the immediate fiscal incidence of higher taxes and applying a broader definition of welfare, recent studies show that tobacco tax can be both progressive and socially desirable. Driven by the empirical assumption that the poor are more responsive to price changes, results suggest that increasing excise taxes on tobacco can be a pro-poor instrument, with health and health-related financial benefits accruing to the bottom 40 percent of the population (Denisova and Kuznetsova 2014, Fuchs and Meneses 2017, Fuchs and others 2017, Fuchs and Meneses 2018, Fuchs and others, 2018, James and others 2017, Postolovska and others 2017, Salti and others 2016, Verguet and others 2015, Verguet and others 2017). This study explores the potential health, financial, and distributional effects of increasing tobacco prices (through higher excise taxes) in the Kyrgyz Republic using extended cost- effectiveness analysis (ECEA) methodology (Verguet and others 2015, Verguet and others 2015b, Verguet and others 2016, Verguet and others 2015c). The Kyrgyz Republic is a lower-middle-income country in the World Bank Europe and Central Asia Region, with a population of approximately 6 million people. In 2016, its per capita gross national income (GNI) was US$1,100, with almost 23 percent of the population living on less than $3.20 (2011 PPP) (World Bank 2016a). Although the government provides and funds health care services under the State Guaranteed Benefit Package (World Bank 2014), out-of-pocket (OOP) expenditures represent an important source of financing, contributing 48.2 percent of current health spending in 2015 (WHO 2018). While the incidence of catastrophic health spending2 is substantially lower than in other countries of the former Soviet Union, almost 9 percent of households spend more than 15 percent of their non-food consumption on health (World Bank 2014). 1 A specific excise tax is a fixed monetary amount per quantity (e.g. per pack), while an ad valorem excise tax is levied as a percentage of the value of tobacco products (e.g. a percentage of the retail price) (Duffy 2006). 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 after tax increases (Bask and Melkersson 2004, Wilson and others 2012). 8 // Introduction Smoking is among the top risk factors in the Kyrgyz Republic, with almost 19 percent of deaths among males attributed to tobacco use (Institute for Health Metrics and Evaluation 2017). Twenty-six percent of men are current smokers, compared to less than 1 percent of women.3,4 The Kyrgyz Republic recently raised the excise tax on tobacco products. Since January 2017, all cigarettes have been subject to a tax of one som per cigarette (or 20 soms per pack). Excise taxes now represent about 40 percent of the average retail price of filtered cigarettes. Further increases of 5 soms per pack every year are expected between 2018 and 2022 (Law No. 65 signed on April 18, 2017), bringing the excise tax to 60 percent of the retail price by 2022 (see Figure 1) (World Bank, forthcoming). While the proposed changes are a significant increase, they still fall short of the WHO-recommended 75 percent rate. Smoking is among the top risk factors in the Kyrgyz Republic, with almost 19 percent of deaths among males attributed to tobacco use (Institute for Health Metrics and Evaluation 2017). Twenty-six percent of men are current smokers, compared to less than 1 percent of women. , The Kyrgyz Republic recently raised the excise tax on tobacco products. Since January 2017, all cigarettes have been subject to a tax of one som per cigarette (or 20 soms per pack). Excise taxes now represent about 40 percent of the average retail price of filtered cigarettes. Further increases of 5 soms per pack every year are expected between 2018 and 2022 (Law No. 65 signed on April 18, 2017), bringing the excise tax to 60 percent of the retail price by 2022 (see Figure 1) (World Bank, forthcoming). While the proposed changes are a significant increase, they still fall short of the WHO-recommended 75 percent rate. 2 Catastrophic health expenditure is defined as out-of-pocket spending exceeding a certain proportion of a household’s income or consumption. 3 It is possible, however, that women are more likely to underreport their smoking due to stigma and other cultural barriers, but similar patterns are observed elsewhere in the region. 4 There are some differences in the estimates of the prevalence of smoking in the Kyrgyz Republic. While estimates from WHO indicate that almost 26 percent of the population smokes, according to the 2015 Kyrgyz Integrated Household Survey (KIHS), only 13 percent of the population are current smokers. Using the definition of ‘ever smoked’, the estimated prevalence from the KIHS at 24 percent is closer to the WHO data. Using the definition of ‘ever smoked’, the estimated prevalence from the KIHS is 48% for men and 2% for women, comparable to corresponding WHO estimates of 50% and 4%, respectively. 9 An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic Figure 1: Share of excise tax in price of filter cigarettes (actual 2011–2017; forecast 2018–22) 100% 69% 64% 61% 49% 46% 53% 46% 41% 36% 32% 29% 26% 80% 15% 15% 15% 15% 60% 59% 15% 56% 15% 15% 53% 15% 49% 15% 45% 40% 39% 39% 15% 15% 36% 33% 15% 24% 20% 21% 16% 0% 2011 2012 2013 2014 2015 2016 2017 2018e 2019e 2020e 2021e 2022e Excise Tax Other Taxes Base Price Source: World Bank (forthcoming). Note: Assuming full tax incidence on consumer. The relative price of cigarettes remains low compared to other countries in the region, signaling the need to harmonize prices across neighboring countries. Given the low price of cigarettes in the Kyrgyz Republic relative to other countries in the region, the Kyrgyz Republic is at low risk of smuggling cigarettes into the country. Latest estimates, however, suggest that more than 28 percent of cigarettes sold were smuggled out of the country (WHO, 2015). The increase in excise taxes is likely to decrease smuggling out of the country as the price gap across countries narrows. 10 // Introduction Figure 2: Cigarette prices in the Eurasian Economic Union (EAEU) and neighboring countries, 2014 (US$ per pack) 3.0 2.59 2.44 2.42 2.5 2.14 1.88 2.0 Cigarette prices (US$ per pack) 1.62 1.55 1.48 1.48 1.45 1.41 1.5 1.15 1.01 0.97 0.98 0.98 0.94 1.0 0.68 0.56 0.55 0.44 0.41 0.4 0.5 0.23 0.0 Kyrgyzstan* Tajikistan China Armenia* Belarus* Uzbekistan Kazakhstan* Russia* Lowest price brand Premium brand Most sold brand Source: World Bank (forthcoming). Note: *Current Eurasian Economic Union (EAEU) Member States 11 THIS STUDY SIMULATES THE IMPACT OF HIGH TOBACCO PRICE RESULTING FROM INCREASES IN TOBACCO EXCISE TAX IN THE KYRG REPUBLIC. TOBAC TAXES ARE AMON 12 // Introduction 2 METHODS Modeling Approach ECEA is a health policy assessment tool that can be used to assess the distributional consequences of a wide range of policies (Verguet and others 2015b, Verguet and others 2016, Verguet and others 2015c). Building on traditional cost-effectiveness analysis, ECEA focuses on quantifying the equity and financial risk protection benefits of policies. Specifically, policy outcomes are examined 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 distributional consequences among the population (e.g., by socioeconomic groups or geographical area). Previous studies modeling the impact of tobacco tax increases have focused on their aggregate impact, often looking exclusively at health gains associated with tax hikes. Given perceived regressivity concerns about imposing high tobacco taxes, an emerging body of literature has begun to estimate the distributional consequences of increased tobacco taxes, including the health and health-related financial consequences. Most recently, ECEA has been applied to study the potential distributional impact of tobacco taxation in China (Verguet and others 2015, Verguet and others 2017), Lebanon (Salti and others 2016), Armenia (Postolovska and others 2017), and Colombia (James and others 2017). In this paper, a previously developed ECEA model is applied (Verguet and others 2017) to estimate the potential consequences of increasing tobacco excise taxes in the Kyrgyz Republic following a single cohort of all men alive in 2015. The analysis focuses on the following four outcomes: (a) averted premature tobacco-related deaths; (b) averted OOP expenditures on tobacco-related disease treatment; (c) government savings resulting from averted treatment costs for those covered under the State Guaranteed Benefit Package; and (d) averted poverty cases as a result of OOP spending. Each outcome is estimated by consumption quintile among Kyrgyz male smokers. This analysis is guided by the conceptual framework presented in Figure 1. Specifically it is assumed that the excise tax is directly passed on to the consumer in the form of higher prices. This would result in reduced consumption of tobacco products, leading to a decline in the incidence of tobacco-related diseases, such as stroke, lung cancer, cardio- vascular disease, and chronic obstructive pulmonary disease (Jha and Peto 2014, Doll and 13 An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic others 2004). This, in turn, would reduce the number of premature deaths associated with tobacco use. In addition, it would lead to lower OOP spending for treatment of tobacco-related diseases, resulting in fewer cases of impoverishment and catastrophic health spending. Given that a large share of health services is financed by the govern- ment, this would also result in government savings on tobacco-related health expenses. Combined with revenue from tobacco taxes, which are expected to increase at least in the short to medium term, these additional resources could be used to finance health and other sectors. Figure 3: Conceptual framework capturing the health, financial, and distributional consequences of increased tobacco taxes among male smokers Averted premature deaths Fewer cases of impoverishment tied to Reduced out-of-pocket Lower expenditures consumption of Lower incidence out-of-pocket tobacco [higher of tobacco-relat- expenditures on reduction among ed diseases tobacco-related the poor] diseases Fewer cases of catastrophic health Government expenditures Higher tobacco Higher tobacco savings on tobacco-related taxes prices diseases Increased Additional financing for governmental health and other revenues sectors Source: Postolovska and others 2017 14 // Methods Estimating Estimating Price Elasticity Estimating Price Price Elasticity Elasticity Estimating A key input Price in Elasticity the model is the price Estimating Price elasticity Elasticity for tobacco. Using 2015 Kyrgyz Integrated Estimating Price Elasticity Estimating A key input Price Household in the Elasticity model Survey is (KIHS) the price data, we estimated the price elasticities for each quintile. First, AA key key Ainput key input inputin in the inmodel the model the model isAis key the theiselasticity the input price price price in elasticity for theelasticity elasticity model tobacco. for for for the istobacco. Using tobacco. tobacco. price 2015 Using Using elasticity UsingKyrgyz 2015 2015 for 2015 Integrated Kyrgyz tobacco. Kyrgyz Kyrgyz Integrated Using Integrated Integrated 2015 Kyrgyz Integrated Household Estimating A Householdkeyin Survey order Price Household input to Elasticity ( KIHS) in Survey estimate the Survey data, model KIHS) the we KIHS) price is data, estimated the data, weofprice cigarettes we estimated the elasticity estimated price faced the elasticities for theby price nonsmokers, tobacco. price elasticities forUsing elasticities each we2015 for quintile. predicted for each Kyrgyz each First, quintile. the Integrated quintile. inFirst, price for First, A key Householdinput in the Surveymodel ( (KIHS) is the ( Household data, price we elasticity Survey estimated ( KIHS) for tobacco. the price data, we elasticities Using estimated 2015 the for each Kyrgyz price elasticities quintile. First, Integrated inin quintile. each in First, order to Household order estimate of order cigarettes to to estimate the Survey estimate price for the of (KIHS) nonsmokers the cigarettes price data, order price of tobased ofwe faced estimate cigarettes estimated cigarettes on by the the faced nonsmokers, individual’s price faced the by price of by cigarettes nonsmokers, nonsmokers, we elasticities consumption predicted faced we for by we the each quintile, nonsmokers, predicted predicted price quintile. region, the ofthe we price First, and predicted price of in of the price of Household order Survey to estimate (KIHS) thedata, price we of cigarettes estimated the faced price nonsmokers, byelasticities for we each predicted quintile. the First, price in of cigarettes A key input order cigarettes for into cigarettes whether nonsmokers the estimate model forthe for nonsmokers nonsmokers is the individual based the priceprice on cigarettes based resided of the elasticity cigarettes based in individual’s for on an nonsmokers the on urbanfor the facedtobacco. individual’s individual’s or consumption by rural based Using nonsmokers, consumption area. on consumption2015 the Following quintile, Kyrgyz individual’s we Hu region, predicted quintile, Integrated consumption quintile, and and region, others the whether region, price (1995) and quintile, of andwhether region, and wh whether Elasticity Estimating order cigarettes Price to estimate Elasticity for the nonsmokers price of based on cigarettes the individual’s faced by nonsmokers, consumption we predicted quintile, theand region, price of and whether the individual Household cigarettes the the Survey individual resided individual (KIHS) for resided an in resided nonsmokersdata, urban in wethe anin or individual estimated based an urbanrural urban onor area. resided the orthe rural Following price individual’s rural area. in an area. urban elasticities Following Hu andor consumption Following others rural for Hu Hu area. each and and(1995)Following quintile. quintile, others others (1995) Adioetomo Hu First, region, (1995) and and in others and (1995) and Adio whether Adioetomo Adioetomo lasticity cigarettes the individual for and nonsmokers Adioetomo resided and in based an others and urban on the (2005), others or rural individual’s (2005),we usedarea. we a consumption used Following two-part a two-part model Hu and quintile, model to others estimate to region, estimate (1995) the and the and whether elasticities. Adioetomo elasticities. In the first part and others order to the and estimate and (2005), individual others others we the (2005), used price resided (2005), we a of wetwo-part cigarettes in used anused aurban two-partmodel faced a two-part or rural to by model estimate nonsmokers, area. model to Following toHu estimate theestimate elasticities. we Hu the predicted and theelasticities. In othersIn elasticities. the the first price (1995) In In the part, of theand first we Adioetomo part, first wewe part, model is the price elasticity A keythe input for andin tobacco. individual others the In themodel resided (2005), Using first is part, in we 2015 the an we Kyrgyz priceused urban estimated a elasticity or estimated two-part Integrated rural the for the area. model tobacco. probability probability Following to of estimate Using ofan an 2015 individual and individual the Kyrgyz others elasticities. being being Integrated (1995) a a smoker smoker and the first Adioetomo ( part, = 1 we ) using the follow estimated cigarettes and estimated the for others estimated probability nonsmokers (2005), the the probability ofbased we an probability usedindividual on of a the antwo-part of an being individual’s individual individual modela smoker being consumption to being a estimate smoker a smoker ( 'quintile, the = 1 ) (=using elasticities. ( = region, 1= the ) using 1 In ) following and the using the whether first the ' part, following following we ymodel (KIHS) data, we estimated is the price elasticity Household and the estimated Survey price others elasticities (2005), (KIHS) the we probability data, for used weeacha of an quintile. estimated two-part logit individual equation: the modelFirst, price in to beingelasticities estimate a smoker the for each elasticities. ( quintile. ' In 1 ' ) the using First,first inthe part, following we thefor logit tobacco. equation: individual estimated using residedUsing the the 2015 in an Kyrgyz following probability logit of urban Integrated equation: or anrural individual area. Following being a smoker Hu and others ' ( (1995) 1) and using Adioetomo the following the price of cigarettes (KIHS) data, we estimated order faced to estimated the estimate by logit price logit nonsmokers, logit equation: equation: the the elasticities equation: probabilityprice we of for predicted cigarettes of anquintile. each the individual faced price First,by being of nonsmokers, in a smoker we predicted (' = 1) the' = using price theof following and others logit (2005), equation: we used a two-part model to estimate the elasticities. In the first part, + we the price based nsmokers on the of cigarettes cigarettes individual’s faced logit by for equation: nonsmokers nonsmokers, consumption we basedquintile, predicted on the region, the individual’s price andof whether consumption+quintile, ( 'region, = 1) = and.(/ whether estimated the probability of an individual being a smoker ( = 1 ) using the following Adioetomo (1) (1) ( = 1 () = + + +,- 0 1234 5674 580) (1) (1) '(= ) )123 1 = 1 = =others + ' 0) ided smokers in an urban based onorthe rural the area. Following individual individual’s resided consumption Hu inand an urban othersor quintile, (1995) rural region, and area. Adioetomo ' Following and whether( ' = 1 ') +,- Hu =.(/ and 0 4 567.(/ 4 58 .(/ 123 (1995) 4123 567 4 567 4 58 and ) 0) 4 58 logit equation: ( = 1 ) = +,- +,- .(/ +,- 0123 0 +0 4 567 4 58 0) 0 (1) ), we used a two-part and model others to (2005), estimate we the used a elasticities. two-part In where the model first (to part, is estimate we the log price of cigarettes ' the elasticities. + .(/ 123 In 0 faced the 4by 0 first part, the individual, Xi is the we (1) vector of socio- ded in an urban or rural area. Following Hu and others (1995) and ' = 1) = +,- .(/0123 Adioetomo ' +,- 4 567 58 ) obability of where is the log price of cigarettes faced by the individual, 4 567 X 4 58 is 0) the vector of socio- aan individual estimated being to a the smoker where probability where is ( of 'theis an= log the individual 1 ) log using price demographic price of thebeing cigarettesfollowing ofcigarettes cigarettes a smoker characteristics, faced faced by by the ( including the = individual, 1 using individual, ) age, XiX sex, the is consumption Xtheitheisfollowing vector the vector ofquintile, socio- of socio- and region, and U where ' the log price of faced by the 'individual, i , we used two-part model estimate where ' the elasticities. In the first part, we i is vector of socio- ' is the 'log price of cigarettes faced by the individual, + is the vector of socio- logit demographic equation: where characteristics, ' is the log including random ( age, error =sex, term. 1 ) = consumption quintile, and region, and U is the(1) demographic demographic (characteristics, )price characteristics, of cigarettes including including age,faced age, sex, bysex, the consumption individual, consumption 4 5674 580)quintile, X i is the quintile, and vector and region, ofand region, 1socio- and U1 is and U is the 1the ' .(/ 0123 bability of an individual being where a smoker demographic demographic ' is the log characteristics, price = 1 ' characteristics, of using cigarettes the including following including faced age, by age, sex, the +,- sex, consumption consumption individual, Xi isquintile, thequintile, vector andand of region, region, socio- U 1 is the random error demographic random random term. error error characteristics, term. term. including age, sex, consumption quintile, and region, and U is the demographic random error characteristics, term. including age, sex, part,consumption quintile, and region, and U 1 is the the amount of where = 1) = (' random is random ' is the the random error+ log price term. error ofterm. In the second cigarettes (' =faced 1) = byper we the (1) used + ordinary individual, Xleast squares i is the vector of socio- (1) regression to 1 estimate In the second error .(/ term. 01234 5674 580) part, cigarettes smoked .(/ day 0123 by 4 567 4 580) smokers (ln(' |' = 1)): current + we used ordinary least squares regression to estimate tothe amount of +,- +,- demographic In theInsecondthe characteristics, second second part, part, we including weused used ordinary age, ordinary sex, least consumption least squares squares regression quintile, regression to and estimate region, estimate the and theamountUamount 1 is the = 1) In (' cigarettes = the part, we used ordinary least squares(1) regression to estimate the amount ofof of random In error cigarettesthesmoked In the .(/ cigarettes term. second second 123 4 per 567 smoked 4 day we part, 58 smoked part, ) by per wepercurrent used used day day by smokers ordinary ordinary by current current ( least least smokers ln ( smokers squares squares (ln| (ln ( = regression regression ( 1 | ) ) : | = to 1= to ) estimate estimate : )) : )socio- the the amount amount of of ):) 1 +,- 0 0 ' ' log price of cigarettes where faced Intheby 'cigarettes the is second the individual, log part, smoked price we X of is i usedper the cigarettes dayvector ordinary by current of faced least socio- by smokers the squares individual, (ln regression ( X' is i | 'to ln 'the ( = 'estimate 1 ') vector ' | ' = the of 1 = amount A ' + of' + A racteristics, including demographic age, sex, cigarettescigarettes smoked smoked perper day day by by current current smokers smokers ( ln (' |' = 1)): log price of cigarettes faced by consumption cigarettes In the the second characteristics, smoked per individual, part, we quintile, Xi day isincluding used the byandvector ordinary region, current age, ln of smokers (least sex,and socio- | consumption U1 (is squares ( = ( ln )the 1| ( = regression ' | ) quintile, =+ ' ) 1) to ): and estimate + region, and U1 is the the amount of (2) (2) (2) ' | 'ln 'ln ' ' = 1= 1= = ' '+ + ' A The total price elasticity ε was calculated as: ' + + A A A ' m. random acteristics, including age,cigarettes error sex, consumption term. quintile, and region, ln and ( U 'is|the = 1 ') = A A A ' + ' + 'A (2) smoked per day by current smokers ln (( 1 ln | ') ' ( = ' | 1 ) ' == 1 ) ) = : G1 + − + ( = 1 )I + A (2) (2) m. The total price elasticity ε was calculated ln('as: |' = 1 = A ' +A' '+ A ' 'A + t, we used ordinary least In the squares second The The total part,The regression total total we priceprice usedtoprice elasticity elasticity ordinary estimateelasticity ε theε where was was least ε was amountδ calculated 1 calculated calculated squares is the coefficient of regressionas: as: as: to estimate the amount for log(price) in eq. 1 and δ 2 is of the coefficient for log(price) in e The ln ( = G1 | − = 1 ) ( = = 1 )I + + + (3) (2) total price :elasticity was calculated εsmokers as: represents the participation G1 G1 A () elasticity. (= )I )I (3) (3) d per t, day by we used currentleast ordinary cigarettes smokers The ( squares lnsmoked total ( price regression The ' | per'elasticity total = day1to price by ))estimate εcurrent was calculated elasticity the ε was amount ' (ln as: = calculated = of ' G1 = ( − − ' as: | − ' ' = 1' ( )='': + 1 ' )I 1 = 1 ' A++ ++ + A + AA A (3) where δ is the coefficient for log(price) in eq. 1 and δ is the coefficient for log(price) in eq. 2. δ δ(3) where where = G1 − ( = 1 )I + δ1δ 1 is δ 1the is the coefficient coefficient for for log(price) log(price) inin in eq. eq.1 and 1 and δis 2 is δ 2the thecoefficient iscoefficient coefficient for for log(price)log(price) inin in2. eq. 2. eq. δ2. δ 1 2 1 d per day by current smokers (total ln where (price | is= the 1 ))coefficient : for log(price) = G1 − eq. ( 1 ' = and 1 δ )I 2 ' ++ the + A for log(price) eq. (3) 1 1 1 The represents where represents the ' represents δ ' elasticity participation 1 is the the the ε was coefficient participation elasticity. participation calculatedfor log(price) elasticity. as: elasticity. in eq. 1 and δ is the A coefficient for log(price) in eq. 2. δ ln(' | where = 1 ) represents= δ1 is the A coefficient + the participation + for ln ( A log(price) | elasticity. = ' in eq. 1 1 ) = and (2) + 2+ (2) is1-0.54 – 1 'δ= 2' is the coefficient for log(price) ' ' ' Table ' 1 presents the A estimated elasticities. ' A The overall price in eq. 2.(3) elasticity δ a finding represents the participation = G1 elasticity. − ( 1)I+ + A ln(' | represents the participation consistent elasticity. with other studies, which (2) 2 is the coefficient for log(price) in eq. 2. δ1 found elasticities ranging from -0.4 to -0.8 (Internati ' = 1)δ where =1 is A the ' + ' + coefficient for A log(price) in eq. 1 and asticity ε was calculated The as: total price elasticity ε estimated was calculated Agency as: for Research onδ Cancer 2011). Reflecting findings from other studies (Internation Table represents 1 presents Table the Table 1 participation 1 the presents presents the the estimated elasticities. elasticity. estimated elasticities. The elasticities. overall The The price overall overall elasticity price price is elasticity -0.54is elasticity aisfinding – -0.54 -0.54 = as: G1 − Table ( = where1 1presents )I + is the the estimated coefficient =AgencyG1 for− for elasticities. Research log(price) ( = in The on (3) 1 eq. )I Cancer overall 1and + 2011, price is the Salti elasticity and coefficient others is -0.54 for2016), –– log(price) awe a – finding finding (3) ain also finding found that price sticity ε was calculated consistent Table ' consistent with consistent 1 other presents + with studies, with A the other other which estimated studies, studies, elasticities found which elasticities. which varied elasticities found ' found across The elasticities ranging + overall elasticities quintiles, A from price ranging with ranging the-0.4 elasticity from poor to from -0.8 -0.4 more is (International -0.54 -0.4 to to -0.8 responsive – -0.8 a finding (International (International to prices than the rich Table consistent 1 presents with other studies, which found elasticities ranging from -0.4 to -0.8 (International oefficient for log(price) = where G1 in− eq. δ11is Agency and the ( for Agency' = δ consistent coefficient eq.2 Agency is 12. Research the )I for the + coefficient forrepresents + on with Research estimated forCancer Research A log(price) other on for the on elasticities. studies, in eq. log(price) participation 2011). Cancer Cancer (elasticities 1 in Reflecting which2011). of The and 2011). found -0.74 eq. δoverall elasticity. 2 Reflecting 2. is findings and(3) δ 1 price the elasticities Reflecting -0.28 coefficient from findingsin elasticity findings other ranging quintiles from for from1 is log(price) studies from other and -0.54(International other 5, -0.4 – ain studies to finding studies respectively). eq. 2. δ1 -0.8 (International (International Participation elasticity w rticipation elasticity. represents consistent Agency the with forother Research studies, on Cancer which 2011).elasticities found Reflecting ranging findingsfrom from-0.4 other to studies -0.8 (International (International efficient for log(price) in eq. Agency Table 1 and 1 for Agency Agency δparticipation presents 2Research is the Agency for for the on coefficient Research for Research elasticity. estimated ResearchCancer on onfor 2011, log(price) elasticities. Cancer estimated Cancer on Cancer Salti2011). 2011, to and in 2011, The be eq. Reflecting Salti others 0.3. 2. overall Salti Due andδ1and 2016), to price others findings data others we elasticity limitations, 2016), also from 2016), we found is other we -0.54 however, also that also aprice – found studies found finding we that (International were that pricenot price able to estimate pric Agency Agency for varied Researchfor Research on Cancer on Cancer 2011). 2011, Reflecting Salti and findings others from 2016), other westudies also found that price (International rticipation elasticity. elasticities consistent Agency elasticities withfor Table elasticities other 1 across Research presents varied varied quintiles, studies, the across on acrosswhich elasticity Cancer estimated quintiles,with found quintiles, the bywith 2011, elasticities. poor elasticities age group. Salti with the more and The the poor Based responsive ranging others overall poor more on more evidence 2016), price from responsive to responsive prices -0.4 we elasticity from to alsoto than -0.8 recent is found to-0.54 prices the (International pricesreviews – that than arichfinding than (International price the therich rich Agency for Agency elasticities for Research varied on across Cancer quintiles, 2011, Salti with and the others poor more 2016), responsive we also found to prices that than price the rich (elasticities Agency for elasticities (elasticities (elasticities of Research consistent -0.74 varied of andon with-0.74 -0.28 Cancer of -0.74 across other and in and quintiles Research 2011). quintiles, -0.28 studies, -0.28 in on Reflecting1with quintiles in overall which and Cancer quintiles found 5, the 1 respectively). 2011, findings poor and 1 5, and elasticities Salti 5, more from and respectively). responsive 5, respectively). ranging other Participation others 2016, studies from to WHO Participation elasticity 2010), (International prices Participation than was we elasticitythe assumed elasticity rich was wasthat those bel the estimated elasticities. Tableelasticities 1 presents The (elasticities overall price the varied of estimatedelasticity across -0.74 and -0.28 elasticities. is -0.54 quintiles, age – in with of a24 quintiles finding The the were poor twice 1 more and price as respectively). elasticity responsive responsive to is to price-0.54prices – -0.4 Participation changes. a finding than to the We -0.8 (Interna- elasticity rich applied these was elasticities in orde estimated Agency for (elasticities estimated to be Research estimated 0.3. to of Due beon to-0.740.3. be to Cancer data and 0.3. Due limitations, 2011, -0.28 Due to data to Salti in data quintilesand limitations, limitations, however, others 1 and however, we 2016), 5, from however, were we respectively). we not also wewere ablefound were to Participation not not estimate that able able price to price toelasticity estimate estimate pricewas price ther studies, which consistent found he estimated elasticities. elasticity elasticities (elasticities The overall estimated with ranging other of tional price -0.74 to studies, Agency be from elasticityand 0.3. Due -0.4 which -0.28 for to Research isDue in -0.54to -0.8 founddata quintiles simulate – on aon limitations, (International elasticities Cancer the finding 1 and potential 5, 2011). however, ranging respectively). Reflecting behavioral we -0.4 findings responses were to Participation not -0.8 from as able a(International other result to elasticityestimate studies of price was price (Interna- increases. elasticities by estimated elasticity age group. varied elasticity by across to age by be ageBased quintiles, 0.3. group. group. on Based evidence to with Based data thelimitations, evidence on from poor evidence recent more from however, from reviews responsive recent recent we (International reviews to prices were reviews not (International (Internationalthan able Agencythe to rich estimate forAgency Agency price for for rch heron Cancer studies, 2011). which Reflecting Agency found forelasticity estimated elasticities findings Research toranging tional be by on from 0.3. Agency age other Cancer fromDue group. for to -0.4 2011).Based studies data to Research on Reflecting (International limitations, -0.8 (International on evidence Cancer findings however, 2011, from Salti fromrecent we and other were reviews others studies not 2016), able(International (International we to estimate also found Agency price that for price Research (elasticities on elasticity Research of Research Cancer -0.74 by on 2011, and age on Cancer -0.28 group. Cancer Salti 2011, in andquintiles Based 2011, Salti others on Salti and 1 2016, evidence and and others others5, WHO respectively). from 2016, 2016, 2010), recent WHO WHO we reviews 2010), assumed Participation 2010), we we assumed that (International assumed those elasticity that below that was Agency those those the forbelow below thethe rch ch on Cancer 2011). on Cancer 2011, Agency Salti Reflecting and for elasticity Research othersResearch by 2016), age onon group. Cancer weCancer also Based2011, 2011, found on Salti that Salti evidence and and price others others from 2016, 2016), recent WHO we reviews also 2010), found (International we assumed that price Agencythat those for below the age estimated offindings 24 Research age wereto age of be elasticities 24 from of twice 0.3. on 24 were otheras Due varied Cancer were responsive twice studies to twice data across 2011, as as (International Salti responsive to limitations, quintiles, responsive price and with to changes. othershowever, price to the price poor 2016, changes. We we more changes. WHOapplied were We responsive 2010), We not these applied ablewe applied elasticities to to these prices assumed estimate these than in that elasticities elasticitiesorder price the those rich in to below order the across ch quintiles, on Cancer 2011, with elasticities Salti the poor Research and simulate others age varied more of on the 24across responsive 2016), Cancer potential were we twice quintiles, 2011, also prices to foundSalti behavioral as withresponsive andthan that theothers responses the price poor to rich price WHO more 2016, as changes. arecent responsive result 2010), of We price to we applied prices assumed increases. these than the that elasticities rich those below in order in order the to to to elasticity age simulate by ofage simulate (elasticities 24 group. thewere potential the of Based twice potential -0.74 as and on evidence responsive behavioral behavioral -0.28 in from to responses quintiles price responses 1 andchanges. as aas 5, reviews a result result respectively). We of (International applied price of price these increases. Participation increases. Agency elasticities elasticity for in was order to 74 and quintiles, across -0.28 in quintiles with (elasticities the 1age andof poor 5,simulate of 24 more respectively). -0.74 were responsive the and twice potential -0.28 Participation as to in quintiles responsive prices behavioral than elasticity to 1the and priceresponses rich 5, was respectively). changes. as a We result applied of price Participation these increases. elasticity was elasticities in order to Research on simulate estimatedCancer the to 2011, potential be Salti 0.3. Due and behavioralto others data 2016, WHO responses limitations, as we a2010), were result not we of ableassumed price to increases. estimate that those elasticity below the .3. Due to data limitations, 74 and -0.28 in quintiles 1 simulate estimated however, andof to 5, 24 be the potential respectively). we 0.3. Due were to not data able limitations, to behavioral responses estimate however,price we as a result of price increases. were not able to estimate price age were twice Participation as responsive elasticity to price changes. was We applied these elasticities in order to group. .3. Due Based to data evidence onlimitations, elasticity fromhowever, by age recent by group. reviews we age were group.Based (International not on evidence Based able on to evidence estimate Agency from from price for recent recent reviews reviews (International (International AgencyAgency for for Research simulate the potential behavioral responses as a result of price increases. er 2011, Salti and others roup. Based on evidence from recent Research 2016, on WHO Cancer on 2010), reviews 2011, Cancer 2011, we Salti assumed (International and others Salti and others that 2016, those Agency 2016, WHO below 2010), the for WHO 2010), we assumed that those below the we assumed that those below the age wice as responsive to er 2011, Salti and others 2016, WHO age price of 24 changes. were We twice 2010), applied as responsive we these assumed to elasticities that price those changes. in order below to We the applied these elasticities in order to of 24 were twice as responsive to price changes. We applied these elasticities in order to ntial ice asbehavioral responsive to simulate responses price changes. as the a result potential We of appliedprice behavioralincreases. these elasticities responses in asorder a result to of price increases. simulate the potential behavioral responses as a result of price increases. 10 10 10 ntial behavioral responses as a result of price increases. 10 10 10 10 10 10 15 10 An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic Table 1: Estimated elasticities, Kyrgyz Republic CONSUMPTION QUINTILE ELASTICITY Quintile 1 (poorest) -0.74 Quintile 2 -0.65 Quintile 3 -0.65 Quintile 4 -0.46 Quintile 5 (richest) Table 1: Estimated elasticities, -0.28 Kyrgyz Republic ties, Kyrgyz public Republic Table 1: Estimated elasticities, Kyrgyz Republic Total Consumption -0.54 Quintile Elasticity mption Quintile city Elasticity Elasticity Consumption Quintile Elasticity Quintile 1 (poorest) -0.74 Source: Authors’ calculations using KIHS 2015. e 1 (poorest) -0.74 -0.74 Quintile 1 (poorest) -0.74 Quintile 2 -0.65 e 2 -0.65 -0.65 Quintile 2 -0.65 Quintile 3 -0.65 e 3 -0.65 Extended -0.65 Cost-Effectiveness Quintile 3 Analysis -0.65 Quintile 4 -0.46 e 4 -0.46 -0.46 Quintile 4 Quintile 5 (richest) -0.46 (corresponding to the additional -0.28 We simulated two price increases: 40 percent excise e 5 (richest) -0.28 -0.28 Quintile 5 (richest) -0.28 Total -0.54 -0.54 tax of -0.5420 soms per pack ifSource: the tax is fully transferred Authors’ calculations using to the consumer) KIHS 2015. and 60 percent Total -0.54 ng KIHS 2015.Authors’ Source: (estimated price calculations increase using KIHS 2015.by 2022 assuming the excise tax is increased by an additional 5 soms per pack per year until 2022) (World Bank, forthcoming). The base price was 45 soms Extended Cost-Effectiveness Analysis s Analysis (approximately US$0.68). Extended Cost-Effectiveness Analysis We simulated two price increases: 40 percent (corresponding to the additional excise tax of 20 creases: cent sponding 40 (correspondingpercentTax to the additional We simulated (corresponding increases to the two price were additional excise tax of soms to estimated increases: the excise 20 40 additional for taxper percent males ofpack 20 if the excise only given (corresponding taxthe tax fully is 20 of transferred disproportionately to the additional the consumer) toexcise tax of 20 and 60 percent (estimated price high smoking dfully to the nsumer) transferred soms consumer) and per 60pack the to and percent if the prevalence consumer) 60 percent (estimated tax among is fully and men price increase 60 (estimated transferred in the percent price Kyrgyz by 2022 to(estimated the assuming consumer) Republic: 26price and 60 percentthe excise of percent tax is males currently increased (estimatedsmoke by an additional 5 soms per pack per year price ds the increased by excise anincrease tax additionalbyby is an increased additional 5 2022 comparedsoms toby per assuming 5 an soms pack theadditional approximatelyper per until year excise pack 5 1tax 2022) soms per percent (World per year is increased pack of femalesBank, by per an forthcoming). year additional (National 5 soms Statistical The base per pack Committee price per of was the 45 soms (approximately US$0.68). year orthcoming). base e was price 45 soms until was 2022)The base 45 (World soms (approximatelyprice was 45 (approximately US$0.68). Bank, forthcoming). soms (approximately US$0.68). The base estimates US$0.68). price wasfrom 45 soms (approximately US$0.68). Kyrgyz Republic 2015). Using population the World Bank’s Health, Nutrition, Tax increases were estimated for males only given the disproportionately high smoking ted nly for given malesthe e disproportionately and only given Population disproportionately highthe disproportionately smoking Statistics high database prevalence smoking (World high among Bank smoking men 2016b) and KIHS 2015, in the Kyrgyz Republic: we divided 26 percentthe male of males currently smoke compared Tax increases were estimated for males only given the disproportionately high smoking the ublic: Kyrgyz 26 rcentprevalence Republic: percent of males currently of 26 males population among men percent currently into smoke of 5-year in the males compared smoke agesto approximately currently compared groups smoke and 1 percent compared consumption quintiles. Kyrgyz Republic: 26 percent of males currently smoke comparedof females We (National simulated Statistical a Gamma Committee of the Kyrgyz Republic t of ational isticalfemales to Statistical Committee (National approximately Committee of the distribution Statistical 1 Kyrgyz percentof of income Committee the Republic Kyrgyz of females 2015). (Kemp-Benedict of Republic Using the (National2001, population Kyrgyz Republic Statistical estimates SalemCommittee and Mount 1974) from of the the World Kyrgyz based on Bank’s Republic the Health, Nutrition, and Population hetimates ank’sWorld2015).from Health,Bank’s the Using World Health, Nutrition, population Bank’s Nutrition, and Health, Population estimates and Statistics Nutrition, Population from theKIHS database and Population World (World BankNutrition, Bank’s 2016b) and KIHS 2015, we divided the male population into 5- consumption quintile cutoffs from 2015 and aHealth, Gini index of 27 (World and Population Bank 2016b). ndBank 5, KIHS we 2016b) 2015, divided and we the KIHS divided male 2015, the we male population dividedinto year population5-the ages maleinto Statistics database (World Bank 2016b) and KIHS 2015, we divided the male population groups population 5- and consumption into 5- quintiles. We simulatedinto 5-a Gamma distribution of income umption quintiles. TheWe agesimulated and quintile-specific a Gamma baseline smoking (Kemp-Benedict distribution 2001, of incomerates Salem were andestimated Mount using the 1974) basedKIHS on the consumption quintile cutoffs from es. We ulated a Gamma yearsimulated ages groups a Gamma distribution distribution of income quintiles. We simulated a Gamma distribution of income and consumption of income d m 1974) onandthe Mount based consumption (Kemp-Benedict 2015 on1974) and based the consumption 2001,were quintile on used Salemthe quintile cutoffs to calculate consumption andfrom KIHS Mount cutoffs the 2015 1974) total and quintile from anumber basedGini cutoffs on theof index smokers from of 27 in each (Verguet age and and consumption quintile cutoffs fromothers 2017). The age and quintile-specific of sand 27 2017). (Verguet others KIHS The 2017). ageconsumption 2015 and and and Theothers a Gini 2017). and ageindex quintile. quintile-specific of The baseline age quintile-specific 27 (Verguet smoking and quintile-specific and others rates 2017). were The estimated using the KIHS 2015 and were used to calculate the age and quintile-specific re ing estimated S 2015the KIHS and were baseline using 2015 used to smoking the and KIHS were 2015 used calculate rates and to were estimated the total were calculate usednumber the to of smokers calculate the using the KIHS 2015 and were used to in each age and consumption calculate the quintile. n on each consumption age quintile. and total number consumption quintile.of smokers Similar quintile. to Postolovskain each age and and consumption others (2017), we calculated quintile. the number of individuals by Similar to Postolovska and others (2017), we calculated the number of individuals by age group others we (2017), calculated age we the group a calculated andthe consumption number aby and ofage quintile consumption individuals q who by age would quintile group quit (from q who the current would adult the quit (from male current adult male smoking population) ed the number Similar to of number individuals Postolovska of individuals and age byothers group(2017), wegroupcalculated the number of individuals by age group e theq who quit (from current a andwouldthe adult quit current smoking male consumption (fromadultthe smoking current male population) quintile q smoking population) who or oradult notwould notmale initiate initiate population) quit smokingsmoking (from population) (among the (among current those those below adult malebelow the agethe smoking ofage 16), of 16), denoted ∆L,N . This would depend denoted population) mong w the of 16), those age ordenoted below of 16), not initiate the denoted age ,N . This ∆Lsmoking of 16), ∆ would This would(among . denoted This depend L ,N depend would thoseon ∆ on below the the . depend L,N This initial initial would age thenumber numberof 16),depend of denoted( of smokers smokers ∆ L ,N ,N . L), theThis would depend participation participation + elasticity (A), price elasticity L,N (per + + + mokers (the e participation tion L ,N ), elasticity on the ), participation (Aelasticity initial price number (A), elasticity ofprice elasticity smokers elasticity ,N (per (L ( ), price,N (per elasticity (per L,Nelasticity + ∆O elasticity (1/2), price Lelasticity age A,N ), the group L participation (perandage group quintile), and and (A), price quintile), relative and price elasticity relative change price ( ): (per L ,N ∆O ∆O O O nd ):relative change age(group price change ): change and O ( ): and relative price change ( ): quintile), ∆O O O + ∆O Q L,N . ∆L,N = P N,L (1) + ∆O + ∆O O A PL,N . ∆Q = P. Q L,N . (1) ∆ (1) ∆O+ L ,N = PA N,L O Q L ,N . (1) (1) A N,L O L,NL,N A N,L O Using estimates from Doll and others (Doll and others 2004, Verguet and others 2017) to model and lsandothers others 2004, Using (Doll Verguet 2004, estimates and others Verguet from Doll2004, and 2017)others andVerguettheand 2017) toothers model changes toothers (Doll and in model expected 2017) others mortality, to model 2004, we calculated Verguet and others 2017)theto model of premature deaths averted number ortality, culated number we the calculated number of premature the changes 16 in of the number premature deathsmortality, // expected Methods of ( premature deaths ∆ averted L,N ) based deathson the avertedage at cessation averted we calculated the number of premature deaths averted ( L ). We assumed that half of all deaths among smokers t cessation We assumed med that (∆ ( half L,N ) of ). that L all deaths based We assumed half of on the all among that deathshalf age at smokers of among all were deaths smokers among cessation (L ). We assumed that half of all deaths among smokersand others 2008, Jha and others attributable tosmokers smoking as shown in equation (2) (Jha ng 2) shown aswere equation (Jha and (2) in (Jha others equation attributable and 2008, to (2) others Jha (Jha 2008, and smoking andJha others as 2013). others and shown 2008, others Jha and in equation others (2) (Jha and others 2008, Jha and others 2013). Similar Similar toto Postolovska Postolovska and and others others (2017), (2017), we we calculated calculated the the number number of of individuals individuals by by age age grou group a and consumption quintile q who would quit (from the current adult male smoking a and consumption quintile q who would quit (from the current adult male smoking population populatio or or not not initiate initiate smoking smoking (among (among those those below below the the age age of of 16), 16), denoted denoted ∆ ∆ L ,N L,N .. This This would would depend depend ++ on on the the initial initial number number of of smokers smokers ( ( L ,N L,N ),), the the participation participation elasticity elasticity ( (), A ), price price elasticity elasticity L,N (per L,N(per A ∆O ∆O age age group group and and quintile), quintile), and and relative relative price price change change (( ):): OO ++ ∆O ∆O ∆ ∆ L ,N= L ,N =PPA A N,L O O L ,N L,N . . N,L QQ ( Using estimates from Doll and others (Doll and others 2004, Verguet and others 2017) to Using Using estimates estimates from from Doll Doll and and others others (Doll (Doll and and others others 2004, 2004, Verguet Verguet and and others others 2017) 2017) toto mod mode model thethe the changes changes changes in expected in in expected expected mortality, we mortality, mortality, calculated we we calculated calculated the number the the number number of premature of of premature deaths premature deaths deaths averted averted (∆ averted(∆ ) L,N L,N ) based based based on onon the the age age at at at cessation cessation cessation ( ( L ). L ). . We We We assumed assumed assumed that that that halfhalf halfof of of all all all deaths deaths deaths among among smokers smokers amongwerewere smokers attributable attributable to smoking to smoking were attributable as shown as shown to smoking as shown in equation in equation (2) (Jha (2) (Jha in equation and and (2) (Jha others others 2008, and others Jha and others 2008, Jha and others 2013). 2013). 2008, Jha and others 2013). + ∆O 1 ∆L,N = PA N,L O Q L L,N . (2) To calculate OOP and government expenditures averted, we distributed averted premature To calculate OOP ∆O and government expenditures averted, we distributed averted deaths to ∆four main + causes: L,N = PA N,L O Q heart L L,N . disease, lung cancer, stroke, and chronic (2) obstructive + ∆O pulmonary premature disease (COPD) deaths (Institute to four ∆ main for causes: L,N Health = P N,L heart Metrics Qdisease, L and L,N lung . Evaluation cancer, stroke,The 2017). andaverage chronic (2) A O+ ∆O + ∆O + ∆O ∆O treatmentobstructive costs for the four main pulmonary causes disease of smoking-related ∆ (COPD) = P L,N (Institute for∆ Q Health deaths L,N . were Metrics + obtained and + Evaluation. from the 2017). (2) A ∆ O = L,N P ∆ ∆ = Q P L Q . ∆O L To calculate OOP and government expenditures averted, we distributed averted N,L premature L,N L A= N,L L,N P AO N =,L AP Q,L NL,N OA N,L L O L,N Q .L,N L L,N . Mandatory Toheart calculate The Health OOP average and Insurance government treatment Fund costs (MHIF). expenditures for the According ∆ four averted, main +to the PA causesN,Lwe ∆O MHIF, distributed of 75 smoking-related L,N L,N . averted percent of the premature deaths O population were (2) + deaths to four main causes: disease, lung + cancer, ∆O stroke, and chronic L,N = obstructive Q L ∆L,N = PA N are covered deaths to∆ four by insurance main= POOP N,Land causes: and Qheart thus, L L,N . them, for disease, lung + out-of-pocket O payments (2) wereobstructive calculated using the Ncancer, stroke, and chronic ∆O pulmonary disease (COPD) (Institute To calculate for obtainedL,N Health Tofrom A calculate MetricsTothe O Mandatory Togovernment and calculate OOP calculate ∆ and Evaluation OOP government = Health OOPexpenditures and PA2017). Insurance and government Qexpenditures The government averted, LFund average .(MHIF). expenditures weaverted, expenditures distributed According averted, we to averted, averted distributed thewe MHIF, we premature distributed 75 averted distributed (2) averted premature averted prema pr average pulmonary co-payment disease (COPD)rate per case. (Institute To calculate For L,N for the Health OOP remaining and ,L Metrics O government 25 and L,N percent Evaluation of expenditures the population, 2017). The averted, we average we assumed distributed averted reatment costs for the four main deaths causes To calculate to of four deaths OOPmain smoking-related to and causes: four deaths governmentmain heart deaths to causes: four disease, were main expenditures heart lung obtained causes: cancer, disease, from heart averted, and To lungstroke, the disease, wecancer, calculate andlung distributed chronic stroke, cancer, OOP and and obstructive averted stroke,chronic government and premature obstructive chronic obstruc that they treatment percent paid costs the for oftotal the the cost deaths population four (MHIF main to four deaths are average causes main covered to four causes: treatment smoking-related of percent mainby insurance heart causes: cost disease, heart and deaths thus, disease, co-payment)lung for were cancer, them, lung obtained stroke, out-of-pocket cancer, out-of-pocket. from and stroke, the chronic The and expenditures obstructive chronic obstr To calculate Mandatory and government OOP Insurance Health Fund pulmonaryexpenditures (MHIF). deaths to disease pulmonary According four averted, main (COPD) to disease the pulmonary causes: we MHIF, distributed (Institute heart(COPD) 75 disease for disease, averted Health (Institute (COPD) of lung premature Metrics the for cancer, Health population (Institute and deaths for stroke, Evaluation Metrics Health to and fourand Metrics chronic2017). Evaluation main andTheEvaluation obstructive causes: average 2017).heart The 2017). disease, average The lung ac To change calculate payments indisease, OOP OOP and spending wereby pulmonary government calculated quintile pulmonary disease expenditures using (∆ the (COPD) disease ) average averted, is calculated (Institute (COPD) co-payment we(Institute as: for distributed Health rate forper Metrics averted Healthcase. of For and Metrics premature Evaluation thepopulation and Evaluation 2017). remaining 2017). The avera Th aredeaths covered to four main causes: by insurance Mandatory and heart thus, treatment for Health them, Insurance lung costs treatment cancer, for out-of-pocket the Fund costs treatmentstroke, four (MHIF). main for payments and the costs chronic causesAccording four N were for mainof the to the obstructive smoking-related causes calculated four main MHIF, of using smoking-related causes 75 the percent deaths of were smoking-related the obtained deaths were from deaths obtained thewere from obtained the deaths (Institute topulmonary four 25for main percent disease causes: treatment the of Metrics (COPD) heart population, treatment (Institute costs disease, we for lung costs assumedthe for fourHealth cancer, for that main the four Metrics stroke, they causes main paid and pulmonary and causes theof Evaluation smoking-related chronic total ofcost disease obstructive smoking-related (MHIF 2017). (COPD)deaths average The average (Institute were deaths obtainedfor were obtain Health from M pulmonary average disease (COPD) co-payment rate per are covered case. Mandatory For by the insurance Health Health Mandatory remaining and Insurance 25 thus, and Health Mandatorypercent for Evaluation Fund them, Insurance of (MHIF). the Health out-of-pocket 2017). Fund population, InsuranceAccording The (MHIF). average we Fund payments to the According assumed(MHIF). MHIF, were to According 75 thecalculated percentMHIF, to of the75 using the percent MHIF, the population75 of the percent populati pulmonary treatment disease costs (COPD) for Mandatory the (Institute four Mandatory main Health for Health causes Insurance Health Metricsof smoking-related Insurance Fund and ∑ (MHIF). Evaluation Fund treatment According (MHIF). deaths IOOP 2017). costs were According to The for the the obtained MHIF, average to four the 75 from main MHIF, percent causes the 75 of percent ofof the smothe po of treatment costs average co-payment rate ∆ per case. = For (1 the − out-of-pocket.)G∑ remaining Δ 25 percent X X,N of the , population, payments we quintile assumed (3) hat they paid thefor total the four cost (MHIF treatment main are causes covered treatment average Mandatory costs for the of smoking-related by are cost treatment Health insurance and covered are Insurance four co-payment) cost are covered main by are and N and covered deaths insurance causesby covered thus, co-payment) Fund by insuranceforNwere insurance of(MHIF).and them, smoking-related by obtained thus, L insuranceand out-of-pocket L,N Accordingfor out-of-pocket. and thus, The and Xfrom them, thus,change for to deaths thus, the for the The Mandatory them, payments in out-of-pocket X them, for MHIF, wereout-of-pocket them, spending 75 Health obtained were out-of-pocket percent out-of-pocket calculated by Insurance payments from were ofpayments the Fund payments using calculated population were werethe (MHIF).calculated were using calcula Accor calcu u Mandatory hange in OOP Health spending Insurance bythat quintilethey Fund average paid (MHIF). ( ∆the total According co-payment average isby cost calculated (MHIF to as: rate co-payment theper as: average MHIF, case. rate 75 For treatment percent per the remaining case. ofForcost the the andpopulation 25 co-payment) percent remaining of 25 the out-of-pocket. percent population, of the we The population, assumed we assum Mandatory are covered Health N ) is calculated Insurance insurance average average Fund average and co-payment co-payment (MHIF). thus, co-payment forrate According them, rate per per to case. out-of-pocket case. rate per the ForMHIF, case. For are the the covered payments remaining 75 For the remaining percent by remaining 25 ofinsurance were the 25 percent percent calculated population and of the of thus, 25 percent of the populatio the using for population, population, thethem, weou are covered by insurancewhere change and thus, in that N isOOP averagefor the they spending them, fraction paid that co-payment the theyof by out-of-pockettotal population quintile paid thatrate cost the they per (∆ payments (MHIFtotalcovered paid case. average cost the N ) is For were by calculated (MHIF total the insurance calculated treatment cost average remaining (MHIFas: in using cost 25 quintile treatment average and percent the ,cost co-payment) treatment of co-payment X is the and contribution theco-payment) cost population, out-of-pocket. rate and per co-payment) wecase. assumed The out-of-pocket. For the out-of rema T are covered by insurance that and they thus, that paid for they the them,paid total cost out-of-pocket the total (MHIF cost average payments (MHIF treatment average were calculated treatment cost and co-payment) using cost and the co-payment) out-of-poc out average co-payment rate (inper %) case.of changedisease For the in OOP to remaining change tobacco-related spending in 25 OOP percent by quintile spending premature of the ( population, ∆ by quintile deaths, ) is calculated ( we ∆ is the assumed ) is as:treatment calculated cost as: for disease , ∆ average = N that (1 co-payment − theyNpaid )G∑ rateLtheΔ change per change total L,N case. ∑ in Xcost change OOP X Forin (MHIF X,N OOP spending in the X remaining OOP I spending , average spending by quintile N by treatment 25by quintileX percent (∆ quintileN cost ( that of (3) (and they ∆ the ) ) is calculated ∆ is co-payment) paid population, calculated the total N ) is calculated as: we as: out-of-pocket. cost assumed as: (MHIF average The treat that they paid the total cost and(MHIF X,N isaverage change the utilization treatment of health cost ∆ = services and (1co-payment) − N for)G∑ disease L Δ ∑ out-of-pocket. per X quintile change N The .NSimilarly, X I , government by quintile (3) (∆ ) is c that they paid the OOP in total spending cost (MHIF N by quintile average as: covered by insurance ( treatment ∆ N ) L,Nis cost calculated X and co-payment) X,N as: in OOP spending out-of-pocket. The N change where Nin isOOP spending the fraction ofsavings by change quintile population ( in OOP (∆ covered _L`'ab_,N spendingN ) is by ) calculated among insurance by quintile those ∆ in(quintile N = (1 ∆ N ) ∆ is ,− Nis calculated )G∑ N the=∆ X ∆ (1 Δ − would L contribution =as: L,N N ∆ N N (1 = ∑)G∑ − (1 X = be: L N − X (1 Δ )G∑ X,N − NL,N L )G∑ Δ X∑ IX )G∑L , L,N Δ X ∑ L,NX,N Δ X ∑ X X∑ X I , X,N X XX,N I , X I , I (3) , in %) of disease to tobacco-related where N is premature the fraction of population deaths, X is the covered treatment by insurance cost for in quintile disease N , X , N is the X L contribution L,N X X X,N X )G∑ ∆ N = (1 − N )G∑ L Δ L,N ∑X X,N X I , ∆ N = (1 (3) − N (in and X,N is the utilization of health services %) ∆ of where disease = where (1 − is to )G∑ the where N for disease is tobacco-related fraction the Δ fraction is ∑ N where of the per population of fraction quintile premature population is the= of I . G∑ , covered fraction covered population Similarly, deaths, Δ of by by ∑insurance covered government is insurance population the treatment by inin covered I (3) quintile. quintile insurance by cost in, insurance for is quintile disease the the in contribution , (4) quintile , is the , contributi is the c N N Lwhere L,N ∆ _ where N is L`'ab_,N X = the X (1 X,N −fraction N N is the X N N )G∑ of fraction L L population Δ L,N L,Nof X ∑ XX population X covered X,N X I by , X X,N X covered by insurance in quintile insurance X in quintile (3) X , X Xis the , Xcont is th avings (_L`'ab_,N ) among and those (in is %) where the covered ofutilization contribution disease (in isby the %)(inof insuranceof to %) fraction health tobacco-related disease of (in disease %) would of services to population of disease d tobacco-related to be: for premature tobacco-relateddisease covered to tobacco-related by per deaths, premature premature insurance quintile where is premature the deaths, deaths, in . Similarly, treatment quintileis the is is fraction deaths, the , government cost treatment treatment is the of for disease contribution population cost for , disease covered b X is the treatment cost cost fo X,N N (in %) of (in disease %) of to disease tobacco-related tois tobacco-related premature X N deaths, premature X XX deaths, is the Xtreatment is the treatment forcosdis where N is the fraction of To savingspopulation calculate and ((in poverty covered is the and )by cases among insurance utilization averted, is the thoseof in we health quintile covered utilization estimated services of by , the insurance health for the contribution number disease services would of for individuals per be: disease quintile for per whom . Similarly, quintile the government . Similarly, governmen where N iscost %)the X,N of for fraction _L`'ab_,N disease disease of and X,N to and ,population X,N and tobacco-related and is X,Nthe is is the covered utilization the is the utilization utilization by premature utilization insurance ofX of of health health health deaths, ofline, in services health services quintile (in services %) services X is for of for the for, disease disease disease X disease is treatment for disease the per tocontribution per per cost quintile quintile quintile tobacco-related for disease .. per quintile . Similarly Similarly,. Similarly, premature , gover go (in %) of disease to tobacco-related simulated savings annual premature ( consumption savings deaths, ( ) among was X is above the X,N those ) given acovered treatment among poverty those cost bythose for insurance covered disease but by would whose , is quintile insurance be: annual would consumption be: (in %) _of and disease L`'ab_,N Similarly, = X,N N is tothe G∑ utilization tobacco-related _L`'ab_,N government L savings Δ savings ∑ L,N savings ( _ of L`'ab_,N X X _ ( health premature X,N ( L`'ab_,N X I services _L`'ab_,N . ) among ) among deaths,for) among disease those is the covered X those covered and covered per (4) treatment by covered X,N bythe insurance byinsurance cost utilization . byinsurance Similarly, for insurance would diseasewouldof government be: health would be: be: , services for d and X,N is the utilizationwould of health have fallen below services for disease the poverty _per line quintile =after . deducting _L`'ab_,N G∑ Similarly,Δ ∑ tobacco-related government I . medical expenditures. (4) and X,Nsavings the utilization iswould (be: _L`'ab_,N of health ) among services L`'ab_,N those for Ncovered disease L by L,N 5 insurance per quintile X X savings X,N would X be:_L`'ab_,N .( Similarly, government ) among those covered by savings (_L`'ab_,N ) among We used those thecovered national by poverty insurance line of US$610 would be: per = year. G∑ TableΔ = 2 presents ∑ G∑ Δ the ∑I inputs . X∑ used X I for . the (4) To calculate poverty averted, casessavings we estimated ( ) the among number those of individuals covered _ L`'ab_,N by for insurance _ whom N L`'ab_,N L would the L,N N = Xbe: L X = N = G∑ X,N L,N G∑ Δ X N N G∑ X Δ L,NL Δ X,N ∑ X X∑ X X I .X I X . I . ECEA To calculatemodel.poverty _L`'ab_,N cases averted, we estimated the number _L`'ab_,N _L`'ab_,N of individuals _L`'ab_,N L for L whom L,NX L,N X,N the XX X,N X,N imulated annual consumption was above a given poverty line, but_L`'ab_,N whose annual = N G∑ consumption L ΔL,N ∑X X X,N X I . _L`'ab_,N (4)= N G∑ would have fallen below the poverty line after simulated To annual calculate = consumption povertyG∑ Todeductingcalculate Δcases was ∑ poverty averted, tobacco-relatedabove cases a we given I . estimated averted, medicalpoverty we the line, estimated expenditures. number but (4) whoseof the individuals number annual of consumption for whom individuals the for whom the Table 2: _ L`'ab_,N Data Inputs N for the L To ToL,N calculate calculate Modeling To X calculate _L`'ab_,N X poverty X,N poverty = Approach, X cases poverty N G∑ cases averted, Δ LKyrgyz cases averted, ∑ L,N Republic we X X X,N we averted, estimated we estimated I . the the number X estimated the number of individuals number of of individuals individuals (4) for for whom who for t w We used the national poverty would line have simulated of fallen US$610 below annual simulated per thesimulated consumption year. 5 poverty annual Table 2 line was consumption presents annual afterabove deducting the consumptionawas given inputs above tobacco-related poverty used was a given for aboveline, the but poverty a given medical whose line,poverty expenditures. annual but whose line, consumption but annual whose consumpti annual c To calculate poverty simulated cases simulated averted, annual we estimated consumption annual consumption was the above number wasTo calculate a given above of individuals a poverty poverty given line, poverty cases forbut whom averted, line,whose the whose but we estimate annual consu annua To calculate ECEA model. poverty cases We used averted, would the we national estimated have fallen would poverty the below have number would line the fallen ofhave US$610 of was poverty below individuals fallen line the per after poverty below year. for the 5 Table whom deducting linepoverty Value after 2 the presents tobacco-related deducting line after the inputs tobacco-related deducting Sourcemedical used for the expenditures. tobacco-related medical expenditu medical e To calculate simulated poverty annual cases would consumption averted, have would fallen we have estimatedbelow above fallen thethe belowa given poverty number the poverty line poverty of simulated afterline, individuals line but deducting afterannual whose for whom deducting consumption annual tobacco-related the consumption tobacco-related was above medical a give expe medica simulated annual consumption ECEA Male model. was We above used population the a We given in national Kyrgyz used poverty poverty Republic the We line, national used line the but of povertywhose nationalUS$610 annual line per poverty of consumption year. US$6102,953,396 line 5 Table of per US$610 2 year. presents 5 perTable year. the 2 5 inputs presents Table 2usedthe presents for inputs the theused for inputs t simulated wouldannual have fallen consumption To calculate poverty cases We below used was We the used poverty national above averted, thewe a line givenpoverty national estimatedafter line poverty deductingof line, the number US$610 would but line of US$610 whoseperhave tobacco-related of individuals year. fallen annual 5 per year. Table below medical consumption for whom 5 2 the presents poverty expenditures. Table 2 presents the inpu the line inputs after d used would have fallen Table 2: Data Inputs for the below the Male poverty Modeling ECEA population line model. Approach, after ECEA deducting distribution, model. Kyrgyz tobacco-related Republic age group (years) medical expenditures. would have We the used fallensimulated the below national theECEA ECEA poverty model. poverty ECEA model. line model. line afterof US$610 deducting per year. tobacco-related 5 We Table used 2 presents the medical national the inputs poverty expenditures. per used line for of US$610 the pe We used the national poverty Tableline 2: Data of US$610 Inputs per forannual the Modeling year. 5 consumption Table 2 Approach, presents was above the Kyrgyzinputs theRepublicnational used for poverty the line of US$610 We used 15 < ECEAthe national model. poverty lineValue of US$610 per year. Source5 Table 32% 2 ECEA presents model. the inputs used for the ECEA model. Table year, 2: Data 5 butTable whoseInputs 2:annual for Data Table the consumption Inputs 2: Modeling Data for the Inputs would Approach, Modeling for have the fallen Kyrgyz Modeling Value Approach, below Republic Kyrgyz Approach, the poverty World Republic Source Kyrgyz Bank line after 2015 Republic Male population in Kyrgyz Republic ECEA 15–24 model. Table 2: Data Table 2,953,396 Inputs for the 2: Data Inputs for the Modeling Approach, Modeling 20% Approach, Kyrgyz Republic Kyrgyz Republic Male deducting population Table 2: Data in tobacco-related Kyrgyz Inputs Republic for the medical Modeling expenditures. Approach, Table 2,953,396 Kyrgyz 2 presents Table Value Republic 2: Data theInputs inputs Value used for Source the forModeling Male 2: Table population Data Inputs distribution, for the Modeling 25–44 age group (years) Kyrgyz Republic Approach, 29% Value Value Value Source Approach Source Sou S Table Male 2: Data population 45–64 the Male Inputs ECEA population model.for the distribution, Male in population Male Modeling KyrgyzMale age population group Republic in Approach, population Kyrgyz (years) in KyrgyzRepublic in Kyrgyz Kyrgyz Republic Republic Republic 16% 2,953,396 Value 2,953,396 2,953,396 Source 2,953,396 < 15 Male32% Value population in Kyrgyz Source Republic 2,953,396 15–24 < 65 ≥ 15 MaleMale population population Male Male distribution, population in KyrgyzMalepopulation 20% age group distribution, Republic population World distribution, (years) age Bank distribution, group 2015 age 32% Value age group(years) 3% Male group 2,953,396 (years) population (years) Source in Kyrgyz Republic Male population in Kyrgyz Republic Male 2,953,396 population distribution, age group (years) Male15–24 smoking population Male < 15 prevalence, in Kyrgyz population < 15 per Republic distribution, age < group 15 age group (years)2,953,396 20% Male 32% World population Bank 32% 2015 32% age group (years distribution, 25–44 < 15 29% < 15 32% 32% Male population distribution, 25–44 age 25–44 group 15–24 (years) 15–24 29% 30% 20% World20% Bank 2015World World Bank Bank 2015 45–64 < 15 Male population < 15 distribution, age 15–24 group 15–24 16% 32% 15–24 (years) 32% < 15 20%20% 20% World World Bank 2015 Ban ≥ 65 45–64 < 15 2015 25–44 We estimated this number using the poverty headcount 15–24 consumption 5 data. 25–44 25–44 25–44 3%25–44 World Bank 2015 16% 32% 29% ratio from the National Statistical Committee of the Kyrgyz Republic of 32.1% in 2015 and KIHS 20% 15–24 29% World Bank 29% 29% 29% 2015 15–24 ≥ 65group 20% 3% Male smoking prevalence, per 15–24 age 45–64 25–44using45–64 45–64 45–64 20% 16% 29% World Bank 25–44 16% 201516% 16% 25–44 5 We estimated this number the poverty headcount 29% 45–64 ratio from the National Statistical Committee of the Kyrgyz Republic 16% 25–44 Male of 32.1% insmoking 25–442015 and ≥ 65prevalence, KIHS 45–64 2015 consumption ≥ 65 per ≥ 65 age data. ≥ 30% group 65 29% 3%16% 45–64 3% 3% 3%3% 45–64 ≥ 16%age group 65 17 25–44 45–64Male smoking ≥ 65 Male prevalence, smoking Male Malesmoking per prevalence, smoking prevalence, per prevalence, age per groupper age 30% 16% age group group 3% ≥ 65 ≥ 65 Male 3% smoking prevalence, per age group poverty ≥ 65 Male 25–44 smoking 25–44 prevalence, 25–44 25–44 per age group 3% Male 30% smoking30% prevalence, 30%30% per 12 age group WeMaleestimated this number smoking prevalence, using the per age group headcount ratio from the National 25–44 Statistical Committee of the Kyrgyz Republic 30% f 32.1% in 2015 and KIHS 2015 consumption 5 We Male estimated smoking data. this number prevalence, 25–44 using the poverty per age group headcount ratio from the National Statistical 30% 25–44 Committee of the Kyrgyz Republic 25–44 of 32.1% in 2015 and KIHS 30% 2015 consumption data. An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic Table 2: Data Inputs for the Modeling Approach, Kyrgyz Republic VALUE SOURCE Male population in Kyrgyz Republic 2,953,396 World Bank 2015 Male population distribution, age group (years) < 15 32% 15–24 20% 25–44 29% 45–64 16% >65 3% National Statistical Male smoking prevalence, per age group Committee of the Kyrgyz Republic 2015 25–44 30% 45–64 38% >65 15% National Statistical Male smoking prevalence, by wealth quintile Committee of the Kyrgyz Republic 2015 Q1 25% Q2 26% Q3 25% Q4 25% Q5 30% National Statistical Daily cigarette consumption 13 Committee of the Kyrgyz Republic 2015 World Health Price per pack of cigarettes (2014 US$) 0.68$ Organization 2015 Mandatory Health Tobacco-related disease treatment costs (2014 US$) Insurance Fund (MHIF) COPD 119 Stroke 129 Heart disease 148 Neoplasm (lung cancer) 161 National Statistical Individual annual expenditure (2014 US$) Committee of the Kyrgyz Republic 2015 Q1 <537 Q2 537–660 Q3 661–817 Q4 818–1070 Q5 >1070 18 // Methods In addition, we calculated the baseline scenario of a 40 percent increase in tobacco prices using the WHO smoking prevalence rates for men. Given that the breakdown was not available by quintile, we applied the relative shares obtained from the KIHS smoking prevalence by quintile. We also estimated the potential impact of increasing cigarette prices by 75 percent – the observed real change in prices between 2015 and 2017. 19 THIS STUDY SIMULATES THE IMPACT OF HIGH TOBACCO PRICE RESULTING FROM INCREASES IN TOBACCO EXCISE TAX IN THE KYRG REPUBLIC. TOBAC TAXES ARE AMON 20 // Methods 3 RESULTS The results from the ECEA analysis are presented in Table 3 and Figures 4–6. Under a 40 percent price increase, our results suggest that the Kyrgyz Republic could avert approximately 55,000 premature deaths, US$1.3 million in OOP expenditures, US$3.8 million in government expenditure on treatment for tobacco-related disease, and 6,400 cases of poverty (using the national poverty line of US$610 per year). As expected, the magnitude of the results is substantially higher under a 60 percent price increase, which could avert approximately 62,900 premature deaths, US$1.5 million in OOP expenditures, US$4.8 million in government expenditure, and 7,300 poverty cases (using the national poverty line of US$610 per year). Table 3: Extended cost-effectiveness analysis results by individual consumption quintile, Kyrgyz Republic TOTAL Q1 (POOREST) Q2 Q3 Q4 Q5 (RICHEST) 40% price increase simulation 55,000 13,000 13,200 8,900 Premature deaths 14,000 5,800 (44,000– (10,000– (11,000– (7,100– averted (11,000–17,000) (4,700–7,000) 66,000) 16,000) 16,000) 11,000) Out-of-pocket 3,800,000 300,000 320,000 230,000 160,000 expenditures for 270,000 (3,060,000– (240,000– (250,000– (180,000– (130,00–190,00) tobacco-related diseases (210,000–320,000) 4,580,000 360,000) 380,000) 280,000) averted (US$) Government 3,800,000 910,000 950,000 690,000 480,000 expenditures for 800,000 (3,060,000– (730,000– (760,000– (550,000– (380,000– tobacco-related diseases (640,000–960,000) 4,580,000) 1,090,000) 1,140,000) 830,000) 580,000) averted (US$) Poverty cases averted 2,600 3,800 (using the national 6,400 0 0 0 (2,100– (3,100– poverty line of US$610 (5,200–7,600) 0 0 3,100) 4,500) per year) 60% price increase simulation 62,900 14,800 15,100 10,200 Premature deaths 16,000 6,700 (50,300– (11,900– (12,100– (8,200 averted (12,800–19,300) (5,300–8,013) 75,500) 17,800) 18,100) -12,300) Out-of-pocket 1,460,000 350,000 360,000 260,000 expenditures due to 300,000 180,000 (1,160,000– (280,000– (290,000– (210,000– tobacco-related diseases (240,000–360,000) (150,00–220,000) 1,750,000) 420,000) 430,000) 320,000) averted (US$) Government 4,370,000 910,000 1,000,000 1,100,000 790,000 expenditures for 550,000 (3,490,000– (730,000– (830,000– (870,000– (630,000– tobacco-related diseases (440,00–660,000) 5,240,000) 1,090,000) 1,250,000) 1,300,000) 940,000) averted (US$) Poverty cases averted 3,000 4,300 0 (using the national 7,300 0 0 (2,400– (3,500– 0 poverty line of US$610 (5,800–8,700) 0 0 3,500) 5,200) 0 per year) Note: Lower and upper bounds are indicated in parentheses and calculated by using ±20 percent variations in price elasticities. An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic Figure 4: Premature deaths averted as a result of tobacco price increases by 40% and 60%, Kyrgyz Republic 16,000 20,000 14,800 15,100 14,000 Premature deaths averted 13,200 13,000 15,000 10,200 8,900 10,000 6,700 5,800 5,000 0 Q1 (poor) Q2 Q3 Q4 Q5 (richest) Consumption quintile 40% price increase 60% price increase Figure 5: Out-of-pocket expenditures due to tobacco-related diseases averted as a result of tobacco price increases of 40% and 60%, Kyrgyz Republic 360,000 350,000 320,000 300,000 300,000 400,000 270,000 Out-of-pocket expenditures 260,000 230,000 due to tobacco-related 300,000 diseases averted 180,000 160,000 200,000 100,000 0 Q1 (poor) Q2 Q3 Q4 Q5 (richest) 40% price increase 60% price increase 22 // Results Figure 6: Government expenditures due to tobacco-related diseases averted as a result of tobacco price increases of 40% and 60%, Kyrgyz Republic 1,100,000 1,000,000 1,200,000 950,000 910,000 910,000 Government expenditures for tobacco- 800,000 790,000 1,000,000 690,000 related diseases averted 800,000 550,000 480,000 600,000 400,000 200,000 0 Q1 (poor) Q2 Q3 Q4 Q5 (richest) 40% price increase 60% price increase The benefits of higher cigarette prices are concentrated among the poorer quintiles, with almost 49 percent of averted deaths and 45 percent of OOP expenditures averted accruing to the bottom two consumption quintiles (Figure 2). However, given the higher smoking prevalence among the richer quintiles, the price increase would also bring about benefits for the highest quintile in terms of reductions in out-of-pocket expenditures related to tobacco-related diseases. Additional results are presented in the Annexes. Notably, if the smoking prevalence among males is assumed to be 48 percent (as estimated by WHO), the number of averted deaths as a result of a 40 percent price increase would amount to 102,900 deaths. In addition, US$ 2.4 million OOP spending would be averted and government savings would represent US$ 7.1 million (see Annex Table 2). 23 THIS STUDY SIMULATES THE IMPACT OF HIGH TOBACCO PRICE RESULTING FROM INCREASES IN TOBACCO EXCISE TAX IN THE KYRG REPUBLIC. TOBAC TAXES ARE AMON 4 CONCLUSION Tobacco taxes are not only one of the most cost-effective interventions for reducing tobacco consumption, they are also the single most important intervention against NCDs (Jamison and others 2013). However, until recently, the Kyrgyz Republic had one of the lowest tobacco excise tax rates in the Europe and Central Asia region, and despite recent steps to substantially increase those taxes by 2022, proposed changes still fall short of the WHO-recommended rate of 75 percent. To estimate the potential effects of increasing the excise tax on tobacco products in the Kyrgyz Republic, an extended cost-effectiveness analysis (ECEA) was applied. The results indicate that higher excise taxes on tobacco in the Kyrgyz Republic would avert a large number of premature deaths and poverty cases (among the bottom three quintiles). In addition, the policy could result in substantial government savings on treatment for tobacco-related diseases. Notably, because our elasticity estimates suggest that the poorer quintiles are more sensitive to price changes, the benefits are concentrated among the poorest 40 percent of the population. This analysis has several limitations. First, no substitution effects (individuals switching to cheaper cigarettes) were modeled. By narrowing the difference between the most and least expensive cigarettes, however, excise taxes encourage individuals to quit smoking altogether rather than to switch to lower-priced cigarettes. Second, it was assumed that changes in the intensity of smoking (number of cigarettes smoked per day) would yield no health benefits. Therefore, our estimates are likely to underestimate the full impact of higher tobacco taxes. Third, due to lack of data on OOP expenditures for each disease case, MHIF costs were used as a proxy for incurred expenditures. In addition, pharmaceutical expenditures or informal payments for tobacco-related diseases were not included as data were not available. Our results are therefore likely to underestimate the expenditures related to tobacco-related diseases and the number of poverty cases averted. By examining the health and financial impact of policies across consumption quintiles, ECEA can serve as a useful tool to evaluate policy proposals, particularly in the era of SDGs, in which there is heightened interest in the distributional impact of reforms. Despite the commonly used regressivity argument against tobacco taxation, and reflecting the findings of other recent studies, a disproportionately higher share of benefits concentrated among the poorer populations was found. Not only can higher tobacco prices result in better health outcomes, they can also improve financial risk protection by preventing OOP medical expenditures on treatment for tobacco-related diseases. In addition to increasing tobacco taxes, however, the Kyrgyz Republic should also strengthen implementation of other tobacco control measures, such as ensuring access to cessation services. 25 THIS STUDY SIMULATES THE IMPACT OF HIGH TOBACCO PRICE RESULTING FROM INCREASES IN TOBACCO EXCISE TAX IN THE KYRG REPUBLIC. TOBAC TAXES ARE AMON REFERENCES Adioetomo S, Djutaharta T, Hendratno. 2005. Cigarette Consumption, Taxation and Household Income: Indonesia case study. World Bank Health Nutrition and Population Discussion Paper (26). Washington, DC: World Bank. Bask M, Melkersson M. 2004. “Rationally Addicted to Drinking and Smoking?” Applied Economics 36(4):373–381. Denisova I, Kuznetsova P. 2014. The effects of tobacco taxes on health: An Analysis Of The Effects By Income Quintile and Gender in Kazakhstan, the Russian Federation, and Ukraine (English). Health, nutrition, and population (HNP) discussion paper. Washington, DC: World Bank. Doll R, Peto R, Boreham J and others. 2004. “Mortality in Relation to Smoking: 50 Years’ Observations on Male British Doctors.” British Medical Journal 328(7455):1519. Duffy M. 2006. “Tobacco Consumption and Policy in the United Kingdom.” Applied Economics 38(11):1235–1257. Fuchs A, Meneses F. 2017. Are Tobacco Taxes Really Regressive? Evidence from Chile. Washington, DC: World Bank. Fuchs A, Meneses F. 2017. Regressive or Progressive? The Effect of Tobacco Taxes in Ukraine. Washington, DC: World Bank. Fuchs A, Meneses F. 2018. Tobacco Price Elasticity and Tax Progressivity in Moldova. Washington, DC: World Bank. Fuchs Tarlovsky, A., Del Carmen, G., & Mukong, A. K. (2018). Long-run impacts of increasing tobacco taxes: evidence from South Africa. Washington, DC: World Bank. Goodchild M, Nargis N, d’Espaignet ET. 2017. “Global Economic Cost of Smoking- Attributable Diseases.” Tobacco Control 27(1):58–64. Hu T-W, Ren Q-F, Keeler TE and others. 1995. “The demand for cigarettes in California and behavioural risk factors.” Health Economics 4(1):7–14. Huang B-N, Yang C, Hwang M and others. 2004. New Evidence on Demand for Cigarettes: A Panel Data Approach.” International Journal of Applied Economics 1(1):81–97. Institute for Health Metrics and Evaluation. 2017. “GBD Compare.” In IHME Viz Hub. http://vizhub.healthdata.org/gbd-compare. International Agency for Research on Cancer. 2011. IARC Handbooks of Cancer Prevention: Effectiveness of Tax and Price Policies in Tobacco Control. Vol. 14. Lyon, France: International Agency for Research on Cancer. 27 An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic James E, Saxena A, Franco Restrepo C and others. 2017. The Distributional Consequences of Increasing Tobacco Taxes on Colombia’s Health and Finances. Washington, DC: World Bank. Jamison DT, Summers LH, Alleyne G and others. 2013. “Global health 2035: A World Converging within a Generation.” The Lancet 382(9908):1898–1955. Jha P, Chaloupka FJ. 1999. Curbing the Epidemic: Governments and the Economics of Tobacco Control. World Bank, Washington, DC. Jha P, Jacob B, Gajalakshmi V and others. 2008. “A Nationally Representative Case–Control Study of Smoking and Death in India.” New England Journal of Medicine 358(11):1137–1147. Jha P, Ramasundarahettige C, Landsman V and others. 2013. “21st-Century Hazards of Smoking and Benefits of Cessation in the United States.” New England Journal of Medicine 368(4):341–350. Jha P, Peto R. 2014. “Global Effects of Smoking, of Quitting, and of Taxing Tobacco.” New England Journal of Medicine 370(1):60–68. Kemp-Benedict E. 2001. Income Distribution And Poverty: Methods for Using Available Data in Global Analysis. PoleStar Tech Notes (4). Lavado R, Meyer M, Postolovska I and others. 2017. “Beyond the Perceived Regressivity of Higher Tobacco Taxes: Turning Short-Term Losers Into Long-Term Winners?” In Tobacco Tax Reform: At the Crossroads of Health and Development. Washington, DC: World Bank. National Statistical Committee of the Kyrgyz Republic. 2015. Kyrgyz Integrated Household Survey [dataset]. Önder Z. 2002. The Economics of Tobacco in Turkey: New Evidence and Demand Estimates. HNP Discussion Paper, World Bank, Washington, DC. Postolovska I, Lavado R, Tarr G, Verguet S. 2017. “The Health Gains, Financial Risk Protection Benefits, and Distributional Impact of Increased Tobacco Taxes in Armenia.” Health Systems and Reform e4:1;30–41. Salem A, Mount T. 1974. “A Convenient Descriptive Model of Income Distribution: The Gamma Density.” Economic Journal of the Economic Society 1115–1127. Salti N, Brouwer E, Verguet S. 2016. “The Health, Financial and Distributional Consequences of Increases in the Tobacco Excise Tax Among Smokers in Lebanon.” Social Science and Medicine 170:161–169. Townsend J, Roderick P, Cooper J. 1994. “Cigarette Smoking by Socioeconomic Group, Sex, and Age: Effects of Price, Income, and Health Publicity.” British Medical Journal 309(6959):923–927. 28 // References Tremblay CH, Tremblay VJ. 1995. “The Impact of Cigarette Advertising on Consumer Surplus, Profit, and Social Welfare.” Contemporary Economic Policy 13(1):113–124. Verguet S, Gauvreau CL, Mishra S and others. 2015. “The Consequences Of Tobacco Tax On Household Health and Finances in Rich and Poor Smokers in China: an Extended Cost- Effectiveness Analysis.” Lancet Global Health 3(4):e206–e216. Verguet S, Laxminarayan R, Jamison DT. 2015b. “Universal Public Finance of Tuberculosis Treatment in India: An Extended Cost-Effectiveness Analysis.” Health Economics 24(3):318–332. Verguet S, Olson ZD, Babigumira JB and others. 2015c. “Health Gains and Financial Risk Protection Afforded by Public Financing of Selected Interventions in Ethiopia: An Extended Cost-Effectiveness Analysis.” Lancet Global Health 3(5):e288–e296. Verguet S, Kim JJ, Jamison DT. 2016. “Extended Cost-Effectiveness Analysis for Health Policy Assessment: A Tutorial.” PharmacoEconomics 34(9):913–923. Verguet S, Tarr G, Gauvreau CL and others. 2017. “Distributional Benefits of Tobacco Tax and Smoke-Free Workplaces in China: A Modeling Study.” Journal of Global Health 7(2). Wilson LM, Avila Tang E, Chander G and others. 2012. “Impact of Tobacco Control Interventions on Smoking Initiation, Cessation, and Prevalence: A Systematic Review.” Journal of Environmental Public Health 7;2012:e961724. World Bank. 2014. Kyrgyz Republic Public Expenditure Review Policy Notes: Health. Washington, DC: World Bank. World Bank. 2016a. World Development Indicators. http://data.worldbank.org/data-catalog/world- development-indicators. World Bank. 2016b. Health, Nutrition and Population Data and Statistics. http://datatopics. worldbank.org/hnp/. World Bank. Forthcoming. Kyrgyz Republic: Fiscal Space Analysis. World Health Organization. 2010. WHO Technical Manual on Tobacco Tax Administration. Geneva: World Health Organization. World Health Organization. 2015. WHO Report on the Global Tobacco Epidemic, 2015. Geneva: World Health Organization. World Health Organization. 2018. Global Health Expenditure Database. http://apps.who.int/nha/database. 29 An Extended Cost-Effectiveness Analysis of Tobacco Price Increases in the Kyrgyz Republic ANNEXES Annex Table 1. Extended cost-effectiveness analysis results (individual consumption quintile) for a 75 percent tobacco price increase, Kyrgyz Republic 75% price increase simulation Premature 68,800 17,600 16,200 16,500 11,200 7,300 deaths averted (55,000–82,500) (14,000–21,100) (13,000–19,500) (13,200–19,800) (8,900–13,400) (5,800–8,700) Out-of-pocket expenditures 1590000 380,000 400,000 290,000 200,000 330,000 due to tobacco- (1,270,000– (300,000– (320,000– (230,000– (160,000– (270,000–400,000) related diseases 1,910,000) 450,000) 470,000) 340,000) 240,000) averted (US$) Government expenditures for 4,780,000 990,000 1,130,000 1,120,000 860,000 600,000 tobacco-related (3,820,000– (800,000– (910,000– (950,000– (690,000– (480,000– diseases averted 5,730,000) 1,200,000) 1,360,000) 1,420,000) 1,030,000) 720,000) (US$) Poverty cases averted (using the national 8000 0 3,300 4,800 0 0 poverty line of (6300–9500) 0 (2,600–3,900) (3,800–5,600) 0 0 US$610 per year) Note: Lower and upper bounds are indicated in parentheses and calculated by using ±20 percent variations in price elasticities. 30 // Annexes Annex Table 2. Extended cost-effectiveness analysis results (individual consumption quintile), 40 percent tobacco price increase (male smoking prevalence of 48 percent), Kyrgyz Republic Q1 (POOR- TOTAL Q2 Q3 Q4 Q5 (RICHEST) EST) Premature 102,900 26,300 24,300 24,700 16,700 10,900 deaths averted (82,300–124,000) (21,000–31,500) (19,400–29,200) (19,800–29,600) (13,300–20,100) (8,700–13,100) Out-of-pocket expenditures 2,381,000 496,000 566,000 591,000 429,000 299,000 due to tobacco- (1,910,000– (400,000– (450,000– (470,000– (340,000– (240,000– related diseases 2,858,000) 680,000) 709,000) 710,000) 515,000) 359,000) averted (US$) Government expenditures for 7,140,000 1,480,000 1,700,000 1,770,000 1,290,000 900,000 tobacco-related (5,720,000– (1,190,000– (1,360,000– (1,420,000– (1,030,000– (719,000– diseases averted 8,575,000) 1,786,000) 2,039,000) 2,129,000) 1,544,000) 1,078,000) (US$) Poverty cases averted (using the national 11,900 0 4,900 7,000 0 0 poverty line of (9,500 -14,200) 0 (3,800–5,800) (5,700–8,000) 0 0 US$610 per year) Note: WHO estimates the male smoking prevalence at 48 percent. Based on this average, we estimated quintile-specific prevalence rates using the patterns observed in KIHS 2015 data. Lower and upper bounds are indicated in parentheses and calculated by using ±20 percent variations in price elasticities. 31