THE DISTRIBUTIONAL IMPACTS OF CIGARETTE TAXATION IN BANGLADESH Giselle Del Carmen Alan Fuchs María Eugenia Genoni Cover photos (clockwise from left): iStock 'Cox's Bazar, Bangladesh - February 09, 2013: Men working in pad- dy field in the suburb of Cox's Bazar, Bangladesh.' iStock Rickshaw Jam, High school girls taking notes. Suapur, Bangladesh.Photo: Scott Wallace/ World Bank THE DISTRIBUTIONAL IMPACTS OF CIGARETTE TAXATION IN BANGLADESH Giselle Del Carmen, Alan Fuchs María Eugenia Genoni The distributional impacts of cigarette taxation in Bangladesh ABSTRACT Despite the obvious positive health impacts of tobacco taxation, an argument raised against it is that poor households bear the burden of the increased prices because of their higher share of spending on tobacco. This note includes estimates of the distributional impacts of price rises on cigarettes under various scenarios using the Household Income and Expenditure Survey (HIES) 2016/17. One contribution of this analysis is to quantify the impacts by allowing price elasticities to vary across consumption deciles. This shows that an increase in the price of cigarettes in Bangladesh has small consumption impacts and does not significantly change the poverty rate or consumption inequality. These findings stem from relatively even cigarette consumption patterns between less and more well- off households. These results hold even if one considers some small substitution through the use of bidis, which are largely consumed by the poor. The short-term consumption impacts are also negligible compared with the estimated gains because of savings in medical costs and the greater number of productive years of life. JEL Codes: H23, H31, I18, O15 II // Executive Summary ACKNOWLEDGMENTS Giselle Del Carmen: Poverty and Equity Global Practice, World Bank (email: gdelcarmen@ worldbank.org); Alan Fuchs: Poverty and Equity Global Practice, World Bank (email: afuchs@worldbank.org); Maria Eugenia Genoni: Poverty and Equity Global Practice, World Bank (email: mgenoni@worldbank.org). Support in the preparation of this report has been provided by the World Bank’s Global Tobacco Control Program, co-financed by the Bill and Melinda Gates Foundation and Bloomberg Philanthropies. The authors are grateful to Urmila Chatterjee, Ruth Hill, Zahid Hussain, Patricio Marquez, Nigar Nargis, Nistha Sinha, and Owen Smith for providing comments and support. III SMOKING HAS The distributional impacts of cigarette taxation in Bangladesh BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 IV // Executive Summary 1. INTRODUCTION Bangladesh is one of the largest consumers of tobacco in the world. Approximately 4 in 10 adults (46 million adults) use some form of tobacco, whether smoked (for example, cigarettes and bidis) or smokeless (for instance, betel, betel nut, tobacco leaf, and so on).2 Tobacco consumption in Bangladesh varies by gender. Men are significantly more likely than women to smoke (40 percent of men versus 1 percent of women above the age of 15).3 The prevalence of smokeless tobacco use is similar among men and women (about 30 percent).4 Smoking is widespread in Bangladesh, but comparable with the prevalence rates in other countries in the region (figure 1). Figure 1. Prevalence of Smoking Any Tobacco Product, Individuals Ages ≥ 15 Years 76.2 39.8 41.4 41.9 37.1 31.6 28.4 20.4 11.1 6.4 3.6 1.9 2.3 3 0.4 0.7 India Sri Lanka Myanmar Nepal Bangladesh Thailand Pakistan Indonesia Female Male Source: WHO 2015a. Information on households from the Household Income and Expenditure Survey (HIES) 2016/17 confirms the high rates of tobacco consumption (both smoked and smokeless). In 2016, approximately 65 percent of households had consumed some type of tobacco product during the month previous to the interview. Comparing across time, the share of households consuming tobacco has been decreasing during the last decade. Using pre- vious rounds of the HIES and comparable measures, one may see that the share of house- holds consuming tobacco was 82 percent in 2005, 71 percent in 2010, and 68 percent in 2016. Yet, this has been accompanied by an increase in the share of households consum- ing cigarettes from about 23 percent in 2005 to 30 percent in 2016. 2 Bidis are popular hand rolled tobacco products smoked in South Asia.. 3 Note that the true smoking prevalence among women may be greater than estimated because evidence suggests that women (and those who proxy-respond for them in surveys) tend, for social reasons, to underreport risky behaviors.. 4 Data on tobacco consumption vary in terms of coverage, period, and quality. These numbers come from the 2009 Global Adult Tobacco Survey Bangladesh. 1 The distributional impacts of cigarette taxation in Bangladesh The high rates of tobacco consumption in Bangladesh impose an increasing health and economic burden on the country. A 2004 epidemiological study found that 1.2 million tobacco-related illnesses and nearly 57,000 deaths attributable to smoking were reported each year (WHO 2007). More current information shows that approximately a quarter of all deaths among men ages 25–69 in Bangladesh are attributable to smoking (Alam et al. 2013). The overall economic cost of tobacco use has been estimated at Tk 110 billion (US$1.85 billion) or over 3 percent of gross domestic product (GDP) (WHO 2007). The government has undertaken efforts to tackle tobacco consumption. It was among the first to sign (June 2003) and ratify (June 2004) the World Health Organization’s Frame- work Convention on Tobacco Control, the world’s first public health treaty, which urged governments to adopt comprehensive policies to limit tobacco use. The country’s partici- pation in the framework convention has led to advances in tobacco control policy, in par- ticular through the 2005 Smoking and Tobacco Usage Act. The new law restricted smok- ing in certain locations, though health care facilities, educational facilities, sport venues, and taxis were the only 100 percent smoke-free environments in Bangladesh at the time.5 Warning labels on cigarette packages and the limited advertising of tobacco products were also mandated as part of the 2005 law. According to the 2009 Global Adult Tobacco Survey and the International Tobacco Control Bangladesh Survey, tobacco consumption rose between 2009 and 2012, despite the government’s actions to restrain it. Among the negatives of the Smoking and Tobacco Usage Act were the low levels of enforcement of non-tax measures, including the advertising ban and smoke-free public places, and relatively low levels of implementation of the warning labels. In April 2013, the National Assembly passed the Tobacco Control Law Amendment Bill and Rules to close many of these loopholes. A major contribution of the amendment was the requirement that packages contain pictorial warnings. In addition, smokeless tobacco was included under the definition of tobacco; restaurants and indoor workplaces were included among completely smoke-free environments; the fines for noncompliance were raised; advertise- ments at points of sale were banned; and the sale of tobacco to minors was prohibited. Although tobacco taxation has been recognized as one of the most effective strategies to reduce smoking, the tobacco tax structure has helped make cigarette prices in Ban- gladesh among the lowest in the region and in the world despite some of the highest tax rates (table 1). There is a variety of taxes on tobacco products in Bangladesh. These include supplementary duties on cigarettes, bidis, and chewing and pipe tobacco, duties on imported tobacco products and on tobacco leaf, and a 15 percent value added tax on all tobacco products.6 Cigarettes are classed into four brand tiers, commonly known as 5 Smoking is also restricted in restaurants, bars, and workplaces. However, these offer designated smoking areas. 6 The value added tax is applied to the retail price. 2 // Introduction slabs in Bangladesh, and the ad valorem excise tax rate is based on the administered retail price, which varies across these four price categories. However, the gaps between brands in the different tier categories create incentives for smokers to substitute less expensive brands in response to price increases and for manufacturers to position brands between price slabs to avoid paying higher taxes (ITC Project 2014). To minimize tax evasion, continuous tier structures were introduced in 2015, allowing brands to be set to a more well-defined tier. By 2017, the lower tier was divided into two new slabs: local brands and international brands. Since July 2017, cigarette excise tax rates have varied between 52 percent for the low-priced category and 65 percent for the premium tier.7 In addition, the tier structure has been gradually shifted toward unified rates. In 2010, the difference between the bottom and top tiers was 25 percentage points; by 2017, it had dropped to 12 percentage points. Table 1. Cigarette Prices and Taxes, Bangladesh and Selected Countries Cigarettes Bidi Chewing Tobacco Taxes, % of Taxes, % of Taxes, % of Country Price Price Price retail price retail price retail price India 9.2 43 0.65 20 - - Sri Lanka 19.6 62 - - - - Nepal 5.7 26 - - - - Bangladesh 3.4 77 0.41 23 1.35 54 Thailand 7.1 74 - - - - Pakistan 2.2 60 - - - - Indonesia 5.2 57 - - - - Source: WHO 2015b, 2017. Note: The table shows the prices for the most widely sold brands in international purchasing power parity U.S. dollars. Cigarette prices are standardized to packs of 20. Bidi prices are standardized to 20 pieces. Chewing tobacco prices are standardized to 20 grams. Taxes are calculated for the most widely sold brands. For Bangladesh, the brand of cigarette reported in the 2017 WHO Report (for year 2016) was not the most widely sold brand. The price of the most widely sold brand was less than half of the price reported in the table. Similarly, the taxes on bidis differentiate between filtered and unfiltered products. The taxes are estimated based on a predetermined tariff value per pack, not the retail price. The tax rate rose from 20 percent in 2011 to 30 percent in 2017 for nonfiltered bidis; for filtered bidis, it increased from 25 percent to 35 percent. This tariff-value-driven tax base 7 The low-priced category was designed to protect cheaper domestic brands and poorer household consumers 3 The distributional impacts of cigarette taxation in Bangladesh actually reduces the effective rate of the supplementary duty and the value added tax. For instance, in 2011, the average retail price for a pack of 25 bidis was Tk 6.0, while the tariff value was Tk 3.2 and Tk 3.4 per pack for nonfiltered and filtered bidis, respectively (Ahmed 2012). Thus, the effective tax rate of the supplementary duty was reduced to 10.5 percent in 2011, not the established 20.0 percent. Bidi manufacturers in South Asia have avoided many of the local and international tobacco regulations and taxes enforced on factory-made cigarettes. Hence, bidis are sold cheaply at various prices and in packag- ing with poorly visible health warnings (Jha 2015; Kostova et al. 2014; Sankaran, Hillamo, and Glantz 2015). Duong et al. (2017) find that, even though bidis have significantly less tobacco content than cigarettes, they are at least as harmful as cigarettes in terms of cardiorespiratory health and mortality. Moreover, bidi smokers in South Asia are typically among the poorest and most vulnerable. In 2017, the minister of finance proposed elim- inating the bidi tariff system and fix prices at Tk 15.0 per pack; nonetheless, parliament reduced the price to Tk 12.5.8 In sum, the current taxation system has neither discouraged consumption nor maximized revenue in Bangladesh. In the last decade, cigarette consumption, particularly the con- sumption of the cheaper brands, has expanded significantly. Since 2011, the volume in the low segment rose by an average of 26 percent, causing the market share to increase. Currently, the market share of the low segment accounts for nearly 77 percent of total volume, from 51 percent in fiscal year 2010–11; yet, it represents only 40 percent of the total revenue from cigarettes (World Bank 2017). As the government considers tobacco-tax reform, particularly in the case of cigarettes, one argument that has been raised against the reform is that tobacco taxation is regres- sive because low-income households assign a larger share of their budgets to tobacco products. This note relies on information from the HIES 2016/2017 to (1) describe recent tobacco consumption patterns among households and (2) shed light on the distribu- tional impacts of an increase in the price of cigarettes. The study is structured as follows. Section 2 presents the data and some descriptive statistics. Section 3 describes the methodology. Sections 4 and 5 present the estimation results, and section 6 concludes 8 “Economy,” Morning News Flash, July 2, 2017, Shanta Securities, Dhaka, Bangladesh, http://print.thefinancialexpress-bd.com/2017/07/02/176701. 4 // Introduction 5 SMOKING HAS The distributional impacts of cigarette taxation in Bangladesh BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 6 // Executive Summary 2. DATA This paper relies on the HIES 2016/17, a comprehensive, nationally representative survey col- lected between April 2016 and March 2017. The survey involved interviews among 46,000 households and gathered detailed information about household expenditure patterns, including on various tobacco products. The survey allows households to be categorized into consumption deciles from the poorest to the richest, thus permitting the simulation of the distributional impacts of a potential reform. Approximately 65 percent of households reported in the HIES 2016/17 that they consumed tobacco products; 42 percent consumed smoked tobacco; and 45 percent consumed smoke- less tobacco products. Many households consume exclusively either smoked tobacco (21 percent) or smokeless tobacco products (23 percent). An average of about 3 percent of total household consumption is allocated to tobacco products. If one compares across consumption deciles (from the poorest 10 percent of households to the richest 10 percent), one finds that consumption patterns vary, though not dramatically (figure 2). The share of households consuming tobacco is not particularly different among the bottom 50 percent of the distribution, and it only declines slightly at higher consumption lev- els; across all deciles, more than one household in two consumes tobacco. Similarly, the share of tobacco expenditures in total consumption is 3.7 percent among the bottom 50 percent and about 3 percent among deciles 6–8, and it declines to 2 percent among the top decile. Figure 2. Tobacco Consumption, by Per Capita Consumption Decile a. Share of households consuming tobacco b. Tobacco consumption in total household consumption 6% 71% 68% 69% 69% 69% 5% 66% 65% 65% 64% 60% 4% 55% 3% 2% 1% 0% 1 2 3 4 5 6 7 8 9 10 All 1 2 3 4 5 6 7 8 9 10 All Per capita household consumption decile Deciles of household per capita consumption All households Conditional on positive consumption Source: Calculations based on data of HIES 2016/17. Note: Tobacco includes cigarettes, bidis, tobacco leaf, gul, betel leaf, betel nut, zorda, lime, khoer, rolled betel leaf, and other smoked or smokeless products. Consumption deciles are defined based on household consumption per capita deflated spatially to control for differences in prices across areas. 7 The distributional impacts of cigarette taxation in Bangladesh To obtain a sense of how different the consumption patterns are relative to the patterns in other countries in the region, one may compare Bangladesh with India, a neighboring coun- try with similar tobacco products, particularly bidis. Information from the India Consumption Expenditure Survey, 2011–12 reveals that the socioeconomic gradient for consuming tobacco is steeper in India than in Bangladesh (figure 3). Among the poorest deciles, the share of households consuming tobacco is similar, but, among the richest decile, a higher share of households consume tobacco in Bangladesh (55 percent) than in India (36 percent). The share of total consumption assigned to tobacco products is also substantially lower in India than in Bangladesh. However, similar to Bangladesh, there is a small gradient in the share of the budget assigned to tobacco products across deciles. These differences become smaller between Bangladesh and the state of West Bengal, highlighting the greater similarity of cul- tural and consumption patterns in the latter region of India. Figure 3. Tobacco Consumption, by Per Capita Consumption Decile, Bangladesh, India, and West Bengal a. Share of households consuming tobacco b. Share of tobacco consumption in total household consumption 80.0% 4.0% 70.0% 3.0% 60.0% 2.0% 50.0% 40.0% 1.0% 30.0% 0.0% 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 India West Bengal Bangladesh India West Bengal Bangladesh Source: Calculations based on data of the Bangladesh HIES 2016/17 and the India National Sample Survey and Consumption Expendi- ture Survey, 2011–12. Note: Tobacco includes cigarettes, bidis, tobacco leaf, gul, betel leaf, betel nut, zorda, lime, khoer, rolled betel leaf, and other smoked or smokeless products. Consumption deciles are defined based on household consumption per capita. The consumption aggregate includes similar broad food and nonfood components; however, they may not be fully comparable due to the differences in survey design, data collection, and methodologies in computing the aggregate. By type of product, one finds additional differences in consumption patterns across deciles (figure 4). The share of households consuming cigarettes is at 26–30 percent for almost the entire distribution, except the poorest decile, among which the share of households consum- ing cigarettes is significantly lower (19 percent). Meanwhile, bidi consumption clearly declines as household resources increase. While a quarter of the poorest households consume bidis, this is only true of 6 percent of the richest households. For smokeless products, consumption is similar, except among the top deciles. This is consistent with other studies describing the socioeconomic gradient in tobacco use; the gradient is not seen in cigarette consumption, 8 // Data but is noticeable in the consumption of bidis and smokeless products (Barkat et al. 2012). On average, the consumption of smokeless products accounts for the largest share, 1.4 percent of the total household budget. Cigarettes follow at an average expenditure share of 1.3 per- cent. The cigarette budget share is even across the distribution. The median quantity of cigarettes consumed was 152 a month per adult man age 15 years or older; the corresponding number of bidis was 348 per month per adult man (figure 5).9 Assuming daily smoking, this is equivalent to an average of 5 cigarettes and 12 bidis per adult man per day. These household figures are consistent with data on individuals from the Global Adult Tobacco Survey among adult men (Barkat et al. 2012). The poorest decile con- sumed about 30 fewer cigarettes per adult man relative to the richest decile (132 compared with 152 cigarettes per adult man per month). However, among deciles 4–10, the median quantities were similar. One may see the opposite pattern in the case of bidis, that is, the richest consume slightly fewer bidis. Figure 4. Tobacco Consumption, by Type of Product and Per Capita Consumption Decile a. Share of households consuming tobacco b. Share of tobacco consumption in total household consumption 60% 2.0% 50% 40% 1.5% 30% 1.0% 20% 0.5% 10% 0% 0.0% 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Cigarettes Bidis Cigarettes Bidis Other Smoked Smokeless Other Smoked Smokeless Source: Calculations based on data of the HIES 2016/17. Note: Consumption deciles are defined based on household consumption per capita deflated spatially to control for differences in prices across areas 9 The data do not allow the identification of the household members who are smoking; so, these results should be interpreted with care. 9 The distributional impacts of cigarette taxation in Bangladesh Figure 5. Quantities and Unit Values, Per Capita for Cigarette and Bidi Consumption, by Decile a. Median monthly number per adult man b. Median unit value (takas), by decile 3.5 400 3.0 350 2.5 300 2.0 250 200 1.5 150 1.0 100 0.5 50 0 0 1 2 3 4 5 6 7 8 9 10 All 1 2 3 4 5 6 7 8 9 10 All Number of cigarettes per adult male Cigarettes Bidis Number of bidis per adult male Source: Calculations based on data of the HIES 2016/17. Note: Consumption deciles are defined based on household consumption per capita deflated spatially to control for differences in prices across areas. Median unit values (per cigarette) are obtained by dividing the total amount consumed or total household budget allocated to purchase tobacco products by the number of units consumed. The median cost was Tk 2.6 per cigarette and Tk 0.5 per bidi, which highlights the signifi- cantly lower cost of bidis.10 The median unit values for cigarettes align with the retail price for a 10-stick pack published by the National Board of Revenue, Ministry of Finance, Bangladesh. For 2016–17, the price for the low-tier was Tk 23 for 10-stick packs. Comparing unit values, one finds that there is a difference of Tk 1 per cigarette between the poorest and richest decile (Tk 2.2 versus Tk 3.2), while the unit values of bidis do not vary across the distribution. One key takeaway from this description is that tobacco consumption is quite widespread across households in Bangladesh and that there is not a major difference in the budget shares spent on tobacco across the consumption distribution. Total tobacco consumption is quite similar across the bottom 80 percent of the population and only smaller among the richest 20 percent. In terms of cigarette consumption, only the poorest 10 percent (households in extreme poverty) tend to consume fewer cigarettes, yet spend a similar budget share rela- tive to the richer deciles. The similarity in consumption patterns is important in assessing the distributional impacts of a potential tax reform. 10 This cost is the median of the unit values reported in the survey, which is obtained by dividing the total amount consumed or total household budget allocated to purchase tobacco products by the number of units consumed. 10 // Data 11 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE Introduction 12 // Executive Summary for the low-tier for for thewas the Tk for low-tier23 the for for low-tier was 10-stick the low-tier was Tk low-tier Tk was 23 23 forpacks. Tk was for 23 10-stickComparing Tk 10-stick for 23 packs. 10-stick packs. unit for 10-stick packs.values, Comparing Comparing packs. one Comparing Comparing unit unit finds unit values, values, that oneonethere unit values, values, one finds finds is a that difference that finds onethere there finds that is is a atha there diffd for the of Tk low-tier 1 the for for per the was low-tier cigarettefor Tk low-tier the for 23 was the for was Tk between oflow-tier low-tier Tk 23 Tk 23 for the 1 was 10-stick for per Tk was packs. 10-stick poorest 23 Tk 10-stick 23 packs. cigarette for and 10-stick for Comparing packs.10-stick between packs. unit packs. Comparing Comparing richest decile the unitComparing Comparing values, one unit (Tk poorest 2.2 finds values, values, and one versus unit that one finds richestunit values, Tk one values, there finds that 3.2), decile is that (Tkaone there while finds finds difference there is the 2.2 a that that is a differ differenc unit versus ther th T of ofTkTk 1 per 1 of perTk 1 per cigarette cigarette cigarette between between between the the poorest the poorest poorest andand richest richest richest decile and decile decile (Tk(Tk2.22.2 (Tk versus versus 2.2Tk versus Tk 3.2), 3.2), Tk while 3.2), while th of ng unit values, one finds thatTk 1of there valuesof per of is Tk Tk acigarette 1 bidis 1 per do of between of Tk Tk1 across per difference not cigarette cigarette vary per 1 per values cigarette cigarette the of poorest between between thethe bidis do between between the and poorest poorest distribution. not thethe richest and vary and acrosspoorest poorest decile richest richest the and (Tk decile richest and richest 2.2 decile (Tkversus (Tk distribution.2.2 decile 2.2decile Tk (Tk 3.2), versus versus (Tk Tk 2.2 2.2 while Tk 3.2), versus versus the 3.2), while Tk unit while Tk the 3 3.2) the un values values ofvalues of of bidis bidis dodo bidis notnot do vary not vary across vary across across thethe distribution. the distribution. distribution. st decile (Tk 2.2 versus values of values Tk 3.2), bidis while values of do ofnot the bidis values values vary bidis unit do doof not of across notbidis varyvary bidis do the do acrossnot across the not varyvary distribution. the across across the distribution. distribution. the distribution. distribution. One key One takeaway One key key One from takeaway One key takeaway this description key takeaway from from takeaway thisfrom this is from description this descriptionthat tobacco this is is description description that consumption that tobaccothatistobacco is tobacco that istobacco quite consumption consumption consumption widespread consumption is is quitequite is across quite widespread qui is wid widesprea One key One households takeaway One key key One from takeaway in Bangladesh One takeaway key key this from from andtakeaway takeaway description this this from from is description description this that this is description is thatdescription tobacco that tobacco is consumption tobacco that is that consumptiontobacco tobacco consumptionis quite consumption is consumption widespread is quite quite widespread is quite is across widespread on wid quite acro aw households households in households households in Bangladeshinthat Bangladesh there in Bangladesh and Bangladesh and isthat that not andthere a there that major and is there isthat not difference not a isthere a not major major is in not the difference adifference major budget a major in in difference the shares difference the in budgetbudget thespent in the shares budget shares bud spsh acco consumption is quitehouseholds in Bangladesh households widespread households in households households across in Bangladesh Bangladesh and that in in andBangladesh Bangladesh there andthat is consumption that there not and there isand that major a notis that not there there is difference a Totala major major is not not difference a major thein in Total differencea major difference difference budgetin the the shares budget budget in in the spent the shares onbudge budget spen s tobacco acrosstobacco tobacco the consumption tobacco across across tobacco across the the distribution. across consumption the consumption the consumption Total distribution. distribution. tobacco distribution. Totalconsumption distribution. tobacco Total tobacco consumption tobacco is consumption quite tobacco similar consumption is is shares across consumption quitequite spent the quite is similar similar is qu sim acroao tobacco major difference in the budget bottom across tobacco shares tobacco 80 the spent across percent tobacco tobacco consumption across onthe of the the across consumption consumption bottom across population 80 the the distribution. and percent consumption consumption distribution. only ofTotal distribution.the distribution. tobacco Total smaller distribution. Total population consumption tobacco among tobaccoand Total the Total consumption richest only tobacco tobacco is 20 consumption smaller quite consumption consumption is similar percent. amongis quite quite similar Inacross the isacross similar terms quite richestis the quite acros of 20sim th bottom bottom bottom 80 80 percent percent 80 of of percent the the of population the population population and and only and only smaller only smaller smaller among among among the the the richest richest richest 20 20 20 percen percent. percent. In Inter 3. MEASURING THE obacco consumption is bottom quite 80 percent bottom similar bottom across 80 80 percentbottom bottom ofpercent the the of 80 80 percent percent population of the population of and population the of onlythe and population population smaller and only only among smaller and and smaller only amongtheonly among smaller smaller richest the the among 20 richest among percent. richest 20 the therichest In 20poverty) percent. richest terms percent. In 20 20perce ofpove In terms per term cigarette consumption, cigarette cigarette cigarette only consumption, consumption,cigarettethe poorest consumption, consumption, only only the 10 only the percent the poorest poorest only the poorest 10 (households 10 percentpoorest percent in 10 extreme percent 10 (households percent (households (households in poverty) (households in extreme extreme in tend extreme poverty)in consume toextremepoverty) tend tend to teo to co ler among the richest 20 cigarette percent.consumption, cigarette In cigarette terms cigarette cigarette consumption, consumption,of only consumption, the consumption, only poorest only the the 10 poorest only percent poorest only 10 the10thepoorest poorest (households percent 10 percent 10 in (householdspercent extreme (households (households in poverty) extreme in tend in extreme poverty) extreme to consumepoverty) tend poverty to con te fewer cigarettes, fewer fewer yet fewer cigarettes, cigarettes,spend fewer yet a yet cigarettes, similar cigarettes, spend yet spend budget a a yet spend similar similar a percent share spend similar relative a similar budget budget (households budget share shareto the budget relative share relative inrelative richer extreme share to to the deciles. relative the topoverty) richerricher the The to similarity the richer deciles. deciles. tend toThe richer The consum deciles. indeci sim similaTh fewer ouseholds in extreme poverty) cigarettes, tend fewer consumption consumption elative to the richer deciles. The fewer to consumption consumption similarity consumption yet cigarettes, consume cigarettes, patterns consumption patterns fewerfewer spend consumption inis patterns isDISTRIBUTIONAL IMPACT OF yet consumption cigarettes, yetacigarettes, spend important consumption patterns consumption important patterns patterns similar spendisaimportant patterns issimilar in patterns is patterns is important in yet important a yet budgetspend similar assessing patterns is important important assessing in spend budget important isassessing in sharea inbudget the is in is important a distributional similar similar relative share important inshare assessing assessing assessing the inthe the distributional in budget budgettorelative relative assessing the in the assessing distributional the share impacts assessingto distributional the distributional assessing distributional the impacts share torelative richer the the the the relative deciles. richer of distributional distributional impacts distributional distributional aricher potential impacts of impacts a potential impacts to of aof the to The deciles. impacts of the deciles. a tax of impacts richer similarity The a potential impacts a potential potential richerThe reform. impacts potential tax reform. of of tax deciles. similarity a potentia atax of deciles. in similari oftax tax a pote potenti refor reform. a T p refo r stributional impacts of a potential tax reform. 3. Measuring 3. Measuring 3. 3. the 3. 3.Measuring the Measuring Measuring 3.A PRICE RISE ON CIGARETTES 3. distributional Measuring Measuring 3. Measuring distributional thethe 3.the Measuring the Measuring distributional impactdistributional distributional the distributional the impact distributional theof the distributional distributional distributional of impact a aimpact impact impact priceimpact price of a of of rise rise price of impact a priceon on a a price impact impact ofcigarettes price of a price rise of a cigarettes riserise on on a of riseon price aon price cigarettes cigarettes cigarettes price rise on cigarettes cigarettes rise riseon rise on cigarettes cigarettes oncigarettes price rise on cigarettes This section This describes This Thisthe section section This sectionpartial describes describes sectionequilibrium the the describes describes partial partial the approach equilibrium partialthe partial equilibrium used equilibrium to simulate equilibrium approach approach approach used used tothe approach to used impact simulate simulate to used on simulate the the to welfare simulate impact impact theon proxied impact onthe welfare welfare impa on p w This section ThisThis This by household describes section section section This consumption This the describes describes describes section by section partial thethe the because household describes describes equilibrium partial partialpartial of the equilibrium equilibrium an the consumption partial approach equilibrium increase partial approach approach in equilibrium equilibrium used approach the because used to priceused ofto of approach approach simulate used simulate an to to simulate cigarettes. increase the simulate the used used impact impact inthe This to the the to simulate on on impact pricesimulate impact approachwelfare wel- on of the on the proxied welfare is impact used cigarettes. impact welfare proxie to on pro Thi o byby household householdby household consumption consumption consumption because because of of becausean an an increase increase of increase in inthe the in price price the of of price of cigarettes. cigarettes. cigarettes. This This approach approachThis appr is u i by h used to simulate the impact household by bywelfare on evaluate farehousehold proxied household the proxiedby consumption first-orderby by household household consumption household consumption because effects evaluate consumption consumption of because ofthe a an Comparing because consumption change increase of first-order of because because an in an because in increase increase prices. effects the of anof in an of price in increase It of the increase an of the relies a increase price cigarettes. in change price theof mainly inin in of pricethe the This cigarettes. of cigarettes. on price prices. price of relies approach cigarettes. householdThis It of cigarettes. This cigarettes. approach approachis used expenditure mainly This is This to on app is used ap us ho t for evaluate evaluate the low-tierevaluate the was the first-order the first-order first-order Tk 23 for 10-stick effects effects packs. effects of of a change a change of a in unit change values, in prices. prices. one finds in It prices. that It relies there relies is It a relies mainlymainly difference on on mainly household on househo household expenexp evaluate the Tk first-order evaluate evaluate evaluate the effects first-order the the offirst-order first-order a change effects in effects of prices. ofa Itchange a change relies in mainlyin prices. prices. on It It relies household relies mainly mainly expenditureon on house househo the price of cigarettes. This approach evaluate is used the to first-order iseffects effects of aa of a change change a in in prices.prices. Itin It relies relies mainly mainly on on household household expend expenditu patterns. This ofThe approach per 1focus cigarette is is onusedbetween the to evaluate the impacts poorest ofthe first-order and rise richest in the effects decile price (Tk of a 2.2 change versus Tk in prices. 3.2), while It relies the mainly unit patterns. patterns. patterns. The The patterns. focus The focus is on The focus on the focus the is on impacts the impacts is on the impacts of of rise inof a impacts of rise in a the cigarettes rise theof price a pricerise the of because ofin price the of cigarettes cigarettes price this of cigarettesbecause becausehas been cigarettes because this the thishas focus because hasthis been been has th the es. It relies mainly on patterns. household values The of focus patterns. on bidis expenditure patterns. of the potential The household patterns. is The focus reform do patterns. not onfocus vary the is expenditureon The is The across impacts on the focus the thefocus of impacts patterns. is a impacts is on distribution. rise Theofonthe in aof focustheimpacts impacts the rise a rise isprice inon in thetheof of the price a of rise price impacts a ofrise cigarettes in ofthe in cigarettes of a the rise price because cigarettes inprice of this because the cigarettes of because price cigarettes hasofthisbeen this has because because the has been focus been this the thishas the focu h f of of the the of potentialtheof potential of tobacco thereform potential reform taxes. potential reform of of reform of tobacco tobacco tobacco taxes. of tobacco taxes. taxes. taxes. rice of cigarettes becauseof the this potential has of of been the the the potential cigarettes One key reform of potential focus because takeaway the of of reformthe reform from potential this potential tobaccoof has this of tobacco been reform taxes. reform tobacco description the taxes.of isfocus tobacco of taxes. that tobacco of the potential tobacco taxes. taxes. consumption reform of tobacco is quite widespread taxes.across To assess households the in Bangladesh distributional and that impact there of the not a major is increase indifference the price in the of budget shares cigarettes, spent the on simulation allows ToTo Toassess assess assess To the the the To assess assess distributional the distributional distributional the distributional impact distributional ofimpact impactthe of impact of the increase theimpact ofincrease increase in the the price of the increase inof in the increase the price in price cigarettes, theof the in price of the cigarettes, price cigarettes, of simulation of cigarettes, the the the for cigarettes, simulation simulation the simula alloa To assess tobacco To To differences the assess across assess in the the To the distributional the Toassess responses assess consumption distributional distributional the impact across differences thedistributional distributional distribution. of impact the impact in the increase of consumption Total of the impact responses impact tobacco the in increase the deciles of increase in acrosstothe of price the consumption in the increase of the reflect increase is quite cigarettes, price consumption price theof in in the similar of price the the cigarettes, cigarettes, fact that price across deciles of the of cigarettes, simulation poorthe to thecigarettes, simulation households reflect allows simulation the the simul the for allows likely fact that sim allowfo p bottom allows differences differences differences 80 forpercent inof differences in the the the in in responses responses population the the responses responses across across and only across consumption across consumption smaller consumption among consumption deciles thedeciles richest deciles to 20 todeciles to reflect reflect percent. reflect to In the reflect the the fact terms fact fact thatthe ofthat that poorfact that poor poor poor ho househo household differences the price of cigarettes, the simulation have in allows differences different the differences in price differences differences responses in for the the responses elasticities have acrossin responses in the the across relative different responses responses consumption acrossconsumption to pricehouseholdsacross across deciles consumption elasticities consumption to consumption deciles deciles with reflect more relative to the to reflect to deciles deciles fact reflect the resources. householdsthat to the fact Theto reflect poor factreflect that with thatthe poor different more the fact households poor that fact poor that likely households households elasticities, resources. poor ho like The havehave cigarette households have different different consumption, likely different priceprice only have the price elasticities elasticities poorest different priceelasticities 10 relative relative percent to relative to households (households elasticities households relative intohouseholds to households extreme with with poverty)more with more tend with resources. to more more resources. consume resources. resources. The The The differ different different el elast es to reflect the fact thathave poor different households have combined have different fewer with price the have different likely cigarettes, have price initial different different elasticities yetprice spend price arelative elasticities elasticities consumption similar price toelasticities relative relative budget elasticities households share patterns to relative relative with households to households relative across to theto more deciles, to households with richer households withresources. more more deciles. explain Thewith with The more resources. resources. whether similaritya in resources. more different The price Theresources. elasticities, different different reform The will The di diffe elastic elasticitie The combined combined differentcombined with combined with the elasticities, the with thewith initial initial combined the consumption initial consumption with initial consumption the consumption patterns initial patterns consumption across patterns across patterns across deciles, deciles, patterns across deciles, explain across explain deciles, explain deciles, whether whether explain whether a a price price abe whether refor price reform with more resources. The combined more consumption with combined different combined elasticities, regressive, the with combined patterns initial morewithcombined the is important consumption the initial neutral, more with initial consumption or withthe in more regressive, initial the consumption initial assessing patterns progressive. more consumption the consumption distributional across patterns patterns neutral, deciles, across or patterns impacts across more patterns of explain deciles, a across deciles, across potential whether explain explain progressive. deciles, tax deciles, reform. explain awhether price whether a whether explain a reform price whether pricewill reform price a pr awill be reform wb moremore explain regressive, more regressive, whether aregressive, more price more reform more neutral, neutral, will be or neutral, or more more more or more progressive. progressive. regressive, progressive. more neutral, or more progressive. s deciles, explain whether more regressive, a price more more reform more more regressive, will regressive, 3. Measuring more beneutral, more theregressive, moreregressive, or neutral, more neutral, distributional more more or neutral, progressive. more orimpact moreneutral, of orrise progressive. a price or more progressive.more progressive. progressive. on cigarettes The lossThis of The The real loss The loss section consumption of real of of The loss describes consumption loss real The real the of arising loss consumption real consumption partial of from real arising consumption equilibrium from arising thethe consumption arising approach price price arising from from used increases the arising toincreases from theprice simulate price in from the the in a aproduct product price increases increases impactthe increases on price in ai product in welfare is a obtained increases obtained in a product product proxied i is ias in is a follows: product i is obtained obtained obtained asias is fo folloob The loss The of The by real loss loss ashousehold consumption of of realThe The real loss consumptionlossof consumption real of arising follows: consumption because of an increase ∆) real consumption consumption from arisingarising the from price from thearising arising increases the price from price from increasesinthe athe increases in the price of cigarettes. This approach price in price product a inincreases a increases i product is product i in obtained is is used to in a i product is aas obtained obtained product asi follows: is i as is obtaine follows: obtai follow ∆)* ∆) ∆)*' (1) ∆)* ncreases in a product i is obtainedevaluate as follows:the first-order effects !of#$ + ∆ a change!#$ ' in ∗ ! #$ +∆) #$ +! ∆ prices. * , It ' ∆ #$ +#$ ! ∗ ' relies∆ ∗#$ , mainly #$ + * ∆ ' , ∗ on #$ ∆) ,∆) ∗ household , (1)(1) (1) (1) expenditure )*,, ∆)* ∆) * * )*,, patterns. The focus is on the ! impacts #$ + ∆ ! of a! + rise #$ #$#$ )' ∆ in ∗ +the * ! ∆ ' , price ∗ #$ #$ ) ) '! of #$ ∗ +, #$ )+ ∆ * cigarettes, *,, )∆ #$ ' ∗ because *,, #$ ' )*,,∗ , (1) this , has (1) been (1) the focus (1)(1) *,, )*,, )*,, *,, *,, (1) where #$where of iswhere the the share potential where is #$ reform of is #$ the product where of the #$ tobacco is share share the #$ ishare of inthe is taxes. of total product product share of household i in i of product in totalproduct total in i in expenditure i household household total total household for household expenditure expenditure household a expenditure forexpenditure for ain for a household a decile household for a household in j;ain aphouseh i,0aisdec decile in where where is where where the share is where the isis where theof the share product share share #$of is of #$ the ofisi productthe inshare product product share total i in in ofhousehold i product of in total total product totalhousehold household i in i in total expenditure household total household expenditure expenditure expenditure household for fora expenditure household for a expenditure for a household a household inin a householdfor a for decile in household aa in jhousehold ; a decilep is decile j; p in j; the initial#$ To price; the the assess #$ ∆ is the #$ initial initial the the initial price; # price; the distributional price ∆ initial price; ∆ impact # is increase; is∆ # the price; the of # is price theprice and ∆ the increase is price increase; the ∆ # increase; in is the price #$increase; the and price and of change increase; ∆ ∆ cigarettes, #$and is is #$ the in ∆and the the the is change #$ consumption ∆ the change simulation change #$in is inthe the the allows change in for of the consumption consumption the in good theof consumption that consump i,0 of the the i,0 gooog d expenditure for a household the initial the depends a in differences the a decile price; decile initial on initial the jin ∆; #p price; price the is∆ price; i,0 is the initial initial thethe price isinitial #∆ elasticity theprice; is price; increase; the ofprice; price ∆ price the # is ∆ and increase; increase; product. the is the is theprice ∆ and price #$ price is andincrease; ∆theincrease; increase; ∆ change is 11 is the andthe and and∆ in change ∆ the 11 change is #$ is consumption in the is theinthe change change the change likely of theof consumption consumption inin thethe in consumption the consumptio goodof the the that good goodtho depends depends the depends onon responses the depends the # on price across the price onconsumption price elasticity elasticity the price # of of elasticitythe elasticity 11 deciles the of product. to product. the #$ of reflect #$the the product. 11 factproduct. that poor 11 #$households he change in the consumption depends of dependsonthe the depends have good consumption price on different on the depends depends that elasticity the price of elasticities pricethe price on good ofthe on the the that elasticity elasticity price relative price product. ofdepends to of the elasticity elasticity the households on 11 product. the product. ofprice with 11 of the 11 more the product. product. elasticity resources. 11 11 of The the product. different 11 elasticities, combined with the initial consumption patterns across deciles, explain whether a price reform will be To estimateToTo more the To estimate estimate estimate regressive,variation To the more To estimate the thein estimate variation variation neutral,cigarette the variation in more cigarette or in consumption the variationin variation cigarette cigaretteconsumption progressive. in cigarette inafter consumption cigarette consumptionafter the consumption the price consumption afterafter price increase, the after the increase, price price the after the price increase, increase, the model model considers theincrease, price the the increase, modelmodel the mode theconside the cons To estimate To To the estimate estimate To variation the To the estimate variationestimate in variation the cigarette in thein variation variation cigaretteconsumption cigarette in cigarette in cigarette consumption consumption after consumption consumption the after price after the increase, the price after price after increase, the the the increase, price model theprice increase, the increase, considers model model the mode the the consider considers moth change in prices change change considers (∆ change in the in # ), prices change the change in (in prices tobacco prices ( ∆ in the # ), # ), ∆ prices prices (∆ theprice # ), tobacco (the the∆ , tobacco elasticity # ), tobacco tobacco the price price ( tobacco elasticity price ) elasticity $price decile for elasticity elasticity price ($( ) j,for elasticity )$for (and for decile decile the for $ )decile ( j$ )share j,for , decile and and and jof decile the, the and share sharej,the cigarette and ofshar of th ci fter the price increase, the model change in expenditure considers prices change The change loss in in inofperiod change the (∆ real prices change ), (0 prices consumptionthe ∆(in in tobacco #∆ ), ( ),prices prices arising the the (∆ from tobaccoprice(the ∆ tobacco ), the # ),price the elasticity #price tobacco price increases tobacco $a)The (in elasticity elasticity price for (price product elasticity (ielasticity )decile $for for for is )obtained (follows: j,inhousehold decile and decile as $j( ) , the $ for )j, for and anddecile share the decile theofj,each share j, of and shareand cigarettethe of the cigaretsha isciga #$ ).inThe change in 0expenditure $ each in each decile inexpenditure in period 0 (The #$ ). change expenditure for househo # # the expenditure share expenditure expenditure of cigarette in periodperiod expenditure0period (0( #$ in ). #$0 ). (* period The The ). #$changechange change . inThe in expenditure change expenditure in expenditure in expenditure for for each eachforfor each household each household household in in each each in de ($ ) for decile j, and expenditure the share presented ofin expenditure expenditure cigarette aspresentedperiod a share expenditure in expenditure in 0 period of ( period total 0). ( in The !0 +in (period ). period change The ∆ ). '∗ The0 ∆) (0 change , in ( ). change The ). expenditure in The in change expenditure change (1) for expenditure ineachin for expenditure expenditure household for each each household for in householdforeach eacheach in household decile inhousehold each each is deciledecin apresented asexpenditure astotal a of share and of averaged total by expenditure decile to and quantify averaged the by overall decile impact, to quantifyas t household inaseach #$ decile #$#$ is share presented #$ #$ *,, a as #$ share #$ of total expenditure and averaged by presented presented as a share share a of oftotal total expenditure expenditure ) expenditureand and and averaged averaged averaged by by decile decile by todecile to quantify to quantify quantify the the the overall overall ove imp im enditure for each household each in 12 presented as where a presented presented decile share as is presented presented a theas is of a share share total share of of as expenditure of total product a as i a share total expenditure in share total of of and expenditure total household total expenditure averaged and expenditure and averaged expenditure by averaged decile for aand by and householdaveraged averaged to by decile quantify decile intoa by to the quantify decile bydecile quantify j; p decile overall is to the to the quantify quantify impact, overalloverall the as the impa impact, ov follows: decile to quantify follows: follows: #$ follows: 12 12 follows:the12 overall impact, as follows: 12 12 i,0 ed by decile to quantify the overall follows: the 12 impact, initial follows: follows: 12 price; 12 as follows: follows: ∆# is the 12 price 12 increase; and ∆#$ is the change in the consumption of the good that M *NO ,$ = ((1 + ∆)!1 + $ (2) depends on theprice ∆ #elasticity of the product.11 ∗ ∆' − 1) ∗ PQR ST)SUV#RWXS NO 11For a detailed 11 discussion 11 For For a 11detailed For a a detailed of11the discussion detailed discussion Formethodology, a detailed discussion ofconsumption the discussion of the methodology, methodology, of the methodology,see Coady see of et the see al. methodology, see Coady Coady (2006); Kpodar et (2006); Coady et al. see(2006). al.the Coady et al. (2006); et Kpodar (2006); Kpodar al. (2006); (2006). Kpodar (2006). (2006). (2006). Kpodar To estimate the11 variation in cigarette after the price increase, model considers 11 12For a11detailed AnotherForexpression 11 For discussion a detailed a detailed Cigarette price Another Another 11 For For mightof Another expressiona12detailed the be discussion discussion Another elasticities expression expression amethodology, might mightdetailed of by discussion of the be discussion =of see expression the ∆ Expenditure methodology, be decile might the of Coady might methodology, ∆C∆P methodology, the see+ be ∆ Expenditureet bemethodology, seeal. ∆CP Coady (2006); Coady+al. ∆ Expenditure et et ∆PC see al..see Kpodar (2006); ? . ∆PC Coady ?? . Coady = et al. (2006). (2006);∆C∆P Kpodar et(2006). al. (2006); (2006); Kpodar + ∆CP + the Kpodar (2006). (2006). . (2006). Kpodar ∆PC ∆ Expenditure ∆ Expenditure = ? ∆C∆P = ∆C∆P =∆CP + ∆C∆P + ? ∆CP ?++ +∆CP ∆PC ∆PC ?. ?+ 12 12 12 ? ? (2006); Kpodar (2006). 12 Another change in prices expression 12 Another 12 Another (∆ 12 Another Another might expression12 be expression # ), might the tobacco expression expression ∆ Expenditure might be price might =might be ∆ Expenditure ∆ Expenditure elasticity be = ∆C∆P be+ ($ ) for decile ∆ Expenditure ∆ Expenditure ∆CP = ∆C∆P ∆C∆P + + ∆PC ∆CP = + ∆CP +.j∆C∆P ,=and ∆C∆P + ∆PC the ∆PC share .+ ∆CP + ?+ . ∆CP ? of+cigarette ? ∆PC ?. ?. ∆PC ?? ? ? ? expenditure in period 0 ( ). The change in expenditure for each household in each decile is ? + ∆PC?. The literature on the demand #$ elasticities to changes in the taxes on and prices of tobacco has largely presented as a share of total expenditure and averaged by decile to quantify the overall impact, as 9 focused on high-income countries. However, recent research also provides estimates for low- and follows:12 9 middle-income countries. Behind the estimates is a variety of methodologies and data sources. However, the overall 9 evidence suggests that the demand for tobacco products is more responsive to For a 11 For 11 detailed discussion a detailed discussion of ofthe methodology, the methodology,see Coady see et al. Coady et (2006); Kpodar al. (2006); (2006).. Kpodar (2006). inAnother price 12 low- and Another 12 middle-income expression might be might countries be ∆ Expenditure than = ∆C∆P + ∆CPin ? high + ∆PC income ?. countries. Table 2 summarizes selected examples. 9 Table 2. Tobacco Price Elasticities, Low- and Middle-Income Countries 13 Country Price elasticities Source China –0.50 to –0.64, cigarettes Hu and Mao (2002) The distributional impacts of cigarette taxation in Bangladesh Cigarette price elasticities by decile The literature on the demand elasticities to changes in the taxes on and prices of tobacco has largely focused on high-income countries. However, recent research also provides estimates for low- and middle-income countries. Behind the estimates is a variety of methodologies and data sources. However, the overall evidence suggests that the demand for tobacco products is more responsive to price in low- and middle-income countries than in high income countries. Table 2 summarizes selected examples. Table 2. Tobacco Price Elasticities, Low- and Middle-Income Countries Country Price elasticities Source China –0.50 to –0.64, cigarettes Hu and Mao (2002) Nepal –0.88, cigarettes Karki, Pant, and Pande (2003) –0.35, cigarettes –0.91, bidis John (2008) –0.88, leaf tobacco India −1.03, cigarettes Guindon et al. (2011) −0.94, bidis Selvaraj, Srivastava, −0.83, the lowest income group; −0.26, the highest and Karan (2015) –0.51 to –0.73 short run, cigarettes Morocco Aloui (2003) –1.36 to –1.54 long run, cigarettes Boshoff (2006); Reekie (1994); van der Merwe −0.5 to −0.87 and Annet (1998); South Africa van Walbeek (2000) −0.53 to −1.5 long-run price, tobacco elasticities −1.39 and −0.81, the poorest and richest income van Walbeek (2000), (2002) quartile, respectively Ali, Rahman, and –0.27, cigarettes Rahaman (2003) −0.5 short run, cigarettes Guindon, Perucic, and –0.7 long run, cigarettes Boisclair (2003) −0.43 to −0.66, cigarettes Nargis, Ruthbah, and Fong Bangladesh −0.22 to 0.64, bidis (2010); Nargis et al. (2014) –0.41 short run, cigarettes Barkat et al. (2012) –0.57 long run, cigarettes −0.49, cigarettes (−0.75, the poorest group; −0.40, the medium group; Nargis et al. (2014) −0.36, the richest group based on the housing index) 14 // Measuring the distributional impact of a price rise on cigarettes Few studies attempt to obtain elasticity estimates across income groups. On Bangladesh, Nargis et al. (2014) estimate total price elasticities of demand using data from wave 1 (2009) and wave 2 (2010) of the International Tobacco Control Bangladesh Survey. They find that the price elasticity of cigarette consumption is higher among people of lower socio- economic status based on a housing index (−0.75 for the poorest group, −0.40 for the medium group, and −0.36 for the richest group). A similar pattern is found in India and South Africa. In countries with a variety of tobacco products, the impact of changes in cigarette prices also depends on the potential substitution by other tobacco products. In the case of Bangladesh, it may be important to assess the potential substitution of bidis, which are also smoked, are significantly cheaper, and are widely consumed by the poor. Anecdotal evidence and cultural features suggest that, once a person starts smoking cigarettes, they will not go back to consuming bidis. Although this may be true among the large majority, there is limited evidence of the extent of this substitution across consumption deciles, particularly among extreme poor households. To complement the existing price elasticity estimates, price elasticities for cigarettes and other tobacco products across consumption deciles are estimated using the HIES 2016/17. Specifically, a system of demands is estimated using a quadratic almost ideal demand system (AIDS) (Banks, Blundell, and Lewbel 1997; Deaton and Muellbauer 1980; Mas-Colell, Whinston, and Green 1995). For trackability, the focus is on estimating a set of demands for five tobacco products (cigarettes, bidis, betel leaf, betel nut, and rolled betel leaf). A few products that are consumed by fewer than 4 percent of households are not included (tobacco leaf, gul, and other smoked tobacco, khoer, lime, and other smokeless products). The models incorporate a set of demographic controls, as well as a control for urban and rural location. To estimate the demand system and the price elasticities, one needs information about the prices of the various tobacco products. However, disaggregated data on prices to esti- mate the elasticities are not available. The HIES 2016/17 unit values obtained by dividing the total values reported by quantities may be used to infer information about prices. As Deaton (1997) points out, unit values not only capture information about prices, but are also affected by household choices on quality. In addition, unit values are measured with errors because they involve a calculation based on a ratio between values and quantities. Partially to address this concern, median values are calculated for the unit values across districts separated by urban and rural areas. This exploits the fact that the survey is repre- sentative across districts and provides a reasonable number of observations to calculate the median values.13 13 In the cases where the number of observations was less than 30, the medians of the full districts are used. 15 The distributional impacts of cigarette taxation in Bangladesh The quadratic AIDS estimates indicate that the price elasticity for cigarettes is −1.3. Compar- ing across deciles, one finds, similar to previous research results, that the elasticities are larger among the poorest (−1.36 for the first decile and −1.23 for the richest deciles). The average price elasticity for bidis is −1.22, and there is not a clear gradient across deciles. The quadratic AIDS own-price elasticities are higher than those estimated in the literature on Bangladesh. One disadvantage of this approach is that the elasticities are estimated using a cross-section with information on household consumption. In addition, this approach does not rely on time variation in prices to identify the elasticities. Another caveat derives from the use of unit values as the source of price information. However, one advantage of this approach is that it models simultaneously the demand for various tobacco products and also considers cross-price effects in the impact analysis. By jointly modeling the demand for various tobacco products, one considers responses for a larger population (65 percent of households consuming tobacco), compared with the more restricted population of cig- arette consumers (29 percent).14 This also explains why the estimated own-price elasticities are higher than in other studies: this estimation captures variations across households that do not include cigarette smokers. In this sense, the estimates are more closely aligned with estimates produced using models that attempt to measure longer-term elasticities, such as those of Aloui (2003) for Morocco and van Walbeek (2002) in South Africa (table 2). Yet, the estimates are still higher than those previously generated on Bangladesh; these elasticities are thus used as an upper bound in the analysis. In terms of cross-price elasticities, there is limited evidence for comparison. John (2008) finds evidence that bidis and cigarettes are complements rather than substitutes in India and that most cross-price elasticities across tobacco products are statistically insignificant. Guindon et al. (2011) find that poor and rural households in India substitute bidis for cigarettes, while richer households tend to consume them together as complements. Elasticity estimates based on HIES data also point to small cross-price elasticities. The cross-price elasticities for cigarettes and bidis are about 10 percent and show almost no variation across deciles. This is consistent with the anecdotal evidence indicating that, culturally, there is not much substitu- tion back to bidis from cigarettes, but that, among some households, there may be a small switch back to bidis if cigarette prices rise substantially.15 Table 3 summarizes the elasticity estimates across deciles used in the simulations. Three scenarios are compared. The first scenario assumes that households do not respond to the changes in prices, which implies a zero-price elasticity for cigarettes. The second scenario 14 The approach still conditions the sample on households that consume some form of tobacco. This exclusion would downward bias the estimated elasticities. 15 The HIES does not allow consumption to be separated by the type of cigarettes. Therefore, we cannot estimate cross-price elasticities by cigarette brands. This is a drawback of the analysis, as there has been significant downward switching to low-tier cigarette brands in the last decade, suggesting substitution of cheaper cigarette brands for expensive ones. 16 // Measuring the distributional impact of a price rise on cigarettes relies on estimates produced by Nargis et al. (2014) using individual data from wave 1 (2009) and wave 2 (2010) of the International Tobacco Control Bangladesh Survey. The last scenario relies on the quadratic AIDS model based on HIES 2016/17 that considers cross-price effects between cigarettes and bidis. Table 3. Quadratic AIDS Own-Price Elasticity Estimates, by Decile Scenario 2: Scenario 1: Nargis et Scenario 3: Quadratic AIDS modela no response al. (2014) estimates Cross-elasticity Cross-elasticity Decile Cigarettes Cigarettes Cigarettes Bidis bidis/cigarettes bidis/cigarettes 1.00 0.00 −0.75 −1.36 −0.11 −1.14 0.04 2.00 0.00 −0.75 −1.33 −0.11 −1.26 0.10 3.00 0.00 −0.75 −1.33 −0.10 −1.18 0.07 4.00 0.00 −0.52 −1.29 −0.09 −1.26 0.10 5.00 0.00 −0.40 −1.33 −0.11 −1.19 0.08 6.00 0.00 −0.40 −1.27 −0.08 −1.18 0.07 7.00 0.00 −0.39 −1.24 −0.07 −1.21 0.11 8.00 0.00 −0.36 −1.25 −0.08 −1.18 0.08 9.00 0.00 −0.36 −1.25 −0.08 −1.27 0.10 10.00 0.00 −0.36 −1.23 −0.08 −1.29 0.20 Average 0.00 −0.50 −1.29 −0.09 −1.22 0.10 Note: Consumption deciles are defined based on household consumption per capita deflated spatially to control for differences in prices across areas. The regressions control for household demographics and urban or rural location. The prices are approximated by the median unit values in a district across rural or urban areas. a. Calculations based on data of the HIES 2016/17. The three scenarios provide upper- and lower-bound estimates of the impacts of a change in the price of cigarettes. Using these bounds allows one to assess the sensitivity of the results to errors in the estimation of the elasticity. Even though the model pre- sented here is not dynamic, another way to motivate the bound analysis is to think about shorter- and longer-term responses from the short-run impact without response to a higher-response, longer-term impact. 17 SMOKING HAS The distributional impacts of cigarette taxation in Bangladesh BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE Introduction 18 // Executive Summary 4. DISTRIBUTIONAL IMPACTS OF A CIGARETTE PRICE INCREASE This section summarizes the results at three levels of cigarette price increases: 25 percent, 50 percent, and 75 percent. These price changes reflect several types of tax reforms that could include eliminating the existing tiered structure and introducing a uniform, specific excise tax. Discussing specific changes in the tax structure and rates is outside the scope of this note. Consumption changes that arise from an increase in cigarette prices are estimated for each decile based on the three elasticity scenarios presented in table 3: (1) the nonresponse case, in which price elasticities are assumed to be zero; (2) the medium-response scenario, in which the average elasticities of Nargis et al. (2014) are used; and (3) the high-response scenario in which quadratic AIDS elasticities are used that are greater than −1 and include changes in consumption arising because of changes in the consumption of bidis. The results of the simulations are summarized in table 4. They measure the average per- centage change in household consumption deriving from the price change, by decile.16 In an extreme scenario where there is no response to the price increase (complete pass- through), the impact across deciles is small. Assuming a price change of 50 percent, the estimated loss arising from a rise in household expenditures would be −0.5 percent for the first and richest deciles. If one allows for some level of response (the medium scenario) for a price change of 50 percent, one finds that the bottom deciles experience a loss of −0.4 percent, compared with −0.2 percent among the richest decile. In the high-response scenario, the impact translates into a gain because households reduce their cigarette con- sumption more than proportionately. For instance, at a 50 percent price increase, the gain for the poorest decile is 0.5 percent, compared with 0.4 percent for the top decile. 16 All simulations assume no changes in the consumption of nontobacco items. 19 The distributional impacts of cigarette taxation in Bangladesh Table 4. Simulations of a Price Increase on Cigarettes percentage change in consumption 25% cigarette price increase 50% cigarette price increase 75% cigarette price increase Complete Complete Complete Decile Medium High Medium High Medium High pass-through pass-through pass-through 1 −0.3 −0.2 0.2 −0.5 −0.4 0.5 −0.8 −0.6 1.0 2 −0.3 −0.2 0.2 −0.7 −0.5 0.6 −1.0 −0.7 1.2 3 −0.3 −0.2 0.2 −0.7 −0.5 0.6 −1.0 −0.7 1.2 4 −0.3 −0.2 0.2 −0.7 −0.4 0.6 −1.0 −0.5 1.3 5 −0.4 −0.2 0.2 −0.8 −0.3 0.7 −1.1 −0.5 1.5 6 −0.3 −0.1 0.2 −0.7 −0.3 0.6 −1.0 −0.4 1.2 7 −0.3 −0.1 0.2 −0.6 −0.2 0.5 −1.0 −0.4 1.1 8 −0.4 −0.1 0.2 −0.7 −0.3 0.6 −1.1 −0.4 1.2 9 −0.3 −0.1 0.2 −0.6 −0.2 0.5 −0.9 −0.3 1.1 10 −0.2 −0.1 0.1 −0.5 −0.2 0.4 −0.7 −0.3 0.8 The impacts on consumption presented in table 4 suggest that the magnitude of the direct impact is small. Even in an extreme case where consumers do not change their levels of consumption of cigarettes and prices increase by 75 percent, the average loss in consumption does not exceed 1.1 percent of total consumption. Table 5 summarizes the impacts in terms of the national upper poverty rate and consumption inequality measured by the Gini coefficient. Overall, there is no statistically significant impact on these indicators. 20 // Distributional impacts of a cigarette price increase Table 5. Changes in Poverty and Inequality Upper poverty rate Gini coefficient Baseline 24.3% 0.31 Complete pass-through 24.6% 0.31 25% cigarette price increase Medium 24.5% 0.31 High 24.1% 0.31 Complete pass-through 24.8% 0.31 50% cigarette price increase Medium 24.7% 0.31 High 23.8% 0.30 Complete pass-through 25.0% 0.31 75% cigarette price increase Medium 24.8% 0.31 High 23.4% 0.30 21 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 22 // Results 5. ADDING THE HEALTH IMPACTS OF A CIGARETTE PRICE INCREASE INTO THE EQUATION The simulations presented so far focus on the direct impact of an increase in cigarette taxes on consumption. However, by reducing tobacco consumption, tobacco taxation can also generate longer-term monetary gains because of the improved health of the population. A large body of literature has documented the economic gains associated with reduced tobacco consumption and a healthier population. This section attempts to monetize some of these gains across deciles to highlight that the monetary impacts of a healthier population largely exceed the direct impacts on consumption. The focus is on two types of gains. First, the savings because of reduced medical costs are considered. The most recent comprehensive estimates of the economic cost of smoking in Bangladesh refer to 2004 (WHO 2007). According to these estimates, the annual direct health care cost attributable to tobacco-related illnesses was Tk 50.9 billion (approximately US$990 million in 2016 prices). Second, the income gains derived from the increased number of working years are con- sidered. Verguet et al. (2015) analyze the health effects of a price increase in China and conclude that a 50 percent rise in prices would result in 231 million life years gained over 50 years and would have a significant impact among the poor. Pichón-Riviere et al. (2014) estimate that tobacco use in Chile would reduce life expectancy by nearly 4.0 years among women and 4.3 years among men. Alam et al. (2013) estimate that deaths among men ages 25 to 69 years result in an average loss of seven years of life per smoker in Bangla- desh. Estimates using data on 2016 indicate approximately 3 million total years of life lost because of premature mortality attributable to smoking.17 Table 6 summarizes back-of-the-envelope estimates of the savings in medical expenses and gains in consumption arising from the increase in working years across consumption deciles. The savings are monetized to represent a percentage change in household con- sumption. The methodology is presented in Fuchs and Meneses (2017, 2018). For medical costs, the exercise distributes the total cost directly attributed to tobacco illnesses across deciles based on the share of tobacco consumers in each consumption decile. To estimate 17 GBD Results Tool (database), Global Burden of Disease Study 2016, Global Health Data Exchange, Institute for Health Metrics and Evaluation, Seattle, http://ghdx.healthdata.org/gbd-results-tool. 23 The distributional impacts of cigarette taxation in Bangladesh the increase in working years, the years of life lost because of smoking are distributed across deciles proportionately to the number of households that smoke tobacco. Subse- quently, the income gain is approximated by increasing consumption proportionally to the number of working years gained and averaged across deciles. Table 6. Simulations Adding Savings in Medical Costs and Increases in Income Because of More Working Years Percentage change in consumption, increase of 25% in the price of cigarettes Medium elasticity High elasticity Savings in Gains due Savings in Gains due Consumption Total Consumption Total Decile medical to more medical to more effect effect effect effect costs working days costs working days 1 −0.2 0.5 2.7 3.0 0.2 0.8 5.0 5.9 2 −0.2 0.4 3.8 3.9 0.2 0.6 6.7 7.6 3 −0.2 0.3 4.2 4.3 0.2 0.6 7.4 8.2 4 −0.2 0.2 3.1 3.2 0.2 0.5 7.8 8.5 5 −0.2 0.1 2.8 2.8 0.2 0.5 9.2 9.9 6 −0.1 0.1 2.7 2.6 0.2 0.4 8.4 9.0 7 −0.1 0.1 2.7 2.6 0.2 0.3 8.6 9.1 8 −0.1 0.1 2.8 2.7 0.2 0.3 9.7 10.1 9 −0.1 0.1 2.7 2.6 0.2 0.2 9.3 9.7 10 −0.1 0.0 2.5 2.4 0.1 0.1 8.5 8.8 The results show that, compared with the direct consumption effects, the gains in health associated with reduced cigarette consumption in response to a price increase are substan- tially larger in magnitude. The main gains arise from the increased years of life, which trans- late into more years working. In a scenario of a 25 percent price increase and medium elastic- ity, these gains are about 3 percent of consumption. The highest direct consumption losses occur in a scenario with high response and a 75 percent price increase. The increased num- ber of productive years because of a 25 percent rise in the price of cigarettes is on the order of 20 times larger than the direct consumption impacts in a medium-response scenario and 24 // Adding the health impacts of a cigarette price increase into the equation 45 times larger in a higher-response scenario. These results are useful in highlighting that the short-term losses because of tobacco taxes are much smaller in order of magnitude than the economic benefits because of better health. Yet, these estimates must be interpreted with care because they are subject to several caveats. First, the estimate of medical costs was calculated 14 years ago and therefore may not reflect the current situation. Second, the estimates are based on a series of assumptions about access to medical care by the population (for example, that 25 percent of people experiencing a disease caused by tobacco would seek inpatient care) and focus on only eight tobacco-related diseases. Third, the previous simulations centered on cigarette consumption, while the medical costs are calculated for all tobacco products. Finally, the simulations are not able to model individual behavioral responses to increases in tobacco prices and do not consider smoking histories; they only rely on differences in the share of households consuming cigarettes by decile and the different price elasticities. The health benefits of reducing tobacco consumption presented in table 6 accrue only from complete abstinence, which would be reflected in a reduction in smoking prevalence and number of smokers. However, the elasticities estimated here include both the effect on smoking prevalence and the effect on cigarettes smoked per day among continuing smok- ers. The epidemiological evidence indicates that continuing smokers do not experience tangible health gains in terms of avoiding diseases or gains in life years. To address this concern, the literature has generally used only the effect on smoking prevalence, which is broadly half the total price elasticity according to the literature (IARC 2011). Even if one reduced the price responses by half, the estimates are significantly larger than the direct impacts in consumption. 25 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 6. CONCLUSION Despite the obvious positive health impacts of tobacco taxation, an argument raised against taxation is that poor households bear much of the burden of the increased prices because they spend more on tobacco. This note summarizes estimates of the distributional impact of increases in the prices of cigarettes based on a partial equilibrium approach using the HIES 2016/17. The note highlights that tobacco consumption is widespread across households in Ban- gladesh. In 2016, approximately 65 percent of Bangladeshi households had consumed some type of tobacco product during the month before the interview; 42 percent had consumed smoked tobacco; and 45 percent had consumed smokeless tobacco products. About 29 percent of households consumed cigarettes. The poorest 10 percent (house- holds in extreme poverty) were less likely to consume cigarettes (19 percent). There is little difference in the budget shares spent on tobacco across the consumption distribution. About 3 percent of total household consumption was allocated to tobacco products. Focusing on cigarettes, the budget share is similar across the consumption dis- tribution (1.3 percent on average). The similarity of consumption patterns is important in assessing the distributional impacts of a potential tax reform. Overall, this note shows that an increase in the price of cigarettes has small consumption impacts and does not significantly change the poverty rate or consumption inequality. These results stem from the fact that poor households are slightly less likely to consume cigarettes and that cigarette budget shares are relatively even across the consumption distribution. Assuming an extreme case whereby households do not change their cigarette consump- tion because of an increase in cigarette prices (complete pass-through), there are small impacts in terms of consumption, poverty, and inequality. For instance, in a scenario in which prices increase by 50 percent, the average consumption loss would be 0.6 percent and similar across deciles. The official upper poverty rate would increase slightly, but this increase is not statistically significant. There is also no impact on consumption inequality measured using the Gini. Assuming households reduce cigarette consumption because of higher prices, the impacts become even smaller. Indeed, in a high-response scenario in which elasticities are larger than −1 and include some substitution by bidis, the impacts translate into gains because households reduce their consumption of cigarettes more than proportionately. In this 27 The distributional impacts of cigarette taxation in Bangladesh upper-bound case, poverty would fall by about a half percentage point, but the change is not statistically significant. There are no changes in consumption inequality. The short-term consumption impacts are also small compared with the estimated gains because of savings in medical costs and the greater number of productive years of life. In particular, the increased number of productive years associated with a 25 percent rise in the price of cigarettes is on the order of 20 times more than the direct consumption impacts in a medium-response scenario and 45 times more in a higher-response scenario. The results presented focus on changes in the price of cigarettes. The link with a specific reform and revenue impacts is outside the scope of this paper and will be addressed in future research. A reform in cigarette taxation that has been advocated by partner orga- nizations for tobacco control would involve (1) the introduction of a specific excise tax replacing the current ad valorem system with a mixed system (a mix of ad valorem and specific excise taxes), (2) greater reliance on specific excise taxes, and (3) a gradual transi- tion to a uniform structure. This reform implies a greater increase in price for low-tier brands, which are largely con- sumed by low-income smokers. This argument has been used in Bangladesh against the reform and in favor of the persistence of the multitiered tax structure. However, the HIES does not identify the type of cigarettes smoked by households so that one might directly address this concern. Nonetheless, the results presented here highlight that tobacco tax- ation is likely to have a small impact on overall consumption and poverty, even if, in some scenarios, the relative impact could be greater among poorer consumers because of the greater consumption of lower-tier cigarettes. This is the result of the combination of the smaller share of poor households consuming cigarettes and the higher response of these households to the change in prices. 28 // Conclusion 29 SMOKING HAS The distributional impacts of cigarette taxation in Bangladesh BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 30 // References REFERENCES Ahmed, N. 2012. “Tobacco Taxation in Bangladesh: Administrative and Policy Constraints.” National Board of Revenue, Ministry of Finance, Dhaka. http://www.who.int/tobacco/economics/ meetings/wctoh_2012_tob_tax_bangladesh.pdf. Alam, D. S., P. Jha, C. Ramasundarahettige, P. K. Streatfield, L. W. Niessen, M. A. Chowdhury, A. T. Siddiquee, S. Ahmed, and T. G. Evans. 2013. “Smoking-Attributable Mortality in Ban- gladesh: Proportional Mortality Study.” Bulletin of the World Health Organization 91 (10): 757–64. Ali, Z., A. Rahman, and T. Rahaman. 2003. “An Economic Analysis of Tobacco Control in Bangladesh.” HNP Discussion Paper, World Bank, Washington, DC. Aloui, O. 2003. “Analysis of the Economics of Tobacco in Morocco.” HNP Discussion Paper, World Bank, Washington, DC. Banks, J., R. Blundell, and A. Lewbel. 1997. “Quadratic Engel Curves and Consumer Demand.” Review of Economics and Statistics 79: 527–39. Barkat, A., A. U. Chowdhury, N. Nargis, M. Rahman, Pk. A. Kumar, S. Bashir, and F. J. Cha- loupka. 2012. “The Economics of Tobacco and Tobacco Taxation in Bangladesh.” Interna- tional Union against Tuberculosis and Lung Disease, Paris. Boshoff, W. H. 2008. “Cigarette Demand in South Africa over 1996–2006: The Role of Price, Income, and Health Awareness.” South African Journal of Economics 76 (1): 118–31. Coady, D. P., M. El Said, R. Gillingham, K. Kpodar, P. A. Medas, and D. L. Newhouse. 2006. “The Magnitude and Distribution of Fuel Subsidies: Evidence from Bolivia, Ghana, Jordan, Mali, and Sri Lanka.” IMF Working Paper WP/06/247, International Monetary Fund, Wash- ington, DC. Deaton, Angus S. 1997. The Analysis of Household Surveys: A Microeconometric Approach to Development Policy. Washington, DC: World Bank. Deaton, Angus S., and J. Muellbauer. 1980. Economics and Consumer Behavior. Cambridge, UK: Cambridge University Press. Duong, MyLinh, S. Rangarajan, X. Zhang, K. Killian, P. Mony, S. Swaminathan, A. V. Bharathi, et al. 2017. “Effects of Bidi Smoking on All-Cause Mortality and Cardiorespiratory Out- comes in Men from South Asia: An Observational Community-Based Substudy of the Pro- spective Urban Rural Epidemiology Study (PURE).” Lancet Global Health 5 (2): e168–e176. 31 The distributional impacts of cigarette taxation in Bangladesh Fuchs, A., and F. J. Meneses. 2017. “Regressive or Progressive? The Effect of Tobacco Taxes in Ukraine.” December 27, World Bank, Washington, DC. ———. 2018. “Tobacco Price Elasticity and Tax Progressivity in Moldova.” World Bank, Washington, DC. Guindon, G. Emmanuel, Arindam Nandi, Frank J. Chaloupka IV, and Prabhat Jha. 2011. “Socioeconomic Differences in the Impact of Smoking Tobacco and Alcohol Prices on Smoking in India.” NBER Working Paper 17580, National Bureau of Economic Research, Cambridge, MA. Guindon, G. Emmanuel, A. M. Perucic, and D. Boisclair. 2003. “Higher Tobacco Prices and Taxes in Southeast Asia: An Effective Tool to Reduce Tobacco Use, Save Lives, and Gener- ate Revenue.” HNP Discussion Paper, World Bank, Washington, DC. Guindon, G. Emmanuel, P. Tobin, M. Lim, and D. Boisclair. 2007. “The Cost Attributable to Tobacco Use: A Critical Review of the Literature.” Tobacco Free Initiative, World Health Organization, Geneva. Hu, T. W., and Z. Mao. 2002. “Effects of Cigarette Tax on Cigarette Consumption and the Chinese Economy.” Tobacco Control 11(2):105–08. IARC (International Agency for Research on Cancer). 2011. Effectiveness of Tax and Price Policies for Tobacco Control. Vol. 14 of IARC Handbooks of Cancer Prevention: Tobacco Control. Geneva: IARC Press, World Health Organization. ITC Project (International Tobacco Control Policy and Evaluation Project). 2014. “Tobacco Price and Taxation Policies in Bangladesh: Evidence of Effectiveness and Implications for Actions.” University of Waterloo, Waterloo, Ontario, Canada. Jha, P. 2015. “Deaths and Taxes: Stronger Global Tobacco Control by 2025.” Lancet 385: 918–20. Jha, P., and R. Peto. 2014. “Global Effects of Smoking, of Quitting, and of Taxing Tobacco.” New England Journal of Medicine 370 (1): 60–68. John, R. M. 2008. “Price Elasticity Estimates for Tobacco Products in India.” Health Policy and Planning 23: 200–09. Karki, Y. B., K. D. Pant, and B. T. Pande. 2003. “A Study on the Economics of Tobacco in Nepal.” HNP Discussion Paper, World Bank, Washington, DC. 32 // References Kostova, Deliana, Frank J Chaloupka, Ayda Yurekli, Hana Ross, Rajeev Cherukupalli, Linda Andes, and Samira Asma.. 2014. “A Cross-Country Study of Cigarette Prices and Affordabil- ity: Evidence from the Global Adult Tobacco Survey.” Tobacco Control 23: e3. Kpodar, K. 2006. “Distributional Effects of Oil Price Changes on Household Expenditures: Evidence from Mali.” IMF Working Paper WP/06/91, International Monetary Fund, Washing- ton, DC. Mas-Colell, A., M. D. Whinston, and J. R. Green. 1995. Microeconomic Theory. New York: Oxford University Press. Nargis, N., U. H. Ruthbah, and G. T. Fong. 2010. “Taxation of Tobacco Products in Bangla- desh: Findings from the 2009 ITC Bangladesh Survey.” ITC Project Working Paper, University of Waterloo, Waterloo, Ontario, Canada. Nargis, N., U. H. Ruthbah, A. K. M. G. Hussain, G. T. Fong, I. Huq, and S. M. Ashiquzzaman. 2014. “The Price Sensitivity of Cigarette Consumption in Bangladesh: Evidence from the International Tobacco Control (ITC) Bangladesh Wave 1 (2009) and Wave 2 (2010) Surveys.” Tobacco Control 23: i39–i47. Pichón-Riviere, Andrés, Ariel Bardach, Joaquín Caporale, Andrea Alcaraz, Federico Augus- tovski, Francisco Caccavo, Carlos Vallejos, et al. 2014. “Carga de enfermedad atribuible al tabaquismo en Chile.” Documento Técnico IECS 8 (April), Instituto de Efectividad Clínica y Sanitaria, Buenos Aires. Poi, Brian P. 2012. “Easy Demand-System Estimation with Quaids.” Stata Journal 12 (3): 433–46. Reekie, W. D. 1994. “Consumers’ Surplus and the Demand for Cigarettes.” Managerial and Decision Economics 15: 223–34. Sankaran, S., H. Hillamo, and S. A. Glantz. 2015. “Implementation of Graphic Health Warn- ing Labels on Tobacco Products in India: The Interplay between the Cigarette and the Bidi Industries.” Tobacco Control 24: 547–55. Selvaraj, S., S. Srivastava, and A. Karan. 2015. “Price Elasticity of Tobacco Products among Economic Classes in India, 2011–2012.” British Medical Journal 5 (12). van der Merwe, R., and N. Annett. 1998. “The Effects of Taxation on Consumption in South Africa.” In The Economics of Tobacco Control in South Africa, Economics of Tobacco Control Project, chapter 4. Cape Town: International Tobacco Initiative, School of Economics, University of Cape Town. 33 The distributional impacts of cigarette taxation in Bangladesh van Walbeek, C. 2000. “Impact of the Recent Tobacco Excise Tax Increases on the Future Government Revenue Potential in South Africa.” In The Economics of Tobacco Control in South Africa, Economics of Tobacco Control Project, Research Release 1. Cape Town: Inter- national Tobacco Initiative, School of Economics, University of Cape Town. ———. 2002. “The Distributional Impact of Tobacco Excise Increases.” South African Jour- nal of Economics 70 (3): 258–67. Verguet, S., C. L. Gauvreau, S. Mishra, M. MacLennan, S. M. Murphy, E. D. Brouwer, R. A. Nugent, K. Zhao, P. Jha, and D. T. Jamison. 2015. “The Consequences of Tobacco Tax on Household Health and Finances in Rich and Poor Smokers in China: An Extended Cost-Ef- fectiveness Analysis.” Lancet Global Health 3 (4): e206–e216. WHO (World Health Organization). 2007. “Impact of Tobacco-Related Illnesses in Bangla- desh.” WHO, Geneva. ———. 2011. “Economics of Tobacco Toolkit: Assessment of the Economic Costs of Smok- ing.” WHO, Geneva. http://apps. who.int/iris/bitstream/10665/44596/1/9789241501576_eng. pdf. ———. 2013. Global Tuberculosis Report 2013. Geneva: WHO. apps.who.int/iris/bitstr eam/10665/91355/1/9789241564656_eng.pdf. ———.2015a. Global Report on Trends in Prevalence of Tobacco Smoking, 2015. Geneva: WHO. http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. ———. 2015b. WHO Report on the Global Tobacco Epidemic 2015: Raising Taxes on Tobacco. Geneva: WHO. ———. 2017. “WHO Tobacco Fact Sheet.” WHO, Geneva. http://www.who.int/mediacentre/fact- sheets/fs339/en/. World Bank. 2017. “Preliminary Recommendations for Effective Tobacco Tax Design in Ban- gladesh.” World Bank Policy Note, World Bank, Washington, DC. 34 // References 35 SMOKING HAS The distributional impacts of cigarette taxation in Bangladesh BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 36 // Annex ———. apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. 2013. ———. ———.Global 2011. 2011. Tuberculosis “Economics“Economics of of Tobacco ReportToolkit: Tobacco Toolkit: 2013. ———. Assessment Assessment 2011. Geneva: of “Economicofthe the Eco W E ———. ———.2015a.———.2015a. WHO, 2013. ———. Geneva. Global ———. Global Global ———.2015a. Report 2011. http://apps. 2011. Report on “Economics “Economics apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. on Tuberculosis Trends who.int/iris/bitstream/10665/44596/1/9789241501576_eng. Global Trends in ofTrends of in Tobacco Report PrevalenceTobacco Prevalence onof Report Toolkit: Trends Toolkit: of Tobacco Tobacco Assessmentin 2013. Smoking, Assessment PrevalenceSmoking, of 2015 of Geneva: of the 2015 Tobacco .the Geneva: Economic .Economic Geneva: Smoking, WHO.Costs WHO. Costs 2015 ofpdf S of ———.2015a. Global Report Global on TrendsReport WHO, WHO, on in Geneva. Geneva. Prevalence in http://apps. Prevalence http://apps. of Tobacco of Tobacco who.int/iris/bitstream/10665/44596/1/ who.int/iris/bitstream/10665/44596/ Smoking, Smoking, 2015. WHO, 2015 . Geneva. Geneva: http:/ W s of Tobacco Toolkit: Assessment Toolkit: of ———.2015a. Assessment ———. Economic of the Economic apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. 2013. http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pd http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. the http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. ———.2015a. Costs WHO, WHO,of Smoking.” Geneva. GlobalGeneva. Costs Global http://apps. Report http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. of http://apps. Smoking.” on Trends ———. Tuberculosis 2011. who.int/iris/bitstream/10665/44596/1/9789241501576_e in Prevalence “Economics who.int/iris/bitstream/10665/44596/1/978924150157 Report of Tobacco of 2013. Tobacco Smoking, Geneva: Toolkit: Geneva: 2015 . 2013. WHO. Assessmen Geneva: WHO W /apps. who.int/iris/bitstream/10665/44596/1/9789241501576_eng. /iris/bitstream/10665/44596/1/9789241501576_eng. http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. ———.2015a. ———. ———. apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. 2015b. 2015b. Global WHO WHO Report ———. Report onon———. pdf. ———. 2015b. the Trends on the Global WHO in Global Tobacco 2013. pdf. Prevalence2013. Report WHO, Tobacco Epidemic on of Global Geneva. Epidemicthe Global Tobacco Global 2015: http://apps. 2015: Smoking, Tobacco Raising TuberculosisTuberculosis Raising Epidemic Taxes 2015 ———. who.int/iris/bitstream/106 Taxes on . Geneva: 2015: on Tobacco Report Tobacco Report .Raising WHO. Geneva: . Taxes2013 20 G o ———. ———. ———. http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. 2015b. WHO 2013. 2013. Report Globalon Global the Global Tuberculosis Tuberculosis Tobacco Epidemic Report 2015: Report Raising 2013. 2013. Taxes on Geneva: Tobacco Geneva: . Ge lobal erculosis Report ———. Tuberculosis WHO. Report http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. 2013. ———.2015a. WHO.2015b. Geneva:WHO 2013. GlobalReport WHO. Report WHO. on theapps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pd Geneva: apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.p Global on Trends Tobacco WHO. ———. in Epidemic Prevalence 2013. 2015: of Tobacco Raising Global Smoking, Taxes on2015 apps.who.int/iris/bitst Tobacco Tuberculosis . . Geneva: Geneva: WHO Rep ———. WHO. apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. apps.who.int/iris/bitstream/10665/91355/1/9789241564656_eng.pdf. 2015b. WHO Report on the Global Tobacco Epidemic 2015: Raising Taxes onTobacco Tobacco Ge .Smok ream/10665/91355/1/9789241564656_eng.pdf. 1355/1/9789241564656_eng.pdf. ———. WHO. ———. 2015b. 2017. WHO 2017. Report http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. ———. “WHO on “WHO———.2015a. ———.2015a. the Global2017. Tobacco Tobacco Tobacco Global Global “WHO Epidemic Report Fact Report apps.who.int/iris/bitstream/10665/91355/1/978924156 Fact 2015: on on Tobacco Trends Sheet.” RaisingTrends Sheet.” in Taxes in PrevalencePrevalence ———.2015a. Fact WHO, on Tobacco WHO, of Sheet.”of Tobacco .2015 Geneva: Geneva. Global Geneva. Smokin Re W ———. ———.2015a. WHO. ———.2015a. 2017. Global Global “WHO Report Report on on Trends Trends Tobacco in inPrevalence Prevalence Fact of Tobacco of Tobacco Sheet.” Smoking, Smoking, WHO, 2015 . Geneva . GeneGe port on Trends in Prevalence of Tobacco Smoking, in Prevalence ———. WHO. of ———. Tobacco 2015 2017. 2015b. Smoking, http://www.who.int/mediacentre/factsheets/fs339/en/. Geneva: http://www.who.int/mediacentre/factsheets/fs339/en/. WHO “WHO 2015 Report “WHO WHO. . http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922 Geneva: http://apps.who.int/iris/bitstream/10665/156262/1/97892415649 Tobacco on the Global Tobacco http://www.who.int/mediacentre/factsheets/fs339/en/. . http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. WHO. ———.2015a. Fact EpidemicFact Global Sheet.” Report 2015: Raising on WHO, Trends http://apps.who.int/ir Taxes on WHO, in Geneva. Prevalence Tobacco. Geneva of T ———. http://apps.who.int/iris/bitstream/10665/156262/1/9789241564922_eng.pdf. 2017. http://www.who.int/mediacentre/factsheets/fs339/en/. Tobacco Sheet.” Ge is/bitstream/10665/156262/1/9789241564922_eng.pdf. ———. http://www.who.int/mediacentre/factsheets/fs339/en/. WHO. 2017. World “WHO ———. Bank. ———. 2015b. Tobacco 2017. 2015b. WHO WHO Report Report http://apps.who.int/iris/bitstream/10665/156262/1/97 Fact on on the Sheet.” the Global Global Tobacco Tobacco ———. WHO, Epidemic Epidemic 2015b. 2015: Geneva. 2015: WHO Raising RaisRe 0665/156262/1/9789241564922_eng.pdf. World World Bank.World Bank. ———. 2017. ———.Bank. 2017. “Preliminary 2015b. 2017. “Preliminary 2015b. http://www.who.int/mediacentre/factsheets/fs339/en/. WHO WHO “Preliminary Recommendations Report Recommendations Report on “Preliminary the on Recommendations Global the Global for Tobacco Recommendations Tobacco for Effective Effective Epidemic Epidemic for Tobacco Effective 2015: Tobacco 2015: for Raising Tax Raising Tobacco Effective Tax Design Taxes Taxes on Tobacco Design Tax Tobacco on in Tobac Desig in. port on the Global al Tobacco Epidemic 2015: Raising Tobacco WorldEpidemic Bank. 2015: 2017. Raising http://www.who.int/mediacentre/factsheets/fs339/en/. ———. Bangladesh.” Taxes Bangladesh.” onWorld 2017. Tobacco “Preliminary World Taxes Bank Bangladesh.” . Geneva: Bank on Policy“WHO WHO. Tobacco Policy WHO. Recommendations Note, . World Note,Geneva:World ———. Tobacco Bank World Bank, 2015b. Policy Bank, for Washington, Fact Note,WHO Effective Washington, World Report DC. Sheet.” Tobacco Bank, on the DC. DC.Tobacco Tax Desig WHO. Global Washington, Tax WHO, Design Tobacco DC. in Epidemic Geneva World Bangladesh.” WHO. Bank. WHO. 2017. World “Preliminary Bank Policy Recommendations Note, World Bank, for Washington, Effective World Bangladesh.” http://www.who.int/mediacentre/factsheets/fs339/en/. Bank. 2017. World Bank “Preliminary ———. Policy ———. Note, Recommendations World 2017. 2017. WHO. Bank, “WHO Washington, for “WHO Effective DC. Tobacco Tobacco Tobacco ———. Tax Fact Fact Design 2017. Sheet.” inSheet ———. ———. World Bangladesh.” 2017. 2017. Bank “WHO Policy “WHO Note, World Tobacco Tobacco Bank, Washington, Fact Fact Sheet.” DC. Sheet.” WHO, WHO, “WHO obacco Tobacco Sheet.” Fact Fact Bangladesh.” World Bank. WHO, Sheet.” World 2017. Bank WHO, “Preliminary Geneva. Policy Note,Geneva.http://www.who.int/mediacentre/factsheets/fs339/en/. http://www.who.int/mediacentre/factsheets/fs339/en/. World ———. Recommendations Bank, Washington, 2017. for Effective DC. “WHO Tobacco http://www.who.int/m Tobacco Tax Design Factin http://www.who.int/mediacentre/factsheets/fs339/en/. http://www.who.int/mediacentre/factsheets/fs339/en/. mediacentre/factsheets/fs339/en/. tsheets/fs339/en/. Bangladesh.” World Bank World World Policy Bank. Bank. Note, World 2017. 2017. “Preliminary “Preliminary http://www.who.int/mediacentre/factsheets/fs339/en/ Bank, Washington, Recommendations Recommendations DC. World Bank. for for Effective 2017. Effective “Pr World World Bank. Bank. 2017. 2017. “Preliminary “Preliminary Recommendations Recommendations forfor Effective Effective Tobacco Tobacco Tax Tax D Annex – Estimation 1Effective Annex of Quadratic1– Bangladesh.” Estimation Bangladesh.” AIDS of World Quadratic World Elasticities Bank Bank Policy AIDS Policy Note, Elasticities Note, World World Bank, Bangladesh.” Bank, Washington,Washington, World DC BD ANNEX 1 – ESTIMATION OF eliminary Recommendations ommendations for Effective Annex 1for – Tobacco Annex Estimation 1 Tax – Bangladesh.” of Tobacco Estimation Design Quadratic World inTax of AIDS Quadratic Bank Design Elasticities World Policy in AIDS Note, Bank. ElasticitiesWorld 2017. Bank, “Preliminary Washington, Recommendations DC. for Annex 1 – Estimation Bangladesh.” of Quadratic World AIDS Bank Policy Elasticities Note, World Bank, Washington, DC. ank e, Policy World Note, Bank, World Bank, Washington, DC. Washington, Annex 1 – DC. Estimation of Quadratic AIDS Bangladesh.” Elasticities World Bank Policy Note, World Bank, Wa Annex We 1consider We consider – Estimation a set a of set of ofconsumer We consumer Quadratic consider demands a AIDS demands set for of Elasticities consumer afor set a of set k of demands tobacco k tobacco for products a set of k and tobacco ingredients products for and which ing set of kproducts tobacco and ingredients for which We consider the consumer the We Annex the consumer consumer consider We aconsider We consumer the 1 a set – has budgeted consider has set QUADRATIC AIDS ELASTICITIES of consumer Estimation has budgeted budgeted of a set of consumer athe consumerhassetconsumer Tk of of consumer Tk Annex budgeted demands m m Quadratic Tk Specifically, .Annex m.has . Specifically, demands Tk 1– 1 demands for Specifically, m for budgeted demands AIDS Estimation – a set Elasticities . Specifically, the k Estimation a the set of of for Tk the for kAIDS ka goods goods k k a of tobacco m. tobacco goods set theof Specifically, Quadratic include of Quadratic k goods include k products include tobacco products cigarettes, AIDS the include cigarettes, cigarettes, AIDS products and and k goods products ingredients bidis, Elasticities Elasticities Annex cigarettes, bidis, ingredients bidis, and ingredients include betel and betel 1– betel bidis, for which for w leaf, ingredients Estimation leaf,which for cigarettes, leaf, betel betel betel leaf, betel for bid of w nut, nut, and and the rolled rolled Annex betel consumer Annex betel 1 leaf. – 1 nut, hasEstimation – leaf. A few Estimation and A budgeted rolled few products of products betel Tk that Quadratic of m Quadratic leaf. . Annex that are consumed Specifically, A are few AIDS consumedproducts Elasticities theconsumed Elasticities by k goods by fewer fewer that are than include consumed than4 percent 4 cigarettes, percent by ofpercentfewer of households households bidis, than betel 4 per leaf, DSQuadraticElasticities AIDS Elasticities and nut,consumer the Werolled nut, consider and hasbetel rolled a set budgetedleaf. of betel Aconsumer few Tk leaf.mproducts . A few demands Specifically, thatproducts areforconsumed the athat k 1set are – Estimation goods of k includeby tobacco fewer of by thanQuadratic products cigarettes, fewer 4 percent and bidis, than AIDS 4 betelof households ingredients Elasticities leaf, forhouseh of betel which are not are not included nut, included and (tobacco rolled are (tobacco betel not leaf, included leaf, leaf. We gul, We A gul, and consider few consider (tobacco and other products aother asmoked set set of leaf, smoked that of consumer gul, tobacco, are consumer and tobacco, consumed demandsotherkhoer, demands khoer, smoked bylime, for fewerfor alime, Weand athan set tobacco, set of andother consider ofk4 tobacco kother percent khoer, smokeless tobacco a othersetsmokeless of lime, products of househ produc cons an are nut, not andtheincluded are consumer rolled We consider not We We betel included (tobaccohas consider consider leaf. a set budgeted A of a(tobacco leaf,set few a consumer setgul, of Tk products of leaf, and consumer mconsumer demands gul, other . Specifically, that and demands are forsmoked demands a other consumed set the of for k smoked tobacco, forgoods a tobacco set by a set of fewer tobacco, khoer, include k of products tobacco kthan lime, tobacco khoer, cigarettes, 4 and and products percent products ingredients lime, other bidis, of and and smokeless betel and households ingredients leaf,smok ingredient bete f umer for a demands set of k tobacco for a set products). of products products). k tobacco are not and products ingredients included products). and for (tobacco ingredients which the the leaf, consumer consumer for gul, We which hasand has other budgeted consider budgeted smoked a TkTk set m. m of tobacco, Specifically, . Specifically, consumer khoer, the demands the the klime, consumer goods k for goods and a include set other has includeof k smok cigare budgete cig toba products). are not nut, products). and included for the which rolledtheconsumer (tobacco the betel consumer consumer leaf. has leaf, A has few budgeted gul, has budgetedand budgeted products other TkTk m.that m smokedSpecifically, are consumed Specifically, . Specifically, tobacco, the the the kkhoer, bygoods k goods goods fewer include lime, include than include and 4 percent cigarettes, cigarettes, cigarettes, other of smokeless bidis, household bidis, betel bete l dcally, Tk m the . Specifically, k goods include the k cigarettes,goods products). include bidis, cigarettes, betel leaf, bidis, nut, betel nut, betel and and rolled leaf, rolled the betelbetel betel consumer leaf. leaf. A has A few few products budgeted products that Tk that m are nut, . are consumed Specifically, and consumed rolled by the by betel fewer k fewer leaf. goods tha At products).The are not bidis, quadratic included nut, nut, betel andAIDS and leaf, (tobacco rolled betel The rolled model betel nut, quadratic leaf, beteland of leaf. gul, leaf. rolled Banks, A AIDS few A and betel few products model Blundell, other leaf. products Aof smoked few and that Banks, that products Lewbel are tobacco, are consumed Blundell, thatconsumed (1997) are khoer, consumed and is by based by lime, fewer Lewbel fewer by on and than than (1997) an other 4 percent indirect 4 is smokeles percentbased of of utility ho on Athat feware products consumed that by arefewer The consumed quadratic than The 4 by AIDS fewer quadratic percent model than of AIDS 4 percent households of Banks, are model are not ofnot Blundell, included households of included Banks, nut, (tobacco and and (tobacco Blundell, Lewbel rolled leaf, and betel leaf, (1997)gul, Lewbel gul, leaf. and is Aand based other few (1997) other on aresmoked productsis an notindirect smoked based indirect that tobacco, includedon tobacco, are an utility khoer, (tobacco consumed indirect khoeu The quadratic products). are AIDS not model included of of (tobacco Banks, Blundell, leaf, gul, and and Lewbel other (1997) smoked leaf, is tobacco,based on khoer, an lime, and utility other leaf, gul, and other smokedfunction function fewer that tobacco, The that leads are than leads quadratic khoer, to 4not apercent to included function system lime, a AIDS system and of households that demands model products). (tobacco of products). other of leads demands of smokeless are to Banks, leaf, expressed a notsystem expressed are notexpressed gul, included Blundell, in and of expenditure included demands in other (tobacco expenditure and (tobacco smoked Lewbel expressed shares. gul, shares. tobacco, (1997) leaf, and Let in other gul, expenditure Let is products). # and khoer, denote smoked based denote other the on lime, shares.the quantity an smoked and Let quantity indirect othe#su tob other smoked tobacco, khoer, function The quadratic function lime, that and leads AIDS products). that other to products). a leads system model smokeless tobacco, khoer, lime, and other smokeless products). to of of a system demands Banks, Blundell, demands expressed and in expenditure Lewbel in (1997) expenditure shares. is based Let shares. on denotean #Let indirect the # denotequantity utility the qua )* [* )* [* # )* [* of good of good i offunction consumed i consumed that by leads of aby good household;ato ai system household; consumed # the of #demands by the price a products). price household; of expressed goodof good i; of and # iin; theand expenditure # = price # *= [*of , thegood shares.the ,based expenditure )* [ and Let i;*expenditure # denote= share share , thefor for qua ex function The that quadratic of good i consumed by a household; good leads i to AIDS consumed a system model ofby a demandsofhousehold; Banks, the expressed price Blundell, the in price # of good i; and # = and expenditure Lewbel good i (1997) ; shares. and) \ , Let the \ is = # expenditure denote , on the anexpenditure the indirect sharequantity \ for utilit shar good of i : good The quadratic iquadratic consumed good AIDS i: by The The a quadratic ofhousehold; # quadratic AIDS AIDS andmodel the price model of of of good Banks, Banks, iLewbeland ; based \ Blundell, Blundell, = The # ) \ and * [ * quadratic and , the Lewbel Lewbel expenditure (1997) AIDS (1997) is is mod shar b good of good good i: good function i: i consumed The that i:The leadsbyquadratic a to amodel household; AIDS system AIDS Banks, model of demands model the Blundell, of price Banks, of expressed #Banks, of good Lewbel Blundell, Blundell, in i; and (1997) expenditure and and is = )Lewbel * [* shares. , on the #(1997) an Let (1997) indirect expenditure\ is # based isdenote based share on the on an quantit for anindireind el of Banks, lundell, and Lewbel Blundell, (1997) and Lewbel is based utility good (1997) on i anisindirect function :function based that leadson utilityfunction an to function aindirect system # that that of leads The utility leads demands to to quadratic a system a system expressed AIDS of of demands in demands # model expenditure \ expressed of) expressed Banks, *[ function shares. in expenditure in Blundell, that expenditure leads and to shares ashar Lewb sys of good function i consumed that thatby leads a leads to household; ato system a system of the demands of demands price of expressed goodexpressed i ; and in expenditure in expenditure = * shares. , the shares. expenditure Let Let # denote denote #share the )fo em of demands expressed in expenditure expressed good shares. i: Let in expenditure Let # denote shares. denote the the Let # denote quantity quantity of of good of good good the i consumed # quantity i consumedfunction consumed by that by byaa household; a leadshousehold; household; to a system # the the the of price price \ demands price of of ofgood good of good )* [ good )expressed i ;consumed [ i ; and i ; and in = expen by = a*[ ) h * # # # # )good i: of good of good )* [ i *consumed i consumed by by a\household; a household; \a h* h* \ the # #\ l the price price l \ of of good good l h* j e i; and i; \ and # l = # = * , the * * , the expenditureexpenditu \ ousehold; * [* i +∑ a ∑ i gkef()) \ g+ gk for i=1…k good i; # thei;price of good ; and , the expenditure he price of good and , # = the ## = # = + expenditure ∑ a the #$ share + expenditure good# for + # good = e#i : #i + :ge share share + gof* e for for+ good j good e j + econsumed :gk # by a household; for i=1…k for f()) good i=1…k i\the : \ (A1.1) price(A1.1) of # = \ a $bc \ #$ $ h l ∑ #$bc a = $bc #$ + ∑ $ $ \ + # ef())g + i()) f()) f()) h i()) j g ef()) + \ gk * f()) j e gk i()) for i=1…k for i=1…k # (A1.1) (A1.1) # = good # + : i:#$ ∑ i$bc $ + f()) i()) f()) good # \ # $bc a #$ $ # \ i()) h * f()) \ l # = # + $bc #$ \ $ + # hgood e g i:+ \ l j e gk for i=1…k (A1.1) # = # + ∑a $bc #$ $ + e) # ( g + * f()) j e i()) gk f()) for i=1…k (A1.1) where where ln ( ln ) ( ) isln () isis where ln f()) is i()) f()) l where ln where () is# = # + ∑a $bc #$ $ + # ef()) g + i()) \ h* j e gk \ \ \ for h* h i=1…k \ \ l l (A1.1) a where ln () is where is a # = # = # + # ∑+ a∑a \ $bc $bc \ #$ f()) #$ h $ h+ $ + # \# e \ le g l+ g + * j e j e gk gk # = # + ∑$bc for fi #$ where lln ( ) is # =# = + ∑ + ∑ a #$ $+ + c e # e gaa + * g + jj * e e c gk f()) gk f()) i()) i()) f()) for f()) i=1…k for i=1…k \ (A1.1) l (A1 h* + # eef()) \\ g+ l h * j e gk \ ln ln () for () = i=1…k = + # ∑ #a+ $bc ∑ a$bc ln (A1.1) ln #$ ln + c () ∑$ + a # = ∑ ∑ aa + f()) ∑ ∑ f()) i()) ###a+$∑ ln i()) a+ f()) ∑ f()) a ∑ a (A1.2) #$ e # g $ (A1.2) \ h* j e gk (A1 a+ = #$ $ $bc + # + (A1.2) c +$i()) j gk f()) wherefor i=1…k ln ln (() ) is (A1.1) ln? () ?#bca = ##bc + # a c # #bc ∑ l ∑#a ln l ##bc ?$bc ∑ #bc $bc #$ # a ∑ #$ $#$ # # $bc $l #bc # $ (A1.2) = ? + ∑#bc # ? ln # + #bc ∑ # $bc i()) #bc f()) i()) f()) f()) $bcl #$ #bc (A1.2) c ln () where = where +ln ∑ ln a ( l( ) is ) ln is + ∑ a ∑ a where ln ( ) is and and () ( is ) where the is( where the ln Cobb-Douglas ln Cobb-Douglas () ln = and( ?) ( + is ( ) ∑ is ) is a price the price ? ln # aggregator: #bc + l ∑a Cobb-Douglas # aggregator: c # #bc ∑$bc #$ # a l price #bc $bc aggregator: $ #$ # $ (A1.2) and () and and is the ) is Cobb-Douglasis theCobb-Douglas the Cobb-Douglas #bc price price aggregator: price where aggregator: aggregator:c ln a (a)# is ln () = + ∑ a ln + ∑ ∑ (A1.2) and () is the Cobb-Douglas ? #bc price # # aggregator: ln l #bc $bc #$ ∑a∑ () ln ()= = + + #a $ ln ln + c c + ∑a ∑ a ∑a ∑ a #$ #$ # # $ln $() = and () is the Cobb-Douglasln price () ln () aggregator: = () = = ? + () ∏ ? ∑a + = a #bc ∑∏ a a q* ln # # q ln# *+ a # c + ? ∑(A1.3) ca ? () #bc ∑#bc a∑ #bc (A1.3) = a $bc #bc ∑ ∏ a # a #$ #bc # #$ # # q # # * l # $ #bc l $ #bc $bc (A1.3) $bc (A1.2) (A1.2) #bc q*= ∏#bc #bc # q $bc (A1.3) (A1.2) () c a () #bc # l# l c (A1.2) * + c ∑ ? +∑#bc aa #bc∑ $bc a # ln + # ∑ a ∑a $ $bc #$ # and ( ) is the Cobb-Douglas price ∏ = aggregator: #bc # (A1.3) ∑ a ln () = ? + #bc # ln # + l #bc $bc #$ # ∑ a ∑ a #$ l# #bc $ () = ∏a q* (A1.3) and and ( () is ) the is the Cobb-DouglasCobb-Douglas #bc # price price aggregator: aggregator: and ( slightly ) slightly is the Cobb-Dou l All lowercase All lowercase Greek All lowercase and and Greek (letters ) Greek ( All is letters ) the is lowercase are letters the are parameters Cobb-Douglas areparameters Cobb-Douglas Greek parameters () = to ∏letters a be priceto estimated, to # be price q *beare estimated, estimated, aggregator: parameters aggregator: except (A1.3) except except to ?, be which estimated, which , which is , set is isset to except set to to ? , which less less is than than set glas price aggregator: the lowest All lowercase All lowercase Greek letters Greek are letters parameters are parameters to #bc be and estimated, ( to ) be is estimated, except ,except which ? is set which to is slightly set to less slightlythan less * the Cobb-Douglas price aggregator: ? gator: the lowest All value value lowercase of of the ln, in Greek lowest ,lowest this in this case letters value case 1.5. are of () 1.5. parameters ln , in = ∏ in a this to#case q be case estimated, 1.5. (A1.3) ? except slightly less ln than the value of #bcthis 1.5. ?, a which is set(A1.3) to slightly less All lowestthe the lowercase value lowest Greek of ln value , in letters ofarethisln case parameters , in 1.5.this case to be 1.5. estimated, () = () ∏ = aexcept ∏ a q* () q*? , which () =∏ (A1.3) =is a #bc∏ (A1.3) set #bc# q * q* to # slightly less (A1.3) than a All (A1.3) the lowercase lowest value Greek of letters ln , in are equations this parameters case 1.5.to be estimated, #bc #bc # # except ? , which in set is = to a slightly less than = ∏#bc # * () = ∏ ∏#bc q q a q #bc # the lowest (A1.3) * Thevalue setof of ln expenditure , in this sharecase 1.5. is estimated using the command quaids () Stata # * (A The The set of set of set expenditure expenditure The share set shareequations of All equations Allexpenditure lowercaselowercaseis estimated is Greek share estimated Greek usingequations letters using letters the are the command are is parameters parameters estimated command quaids to using quaids tobe in be All the Stata estimated, in estimated, lowercase command Stata(Poi (Poi2012). except Greek except quaids 2012). As , 2012 in As whic ?letters,w S The set the of The expenditure lowest (Poi All 2012). Allof lowercase value lowercase As expenditure of share part lnof Greek the,Greek equations share inmodel, letters this letters case equations is are some estimated 1.5.are parameters is parameters demographic estimated using to the controlsbe to command using estimated, be estimated, the are incorporated, command except quaids except in Stata quaids , which ? including ? , which (Poiis in Stata set 2012). is set (Poi to As slightly to sligh? are to be parameters estimated, to be estimated, except ? , part part of which the of The except the model, is set model, set to some ?of, slightly which some part demographic expenditure isless setthe of demographic the to than share themodel, lowest slightly controls lowest equations some controls less value All valueare than demographic of is are incorporated, lowercase oflnestimated lnincorporated, , in ,Greekthis in controls using this including case letters case including the 1.5. are command 1.5. are incorporated, household household parameters thequaids lowest size; including to size; in thebe value Stata the number estimated, of househo number (Poi ln 2012 ,exin part The setthe of ofpart model, household the expenditure of the the some lowest model, lowest size; demographic value the share some value number of equations of ln demographic ln of controls , men in, estimated this is in ages this are casecontrols more case incorporated, 1.5. than using 1.5. are 15; the incorporated, the command including age, gender, quaids including household religion, in Stata household and size; (Poi the 2012). numbersize; As the nu n this case 1.5. part of the model, some demographic the controls lowest are incorporated, value of ln , in this including case 1.5. household size; the nu part of The set theeducational model, of expendituresome demographic attainment share of the equations controls household isare estimated head, incorporated,and whether using the including the commandhousehold household quaids is in a in size; rural Stata the(Poi number 2012). A The The setset of of expenditure expenditure share share equations equations is estimated is estimated The using set using ofthe expenditure the command command sha qu part of or urban the The model, The setset area. of Thesome expenditure of results demographic expenditure are robust share share controls equations to equations the inclusion are incorporated, is estimated isof estimated additional using including using demographic thethe command household command and quaids size; quaids in the Stata in Stata numbe 21 (Poi (P2 21 re equations estimated using is estimated the command usingquaids the command in Stata quaids (Poi 2012). in Stata partAs part of (Poi of the the 2012). model, model, The As some set of demographic some expenditure demographic controls share controls equations are part are incorporated, incorporated, of is the estimated model, including some includin using 21 size; deh the geographic part part of controls. the of themodel, model, After some some the demographic parameters demographic are controls controls estimated, are the areincorporated, quaids incorporated, postestimation including including household household size; the mographic rols are incorporated, controls are incorporated, including household including size;household the number size; the number part of the model, some demographic controls are incorporat 21 commands allow the estimation of own-price and cross-price elasticities for all goods and different population groups. Uncompensated price elasticities are estimated for 2 individual observations in the data and are summarized by decile using the household 21 21 survey weights. Table A1 presents the estimated coefficients from which the elasticities presented in table 3 have been calculated. 37 The distributional impacts of cigarette taxation in Bangladesh Table A1. Quadratic AIDS model Number of observations 29,971 Number of demographics 7 Alpha_0 1.5 Log-likelihood 16517 Coeffi- Standard z P>|z| [95% Conf. Interval] cient Error Alpha Cigarettes −0.19 0.03 −5.57 0.00 −0.25 −0.12 Bidis −0.03 0.03 −1.02 0.31 −0.08 0.03 Betel leaf 0.09 0.02 5.41 0.00 0.06 0.13 Betel nut 0.18 0.02 12.13 0.00 0.15 0.21 Rolled betel leaf 0.94 0.04 22.92 0.00 0.86 1.02 Beta Cigarettes 0.16 0.01 20.18 0.00 0.15 0.18 Bidis 0.16 0.01 20.78 0.00 0.14 0.17 Betel leaf −0.08 0.01 −14.30 0.00 −0.09 −0.07 Betel nut −0.06 0.00 −13.19 0.00 −0.07 −0.05 Rolled betel leaf −0.18 0.01 −21.97 0.00 −0.20 −0.16 Gamma Cigarettes-cigarettes −0.15 0.01 −16.62 0.00 −0.17 −0.13 Bidis - cigarettes 0.00 0.01 0.60 0.55 −0.01 0.02 Betel leaf - cigarettes 0.03 0.00 7.06 0.00 0.02 0.03 Betel nut - cigarettes 0.00 0.00 1.21 0.23 0.00 0.01 Rolled betel leaf - cigarettes 0.12 0.01 15.13 0.00 0.10 0.13 Bidis-bidis −0.10 0.01 −14.63 0.00 −0.11 −0.09 Betel leaf - bidis 0.01 0.00 1.94 0.05 0.00 0.01 Betel nut - bidis 0.03 0.00 9.93 0.00 0.02 0.03 38 // Annex Gamma (continued) Rolled betel leaf - bidis 0.06 0.01 9.74 0.00 0.05 0.08 Betel leaf- betel leaf −0.04 0.00 −16.76 0.00 −0.05 −0.04 Betel nut- betel leaf −0.03 0.00 −15.46 0.00 −0.04 −0.03 Rolled betel leaf- betel leaf 0.05 0.00 12.54 0.00 0.04 0.05 Betel nut- betel nut 0.00 0.00 −1.48 0.14 −0.01 0.00 Betel nut- Rolled betel leaf 0.00 0.00 1.14 0.25 0.00 0.01 Rolled betel leaf- −0.23 0.01 −20.30 0.00 −0.25 −0.21 Rolled betel leaf Lambda Cigarettes 0.01 0.00 10.62 0.00 0.01 0.01 Bidis −0.02 0.00 −17.03 0.00 −0.02 −0.01 Betel leaf −0.01 0.00 −17.07 0.00 −0.01 −0.01 Betel nut −0.01 0.00 −15.02 0.00 −0.01 −0.01 Rolled betel leaf 0.02 0.00 27.54 0.00 0.02 0.02 Eta Cigarettes: Household size −0.01 0.00 −8.99 0.00 −0.01 −0.01 Bidis: Household size 0.00 0.00 4.38 0.00 0.00 0.00 Betel leaf: Household size 0.01 0.00 11.46 0.00 0.00 0.01 Betel nut: Household size 0.00 0.00 11.95 0.00 0.00 0.01 Rolled betel leaf: Household size 0.00 0.00 −8.95 0.00 −0.01 0.00 Cigarettes: Number of males 15+ 0.01 0.00 4.23 0.00 0.00 0.01 Bidis: Number of males 15+ 0.00 0.00 −2.54 0.01 −0.01 0.00 Betel leaf: Number of males 15+ 0.00 0.00 −4.43 0.00 −0.01 0.00 Betel nut: Number of males 15+ 0.00 0.00 −4.17 0.00 0.00 0.00 Rolled betel leaf: 0.00 0.00 2.59 0.01 0.00 0.00 Number of males 15+ Cigarettes: 0.00 0.00 12.34 0.00 0.00 0.01 Education of household head Bidis: 0.00 0.00 −20.28 0.00 −0.01 0.00 Education of household head 39 The distributional impacts of cigarette taxation in Bangladesh Eta (continued) Betel leaf: 0.00 0.00 −2.44 0.02 0.00 0.00 Education of household head Betel nut: 0.00 0.00 −4.17 0.00 0.00 0.00 Education of household head Rolled betel leaf: 0.00 0.00 7.09 0.00 0.00 0.00 Education of household head Cigarettes: −0.05 0.00 −11.49 0.00 −0.05 −0.04 Gender of household head Bidis: Gender of household head −0.02 0.00 −6.29 0.00 −0.03 −0.02 Betel leaf: 0.06 0.00 20.66 0.00 0.05 0.07 Gender of household head Betel nut: 0.05 0.00 21.45 0.00 0.05 0.06 Gender of household head Rolled betel leaf: −0.04 0.00 −13.69 0.00 −0.05 −0.04 Gender of household head Cigarettes: 0.00 0.00 −16.71 0.00 0.00 0.00 Age of household head Bidis: Age of household head 0.00 0.00 9.10 0.00 0.00 0.00 Betel leaf: 0.00 0.00 22.21 0.00 0.00 0.00 Age of household head Betel nut: Age of household head 0.00 0.00 20.07 0.00 0.00 0.00 Rolled betel leaf: 0.00 0.00 −19.88 0.00 0.00 0.00 Age of household head Cigarettes: 0.00 0.00 0.59 0.55 0.00 0.01 Religion of household head Bidis: Religion of household head 0.00 0.00 1.92 0.06 0.00 0.01 Betel leaf: −0.01 0.00 −5.66 0.00 −0.01 0.00 Religion of household head Betel nut: 0.00 0.00 −3.34 0.00 −0.01 0.00 Religion of household head Rolled betel leaf: 0.01 0.00 4.62 0.00 0.00 0.01 Religion of household head Cigarettes: 0.06 0.00 14.74 0.00 0.05 0.06 Urban/rural household Bidis: Urban/rural household −0.04 0.00 −17.70 0.00 −0.04 −0.03 Betel leaf: Urban/rural household −0.02 0.00 −12.74 0.00 −0.02 −0.02 40 // Annex Eta (continued) Betel nut: Urban/rural household −0.02 0.00 −14.38 0.00 −0.02 −0.02 Rolled betel leaf: 0.02 0.00 14.78 0.00 0.02 0.03 Urban/rural household Rho Household size 0.03 0.03 1.09 0.28 −0.02 0.08 Number of males 15+ 0.12 0.06 1.84 0.07 −0.01 0.24 Education of household head 0.11 0.02 5.52 0.00 0.07 0.15 Gender of household head 0.34 0.13 2.61 0.01 0.08 0.59 Age of household head 0.01 0.00 2.68 0.01 0.00 0.01 Religion of household head −0.21 0.06 −3.73 0.00 −0.32 −0.10 Urban/rural household 0.21 0.11 1.94 0.05 0.00 0.43 Source: Calculations using data from HIES 2016/17. 41 The distributional impacts of cigarette taxation in Bangladesh