THE ECONOMICS OF TOBACCO TAXATION AND EMPLOYMENT IN INDONESIA Health, Population, and Nutrition Global Practice Cover photos (clockwise from left): Aerial view of forest and fields, Indonesia. Photo by Curt Carnemark/ World Bank. A child from Jakarta. Project: JEDI. Photo by Farhana Asnap/World Bank. Home and boats on the water. Photo by Curt Carnemark/World Bank. Jakarta city landmark. Photo by Jerry Kurniawan/World Bank. THE ECONOMICS OF TOBACCO TAXATION AND EMPLOYMENT IN INDONESIA Health, Population, and Nutrition Global Practice Revised Draft 25 September 2017 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 LIST OF ACRONYMS BPJS Badan Penyelenggara Jaminan Sosial FGD Focus Group Discussion FT Full-time FTE Full-time Equivalent GDP Gross Domestic Product KIS Kartu Indonesia Sehat KKS Kartu Keluarga Sejahtera KPS Kartu Perlindungan Sosial NCD Noncommunicable Diseases PBI Penerima Bantuan Iuran PPP Purchasing Power Parity Rp Indonesian Rupiah SKM Machine-made kretek SKT Hand-made kretek SPM White cigarettes EXECUTIVE SUMMARY Indonesia has one of the most complex cigarette excise tax structures in the world. The current cigarette excise tax is divided into 12 tiers, which are based on manufacturers’ type of cigarettes, the number of cigarette productions, and per-unit retail price. The tiers in the cigarette excise tax structure aim to accommodate small-scale cigarette firms, espe- cially hand-rolled kreteks (SKT) firms. The rationale for such a structure is that smaller SKT firms account for more than half of the total factories in the tobacco industry. Moreover, these firms are responsible for employing a significant share of the workers in tobacco manufacturing. Such a system favors downward substitution to lower priced products and has a limited impact on smoking reduction. This report aims to contribute to the policy debate over the tobacco excise tax reform, and specifically, any effects it might have on employment in Indonesia. It uses data from the Central Bureau of Statistics to observe the trends in employment and output in the tobacco sector and estimate the potential impact on tobacco employment (gross effect) from raising cigarette taxes in Indonesia. The report shows that tobacco manufacturing represents only a small share of the economy-wide employment (0.60 percent). Additionally, the contribution of tobacco manufacturing to employment in the manufacturing sector is quite small (5.3 percent) in comparison to the contribution of the food (27.43 percent), garment (11.43 percent), and textile (7.90 percent) sectors. Although small, tobacco manufacturing jobs are heavily concentrated, with about 94 percent of tobacco manufacturing workers in Central Java, East Java, and West Nusa Tenggara. In these provinces, several districts are quite depen- dent on tobacco sector employment, for example, Kudus (30 percent), Temanggung (27.6 percent), and Kediri (26 percent). We found that most tobacco manufacturing workers are females and unskilled. The share of female workers in tobacco manufacturing is 66 percent, which is the highest in the manufacturing industry. Other sectors in which most workers are female are the garment sector (63 percent) and textiles sector (56 percent). About 69 percent of total workers in the tobacco sector completed at most, junior high school. Among tobacco households, the share of tobacco households with female tobacco workers is 78 percent. Average years of schooling completed by tobacco manufacturing workers is among the lowest at 8.22 years, which is comparable to the average years of schooling completed by workers in the food processing sector (8.50 years), manufacturing of wood products sec- tor (7.31), nonmetallic metal products sector (7.42), furniture sector (8.77), and recycling sector (5.20). Tobacco households are partially dependent on income from tobacco employment, which represents 60 percent of household income, on average. Additionally, among these households, only 9 percent have a female as the primary income earner. These find- 4 // Executive Summary ings from secondary data (the Sakernas data) are consistent with the ones from the World Bank/ACS kretek workers survey in Indonesia—the proportion of wage income from kretek is 54 percent and the share of female tobacco workers as primary earner is about 10 per- cent (World Bank, 2017a). Productivity of tobacco sector workers, measured by the output per worker, is relatively low in comparison to productivity of workers in the comparable sector. A typical worker in the medium and large tobacco firms can produce IDR 104 million (US$7,761) worth of products annually. For comparisons, a typical worker in the food and drink, and textile industries can produce IDR 265 million (US$19,776) and IDR 300 million (US$22,388) worth of products respectively. The estimated output elasticity of labor demand for the cigarette manufacturing sector is 0.160. This means that a 1 percent decrease in output results in a 0.16 percent decrease in employment in the cigarette manufacturing sector. On the other hand, the estimate for the tobacco processing sector implies that a 1 percent decrease in output corresponds to a 0.092 percent decrease in employment in the tobacco processing sector. Our estimates predict that raising cigarette taxes by an average of 47% and sim- plifying the cigarette tax structure to 6 tiers will reduce cigarette demand by 10.4 percent, increase government revenue by 8.4 percent, and reduce gross employ- ment in the tobacco manufacturing sector by 9.13 percent. Our estimates suggest that the consumption of cigarette is predicted to decrease by 36 billion sticks. Higher cig- arette taxes are expected to increase government revenue by IDR 12,875 trillion. In terms of employment, we estimate a reduction of 32,132 tobacco manufacturing jobs, most of them in the SKT industry (24,710 fewer jobs). This implies a 4.79% reduction in tobacco manufacturing jobs and an 8.60% reduction in SKT industry jobs. The evidence presented in this report clearly shows that the gross employment impact of reforming tobacco excise taxes and structure in Indonesia is not as big as previously thought. Analyses presented here are complemented by other evidence pre- sented in the World Bank/ACS Indonesia Tobacco Studies, which highlighted the economic and social costs of tobacco workers’ and farmers’ livelihoods. Given the additional reve- nues the government will obtain with the reform (IDR 12,875 trillion), there is an oppor- tunity to implement measures to reduce the impact on the tobacco workers’ livelihoods (such as cash transfers or expanded access to social safety nets) or to find alternative occu- pations for the workers affected (retraining programs, educational grants, etc.). It is important to note that these estimates represent the gross employment effect of lower cigarette consumption. When prices of cigarettes increase, consumers may shift their consumption to other goods and services which will create jobs in these sec- tors. Evidence has shown that the job losses in the tobacco sector (gross effect) are usu- ally compensated with job creation in the other sectors (net effect). For Indonesia, Ahsan and Wiyono (2007) estimated positive effects varying from 84,340 to 281, 135 jobs with tax increases of 25 percent and 100 percent, respectively (Ahsan and Wiyono, 2007). 5 • For kretek hand-rollers o First, the most vulnerable groups in the affected population who would need immediate income support in the event of job loss include the workers who are less educated, older, heads of their households, and who contribute a significant pro- portion of total household income from kretek rolling. The government can provide income support to these workers with less than 2% of the revenue gained from a tax increase; and o Second, the additional revenue or part thereof can be allocated for re-skilling and redeployment of laid-off kretek workers to smooth their transition to an alternative source of employment in the short to medium term. • For (tobacco and clove) farmers: o The government should help to improve supply chains and value chains for other goods in tobacco-growing areas. Many former tobacco farmers are making a better living growing other common, locally grown crops (e.g., corn, sweet potato, and green vegetables), an outcome that could be further enhanced with even small investments by governments in improved supply chains for these products. Results from the World Bank/ACS survey suggest that current tobacco farmers are already growing many of these crops, so it is an issue of shifting their factors of production to maximize economic opportunity. o The government should help to facilitate access to credit for tobacco farmers. Greater access to capital through improved credit schemes could help to improve the possibilities for tobacco farmers to cultivate other crops and/or develop other nonagricultural economic enterprises. That could be in the form of grants or low-interest loans to farmers willing to move away from tobacco cultivation; and o Specifically, on clove farmers, it is important to emphasize that clove farming is not particularly profitable for most clove-producing households in at least a couple of major clove-producing districts. The government needs to research which alternatives could be viable and target the least profitable areas for switching and help these farmers make successful transitions to growing other crops and/or economic activities. 6 // Executive Summary ACKNOWLEDGMENTS The preparation of this report was carried out under the World Bank Global Tobacco Control Program coordinated by Patricio V. Marquez, with the support of the Bill & Melinda Gates Foundation and the Bloomberg Foundation. The report is part of the Indonesia Tobacco Employment Studies and was prepared by a team comprised of: Gumilang Aryo Sahadewo (Research Faculty Universitas Gadjah Mada and Consultant World Bank); Roberto Iglesias (Director, Center of Studies on Integration and Development, CINDES, Brasil); Edson C. Araujo (Senior Economist and Task Team Leader, World Bank); Nigar Nargis (Director, American Cancer Society); Pandu Harimurti (Senior Health Specialist and Task Team Leader, World Bank); Jeffrey Drope (Vice-President, American Cancer Society); Qing Li (Senior Data Analyst, American Cancer Society); Josefine Durazo (Survey Specialist, World Bank); Firman Witoelar (Director of Research, SurveyMeter); and Bondan Sikoki (Chairperson, SurveyMeter). The report greatly benefited from comments, inputs, and advice provided by Violeta Vulovic (Economist, World Bank), Frank Chaloupka (Professor, Institute for Health Research and Policy, University of Illinois, Chicago), Jeremias N. Paul (Coordinator of Tobacco Control, WHO) and Michael Weber (Senior Economist, World Bank). 7 TABLE OF CONTENTS List of Acronyms 3 Executive Summary 4 Acknowledgments 7 List of Tables 9 List of Figures 9 1. Introduction 11 2. The Economics of Tobacco Taxation 15 2.1 — Tobacco Consumption, Health Outcomes and Economic Costs 15 2.2 — Tobacco Taxes and Tobacco Consumption 17 2.3 — Tobacco Taxation and Tax Revenues 18 2.4 — Employment Impacts of Tobacco Taxation 19 3. Cigarette Tax Policies in Indonesia 25 3.1 — Effects of Cigarette Consumption, Revenues, and Employment 26 4. Employment in the Indonesian Tobacco Sector 33 4.1 — Data Sources 33 4.2 — 23 Employment in the Indonesia Tobacco Farming Sector 34 4.3 — Employment in the Tobacco Manufacturing Sector 36 5. Simulations on the Effects of Raising Cigarette Taxes on Employment 49 6. Conclusion 55 References 57 Annex I. Estimation of the Wage Equation 61 Annex II. Sample Selection Procedure for the Manufacturing Industry Data 62 Annex III. Estimation of Output Elasticity of Labor Demand 64 Annex IV. Estimation of Price Elasticity of Demand 66 Annex V. Simulation of the Effects of Raising Cigarette Prices on Employment 68 8 // Table of Contents List of Figures Figure 1. Trends of the cigarette excise tax in Indonesia in real terms, 2010–2017 27 Figure 2. Revenue from the tobacco excise, 2005–2015 29 Figure 3. Trends of cigarette production in Indonesia, 2010–2016 30 Figure 4. Number of tobacco farmers and share of tobacco farmers to total workers, 1990–2011 35 Figure 5: Number of medium and large tobacco manufacturers, 2000–2014 39 Figure 6. Share of tobacco employment to total manufacturing employment, 2000–2014 40 Figure 7. The concentration of tobacco manufacturing workers, 2014 41 Figure 8. Number of workers per firm by industry 42 Figure 9. Productivity of workers, 2000–2014 46 Figure 10. Relationship between output and labor in medium and large industry, 2000–2014 46 List of Tables Table 1. Selected studies on net employment impact of tobacco control policies 21 Table 2. Type of external tobacco trade and likely net employment impacts 23 Table 3. Cigarette excise tax in Indonesia, 2017 26 Table 4. Distribution of tobacco farmers across regions, 2014 36 Table 5. Distribution of clove farmers across regions, 2014 37 Table 6. Employment structure of the Indonesian tobacco industry, 2014 38 Table 7. Concentration of workers in selected districts, 2014 41 Table 8. Characteristics of tobacco sector workers, 2011, 2013, & 2015 43 Table 9. Estimation results of the wage equation, 2001–2015 45 Table 10. Estimation of own-wage and output elasticity of labor demand, 2000–2014 47 Table 11. Estimated number of workers in each tax tier 50 Table 12. Proposed cigarette tax increase and new cigarette excise tariffs 51 Table 13. The effects of raising cigarette prices on employment in the manufacturing sector 51 Table 14. Simulation on the effects of raising cigarette taxes on employment 53 Table 15. Summary statistics, National Labor Force Survey Sample 61 Table 16. Selected sample, Annual Survey of Manufacturing Industry 62 Table 17. Summary statistics, Annual Survey of Manufacturing Industry 65 Table 18. Estimation of price elasticity of cigarette demand, 2015 67 9 SMOKING HAS The Economics of Tobacco Taxation and Employment in Indonesia BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 10 // Table of Contents 1 INTRODUCTION Smoking has been a major contributor to the disease burden in Indonesia. Smoking prevalence among working-age individuals has exceeded 30 percent since 2001. In 2013, smoking prevalence among males was 66%, while among females was 6.7%. Smoking prevalence among children ages 10–14 in 2013 was 3.7%, twelve times higher than in 1995 (IAKMI, 2014; Ahsan, 2015). Diseases attributable to smoking include hypertension, acute respiratory infection, coronary heart disease, cardiovascular diseases, selected can- cers, and perinatal disorders (IAKMI, 2014; Kosen et al., 2012; Kristina et al., 2015). In 2013, healthy years lost at the population level due to smoking-induced diseases was estimated to be 6.2 million disability-adjusted life years (DALY) (IAKMI, 2014). The government of Indonesia continues its efforts to reform the cigarette excise tax system. The main objectives of the excise tax reform are to reduce smoking preva- lence and to increase tax revenues. Studies suggest that a 10 percent increase in ciga- rette excise tax would lead to a reduction in cigarette consumption by 0.9 to 3 percent and an increase in government revenue by 6.7 to 9 percent (Hidayat and Thabrany, 2010; Setyonaluri et al., 2008). Currently Indonesia has one of the most complex cigarette tax structures in the world, which favors downward substitution to lower priced cigarettes (World Bank, 2015). Cigarette prices across all tiers increased at a modest rate between 2010 and 2017 as the government continued to increase cigarette taxes; nevertheless cigarette prices were more affordable in recent years than they were in 2000 owing to a faster income growth (NCI-WHO, 2017). The main argument against adopting such a complex cigarette tax structure was to protect employment by differentiating firms with different production scales with those that employ more workers, such as hand-rolled kreteks (SKT) workers. Despite the concerns, tobacco manufacturing represents only a small share of the economy-wide employment (0.60 percent). Additionally, the contribution of the tobacco manufacturing to employment in the manufacturing sector is quite small (5.3 percent) in comparison to the contribution of the food (27.43 percent), garment (11.43 percent), and textile (7.90 percent) sectors. Although small, tobacco manufacturing jobs are heavily concentrated in Central Java, East Java, and West Nusa Tenggara, where about 94% of tobacco manufacturing workers are employed. In these provinces, several districts are quite dependent on sector employment, as for example Kudus (30 percent), Temang- gung (27.6 percent), and Kediri (26 percent). A tax change shock that affected the tobacco sector would affect these districts most. 11 The Economics of Tobacco Taxation and Employment in Indonesia The objective of this report is to analyze the recent employment trends in the Indonesian tobacco industry and estimate the potential effects of raising cigarette taxes on employment in the tobacco manufacturing sector. The report provides new evidence to contribute to the ongoing debate about the effects of raising cigarette taxes on tobacco sector employment. It complements the current analytical work conducted by the World Bank, in partnership with the American Cancer Society, to explore the employment conditions and livelihoods of tobacco and clove farmers and kretek rollers in Indonesia. The report is part of the World Bank technical assistance to the government of Indonesia in the areas of revenue policy reform and health systems reform. This report is structured as follows: Section two provides a review of the global evi- dence on the impacts of raising cigarette taxes on population health outcomes and on the economy. It briefly reviews the economic issues around tobacco taxation, including a summary of the global evidence on the effects of taxation on smoking reduction, fiscal revenues, tobacco production and employment in the sector. Section three presents an overview of the cigarette tax reforms in Indonesia in recent years and discusses the government of Indonesia’s plans for reforming the cigarette tax structure. Section four discusses the employment trends in the tobacco industry in Indonesia, analyzes workers’ characteristics and compares workers to similar sectors and socio-demographic profiles. The section also discusses the potential impacts of raising cigarette taxes on employment by presenting results of simulations. The final section, section five, discusses the results in light of the current debate over cigarette tax reform in Indonesia and provides policy recommendations on the employment aspects of the reform. 12 // Introduction 13 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO PREVALENCE AMONG WORKIN AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 THE ECONOMICS OF TOBACCO TAXATION Raising cigarette prices through taxes is a cost-effective way to reduce cigarette consumption as recommended by the World Health Organization Framework Convention on Tobacco Control (WHO FCTC, 2003). The objective of raising cigarette taxes is to decrease consumption of cigarettes through higher prices.1 Additionally, rais- ing cigarette taxes can increase government’s tax revenues, which can be allocated to finance complementary tobacco control policies and social investments such as health and education. Experiences in both high and low-and middle-income countries (LIMCs) show how tobacco tax revenues can be used to expand health care coverage. In France, for example, tobacco tax revenues are used to supplement the funding for public health services. In Philippines, higher tobacco and alcohol tax revenues have been used to expand health coverage for the poor. On the other hand, the tobacco industry argues that lower consumption of tobacco products due to higher tobacco tax can have neg- ative effects on production and, consequently, on employment. This section presents a summary of the global evidence on the impacts of tobacco consumption on population’s health and the impacts of tobacco taxation on fiscal revenues and employment. 2.1 Tobacco Consumption, Health Outcomes, and Economic Costs Smoking is a major cause of morbidity and premature mortality. Tobacco consump- tion has been directly linked to diseases of the circulatory system (e.g., ischemic heart and cerebrovascular diseases); cancers of the trachea, bronchus and lung, esophagus, oropharynx, larynx, stomach, liver, pancreas, kidney and ureter, cervix, bladder, colon/ rectum, as well as acute myeloid leukemia; chronic respiratory diseases (e.g., asthma, chronic obstructive pulmonary disease); and metabolic diseases such as diabetes mel- litus (U.S. Department of Health and Human Services, 2014). Besides the harm of direct consumption, it has been proven that secondhand smoking (sometimes referred to as passive smoking, environmental tobacco smoke, or tobacco smoke pollution) also has damaging health consequences. Secondhand smoking is quite dangerous because there are at least 50 carcinogenic chemicals inhaled by those who are around smokers, and the scientific evidence shows that there is no safe level of exposure to secondhand smoking 1 We use the term cigarette taxes and tobacco taxes interchangeably because cigarettes are a major form of tobacco product in Indonesia. 15 The Economics of Tobacco Taxation and Employment in Indonesia (US NCI-WHO, 2016). The World Health Organization (WHO) estimates that approximately 7.2 million deaths per year worldwide are attributable to smoking, more than 6 million of those deaths are the result of direct tobacco use, while around 900 000 are the result of non-smokers being exposed to second-hand smoke (WHO, 2017a). Half of these deaths occurred in LMICs. By 2030, the annual death toll could reach 10 million if no tobacco control measures are taken (WHO, 2008). Smoking is one of the major risk factors for noncommunicable diseases (NCD) deaths. NCDs kill 40 million people each year, which is equivalent to 70% of all deaths globally. Each year, 15 million people die from a NCD between the ages of 30 and 69 years; over 80% of these premature deaths occur in LMICs (WHO, 2017b). Worldwide, approximately 14% of adult deaths from NCDs are attributed to tobacco use, includ- ing 10% of all adult deaths from cardiovascular diseases (14% among men, 6% among women), and 22% of all adult deaths from cancer (32% among men, 11% among women). The clear majority (71%) of adult lung cancer deaths (78% among men, 53% among women) were attributable to tobacco. In addition, 36% of all adult deaths from diseases of the respiratory system were attributable to tobacco (42% among men, 29% among women) (WHO, 2012). Tobacco smoking is also an important risk factor for chronic obstructive pulmonary disease (COPD). In 2004, about 49% of the COPD deaths among adult men and 34% of COPD deaths among adult women were attributable to tobacco (WHO, 2012). NCDs have become a major public health concern in Indonesia. The WHO estimates that the proportional mortality due to NCDs has increased from 50.7% in 2004 to 71% in 2014 (WHO, 2014). Tobacco smoke is the fourth risk factor that contributes to most death and disability combined from NCDs, after dietary risks, high blood pressure, and high fast- ing plasma glucose (IHME, 2017). In 2012, NCDs accounted for more disability-adjusted life years (DALYs) than communicable diseases—approximately 476 million and 240 mil- lion DALYs, respectively (WHO, 2014). Smoking is also responsible for deaths from communicable diseases. Approximately 5% of global deaths from communicable diseases are attributed to tobacco, including 7% of all deaths due to tuberculosis (TB) and 12% of deaths due to lower respiratory infections (WHO, 2012). A systematic review of the literature found a significant positive relationship between exposure (passive or active) to tobacco smoke and TB infection and disease, independent of various potential health issues including alcohol use and socio- economic status (Salma et al., 2007). Recurrent TB and mortality resulting from TB were also associated with active smoking. This relationship is particularly important in Indo- nesia, given that the country has one of the highest TB infection rates in the world and where tuberculosis is one of the top causes of death (USAID, 2009). 16 // The Economics of Tobacco Taxation The risk of death due to smoking declines with the length of time from an individ- ual quits smoking. In other words, the sooner one quits smoking, the longer she/he will live. Studies have been conducted for the United States (Burns et al., 1997), India (Jha et al., 2008), and Germany (Neubauer et al., 2006) that have provided a quantitative relation- ship between the length of smoking, smoking cessation, and benefits of quitting smok- ing concerning mortality, showing that the reductions in relative risk of heart disease and stroke are more immediate than the effects on respiratory disease and cancer. Due to the lengthy time lags for the development of cancers and chronic respiratory diseases asso- ciated with tobacco smoking, deaths from these illnesses in LMICs may continue to rise, even if smoking prevalence remains the same or decreases (US NCI and WHO, 2016). Smoking imposes a substantial economic burden on countries due to increased health care costs and worker’s productivity losses. Goodchild and colleagues (2016) measured the economic costs of smoking in 152 countries, representing 97% of the worldwide smoking population. They considered direct costs related to health care treatment and indirect costs from productivity losses due to tobacco-related premature mortality and morbidity. The estimated total economic cost of smoking was equivalent in magnitude to 1.8% of the world’s annual gross domestic product (GDP) in 2012 and 40% in LMICs. Of those, 76% are indirect costs related to productivity losses due to mor- bidity (35%) and premature mortality (65%). The direct health care costs were estimated at purchasing power parity (PPP) $467 billion (PPP international dollars), which is equiv- alent to 5.7% of global total health expenditures. The findings from this study highlight the urgent need to implement comprehensive tobacco control measures to reduce the economic costs of smoking. 2.2 Tobacco Taxes and Tobacco Consumption Increasing tobacco and cigarette taxes reduces consumption and, consequently, can reduce smoking-attributable mortality and morbidity. The impact of tobacco taxation on the reduction of mortality depends on (i) the magnitude of the price increase resulting from a tobacco tax increase, (ii) the reaction of consumers to price changes, that is the price elasticity of demand,2 which is related to smoking behavior (initiating, reducing intensity, or quitting), and (iii) the relationship between mortality and quit- ting smoking. The price of tobacco products in relation to income, i.e. affordability, also matters on initiation decisions, intensity, or quitting.3 Empirical studies in high- and low- and middle-income countries have found a negative relationship between cigarette prices and smoking. Once prices increase, smokers adjust their decision of consuming 2 Technically, the price elasticity of demand is the percentage change in the consumption of a product in response to a 1% change in the price of the product, with all else remaining constant. 3 To be effective in reducing tobacco demand, tax and price increases need to be significant to counteract the effect of income growth on tobacco demand and reduce affordability. 17 The Economics of Tobacco Taxation and Employment in Indonesia through quitting or smoking reduction. Additionally, higher prices, due to a higher tax, act as a deterrent for new smokers, particularly among the youth or the poor. In high-income countries (HICs), price elasticity estimates are clustered around –0.4 percent (IARC, 2011). In LMICs also show a negative price-elasticities of tobacco demand, ranging from –0.1 to –1.0, with estimates clustered around –0.5 percent (John et al., 2010; Jha and Chaloupka, 2012; Szabo et al., 2016). In other words, in HICs, a 10 percent increase in the price of tobacco is expected to decrease tobacco consumption by 4 percent. In LMICs, a 10 percent increase in price would be expected to decrease tobacco consumption by 5 percent (IARC, 2011). A recent global simulation study shows that an 80 percent increase in excise per pack may lead to 42% increase in price, reduce global annual cigarette consumption by 18 percent and global smoking prevalence by 9 percent (Goodchild, Perucic, and Nargis, 2016). Studies also show that smoking reduction after a tax increase has positive impacts on individual and public health. Several studies show the health benefits of quitting smoking due to tax increases (John et al., 2010; Blakely et al., 2015; Goodchild, Perucic and Nargis, 2016; Szabo et al., 2016). A recent systemic review of studies, published in English between 2000 and 2012, also concludes that tobacco taxation is a highly cost-effective policy, because the costs of intervention are minimal and significantly save health care costs after the tax increase implementation, although many health benefits of quitting take time to materialize (Contreary et al., 2015). 2.3 Tobacco Taxation and Tax Revenues Tobacco and cigarette excise taxation can also be efficient sources of fiscal revenue. Given that tobacco demand is relatively inelastic, due to consumer addiction and the lack of close substitutes, tobacco taxes can generate considerable amounts of tax revenues, particularly if sales are large. Tobacco taxes may also create fewer distortions in the mar- kets than would result from taxes on goods and services with more elastic demand. Also, given the small number of producers, tobacco taxes are relatively easy to collect at low administration and enforcement cost, as compared to general consumption and income taxes. Experiences in numerous countries indicate that an increase in tobacco taxes will increase nominal (as well as real) tax revenues in the short to medium terms. The magnitude of tobacco and cigarette tax revenue that a government can gen- erate largely depends on the tax system and the demand characteristics. Given a certain income level, per capita or total, a country’s tobacco excise revenue depends on: (i) the level of taxation per unit of tobacco product (either as percent of the price or abso- lute amount of tax per pack); (ii) the number of different tax tiers; (iii) the price elasticity of tobacco demand; and (iv) the volume of tobacco sales. Normally, low levels of tobacco excise revenues are associated with low levels of taxation per unit of product (Chaloupka et al., 2012). Different tax tiers, such as in Indonesia, allow producers to reduce their tax 18 // The Economics of Tobacco Taxation burden, expanding production through less taxed products. Also, smokers could avoid higher rates by switching to lower taxed cigarettes. The result may be lower revenues than potentially anticipated. The more inelastic the demand, the more government rev- enues can be generated with a certain tax rate increase. With an inelastic demand, the proportional reduction in cigarettes purchased by the consumer after the tax increase is smaller than the proportional increase in tax revenue. As the price elasticity of demand increases in absolute value, the possibility to raise revenues for a given tax rate change decreases. Finally, the volume of tobacco sales will determine the possibility of revenue expansion, given certain demand elasticity and percent changes on the taxation level. Tobacco taxes are likely to remain high after a significant tax increase, even with a considerable decline in tobacco use. Chaloupka et al. (2012) argue that over time, inflation will erode the value of tobacco tax revenues, unless those taxes are increased often enough to keep pace with inflation. Similarly, as tobacco use declines in response to other tobacco control efforts, revenues from tobacco taxes will also decline, unless taxes are increased periodically. Nevertheless, it is possible that tax revenues may remain higher many years after a significant tax increase than they were before, even in the wake of a considerable decline in tobacco use. For example, in the case of California, tax rates increased by 770% between 1989 and 1999, while cigarette sales declined by more than 60% between 1989 and 2010, and tobacco excise revenues increased from US$250 mil- lion before 1989 up to US$845 million in 2010 (Chaloupka et al. 2012). In Brazil, the excise tax amount per cigarette pack was increased in real terms by 81.4% between 2011 and 2015, while total cigarettes sales decreased by 35% in the same period. However, real excise revenues in 2015 were still 17% higher than in 2011 (Iglesias, 2016). 2.4 Employment Impact of Tobacco Taxation Despite its effectiveness in reducing tobacco consumption and increasing tax rev- enues, there is often a debate over the effects of tobacco tax employment in the tobacco industry. The tobacco industry generates jobs in diverse parts of the economy, including farming, manufacturing, and wholesale sectors. However, it is important to differentiate the employment that is from core-tobacco sectors (directly dependent on tobacco production, such as farming and manufacturing) compared to tobacco-related employment (jobs that are just partially dependent on tobacco, such as retail). Historic analyses of the tobacco industry show that the industry has significantly reduced employ- ment because, over time, the industry has become more capital intensive and farming has become more efficient, so that job losses have occurred even in the absence of tobacco control measures (NCI-WHO, 2017). The tobacco industry has sponsored studies to document the employment contribution of the sector. Usually these studies argue that tobacco control measures, such as higher taxation, would result in job losses in the tobacco industry and, consequently, increase unemployment (Zhang, 2002). 19 The Economics of Tobacco Taxation and Employment in Indonesia In contrast, most academic studies have shown that tobacco control policies, such as taxation, have an overall neutral or positive impact on employment. Zhang (2002) argues that the (industry-sponsored) studies use unrealistic assumptions about impacts of sales drop, overestimate the number of jobs associated with the tobacco industry and, consequently, overestimate the possible negative impact of tobacco con- trol measures on overall employment. However, the main problem is that those studies do not consider: (i) the expansionary employment effect of consumption substitution of smokers who redirect their expenditure toward other products after the tax increase— tobacco expenditures do not disappear from the economy; rather, they are redistributed to the consumption and production of other goods and services; and (ii) the expansionary employment effect of higher public expenditure after the tax increase.4 Studies simulating the impacts of tobacco control policies on employment depend on key assumptions. For example, studies applying input–output models first estimate the change in final consumer demand for goods and services resulting from a tobacco control policy (US NCI and WHO, 2016). The change in demand is composed of two com- ponents: (i) the reduction of tobacco consumption, and (ii) the expansion of expenditures in other products, according to consumers (smokers) preferences. These studies then calculate the induced changes in outputs based on input–output tables that describe the flow of goods and services within the economy. Finally, changes in outputs are converted into changes in employment to obtain the employment impacts. The critical assump- tions these studies rely on are: the impact of tobacco control measures—estimated price and income elasticities in the case of tax increases; and, more importantly, the type of consumption substitution that smokers display after the tax increase or tobacco control measure. The normal assumption is that ex-smokers would follow the average expendi- ture pattern or the most recent quitter expenditure pattern. If other selected goods and services have a larger direct and indirect employment input than tobacco product production, the net employment effect is positive. The challenge to measure the employment impacts of higher tobacco taxes is that employment losses could be relatively concentrated, whereas employment gains tend to spread throughout the economy. Table 1 presents a selection of recent independent studies in low- and middle-income countries (LMICs). Under the assump- tions discussed above, four out of the six studies resulted in net employment gains after tobacco control policies. Generally, the net effects were not significant, except in the case of Bangladesh and Egypt. Reductions occurred in core tobacco sectors, including tobacco farming and manufacturing; in tobacco-related sectors, such as wholesaling and retailing; and in ancillary sectors, such as the paper and pesticide industries. The net gains in employment depended on several factors, such as: the assumed structure of 4 This expansionary effect depends on fiscal policy decisions; greater public revenues do not mean automatically higher public expenditures. 20 // The Economics of Tobacco Taxation Table 1: Selected studies on net employment impact of tobacco control policies STUDIES MODEL AND ASSUMPTIONS CONCLUSIONS South Africa Static input–output model Domestic tobacco control Van der Merwe Domestic consumption expenditures were eliminated, and the rate of policies may have negative net and Abedian, consumption decline in 1995 doubled. employment effects. Smoking 1999 Expenditures were allocated by recent quitter and average prevalence and size of control expenditure pattern. policy could be important for absolute impact on net Government spending was reduced or kept at the same level by employment. increasing other taxes. Zimbabwe Static input–output model Van der Merwe, Domestic consumption expenditures and tobacco production in 1980 were eliminated. Net loss of 87,798 jobs in 1998 1980, and 47,463 jobs when Average input–output pattern changed, and all tobacco production all output went to alternative was shifted to alternative agriculture products. agriculture products Because of increases in other taxes, no change in government spending occurred. Bangladesh Static input–output model Van der Merwe, Domestic consumption expenditures and all tobacco production for 1998 tobacco products and bidis in 1994 were eliminated. Net gain of 10,989,192 jobs Average input–output pattern changed, and all tobacco production in 1994 was shifted to alternative agriculture products. Because of increases in other taxes, no change in government spending occurred. Bulgaria Static input–output model Petkova and Domestic consumption expenditures and tobacco production in 1999 colleagues, were eliminated. 2003 Net loss of 5,567 jobs in 1999 Average input–output pattern changed, and all tobacco production was shifted to alternative agriculture products. Because of increases in other taxes, no change in government spending occurred. Egypt Static input–output model Nassar and Domestic consumption expenditures and tobacco production in 1999 Metwally, 2003 were eliminated. Net loss of 5,567 jobs in 1999 Average input–output pattern changed, and all tobacco production was shifted to alternative agriculture products. Because of increases in other taxes, no change in government spending occurred. Indonesia Static input–output model Net gain of 84,340 jobs with a 25% tax increase; net gain Ahsan and Percentage increases of 25%, 50%, and 100% occurred in the of 140,567 jobs with a 50% Wiyono, 2007 cigarette tax. tax increase; and net gain of 281,135 jobs with a 100% tax Expenditures were allocated by the average expenditure pattern. increase Source: National Cancer Institute and World Health Organization (2017), Table 15.3, page 560. 21 The Economics of Tobacco Taxation and Employment in Indonesia population (smoker) consumption; the production structure of the economy, i.e., the extent to which final products, inputs, and services were produced domestically or imported in the tobacco industry and in the industries where ex-smokers would spend their money; and the labor intensity of tobacco growing/manufacturing versus the rest of the industries composed of the average consumer expenditures. For example, Ahsan and Wiyono (2007) found that, in Indonesia, the top five sectors that would experience increased employment include rice, tea, coffee, sugarcane, and root crops, which have higher labor intensity than tobacco growing (Ahsan and Wiyono, 2007) The possible employment impacts varied depending on whether the country was a net exporter or net importer of tobacco leaf. When tobacco control policies reduced the demand for cigarettes, a country was likely to have lower employment losses if that country imported a significant percentage of the cigarettes smoked and/or leaf used to make them, and domestically produced a large portion of the rest of the products included in the average consumption expenditure. Conversely, the more the tobacco leaf and other inputs and cigarettes were nationally grown and/or produced relative to the local content of the things people buy instead, the greater is the likelihood that there was some employment losses locally. The higher the production diversification of the economy, the higher was the domestic employment created by the demand switch from tobacco toward other products. In contrast, the higher the expenditure on tobacco and the lower the sectoral diversification of domestic production, the smaller (or even neg- ative) was the net employment impact. The United States National Cancer Institute (US NCI) and the WHO published a report that presented a classification of countries and the likely employment impacts, based on a country’s situation in terms of tobacco trade: net exporter, net importer, balanced economy and mixed situation (US NCI and WHO, 2016). As shown in Table 2, domestic tobacco control policies likely had a larger impact in coun- tries with a balanced tobacco economy—self-sufficient in tobacco, because the employ- ment destruction reduction of tobacco demand could not be offset. Decline in tobacco consumption as a result of taxation or tobacco control policies may occur gradually. Although most of the studies presented in Table 1 assume a sharp and total reduction in cigarette consumption, smoking prevalence reduction occurred gradually even when a significant tax increase was implemented in the short term.5 Tobacco control policies normally gradually reduce smoking, distributing adjustment costs through time and diluting them in decades (US NCI and WHO, 2016). The adjust- ment costs of labor-intensive segments of the industry, such as tobacco farming, depend on the existence of viable alternatives. In many LMICs, such as Brazil and the Philippines, tobacco farmers are diversified, producing other crops, which may facilitate the transition to other crops. The result from the World Bank/ACS survey on the economics of tobacco 5 The Philippines, for example, increased the excise taxes for the cheapest cigarettes by 341% in a one-year period that resulted in a decline of tobacco use prevalence by approximately 20% Global Adult Tobacco Survey (GATS). 22 // The Economics of Tobacco Taxation Table 2: Type of tobacco external trade and likely net employment impacts TYPE OF EXTERNAL MEANING LIKELY EMPLOYMENT IMPACTS TOBACCO TRADE THAT A COUNTRY HAS Net exporter of tobacco Production of tobacco leaf and Domestic tobacco demand is not the only determi- products cigarettes is higher than domestic nant of production/employment in the core sectors. consumption. Drop of domestic tobacco demand could be com- pensated with production directed to external mar- Tobacco employment distribution kets and employment effects could be minimized. in domestic sales or exports would The relative effects of global and domestic policies depend on the share of exports to will depend on the share of production that is total sales exported Balanced tobacco econ- Domestic production of tobacco Domestic tobacco control policies may have nega- omy leaf or cigarettes is used primarily tive net employment effects. Smoking prevalence for local consumption. Self-suffi- and size of control policy could be important for cient in tobacco absolute impact on net employment. Net importers of tobacco Produce less tobacco leaf or ciga- Tobacco control policies in countries with small products rettes than they consume. tobacco/cigarette production may increase overall employment. Not affected by global tobacco demand. Mixed tobacco economy Significant grower and producer US is an example: tobacco leaf producer, importer and also imports and/or exports a and exporter of large amounts of tobacco leaf. substantial share of tobacco leaf Changes in both domestic and global tobacco con- and tobacco products trol policies would affect employment. Source: Own elaboration based on US NCI and WHO (2016). farming unequivocally demonstrated that former tobacco farmers were growing many of the same crops as current tobacco farmers, and were simply increasing production of these crops. Moreover, these former tobacco farmers were typically generating higher revenues and incurring lower costs than their neighbors and peers who continued to grow tobacco. The employment issue has lost preeminence in the debate of tobacco control policies in many countries. Since the first half of the last decade, there have been few studies about the employment effect of tobacco control policies from either side of the debate; instead the debate has concentrated on other issues, such as illicit trade in tobacco products. For example, in the recent Brazilian experience of large increases of tobacco taxes (2011–2016), effects on employment were not an issue, because the export share of leaf production in the country increased. In the Philippines (2012 onward) poten- tial employment impact was a big issue during Congressional deliberations, but since 80% of tobacco leaf was exported, the Ministry of Finance recognized that the employ- ment impact risk on tobacco farmers was low. Furthermore, a budget was also allocated to help those that may be adversely affected—15% of the incremental tobacco tax reve- nues was allocated to tobacco farmers to encourage them to shift to alternative crops. 23 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO V LENCE PREVA AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 3 CIGARETTE TAX POLICIES IN O INDONESIA Indonesia has one of the most complex cigarette excise tax structures in the world. N The current cigarette excise tax system has 12 tiers, which are based on manufacturers’ types of cigarettes, the scale of cigarette productions, and per unit retail price.6 There are three types of cigarettes: machine-made kreteks (SKM), machine-made white cigarettes (SPM), and hand-rolled kreteks (SKT).7 Manufacturers of either SKM or SPM are considered as a Class I if they produce more than 3 billion cigarettes annually and as a Class II if they O produce less than the 3 billion cigarettes. On the other hand, manufacturers of SKT are considered a Class I if they produce more than 2 billion cigarettes annually, a Class II if they produce between 500 million to 2 billion cigarettes annually, and a Class III if they produce less than 500 million cigarettes annually. Excise taxes for cigarettes produced by larger manufacturers increased at a modest rate between 2010 and 2017 (Figure 1). As shown in column 4 of Table 3, the excise taxes for machine-made kreteks (SKM, Class I) and white cigarettes (SPM, Class I) increased by 27 and 46 percent, respectively in real terms between 2010 and 2017. Note that changes in the tariffs for the small-scale manufacturers were relatively smaller compared to changes for the Class I SKM and SPM manufacturers. For example, excise taxes for Class II SKM manufacturers increased 24 percent while excise taxes for Class II SPM manufactur- ers increased 35 percent. It is important to note that tax increases that occurred have not been enough to reduce cigarette affordability. The US NCI and WHO show that cigarettes in Indonesia were much more affordable in 2013 than they were in 2000 (US NCI and WHO, 2016). L 3 6 Ministry of Finance Decree Number 43/PMK.04/2005. 7 In addition to SKM, SPM, and SKT, there are hand-rolled white cigarettes with filters. The structure of the excise tax for these products are identical to the structure for SKM.. 2 25 The Economics of Tobacco Taxation and Employment in Indonesia Table 3: Cigarette excise tax in Indonesia, 2017 TYPE OF CATEGORY PER-UNIT % NOMINAL NUMBER CIGARETTE (PRODUCTION) TARIFFS CHANGE OF (TIERS), IN (REAL), FACTORIES RUPIAH 2010-2017 (2015) Machine-made kreteks (SKM), hand- Class I, > 3 billion 530 78 (27) 14 rolled kreteks with filter (SKTF), sticks hand-rolled white cigarettes with filter (SPTF) Hand-rolled white cigarettes with Class II, ≤ 3 billion 335-365 75 (24) 232 filter (SPTF) sticks Machine-made white cigarettes Class I, > 3 billion 555 106 (46) 1 (SPM) sticks Class II, ≤ 3 billion 290-330 90 (35) 25 sticks Hand-rolled kreteks (SKT) Class I, > 2 billion 265-345 66 (18) 16 sticks Class II, 500 million–2 155-165 64 (15) 21 billion sticks Class III, ≤ 500 million 80-100 38 (-2) 404 sticks Source: Peraturan Menteri Keuangan (Ministry of Finance Decree) abbreviated as PMK: PMK Number 181 2009, PMK Number 190 2010, PMK Number 167 2011, PMK Number 205 2014, PMK Number 198 2015, PMK Number 147 2016. Note: For parsimony, per-unit tariffs for Class II and Class III categories are simplified from per-unit tariffs for Class IIA, Class IIB, Class IIIA, and Class IIIB. Real change in per-unit tariffs accounts for changes in prices of goods and services in the economy. The tiers in the cigarette excise tax structure aimed to accommodate small-scale cigarette firms, especially SKT firms. The rationale for such a structure was to protect smaller SKT firms that accounted for more than half of total factories in the tobacco indus- try (column 5 of Table 3). Moreover, these firms were responsible for employing a signifi- cant share of the workers in tobacco manufacturing. For example, the per-unit tariffs for a SKM produced by Class I manufacturers and sold for a minimum of IDR 1,120 per unit was IDR 530 (47.3 percent of retail sale price), while the per-unit tariffs for a SKT produced by Class I manufacturers and sold for a minimum of IDR 1,215 per unit was only IDR 345 (28.4 percent of retail sale price). These numbers imply that per-unit tariffs for SKT was half of the per-unit tariffs for SKM of a similar tier. The ratio between the highest and the lowest per-unit tariffs in the 2017 tax structure was quite high at 6.7. Changes in real excise tax for SKT between 2010 and 2017 was also quite low at 18% or lower. Additionally, SKT pro- duced by Class III producers were cheaper, in real terms, in 2017 than they were in 2010. 26 // Cigarette Tax Policies in Indonesia Figure 1. Trends of the Cigarette Excise Tax in Indonesia in Real Terms, 2010–2017 400 400 EXCISE TARIFF PER STICK, IN REAL TERMS 300 300 200 200 100 100 2010 2011 2012 2013 2014 2015 2016 2017 2010 2011 2012 2013 2014 2015 2016 2017 SKT Class I, layer 1 SKT Class I, layer 2 SKT Class II, layer 1 SKT Class II, layer 2 SKM Class I SPM Class I SKM Class II, layer 1 SKM Class II, layer 2 SPM Class II, layer 1 SPM Class II, layer 2 Source: Peraturan Menteri Keuangan (Ministry of Finance Decree) abbreviated as PMK: PMK Number 181 2009, PMK Number 190 2010, PMK Number 167 2011, PMK Number 205 2014, PMK Number 198 2015, PMK Number 147 2016. Note: The pre-2009 tariffs were calculated based on the current 12 layers. Despite the low tax rates, there was a downward trend in the sale of SKT in the market. In 2001, the share of SKT in the market was about 40% (Ministry of Industry, 2009). The share decreased to 35.5% in 2005, to 30% in 2011, and to just 26% in 2013. We argue that the shift away from SKT was not likely the result of higher cigarette taxes. First, as shown in Table 3, taxes imposed on SKT were among the lowest in the cigarette industry, and the changes of the taxes in real terms were quite small. Second, despite the decreasing share, productions of SKT were still increasing over time at least up to year 2010 (Tobacco Control Support Center, 2014). The shift away from SKT have been driven by changing preferences among smokers for machine-made products due to income growth. It may also have reflected substitution of more capital-intensive technology away from labor-intensive hand-rolling in cigarette production. 3.1 Effects on Cigarette Consumption, Revenues, and Employment Previous evidence suggests that raising cigarette prices through taxes may have potential to reduce consumption in Indonesia. A review by Setyonaluri and col- leagues (2008) reported price elasticities of demand ranging from –0.26 to –0.76. These estimates suggest that raising prices through taxes can reduce consumption. Ahsan (2011) suggested that a 16 percent increase in cigarette tax would reduce consumption by 4.7 percent, implying a price elasticity of –0.29. The impacts of raising cigarette tax on health were also potentially significant. A study suggested that raising cigarette taxes to 27 The Economics of Tobacco Taxation and Employment in Indonesia 50 percent of retail prices can reduce expected mortality rate by 2 to 5 percent, depend- ing on the price elasticity of demand (Setyonaluri et al., 2008). This report updates earlier elasticity estimates. To do so, we used the 2015 National Socioeconomic Survey (Susenas) to estimate the price elasticity of cigarette demand. The NSS contains data on whether an individual smoked kreteks or white cigarettes, the number of cigarettes smoked in the past week, and total expenditure for cigarettes. The limitation of our estimation was that we could not differentiate between hand-made and machine-made kretek. However, obtaining separate estimates for kreteks and white cigarettes offered a significant improvement over the existing literature. This information allowed estimating price elasticity of demand for SKT and SKM separately. Results showed variation in price elasticities of demand between kreteks and white cigarettes. The esti- mated price elasticity of demand for kreteks was about –0.42 while for white cigarettes was about –0.51 (within the range of –0.26 and –0.76 reported in previous studies). We acknowledge our price elasticities may be underestimated because they do not account for substitutions to cheaper products. For example, individuals can easily switch to cheaper cigarettes if cigarette prices increase because of higher excise taxes. An important aspect is to observe the impact of tobacco excise tax across income groups. The global evidence suggests that cigarette tax was regressive and showed that the young and the poor were more responsive to price changes. For example, Nasrudin and colleagues (2013) estimated price elasticities of –0.15 to –0.16 among individuals in the first two income deciles and price elasticities of –0.20 to –0.28 among individuals in the third to ninth deciles. The authors estimated that the tax burden as a percentage of income of individuals in the first two deciles was between 3 to 10.63 percent and the tax burden of individuals from the other income groups was below 2 percent (Nasrudin et al., 2013). These results may be explained by the policy of keeping kreteks cheap. However, the study did not consider the benefits of raising cigarette taxes. A more recent study showed that once indirect benefits (such as lower health care expenditures and higher productivity) of raising cigarette taxes were accounted for, a tobacco tax increase was actually progressive (Fuchs and Meneses, 2017). In Indonesia, annual excise tax revenues from tobacco products had been increas- ing monotonically since 2005 (Figure 2). In 2007, the realized excise tax revenue was IDR 43.54 trillion (US$4.87 billion). The figure grew to IDR 55.38 trillion (US$5.33 billion) in 2009, IDR 103.6 trillion (US$9.90 billion) in 2013, and IDR 139.5 trillion (US$11.76 billion) in 2014 (IAKMI, 2014; Dwi Kurnaini, 2016).10 Historically, tobacco excise accounts for more than 95 percent of total excise revenue. Simulation studies suggest that a 10 percent increase in cigarette tax would increase excise revenues by 6.7 to 9.0 percent (Setyonaluri 10 The exchange rates were IDR 8,938/US$ in 2004, IDR 10,389/US$ in 2009, IDR 10,461/US$ in 2013, and IDR 11,865/US$ in 2014 (OECD, 2017). 28 // Cigarette Tax Policies in Indonesia Figure 2. Revenue from the tobacco excise, 2005–2015 1.21 TOBACCOV EXCISE REVENUE, IN TRILLION IDR 140 140 1.18 SHARE OF EXCISE REVENUE TO GDP 120 1.20 113 1.14 107 100 104 1.15 1.11 1.10 1.10 91 92 89 80 82 1.10 73 1.18 1.07 63 60 58 70 1.05 55 63 52 47 48 50 55 40 1.01 0.99 1.00 44 0.98 0.99 37 33 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 realization in real term share to GDP Source: Data collected from Tobacco Control Support Center—IAKMI (2014) and Dwi Kurnaini (2016) for post 2010 data. et al., 2008). Another simulation suggests that a 16 percent increase in cigarette tax would boost excise revenue by 30.5 percent (Ahsan, 2011). These studies suggest that the tax elasticity of revenue is between 0.67 to 1.90. Raising cigarette taxes did not seem to affect cigarette production as production exhibited an increasing trend over the years. As shown in Figure 3, the tobacco indus- try produced 235.5 billion cigarettes in 2005 and increased that number for the next 10 years. Cigarette production hit the 300 billion mark in 2011, peaked at 348 billion in 2015, and decreased to 342 billion sticks in 2016 (Ahsan, 2015; Dwi Kurnaini, 2016). A positive trend in cigarette production over the years could have been driven by growing cigarette affordability and the ineffectiveness of low excise taxes in reducing affordability. Never- theless, it is important to note that growth of cigarette production exhibited a decreasing trend post 2010. 29 The Economics of Tobacco Taxation and Employment in Indonesia Figure 3. Trends of cigarette production in Indonesia, 2010–2016 CIGARETTE PRODUCTION 8.12 400 8.00 PRODUCTION GROWTH 7.30 6.62 346 344.52 348.12 342 326.8 310.2 5.88 289.1 6.00 300 267.4 5.35 245.4 241.5 235.5 250.8 4.20 3.85 4.00 200 52.50 58.25 57.82 56.77 57.87 2.00 57.52 1.04 36.34 34.94 35.1 100 35.5 35.95 35.32 0.00 -1.59 6.81 6.84 -0.43 6.95 6.23 6.81 0 6.9 -1.76 -2.00 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 overall, billion sticks SKM (%) SKT (%) SPM (%) growth (%) Source: Ministry of Finance as cited in Ahsan (2015) and Dwi Kurnaini (2016) for post 2013 data. Note: production data by types of cigarette are not available for public after 2010. Despite a positive production trend, the number of cigarette firms decreased quite significantly in the past decade. Dwi Kurnaini (2016) reported that the number of cig- arette factories decreased from 4,699 in 2007 to just 713 in 2015. There were 246 factories in the SKM industry (34.5 percent), 441 factories in the SKT industry (61.9 percent), and 26 factories in the SPM industry (3.6 percent) in 2015 (Dwi Kurnaini, 2016). Moreover, the Class III SKT industry operated 404 factories in 2015 which accounted for 56.6 percent of total factories in the cigarette industry (Dwi Kurnaini, 2016). The reduction in the num- ber of factories was partly due to closing a tax loophole. Larger companies formed small cigarette firms to take advantage of a lower tax rate for small-scale firms. The government closed this loophole by the requirement that there should be no ownership link between the small and the large cigarette firms. Previous evidence showed that raising cigarette taxes had a net positive impact on employment in Indonesia. A study by Ahsan and Wiyono (2007) investigated three policy scenarios using an input–output (IO) model: a 30%, 50%, and 100% increase in cigarette tax. These increases corresponded to higher cigarette prices by 8%, 13%, and 26%, respectively. The authors found overall employment gains in all scenarios, although employment in cigarette manufacturing, and tobacco and clove farming decreased (which meant net employment effect was positive). The analysis suggested that work- ers would have shifted to other agricultural sector such as other food crops, paddy, tea, coffee, sugarcane, and root crops. The net effect on employment of a 100% increase in 30 // Cigarette Tax Policies in Indonesia cigarette taxes was an additional 281,135 workers. This was equivalent to a 0.3% increase in total employment (Ahsan and Wiyono, 2007). Another study by Marks (as cited in Ahsan and Wiyono, 2007; Setyonaluri et al., 2008) showed that an increase in real price of cigarettes by 80% resulted in a gross reduction of jobs in the SKT sector by 86,000. A more recent study by Nasrudin and colleagues (2013) showed that raising cigarette taxes resulted in gross employment reduction in the cigarette manufacturing sectors. Specifi- cally, the study estimated a decrease in employment by 3.27 to 3.46 percent among small SKT producers, by 6.66 to 6.78 percent among medium and large SKT producers, and by 5.85 to 6.04 percent among the machine-made cigarette sector (Nasrudin et al., 2013). 31 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO V LENCE PREVA AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 4 EMPLOYMENT IN THE O INDONESIAN TOBACCO SECTOR N This section presents an overview of the employment trends in the Indonesian tobacco industry and estimates the effects of raising cigarette taxes on gross employment in the tobacco sector. It reviews the recent trends in employment in the tobacco sector, such as the number of workers, the share of tobacco manufacturing O employment to total manufacturing employment, and labor productivity. The section also discusses the concentration of tobacco workers across provinces in Indonesia and compares employment trends to other similar sectors. The core Indonesian tobacco sector includes workers in the tobacco manufactur- ing sector, tobacco farmers, and clove farmers. The tobacco manufacturing sector is divided into three main industries: kretek, white cigarette, tobacco processing and tobacco/clove farming. In 2014, there were 307,793 workers in the kretek industry, 10,598 workers in the white cigarette industry, 352,086 workers in tobacco processing, and 16,529 workers in the non-cigarette industry.11 There were also approximately 761,310 tobacco farmers in 2011 and 1.04 million clove farmers in 2013.12 However, evidence suggests these farmers also dedicated some percentage of their time to other crops and, therefore, the full-time equivalent of the number of workers should be significantly lower (World Bank, 2017a; 2017b; 2017c). Clove farmers were considered as the core employ- ment because 90 percent of annual clove production was purchased by cigarette compa- nies to produce kreteks even though a clear majority of clove farmers made just a fraction of their income from clove (World Bank, 2017c). L 4.1 Data Sources To provide a picture of the size, composition, and trends of the tobacco sector workforce, this report utilized firm- and household-level micro data. We used the annual survey of micro and small industry (SIMK), aggregated statistics of micro and small 3 industry, the annual survey of manufacturing industry (SI), and the National Labor Force 11 Non-cigarette products included cigars, kelembak menyan, and tobacco flavoring. 12 The numbers were calculated or obtained from the 2014 Annual Survey of Manufacturing Industry (SI), the 2014 Annual Survey of Micro and Small Manufacturing Industry (SIMK), Table 4.8 of Tobacco Control Support Center—IAKMI (2014), and the 2014 Clove Farming Statistics. 2 33 The Economics of Tobacco Taxation and Employment in Indonesia Survey (Sakernas), all published by the Central Bureau of Statistics.13 The firm-level data included, among others, variables such as the number of workers, wages, materials, costs of inputs, outputs, and value added. We used these data to generate the employment trend in the tobacco industry and to estimate output elasticity of employment of output. The Sakernas included employment variables such as years of schooling, hours of work, type of jobs, employment status, and occupation category. We used these data to describe the labor market characteristics of workers in the tobacco manufacturing sector.14 For the tobacco manufacturing sector, we focused our analysis on the kretek, white cigarette, and tobacco processing industry. We acknowledged that we could not iden- tify in SIMK or SI whether kretek firms produced SKT or SKM. This information would have been very useful for estimations of output elasticity of labor since SKT firms tend to be more labor intensive than SKM firms. Moreover, we could not identify tobacco and clove farmers in the Sakernas dataset. Therefore, we use aggregated data of tobacco and clove farmers from Tobacco and Clove Farming Statistics. The analysis complemented the household level data collected under the Indone- sian Tobacco Employment Studies, which focused on small-holders’ tobacco and clove farmers and kretek rollers. Results from these surveys are reported in the accom- panying reports (World Bank, 2017a, b and c). 4.2 Employment in the Indonesia Tobacco Farming Sector The share of tobacco farmers to total farmers in the agricultural sector fluctuated around 1.6 percent in recent decades. Additionally, the share to total workers in the economy fluctuated around 0.7 percent (see Figure 4). Furthermore, tobacco was just a part of what the farmers did economically. As reported by the World Bank/ACS report on tobacco farming in Indonesia, tobacco farmers only dedicated a portion of their land to cultivate tobacco and only a minority of tobacco-farming households relied on tobacco farming as their major income-earning activity (World Bank, 2017c). We can observe in Table 4 that tobacco farmers were concentrated in Java, particularly in East Java and Cen- tral Java. The combined number from these two provinces made up about 84% of total tobacco farmers in the nation. There was also a non-negligible number of tobacco farm- ers in Nusa Tenggara Barat, which accounted for about 7% of total Indonesian tobacco farmers. These statistics underlined the phenomena that the tobacco industry, particularly tobacco manufacturing and tobacco farming, was concentrated in few provinces. 13 The annual survey of micro and small industry was available from 2010. 14 When using the National Labor Force Survey (NLFS), we acknowledge that estimation of statistics using 3-digit International Stan- dard Industrial Classification (ISIC) yields high relative standard error owing to the sampling design. 34 // Employment in the Indonesian Tobacco Sector Figure 4: Number of tobacco farmers and share of tobacco farmers to total workers, 1990–2011 913 900 894 809 800 761 TOBACCO FARMERS, IN THOUSANDS 715 694 700 684 680 669 665 636 628 598 600 582 500 512 400 400 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 share to total agriculture workers share to total workers in the economy 2.6 2.5 2.3 2 2 1.8 1.8 1.7 1.7 1.7 1.6 SHARE, IN % 1.6 1.6 1.5 1.4 1.5 1.4 1.2 1 1 1 1 .9 .8 .8 .7 .7 .7 .7 .7 .6 .6 .6 .6 .5 .5 .5 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Source: Data collected from Tobacco Control Support Center—IAKMI (2014). 35 The Economics of Tobacco Taxation and Employment in Indonesia Table 4. Distribution of tobacco farmers across regions, 2014 PROVINCE FARMER, 2014 SHARE, 2014 (%) PRODUCTIVITY (TON/FARMER) East Java 351,217 61.87 0.308 Central Java 125,154 22.05 0.26 NTB 38,336 6.75 0.967 West Java 26,319 4.64 0.31 Sumatera (all provinces) 11,188 1.97 0.671 DIY 8,888 1.57 0.123 Sulawesi 2,874 0.51 0.546 NTT 2,563 0.45 0.509 Bali 1,098 0.19 0.853 Indonesia 567,637 100 0.349 Source: Indonesian Plantation Statistics: Tobacco (2014). Similar to tobacco farmers, clove farmers were concentrated in East, Central, and West Java (Table 5). Data from the Indonesian Plantation Statistics (2015) suggest that there were around one million clove farmers in Indonesia. This suggests that about 2.69 percent of all agricultural workers grew cloves, or 0.95 percent of total workers in the economy. However, the World Bank/ACS survey showed that clove farming was just but one of the crops that these farmers grew. As reported in the World Bank/ACS report, farmers did not need to tend the plants for much of the year (World Bank, 2017c). Clove farming also contributed a small part to household income. Using the aggregated statistics, we found that the productivity of clove farmers in Java was quite low, which is similar to tobacco farmers in Java. Nevertheless, the calculation of productivity was crude because it didn’t consider many aspects of clove farming reported in the World Bank/ACS report (World Bank, 2017c). 4.3 Employment in the Tobacco Manufacturing Sector The tobacco manufacturing sector employed approximately 692,000 workers in 2014. As shown in Table 6, employment in the kreteks and tobacco processing industry accounted for 96 percent of tobacco employment. The kretek industry employed 307,793 workers in 2014, while the tobacco processing industry employed 352,086. Tobacco pro- cessing usually involved farming households that dried, cured, and sometimes cut the tobacco leaves for further processing, such as for cigarette manufacturing. Most kreteks workers worked in large firms (97 percent) while most tobacco processing workers 36 // Employment in the Indonesian Tobacco Sector Table 5. Distribution of clove farmers across regions, 2014 PROVINCE FARMERS, 2013 SHARE, 2013 (%) PRODUCTIVITY (TON/ FARMER) East Java 238,100 22.62 0.045 Central Java 189,527 18.00 0.033 West Java 143,249 13.61 0.046 Sumatera (all provinces) 76,109 7.00 0.115 North Sulawesi 72,284 6.87 0.130 Maluku 65,352 6.21 0.212 South Sulawesi 61,114 5.81 0.286 Bali 53,233 5.06 0.058 Central Sulawesi 44,629 4.24 0.309 East Nusa Tenggara 24,725 2.35 0.078 North Maluku 19,749 1.88 0.227 Southeast Sulawesi 17,826 1.69 0.368 Banten 17,618 1.67 0.267 DI Yogyakarta 12,591 1.20 0.029 Gorontalo 6,360 0.60 0.119 West Nusa Tenggara 3,344 0.3=2 0.038 West Sulawesi 2,705 0.26 0.146 Kalimantan (all provinces) 2,490 0.24 0.145 Papua (all provinces) 1,639 0.16 0.037 Total 1,052,644 100 0.104 Source: Indonesian Plantation Statistics: Clove (2015). worked in micro or small firms (92 percent).15 As pointed out above, the available data do not allow to distinguish between firms that produced SKM and SKT. Therefore, we assumed that most of the large firms were firms producing SKT as they required a larger number of workers in the production process. 15 We used the government’s definition to define the production scale of a firm. Firms that employed less than 5 workers were con- sidered micro firms, between 5–19 workers were considered small firms, between 20–99 workers were considered medium firms, and more than 99 workers were considered large firms. 37 The Economics of Tobacco Taxation and Employment in Indonesia Table 6. Employment Structure of the Indonesian Tobacco Industry, 2014 PRODUCTION KRETEKS WHITE PROCESSING OTHERS SCALES Worker Firms Worker Firms Worker Firms Worker Firms Micro 828 368 119,400 36,500 5,288 4,334 Small 1,748 92 203,067 21,363 558 54 Medium 6,132 148 448 6 13,094 399 1,606 44 Large 299,085 209 10,150 11 16,525 32 9,077 13 Total 307,793 817 10,598 17 352,086 58,294 16,529 4,445 Source: Calculated using data from ASMSI & ASMI (2014). Employment in large and medium tobacco manufacturing firms exhibited positive trend in the years 2000–2014. As shown in Panel B of Figure 5, the number of tobacco manufacturing workers in large and medium firms was about 245,000 in 2000 and increased by 68 percent to 356,000 in 2014. In Panel A of Figure 5, we can observe that the total number of workers in the tobacco manufacturing sector was approximately 692,000 in 2014. The tobacco sector employed a high share of female production workers, which was similar to the food, textile, and garment sectors.16 The contribution of the tobacco sector to employment in the manufacturing sector was quite small relative to similar manufacturing sectors. Tobacco manufacturing employment represented 5.13 percent of total manufacturing employment in Indonesia, while industries such as food employed 27.43 percent, garment employed 11.43 percent, and textile employed 7.90 percent. Furthermore, the contribution of tobacco manufac- turing employment to economy-wide employment was also quite low at 0.60 percent in 2014. In Panel B of Figure 6, we can observe the long-run trend of employment contribu- tion by each sector in the medium and large manufacturing industry. The employment share of medium and large tobacco firms increased steadily from 6.36 percent in 2001 to about 8 percent in 2008. However, the share exhibited a decreasing trend post 2008 and reached 6.93 percent in 2014. 16 We acknowledged that there was a drop in the tobacco manufacturing employment in 2011. However, we were not able to find explanations about the drop during the period of the study. 38 // Employment in the Indonesian Tobacco Sector Figure 5. Number of Medium and Large Tobacco Manufacturers, 2000–20142014 PANEL A: ALL SCALES 4,198,921 3,702,756 4,000,000 3,390,846 3,446,568 2,828,211 3,000,000 2,237,451 2,000,000 1,841,062 1,722,296 1,542,540 1,185,421 1,101,802 1,066,129 928,264 895,975 1,000,000 910,796 691,625 692,795 556,878 381,001 382,411 2010 2011 2012 2013 2014 PANEL B: MEDIUM & LARGE FIRMS ONLY 1,000,000 881,769 899,945 875,739 800,000 726,178 739,832 690,746 659,821 675,230 581,089 622,442 636,026 583,523 650,270 598,672 600,000 564,974 560,594 571,008 572,861 551,675 525,011 536,933 528,977 528,772 524,698 527,461 570,726 546,765 494,432 495,382 484,154 483,808 481,914 519,224 507,056 486,813 474,151472,806 475,303 482,100 476,684 477,037 442,888 428,924 424,259 400,000 362,933 356,117 346,042 329,877 334,194 331,590 324,614 299,746 304,243 263,943 316,991 265,378 242,930 272,343 254,728 200,000 189,652 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 tobacco food textile garmet Source: Calculated using data from ASMSI (2000-2014) and aggregated statistics of micro and small industry (2014). 39 The Economics of Tobacco Taxation and Employment in Indonesia Figure 6. Share of tobacco employment to total manufacturing employment, 2000–2014 PANEL A: ALL SCALES PANEL B: MEDIUM & LARGE FIRMS ONLY 30.00 28.57 27.43 20.00 26.91 18.09 18.14 25.93 24.21 16.57 15.72 16.32 17.05 15.26 14.84 14.84 20.00 15.00 14.55 14.23 15.39 15.45 15.14 15.98 14.30 13.60 13.48 13.75 13.15 12.87 13.45 13.10 12.53 12.46 12.76 12.47 12.38 13.12 12.28 11.43 10.87 12.29 11.45 12.23 11.82 11.45 IN PERCENTAGE 11.59 11.16 11.50 11.40 10.80 10.00 10.52 8.35 10.64 7.50 7.90 10.00 10.21 7.25 10.04 10.09 9.89 6.40 9.61 6.34 8.00 7.88 5.13 7.39 7.31 7.45 6.84 6.64 6.86 6.93 6.71 3.79 6.25 6.36 6.62 6.66 6.21 2.97 2.73 5.00 2000 2001 2002 2004 2005 2006 2008 2009 2011 2012 2013 2014 2003 2007 2010 2010 2011 2012 2013 2014 tobacco food textile garmet tobacco food textile garmet Source: Calculated using data from ASMSI (2000–2014) and aggregated statistics of micro and small industry (2014). Tobacco manufacturing jobs were heavily concentrated in a few regions of Indonesia. Most of the tobacco manufacturing workers were concentrated in Central and East Java (Figure 7). Our estimates from the 2014 annual industry survey showed that the combined share of tobacco manufacturing workers in Central and East Java was about 76 percent of total tobacco manufacturing workers in Indonesia. Another 18 per- cent of manufacturing workers were in West Nusa Tenggara. This was a higher concentra- tion when compared to other sectors, for example, 40 percent of the workers in the food and drink sector were concentrated in Central Java (21 percent), East Java (19 percent) and 18 percent in West Java. In the garment sector 42 percent of workers were in West Java and 29 percent in Central Java. This concentration of tobacco workers meant that some districts were dependent on tobacco sector employment. In Kudus, Temanggung, and Kediri, employment in the tobacco sector accounted for more than 60 percent of local manufacturing employment and more than 25 percent of overall local employment (Table 7). The type of tobacco industry varied across districts. For example, Kudus and Kediri relied quite heavily on the kreteks industry. In Kediri, the share of kreteks workers to total employment in the manu- facturing industry was about 67.6 percent. On the other hand, Temanggung relied heavily on the tobacco processing industry. Any intervention to alleviate possible employment impacts of tobacco consumption shocks should be focused on these districts. 40 // Employment in the Indonesian Tobacco Sector Figure 7. The Concentration of Tobacco Manufacturing Workers, 2014 50 48.50 SHARE, IN PRECENAGE 40 30 27.18 20 18.37 10 3.75 0.70 0.63 0.35 0.21 0.14 0.07 0.04 0.02 0.01 0.01 0.00 0 CENTRAL JAVA EAST JAVA WEST NT WEST JAVA NORTH SUMATERA DI YOGYAKARTA SOUTH SULAWESI BALI NAD WEST SUMATERA RIAU ISLANDS LAMPUNG RIAU CENTRAL SULAWESI BANTEN Source: Calculated using data from ASMSI (2000–2014) and aggregated statistics of micro and small industry (2014). Table 7: Concentration of Workers in Selected Districts, 2014 SHARE TO TOTAL CENTRAL KUDUS TEMANG- EAST KEDIRI SRA-BAYA MALANG WEST NT LOCAL JAVA GUNG JAVA EMPLOYMENT IN MANUFACTURING INDUSTRY Kreteks 5.47 57.98 - 5.99 67.6 12.29 34.55 - White ciga- 0.02 0.13 - 0.23 - - 0.61 - rettes Tobacco 6.6 2.6 66.48 0.66 - 0.19 - 39.18 processing Total tobacco 12.18 60.72 66.48 7.32 67.6 12.49 35.2 39.18 Food 31.08 8.99 9.83 28.91 17.54 16.71 17.24 13.9 Wood products 12 0.58 18.26 15.83 0.36 2.15 0.04 17.55 Garment 11.64 6 1.29 3.64 0.2 6.42 6.06 3.53 Textiles 9.39 2.18 0 5.2 0.1 2.56 2.5 5.54 Furniture 6.23 4.06 0.4 4.93 0.36 1.97 4.49 1.88 Others 17.48 17.47 3.71 34.17 13.84 57.7 34.47 18.42 Share of tobac- 2.01 30.13 27.65 0.97 26.29 2.12 5.07 6.03 co employment to local employment (%) Source: Calculated from ASMI & ASMSI (2014) and NFLS (2014). 41 The Economics of Tobacco Taxation and Employment in Indonesia The Indonesian tobacco manufacturing sector exhibited a high worker per firm ratio relative to other sectors, and that was mostly due to kretek manufacturing. Available data only focused on the medium and large industries in which we observed variations across firms (Figure 8). In the food and textile industries, typical firms employed on average 157 and 214 workers in 2014, respectively. In contrast, a typical tobacco manufacturer employed, on average, 414 workers in 2014. Note that we could not distinguish whether firms produced SKT or SKM. Nevertheless, a recent report by the Tobacco Control Support Center—IAKMI showed that the SKT industry was the most labor-intensive industry in the tobacco manufacturing sector, with a typical SKT firm employing 662 workers and a typical machine-made kreteks firm employing just 84 workers (IAKMI, 2014). Figure 8. Number of workers per firm by industry PANEL A: ALL SCALES PANEL B: MEDIUM & LARGE FIRMS ONLY 13 419 413 12 11 11 400 351 339 336 10 326 320 344 9 317 315 326 330 306 299 300 302 317 305 308 8 7 269 277 277 259 234 229 232 246 237 255 6 6 219 5 229 231 211 215 214 207 5 216 212 200 200 202 209 4 4 4 4 4 4 4 177 176 180 3 156 156 3 4 4 144 142 137 136 147 133 3 3 129 121 117117 116 118 2 100 104 2010 2011 2012 2013 2014 2000 2001 2004 2005 2006 2008 2009 2011 2012 2013 2014 2002 2003 2007 2010 tobacco food textile garmet tobacco food textile garmet Source: Calculated using data from ASMSI (2000–2014) and aggregated statistics of micro and small industry (2014). Most tobacco sector workers were females and unskilled. The share of female work- ers in the tobacco industry was 66 percent, which was the highest in the manufacturing industry (Table 8). Other sectors in which most workers were female were the garment sectors (63 percent) and textiles sectors (56 percent). About 69 percent of total workers in the tobacco sector completed at most junior high school. Among tobacco households, the share of tobacco households with female tobacco workers was 78 percent. Average years of schooling completed by tobacco sector workers was among the lowest at 8.22 years, which was comparable to average years of schooling completed by workers in the food processing sector (8.50 years), manufacturing of wood products (7.31), non-metallic metal products (7.42), furniture sectors (8.77), and recycling sectors (5.20). 42 // Employment in the Indonesian Tobacco Sector Table 8. Characteristics of Tobacco Sector Workers, 2011, 2013, and 2015 VARIABLE 2011 2013 2015 1 if rural 52.50% 55% 60% Age 36.2 35.9 37.1 1 if female 72% 75% 66% 1 if production workers 93% 91% 92% Hours worked 46.7 34.6 44.1 Years of Schooling 7.79 8.12 8.21 1 if less than elementary school 16% 15% 15% 1 if completed elementary school 32% 31% 30% 1 if completed junior high school 27% 25% 24% 1 if completed high school 25% 26% 26% 1 if working part-time i.e. <30 hours pw 7.96% 33.04% 13.73% 1 if wage is below regional minimum wage 40.51% 44.86% 50.10% 1 if wage is below regional poverty line 5.01% 5.51% 3.62% Number of workers17 518,328 504,726 465,236 Source: Statistics Indonesia, the August National Labor Force Survey, 2011, 2013, 2015. Note: The statistics are based on population-weighted estimates. The sample includes only working-age individ- uals who were currently employed or those who were temporarily out of their jobs during the survey. Tobacco sector workers were those whose firms were classified as ISIC 160. On average, the share of tobacco income to overall household income was 60 percent. One interesting aspect to note was that the share of female tobacco workers as primary earners was only 9 percent.18 The findings from the Sakernas data were quite consistent with the ones from the World Bank/ACS kretek workers survey in Indonesia (World Bank, 2017b). In this survey, we found that the proportion of wage income from kretek was 54 percent. We also found that the share of female tobacco workers as primary earners was about 10 percent. Unconditional wages in the tobacco industry were relatively low, reflecting the fact that the industry employed a high share of low-skilled workers.19 In 2015, the average and median real monthly wage of tobacco production workers was IDR 758,859 17 The firm- and household-level estimates of total tobacco workers were quite different. This discrepancy can be attributed to a high relative sampling error when we estimate industry-level statistics using the household-level data. 18 A female worker was considered a primary earner in the household if the only source of income in the household came from the female worker. 19 Unconditional wages refers to average wages for every worker in a sample. Wages, however, were determined by many factors such as gender, age, years of working experience, years of schooling completed, employment status, and other characteristics. It is important to note that there were determining factors when we discuss unconditional wages. 43 The Economics of Tobacco Taxation and Employment in Indonesia (US$56.6) and IDR 556,994.6 (US$33.4) respectively. For comparisons, the average and median real monthly wage of workers in the economy was IDR 1,080,530 (US$80.6) and IDR 756,107 (US$56.4), respectively. The ratio of average and median wage to minimum wage was 1.06 and 0.78.20 Nevertheless, about half of tobacco manufacturing workers earned less than the minimum wage. Between 2011 and 2015, the average annual growth of real monthly wage of these workers was 4.23 percent. During the same period, the average annual growth of minimum wage was 7.4 percent while the average annual wage growth in the economy was 3.23 percent. Regression analysis showed that female production workers earned 25% lower wages than male production workers.21 One possible explanation for the male-female wage gap in the tobacco industry was the education gap between male and female tobacco workers. Female production workers in this sector completed 7.7 years of schooling while their male counterparts completed 8.4 years of schooling.22 We conjec- ture that male production workers worked in high-productivity tobacco manufacturing firms (machine operators), while female production workers worked in low-productivity tobacco manufacturing firms (hand rollers). However, our data did not have the relevant variables to estimate this conjecture. Note that the male-female education gap in the tobacco sector was wider than the male-female education gap in the economy. On aver- age, male workers completed 8.7 years of schooling while female workers completed 8.4 years of schooling. To investigate relative wages of tobacco workers, we included a dummy variable for the tobacco sector in the regression specification and we presented it in column 1 to 3 (Table 9). We found that tobacco workers earned higher wages than workers in other manufacturing workers, even those in comparable sectors such as the food and drink, garments, and textile industries. We found that, on average, tobacco work- ers earned about 15 percent higher wages than other manufacturing workers. 20 For the ratio, we used the average regional minimum wage of Central Java, East Java, DIY, and West Nusa Tenggara. 21 Annex I provides a detailed description of the econometrics model for this estimation. 22 These statistics suggest a gender segmentation in the tobacco manufacturing sector. In developing countries, female workers were more likely to be employed as informal workers while male workers were more likely to be employed as formal workers, thus gender segmentation (Chen, 2005, 2012) 44 // Employment in the Indonesian Tobacco Sector Table 9. Estimation results of the wage equation, 2001–2015 DEP. VARIABLE: LOG 1: 3: 4: 5: OF WAGE MANUF. WITH PRODUCTION TOBACCO INDUSTRY COMPARABLE WORKERS SECTORS ONLY 1 if female -0.276*** -0.248*** -0.304*** -0.088 (0.024) (0.026) (0.026) (0.069) Years of schooling 0.060*** 0.055*** 0.054*** 0.044*** (0.002) (0.002) (0.002) (0.004) 1 if production worker -0.315*** -0.366*** -0.191 (0.017) (0.028) (0.105) 1 if tobacco 0.155*** 0.162** (0.025) (0.061) Female x production -0.247** (0.082) Observations 152,377,119 69,493,078 131,981,606 6,437,039 R-squared 0.54 0.51 0.51 0.55 Year 2001-15 2001-15 2001-15 2001-15 Clustering of S.E. District District District District Controls Y Y Y Y Source: Calculated from the August National Labor Force Survey, 2001–2015. Productivity of tobacco sector workers, measured by the output per worker, was relatively low in comparison to productivity of workers in the comparable sector.23 In 2014, a typical worker in the medium and large tobacco industry produced IDR 104 million (US$7,761) worth of products annually (Panel B of Figure 9). For comparisons, a typical worker in the food and drink and textile industries produced IDR 265 million (US$19,776) and IDR 300 million (US$22,388) worth of products, respectively. However, there were variations within the tobacco sector. The productivity of workers in the kreteks industry was IDR 152 million (US$11,343), while the productivity of workers in the white cigarette industry was IDR 421 million (US$31,417).24 We could not distinguish the pro- ductivity of the SKT and SKM industries. However, we expected that the productivity of the SKT industry was far below the productivity of the SKM industry. Figure 10 shows that output and labor were positively correlated. However, this correlation masked important variations across manufacturing firms within the sector. 23 The Central Bureau of Statistics, Indonesia does not publish the quantity of goods produced in the Annual Survey of Manufacturing Industry. The Bureau published the value of output produced, which may include the value of goods, the value of electricity sold, the value of services sold, and the value of the difference in the inventory of intermediate goods. We used output in the text for simplicity.. 24 Note that we could not differentiate productivity of workers in the hand-rolled kreteks industry and that in the machine-made kreteks industry. 45 The Economics of Tobacco Taxation and Employment in Indonesia Figure 9. Productivity of workers, 2000–2014 PANEL A: ALL SCALES PANEL B: MEDIUM & LARGE fiRMS ONLY 150 150 300 300 134 238 238 268 119 119 200 195 155 178 98 142 100 136 89 122 144 104 78 100 96 108 105 72 76 92 71 72 73 70 74 72 66 69 83 67 65 67 72 in million rupiah 54 57 49 51 51 54 51 43 52 50 52 43 46 50 31 34 28 32 23 22 20 37 36 17 18 33 14 14 32 16 16 12 13 25 26 25 23 23 7 2010 2011 2012 2013 2014 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 tobacco food textile garmet tobacco food textile garmet Source: Calculated using data from ASMSI (2000–2014) and aggregated statistics of micro and small industry (2014). Figure 10. Relationship between output and labor in medium and large industry, 2000–2014 60,000 400 50,000 REAL OUTPUT, IN BILLION IDR TOTAL LABOR, IN THOUSANDS 40,000 350 30,000 300 20,000 250 10,000 200 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 real output total labor Source: Calculated using data from ASMSI (2000–2014). 46 // Employment in the Indonesian Tobacco Sector We estimated the output elasticity of labor demand for the cigarette manufac- turing sector at 0.160. This estimate implied that a one percentage decrease in output corresponded to a 0.160 percent decrease in employment in the cigarette manufacturing sector (Table 10). On the other hand, the estimate for the tobacco processing sector implied that a one percentage decrease in output corresponded to a 0.092 percent decrease in employment in the tobacco processing sector.25 Table 10. Estimation of own-wage and output elasticity of labor demand, 2000–2014 DEPENDENT FE, ALL FE, FE, FE, PRO- OLS, PRO- VARIABLE: LOG OF CIGARETTE KRETEKS CESSING CESSING EMPLOYMENT Log of wage -0.045*** -0.107*** -0.107*** -0.058*** -0.157*** (0.001) (0.013) (0.013) (0.007) (0.019) Log of output 0.157*** 0.160*** 0.162*** 0.092*** 0.177*** (0.003) (0.017) (0.017) (0.022) (0.043) N 275,189 4,135 4,031 6,471 1,557 Clusters 27,580 458 453 762 - Year 2000-14 2000-14 2000-14 2000-14 2010, 2011, 2014 Clustering of S.E. Firm-Level Firm-Level Firm-Level Firm-Level Robust SE Scale M&L M&L M&L M&L Mi & S Source: Estimated from the Annual Survey of Manufacturing Industry, 2000–2012, and the Annual Survey of Micro and Small Industry, 2010, 2011, and 2014. Note: ***, **, * indicate significance at the 1, 5, and 10 percent levels. The regression sample included an unbal- anced panel of manufacturing firms in Indonesia as explained in Annex II. Wage, output, and other nominal variables were adjusted by wholesale price index. The control variables included price-adjusted capital, price-ad- justed material expenditures, and price-adjusted energy expenditures, the share of production workers, scale of production, investment status, dummies for districts, dummies for industries, year dummies, and the interaction between industry and year dummies. The standard errors were clustered at the firm level. 25 We noted several advantages and drawbacks of using a fixed-effect model to estimate output elasticity of labor demand. First, a fixed-effect model accounted for firm-specific unobserved heterogeneity which may bias the estimate of the elasticity. On the other hand, a fixed-effect model may not account for a possible simultaneity bias between output and labor demand. 47 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO V LENCE PREVA AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 5 SIMULATIONS ON THE EFFECTS O OF RAISING CIGARETTE TAXES ON EMPLOYMENT N To estimate the impact of raising taxes on employment, we needed to estimate the number of workers in each production tier. Given that available data did not allow to directly identify the number of workers in each tier, we used secondary data and applied a few assumptions to estimate the number of workers in each tier. First, we obtained O the number of factories by cigarette type and tier using 2015 data from Dwi Kurnaini (2016); secondly, we used the average workers per firm in 2011 reported in Tobacco Control Support Center and IAKMI (2014). These are shown in Column B and C of Table 11, respectively; thirdly, we calculated the estimated total workers per tier by multiplying the number of factories and the average workers per firm; and finally, we summed these numbers by subindustries to obtain total workers in the kretek and white cigarette indus- tries, which are shown in Column D of Table 11. Note that estimated total workers in kretek and white cigarette industries were consistent with the actual total workers.26 Next, we calculated the estimated share of workers in each tier to total employment in the kreteks and white cigarette industries, and we show these shares in Column E. Using these shares, we calculated the implied number of workers by multiplying the shares with the actual number of workers in the kretek and white cigarette industries calculated using SIMK and SI. We show the implied number of workers that we used for simulations in Column F of Table 11. L 3 25 However, we acknowledge that the estimated total workers in the white cigarette industry was underestimated. 2 49 The Economics of Tobacco Taxation and Employment in Indonesia Table 11: Estimated number of workers in each tax tier TYPE A: B: C: D: E: F: TIER NUMBER OF WORKERS ESTIMATED ESTIMATED IMPLIED FACTORIES PER FIRM NUMBER OF SHARE (%) NUMBER OF (2015) (2011) WORKERS WORKERS (2014 ESTI- MATE) SKM I 14 85 1,190 0.38 1,167 IIA 84 85 7,140 2.28 7,005 IIB 148 85 12,580 4.01 12,342 SKT IA 1 664 664 0.21 651 IB 15 664 9,960 3.17 9,771 IIA 6 664 3,984 1.27 3,909 IIB 15 664 9,960 3.17 9,771 IIIA 88 664 58,432 18.62 57,326 IIIB 316 664 209,824 66.88 205,851 SPM I 1 31 31 3.85 408 IIA 7 31 217 26.92 2,853 IIB 18 31 558 69.23 7,337 Tobacco n/a 58,294 352,086 processing Estimated total workers, kretek 313,734 Estimated total workers, white 806 Actual total workers, SKM & SKT (2014) 307,793 Actual total workers, SPM (2014) 10,598 Actual total workers, processing (2014) 352,086 Source: We obtained the numbers of factories in 2015 from Dwi Kurnaeni (2016), the average workers per firm from Tobacco Control Support Center—IAKMI (2014), and the actual total workers in each tobacco subindustry from the 2014 SIMK and SI. We observed that most kretek workers worked in the small kretek firms. Specifically, small SKT firms employed about 268,000 workers or 86 percent of total kretek workers in Indonesia. On the other hand, workers in SKM firms employed only 6.7% of total kretek workers in Indonesia. The World Bank has been providing inputs for the tobacco tax excise reform since 2015, in which the reform proposal was to streamline the excise tax structure and increase disproportionately the average excise tax rate across tiers. Specifically, tiers with lower cigarette taxes would experience higher tax increase. Under the proposed pol- icy, the number of tiers was reduced by half to just 6 tiers. For example, there are currently five tiers for SKT and the policy proposes to reduce the tiers to just two. The proposed reform would reduce cigarette demand by 10.4 percent, but increase government revenue by 8.4%. We show our prediction on the effects of raising ciga- rette taxes on government revenue in Table 13. Results show that the consumption of tier 1 SKM is predicted to decrease by 10 billion sticks but the revenue is predicted to increase by IDR 7,262 billion. Overall, higher cigarette taxes are expected to increase government revenue by IDR 12,875 billion. 50 // Simulations on the Effects of Raising Cigarette Taxes on Employment Table 12. Proposed cigarette tax increase and new cigarette excise tariffs TYPE A: B: C: D: F: H: TIER CONSUMPTION EXCISE BASE PROPOSED NEW (BILLION OF TARIFF 2017 REVENUE TAX TARIFFS STICKS) (RUPIAH) (BILLION INCREASE (RUPIAH) RUPIAH) (%) SKM I 212 530 112,360 12 594 IIA 17 365 6,205 40 511 IIB 21 335 7,035 53 511 SKT IA 13 345 4,485 16 400 IB 40 265 10,600 51 400 IIA 5 165 825 21 200 IIB 5 155 775 29 200 IIIA 5 100 500 100 200 IIIB 7 80 560 150 200 SPM I 16 555 8,880 10 611 IIA 2 330 660 30 429 IIB 2 290 580 48 429 Total 345 153,465 Source: World Bank team calculation (2015–2017). Table 13. The effects of raising cigarette taxes on government revenue TYPE A: B: E: G: H: I: G: CHANGE TIER CONSUMP- % CHANGE ESTIMATED NEW ESTIMATED IN ESTI- TION IN CONSUMP- TARIFFS REVENUE MATED (BILLION OF DEMAND TION (RUPIAH) (BILLION REVENUE STICKS) RUPIAH) (BILLION RUPIAH) SKM I 212 -4.94 202 594 119,622 7,262 IIA 17 -16.48 14 511 7,255 1,050 IIB 21 -21.84 16 511 8,388 1,353 SKT IA 13 -6.59 12 400 4,860 375 IB 40 -21.01 32 400 12,643 2,043 IIA 5 –8.65 5 200 912 87 IIB –11.95 200 5 4 880 105 IIIA –41.20 200 5 3 588 88 IIIB –61.80 200 7 3 535 -25 SPM I 16 –5.02 15 611 9,278 398 IIA 2 –15.06 2 429 729 69 IIB 2 –24.10 2 429 652 72 Total 345 309 166,340 12,875 Source: Authors' estimates. 51 The Economics of Tobacco Taxation and Employment in Indonesia The proposed cigarette tax increase is predicted to reduce tobacco employment by 32,132. We summarized the effects of the higher cigarette taxes on employment in the cigarette industry in Table 14. The largest employment impact is expected to happen in the SKT firms with a reduction of 8.6 percent or about 24,710 posts. Most of the impact will happen in the small-scale SKT firms (20,355 job losses) given the fact those are more labor intensive within the industry. Note that we assume identical output elasticity of employment in the SKT and SKM firms owing to data limitations. Unfortunately, we do not have a strong prior statistic on the difference in the elasticities between SKT and SKM firms owing to a lack of information from previous studies. Nevertheless, if output elas- ticity of employment is relatively more elastic in the SKT firms, then we should expect a higher negative impact on future employment in the SKT firms. It is important to note that these analyses qualified as the effects of raising ciga- rette taxes on gross employment. Consumers may shift their consumption to other goods and services when prices of cigarettes increase. Over time, a higher demand of other goods and services leads to a higher demand in labor. Workers laid off from the cig- arette industry due to higher cigarette taxes can be employed in another sector. Former cigarette industry workers may fulfill the higher labor demand. However, a subsequent study is required to identify these sectors, and whether former cigarette industry workers have the skills to work in these sectors. While it is more ideal to evaluate the effects on net employment, this study provides an estimate to the government of the number of workers who would need immediate income support and training programs during the transitional period. 52 // Simulations on the Effects of Raising Cigarette Taxes on Employment Table 12: Simulations on the effects of raising cigarette taxes on employment TYPE A: B: C: D: E: F: G: H: TIER PROPOSED NUMBER PRICE CHANGE LABOR CHANGE LOSS TAX OF ELASTIC- IN ELASTIC- IN OF INCREASE WORKERS ITY OF DEMAND ITY OF EMPLOY- EMPLOY- (%) (2014 DEMAND (%) OUTPUT MENT MENT ESTIMATE) (%) SKM I 12 1,167 -0.412 -4.94 0.160 -0.79 -9 IIA 40 7,005 -0.412 -16.48 0.160 -2.64 -185 IIB 53 12,342 -0.412 -21.84 0.160 -3.49 -431 Total: A -625 SKT IA 16 651 -0.412 -6.59 0.160 -1.05 -7 IB 51 9,771 -0.412 -21.01 0.160 -3.36 -329 IIA 21 3,909 -0.412 -8.65 0.160 -1.38 -54 IIB 29 9,771 -0.412 -11.95 0.160 -1.91 -187 IIIA 100 57,326 -0.412 -41.20 0.160 -6.59 -3,779 IIIB 150 205,851 -0.412 -61.80 0.160 -9.89 -20,355 Total: B -24,710 SPM I I 408 -0.502 -5.02 0.160 -0.80 -3 IIA IIA 2,853 -0.502 -15.06 0.160 -2.41 -69 IIB IIB 7,337 -0.502 -24.10 0.160 -3.86 -283 Total: C -355 Tobacco 352,086 -19.89 0.092 -1.83 -6,442 processing: D Total loss of -32,132 employment (A+B+C+D) Source: Authors' estimates. 53 SMOKING HAS BEEN A MAJOR CONTRIBUTOR O DISEASE BURDEN INDONESIA. SMO V LENCE PREVA AGE INDIVIDUAL HAS EXCEEDED 3 PERCENT SINCE 2 5 CONCLUSION O This report provided an overview of the main economic issues related to tobacco taxation and employment to inform current debate over tobacco tax reform in Indonesia. Overall, there is consistent global evidence suggesting that raising tobacco N taxes has a positive effect on government revenues and a small negative effect on employment in the tobacco sector. However, research has demonstrated that the job losses in the tobacco sector (gross effect) are usually compensated with job creation in the other sectors (net effect).27 This report used data from the Central Bureau of Statistics to estimate trends in employ- O ment and output in the tobacco sector and estimated the potential impact on tobacco employment (gross effect) from raising cigarette taxes in Indonesia. The share of tobacco employment to total manufacturing and economy-wide employment was quite low at 5.13 and 0.60 percent in 2014, respectively. Tobacco jobs were heavily concentrated in Central Java, East Java, and West Nusa Tenggara—about 94% of tobacco manufacturing workers and about 91% of tobacco farmers were concen- trated in these three provinces. In these provinces, several districts were quite dependent on the tobacco sector. For example, the share of tobacco employment to local employ- ment was 30 percent in Kudus, 27.6 percent in Temanggung, and 26 percent in Kediri. Any effect from tobacco taxation to the tobacco sector affected these districts more. Most tobacco manufacturing workers were female and production workers. Additionally, there was a considerable male-female wage gap in the tobacco industry (female production workers earned 25% lower wages than male production workers). We found that about 43 percent of tobacco households were poor. The World Bank/ACS survey among kretek workers provided more details on the livelihoods of these workers L and how they would be affected by an increase in cigarette taxes (World Bank, 2017b). We predict that raising cigarette taxes by an average of 47% and simplifying the cigarette tax structure to 6 tiers will reduce cigarette demand by 10.4 percent, increase government revenue by 8.4 percent, and reduce gross employment in tobacco manufacturing sector by 4.79 percent. That means that a reduction of 32,132 3 tobacco manufacturing jobs, most of them in the SKT industry (24,710 less jobs). Given the additional revenues government will obtain with the reform (IDR 12,875), there is 27 For Indonesia, Ahsan and Wiyono (2007) estimated positive net effects of 84,340 jobs (25 for a percent tax increase), 140,567 jobs (50 percent tax increase), and 281,135 jobs (100 percent tax increase) (Ahsan and Wiyono, 2007). 2 55 The Economics of Tobacco Taxation and Employment in Indonesia scope to implement measures to reduce the impact on the tobacco workers’ livelihoods (such as cash transfers or expanded access to social safety nets) or to find alternative occupations for the workers affected (retraining programs, educational grants, etc.). In line with the other reports of the World Bank/ACS Indonesia Tobacco Studies, this report recommends: • For kretek hand-rollers o First, the most vulnerable groups in the affected population who would need immediate income support in the event of job loss include the workers who are less educated, older, heads of their households, and who contribute a significant pro- portion of total household income from kretek rolling. The government can provide income support to these workers with less than 2% of the revenue gained from a tax increase; o Second, the additional revenue or part thereof can be allocated for re-skilling and redeployment of laid-off kretek workers to smooth their transition to an alternative source of employment in the short to medium term. • For (tobacco and clove) farmers: 1) The government should help to improve supply chains and value chains for other goods in tobacco-growing areas. Many former tobacco farmers are making a better living growing other common, locally grown crops (e.g., corn, sweet potato, and green vegetables), an outcome that could be further enhanced with even small investments by governments in improved supply chains for these products. Results from the World Bank/ACS survey suggest that current tobacco farmers are already growing many of these crops, so it is an issue of shifting their factors of production to maximize economic opportunity. 2) The government should help to facilitate access to credit for tobacco farmers. Greater access to capital through improved credit schemes could help to improve the possibilities for tobacco farmers to cultivate other crops and/or develop other nonagricultural economic enterprises. Access could be in the form of grants or low-interest loans to farmers willing to move away from tobacco cultivation. 56 // Conclusion REFERENCES Ahsan, A. (2011) Cigarette Tax and Price: Affordability and Impacts on Consumption and Rev- enue. Ahsan, A. (2015) Ekonomi Tembakau di Indonesia. doi: 10.1596/978-0-821-36179-5/Chpt-46. Ahsan, A., and Wiyono, N. (2007) An Analysis of the Impact of Higher Cigarette Prices on Employment in Indonesia. Bader, P., Boisclair, D., and Ferrence, R. (2011) “Effects of tobacco taxation and pricing on smoking behavior in high risk populations: a knowledge synthesis,” International journal of environmental research and public health. Multidisciplinary Digital Publishing Institute (MDPI), 8(11), pp. 4118–39. doi: 10.3390/ijerph8114118. Blakely, T., Cobiac, L. J., Cleghorn, C. L., Pearson, A. L., van der Deen, F. S., Kvizhinadze, G., Nghiem, N., McLeod, M., and Wilson, N. (2015) “Health, Health Inequality, and Cost Impacts of Annual Increases in Tobacco Tax: Multistate Life Table Modeling in New Zealand,” PLOS Medicine. Edited by T. E. Novotny. Public Library of Science, 12(7), p. e1001856. doi: 10.1371/journal.pmed.1001856. Burns D, Lee L, Shen L et al. (1997). Cigarette smoking behavior in the United States. In: Burns D, Garfinkel L, Samet J, eds., Changes in Cigarette-Related Disease Risks and Their Implications for Prevention and Control. Smoking and Tobacco Control Monograph N°2. Bethesda, MD, U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute, NIH Pub No. 97-4213 13-112. Chaloupka, F.J., Yurekli, A. and G.T. Fong (2012). Tobacco taxes as a tobacco control strat- egy. Tobacco Control. 21:172e180. doi:10.1136/tobaccocontrol-2011-050417 Chen, M. A. (2005) “Rethinking the Informal Economy,” 2005(30). Chen, M. A. (2012) “The informal economy: Definitions, theories and policies,” Women in informal economy globalizing and organizing: WIEGO Working Paper, 1. Contreary, K. A., Chattopadhyay, S. K., Hopkins, D. P., Chaloupka, F. J., Forster, J. L., Grimshaw, V., Holmes, C. B., Goetzel, R. Z., and Fielding, J. E. (2015) “Economic Impact of Tobacco Price Increases through Taxation: A Community Guide Systematic Review,” American Journal of Preventive Medicine, 49, pp. 800–808. doi: 10.1016/j.amepre.2015.04.026. De Loecker, J., and Goldberg, P. K. (2014) “Firm Performance in a Global Market,” Annual Review of Economics, 6(1), pp. 201–227. doi: 10.1146/annurev-economics-080113-104741. Dwi Kurnaini, Z. (2016) Kebijakan Cukai Hasil Tembakau. Available at: http://www.fkm.ui.ac. id/wp-content/uploads/2016/10/KEMENKEU-Paparan-Kebijakan-Cukai.pdf (Accessed: November 1, 2016). 57 The Economics of Tobacco Taxation and Employment in Indonesia Fuchs, A., and Meneses, F. (2017) Are Tobacco Taxes Really Regressive? Evidence from Chile. Washington DC. Goodchild, M., Perucic, A.-M., and Nargis, N. (2016) “Modelling the impact of raising tobacco taxes on public health and finance,” Bulletin of the World Health Organization. World Health Organization. Hamermesh, D. S. (1986) “The Demand for Labor in the Long Run,” in Ashenfelter, O. and Layard, R. (eds.) Handbook of Labor Economics. Elsevier Science Publishers. Hidayat, B., and Thabrany, H. (2010) “Cigarette smoking in Indonesia: examination of a myopic model of addictive behaviour,” International journal of environmental research and public health. Multidisciplinary Digital Publishing Institute (MDPI), 7(6), pp. 2473–85. doi: 10.3390/ijerph7062473. IAKMI—Tobacco Control Support Center (2014) Buku Bunga Rampai—Fakta Tembakau dan Permasalahannya [Anthologies Books: Facts about Tobacco and Its Problems]. Jakarta Pusat. Available at: http://www.tcsc-indonesia.org/wp-content/uploads/2016/06/ Buku-Fakta-Tembakau-2014__Web-Version.pdf (Accessed: November 1, 2016). IARC Handbooks of Cancer Prevention, Tobacco Control, Vol. 14: Effectiveness of Tax and Price Policies for Tobacco Control (2011: Lyon, France) Iglesias, RM (2016). Increasing excise taxes in the presence of an illegal cigarette market: the 2011 Brazil tobacco tax reform. Rev Panam Salud Publica. 2016;40(4):243–9. Institute for Health Metrics and Evaluation (IHME), (2017), Indonesia. Available at: http:// www.healthdata.org/indonesia Jha, P., and Chaloupka, F. (2012) Tobacco Control in Developing Countries. Edited by P. Jha, and F. Chaloupka. Washington DC. Jha P, Jacob B, Gajalakshmi V et al. RGI- CGHR Investigators (2008). A nationally represen- tative case-control study of smoking and death in India. N Engl J Med, 358:1137–1147. doi:10.1056/NEJMsa0707719 PMID:18272886 John, R. M., Rao, M. G., Deshpande, R. S., Selvaraj, S., Rao, R. K., Moore, J., Sengupta, J., Cha- loupka, F. J., and Jha, P. (2010) The Economics of Tobacco and Tobacco Taxation in India. Paris. Kosen, S., Hardjo, H., Kadarmanto, Sinha, D. N., Palipudi, K. M., Wibisana, W., and Tarigan, I. (2012) Global Adult Tobacco Survey: Indonesia Report. Kristina, S. A., Endarti, D., Prabandari, Y. S., Ahsan, A., and Thavorncharoensap, M. (2015) “Burden of Cancers Related to Smoking among the Indonesian Population: Premature Mortality Costs and Years of Potential Life Lost,” Asian Pacific journal of cancer prevention : APJCP, 16(16), pp. 6903–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/26514465 (Accessed: March 5, 2017). 58 // Conclusion Ministry of Industry (2009) Roadmap Industry Pengolahan Tembakau. Available at: https:// www.bps.go.id/Subjek/view/id/9. Nasrudin, R., Trialdi, L., Hartono, D., and Ahsan, A. (2013) Tobacco Economic of Indonesia: Poor Households’ Spending Pattern, Tax Regressivity and Economic Wide Impact of Tax Simpli- fication. 2. Available at: http://econ.fe.ui.ac.id/workingpage (Accessed: February 19, 2017). (NCI-WHO) National Cancer Institute and World Health Organization (2017) NCI Tobacco Control Monograph Series 21: The Economics of Tobacco and Tobacco Control, WHO. World Health Organization. Available at: http://www.who.int/tobacco/publications/economics/ nci-monograph-series-21/en/ (Accessed: September 5, 2017). OECD (2017), Exchange rates (indicator). doi: 10.1787/037ed317-en (Accessed on 25 Sep- tember 2017). Raghunathan, T. E., Lepkowski, J. M., Van Hoewyk, J., Solenberger, P., and van Hoewyk, J. (2001) “A Multivariate Technique for Multiply Imputing Missing Values Using a Sequence 001). Available at: http://citeseerx.ist. of Regression Models,” Survey Methodology, 27(12­ psu.edu/viewdoc/download?doi=10.1.1.405.4540&rep=rep1&type=pdf (Accessed: May 5, 2017). Salma K, Chiang CY, Enarson DA, Hassmiller K, Fanning A, Gupta P, et al. Tobacco and tuberculosis: a qualitative 
systematic review and meta-analysis. Int J Tuberc Lung Dis. 2007;11(10):1049-61. Setyonaluri, D., Adioetomo, S. M., Barber, S., and Ahsan, A. (2008) Tobacco Economics in Indonesia. Sumartono, W., Sirait, A. M., Holy, M., and Thabrany, H. (2007) “Smoking and Socio-Demo- graphic Determinant of Cardiovascular Diseases among Males 45+ Years in Indonesia,” Int. J. Environ. Res. Public Health International Journal of Environmental Research and Public Health, 8, pp. 528–539. doi: 10.3390/ijerph8020528. Szabo, A., Lazar, E., Burian, H., Rogers, T., Foley, K., Abram, Z., Meghea, C., Ciolompea, T., and Chaloupka, F. J. (2016) The Economics of Tobacco and Tobacco Taxation in Romania. Tirgu Mures. Tobacco Control Support Center - IAKMI (2014) Buku Bunga Rampai - Fakta Tembakau dan Permasalahannya [Anthologies Books: Facts about Tobacco and Its Problems]. Jakarta Pusat. Available at: http://www.tcsc-indonesia.org/wp-content/uploads/2016/06/Buku-Fakta-Tem- bakau-2014__Web-Version.pdf (Accessed: November 1, 2016). USAID Health: Infectious Diseases, Tuberculosis, Countries, Indonesia." U.S. Agency for Inter- national Development. May 2009. Web. 07 Dec. 2010. http://www.usaid.gov/our_work/global_ health/id/tuberculosis/countries/asia/indonesia_profile.html. 59 The Economics of Tobacco Taxation and Employment in Indonesia U.S. Department of Health and Human Services (2014). The health consequences of smoking—50 years of progress: a report of the Surgeon General. Atlanta: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Cen- ter for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2014. Available from: http://www.surgeongeneral.gov/library/reports/50-years-of-progress. 
 U.S. National Cancer Institute and World Health Organization, 2016, The Economics of Tobacco and Tobacco Control. National Cancer Institute Tobacco Control Monograph 21. NIH Publication No. 16-CA-8029A. Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute; and Geneva, CH: World Health Organization. World Bank (2017a). The Economics of Kretek Rolling in Indonesia. Indonesia Tobacco Employment Studies. World Bank, Washington, DC. World Bank (2017b). The Economics of Tobacco Farming in Indonesia. Indonesia Tobacco Employment Studies. World Bank, Washington, DC. World Bank (2017c). The Economics of Clove Farming in Indonesia. Indonesia Tobacco Employment Studies. World Bank, Washington, DC. World Health Organization. WHO global report: mortality attributable to tobacco. Geneva: World Health Organization; 2012. Available at: http://apps.who.int/iris/bitstr eam/10665/44815/1/9789241564434_eng.pdf. 
 World Health Organization (2008). WHO report on the global tobacco epidemic 2008: the MPOWER package. Geneva, World Health Organization. World Health Organization, Noncommunicable Diseases (NCD) Country Profiles, Geneva: World Health Organization; (2014). Available at: http://www.who.int/nmh/countries/idn_en.pdf. World Health Organization, Fact sheet on tobacco, Geneva: World Health Organization; (2017a). Available at: http://www.who.int/mediacentre/factsheets/fs339/en/. World Health Organization, Fact sheet on Noncommunicable diseases; Geneva: World Health Organization; (2017b). Available at: http://www.who.int/mediacentre/factsheets/fs355/en/. Zhang, P. (2002). Understand and Evaluate the Impact of Tobacco Control Policies on Employment, Word Bank Tobacco Control Toolkit, Tool 5, World Bank. 60 // Annex Annex I. Estimation of the wage Wage equation We use the basic Mincer model to estimate the wage equation. Let wist indicates the real monthly wage of worker i in industry s at time t. The specification of the model is: logwist = b0 + b1educi + b2femalei + b3productioni + BXist + uist where educ indicates the years of schooling completed, female takes a value of one for female workers, and production takes a value of one for production workers. The vector X includes age, age squared, a dummy for urban status, dummies for employment status (employee, casual workers, etc.), dummies for type of work (managers, administrative, sales, services, etc.), dummies for industry, dummies for districts, and year dummies. We cluster the standard errors for the district level and we use the sample weights for the estimations. We estimate the model using the August National Labor Force Survey from year 2000 to 2015. The August National Labor Force Survey data is an individual-level data system that includes workers’ characteristics and labor market variables. The availability of survey data for multiple years allow us to construct a pooled cross cross-section data for the estimation. We present the summary statistics of variables used in the estimation in Table 815. Table 15: Summary statistics, National Labor Force Survey Sample VARIABLE MEAN SD MIN MAX N Log of wage 13.416 0.802 7.853 18.449 152,377,119 1 if female 0.398 0.490 0 1 152,377,119 Years of schooling completed 8.959 3.436 0 16 152,377,119 1 if production worker 0.866 0.340 0 1 152,377,119 Industry code 21.644 6.540 15 37 152,377,119 Type of work 6.501 1.349 1 8 152,377,119 Age 33.321 11.432 15 98 152,377,119 1 if lives in urban area 0.678 0.467 0 1 152,377,119 Employment Status 3.719 1.209 1 6 152,377,119 Hours worked in a month 178.241 50.421 0 392 152,377,119 Province ID 33.968 11.041 11 94 152,377,119 District ID 400.927 139.346 1 1001 152,377,119 Year 2008.514 4.385 2001 2015 152,377,119 Source: Calculated using data from the August 2001–2015 National Labor Force Survey. Notes: Sample weights are used for the estimation. 61 The Economics of Tobacco Taxation and Employment in Indonesia Annex II. Sample Selection Procedure for the Manufacturing Industry Data We do not obtain data on the quantity of production outputs and inputs such as capital, machine, materials, and energy. This is quite common when using survey data. Thus, we use sales data for the outputs and expenditure data for the inputs. A study shows that estima- tions of production functions using sales and expenditure data are sensible (De Loecker and Goldberg, 2014). We also face thorny issues of unbalanced panel data and missingness in our data. In this section, we discuss the procedure that we use to deal with the unbalanced panel data and missingness. Our data include 48,739 manufacturing firms with 332,360 firm-level observa- tions from 2000 to 2014 (Table 16). However, many firms have gap observations with undis- closed reasons which leads to unbalanced panel data. Table 16: Selected sample, Annual Survey of Manufacturing Industry INITIAL SAMPLE SELECTED SAMPLE SAMPLE KEPT (IN PERCENTAGE) All Manufacturing Industry Number of firms 48,739 27,580 56.69 Number of observations 332,360 275,280 82.83 Tobacco Industry Number of firms 2,322 1,208 52.02 Number of observations 14,101 11,270 79.92 Kretek Industry Number of firms 764 453 59.29 Number of observations 4,757 4,031 84.74 Source: Calculated using the 2000–2014 Annual Survey of Manufacturing Industry. Thus, we create a selection criteria based on observation gaps within each firm and the length of observations. First, we drop firms with more than three unbroken chains of observations. For example, we keep firms with three unbroken chains of observations between 2000–2006, 2009–2011, and 2013–2014. Second, we drop firms with less than 5 total observations in the data. As shown in Table 10, we keep about 57 percent of initial firms but these firms account for 83 percent of initial observations. This selection proce- dure is important because we rely on a fixed-effects model to estimate the output elastic- ity of labor demand. 62 // Annex The second issue that we face is data missing, particularly in terms of capital, machine, energy, and materials. In the data, several firms only have one, or just few, observations of capital and machine across periods. First, we assume that firms without any observation of capital and machines do not use capital and machines in their production processes. Thus, we impute zeros to the missing observations. Second, for firms with one or just a few observations of capital and machines, we impute the maximum value to the missing observations. We use univariate imputation using linear regressions for the energy and materials variables (Raghunathan et al., 2001). We choose linear regressions for the imputations because energy and materials are continuous variables. For the imputation, we regress the variables on log of output, log of wage, log of production workers, share of produc- tion workers, a dummy for firm scale, year dummies, and industry dummies. We simulate the imputations for 30 times for each variable. We then use the 30 sets of imputed vari- ables to estimate the output elasticity of labor demand. 63 The Economics of Tobacco Taxation and Employment in Indonesia Annex III. Estimation of Output Elasticity of Labor Demand We assume that tobacco manufacturers follow a constant-elasticity of substitution (CES) production function. Given this production function, we can derive the labor demand function for the estimation of the output elasticity of labor demand. The derivation of the labor demand function is beyond the scope of this paper but it is available in previous publications (Hamermesh, 1986). Let eist be the number of workers that firm i in industry s employ at time t. The specifica- tion of the model is: logeist = b0 + b1logwist + b2logqist + BXist + uist where w indicates wage rate, and q indicates output, proxied by the value of output. The vector X includes capital, material expenditures, energy expenditures, the share of pro- duction workers, scale of production, investment status, dummies for districts, dummies for industries, year dummies, and the interaction between industry and year dummies. All nominal variables are adjusted with the industry-specific wholesale price index. The standard errors are clustered at the firm level to account for autocorrelations of the unobservables within each firm. We use the 2001 to 2014 Annual Survey of Manufacturing Industry published by the Indo- nesian Central Bureau of Statistics to estimate the model. The survey collects firm-level data including, but not limited to, production, employment, capital, material and energy expenditures, ownership status, and type of industry. Note that the survey data allow us to construct an unbalanced panel data of firms. In Table 17, we present summary statistics of the variables used in the estimation. 64 // Annex Table 17: Summary statistics, Annual Survey of Manufacturing Industry VARIABLE MEAN SD MIN MAX N Log of labor 4.046 1.228 0 10.893 275,280 Log of wage rate 12.541 1.179 –1.531 19.475 275,280 Log of output 14.501 2.224 6.480 24.846 275,280 Log of capital 12.480 4.991 –0.550 29.692 275,280 Log of material expenditures 13.376 0.006 –0.106 24.230 275,280 Log of energy expenditures 9.917 0.005 –0.448 21.553 275,280 % of production workers 84.350 16.114 0.370 100 275,280 Investment status 2.456 0.910 0 3 275,280 Industry 17.732 7.057 10 33 275,280 District unique ID 199.659 67.205 1 444 275,280 Year 2008 4 2000 2014 275,280 Source: Calculated using the 2000–2014 Annual Survey of Manufacturing Industry.. Notes: Statistics for log of material and energy expenditures are estimated using univariate imputation methods with 30 replications. 65 The Economics of Tobacco Taxation and Employment in Indonesia Annex IV. Estimation of Price Elasticity of Demand We estimate price elasticity of cigarette demand using the 2015 National Socioeconomic Survey conducted by the Indonesian Central Bureau of Statistics. Let qi be the number of cigarette sticks consumed by individual i. The specification of the model is: logqist = b0 + b1logpricei + b2kreteki + b3logpricei•kreteki + BXist + uist where price indicates the price per cigarette stick. The price is estimated by dividing total cigarette expenditure to total cigarette sticks consumed. We include a dummy for kretek cigarettes and the interaction between log of price and this dummy. The interaction terms allow us to investigate whether price elasticities of demand between kreteks and white cigarettes differ significantly. The vector X includes age, a dummy for female, a dummy for marital status, a dummy for urban status, dummies for highest schooling completed, a dummy that indicates working status, a dummy that indicates poverty, dummies for occupation sector, dummies for dis- tricts, and a dummy for possession of health insurance. We include per-capita expenditure in the vector X which is a proxy for income. We use the sample weights for the estimations. Table 18 presents the estimate of the price elasticity of cigarette demand in 2015. The estimates show that price elasticity of demand significantly differs by cigarette type. As implied by the estimates in column 1 of Table 18, the price elasticity of demand for kreteks is about –0.42, while the price elasticity of demand for white cigarettes is about –0.51. These estimates are quite close to previous estimates which range from –0.29 to –0.67 (Setyonaluri et al., 2008). 66 // Annex Table 18: Estimation of price elasticity of cigarette demand, 2015 DEPENDENT ALL ALL ALL KRETEKS KRETEKS WHITE VARIABLE: LOG KRETEKS W/ FILTER W/O OF QUANTITY FILTER (CIGARETTE STICKS) Log of price -0.421*** -0.510*** -0.416*** -0.476*** -0.376*** -0.502*** (0.000278) (0.00110) (0.000285) (0.000391) (0.000385) (0.00103) Log of price x 0.0945*** kreteks (0.00112) N 53,584,645 53,584,645 50,113,104 29,553,444 25,065,756 5,477,027 R-squared 0.158 0.158 0.160 0.170 0.154 0.185 Year 2015 2015 2015 2015 2015 2015 District-FE Y Y Y Y Y Y Controls Y Y Y Y Y Y Source: Calculated using data from the 2015 National Socioeconomic Survey. Note: ***, **, * indicate significance at the 1, 5, and 10 percent level. The control variables include per-capita expenditure, age, a dummy for female, a dummy for marital status, a dummy for urban status, dummies for high- est schooling completed, a dummy that indicates working status, a dummy that indicates poverty, dummies for occupation sector, dummies for districts, and a dummy for possession of health insurance. Sample weights are used for the estimations. 67 The Economics of Tobacco Taxation and Employment in Indonesia Annex V. Simulation of the Effects of Raising Cigarette Prices on Employment The simulation of the effects of raising cigarette prices on employment in the tobacco indus- try consists of three steps. The first step is an estimation of the change in cigarette consump- tion followed by an estimation of the change in output. Lastly, we estimate the change in employment in the tobacco industry. We discuss these steps in more detail below. Step 1: Estimation of the change in cigarette consumption We use the estimates of price elasticities of demand to predict the change in cigarette con- sumption. Let qj be consumption of type j cigarette, Eqp,j be the price elasticity of demand for type j cigarette, and Pj be the price of type j cigarette. The estimated change in consumption is: %∆Cj = Eqp,j ˙ %∆Pj Step 2: Estimation of the change in the value of output (sales) We do not have evidence about the elasticity between cigarette consumption and produc- tion, and between cigarette consumption and the value of output. Therefore, we assume that changes in consumption will be reflected in equivalent changes in cigarette production and the value of output (sales) of cigarette firms. This is a strong assumption but it is neces- sary given the lack of data about consumption and production. Let Qj be the output of firms that produce type j cigarette. Then: %∆Cj = %∆Qj We also need to make an additional assumption that changes in cigarette consumption will be reflected in equivalent changes in output of tobacco processing firms. We use price elasticity of demand for all types of cigarette (column 1 of Table 18) to estimate the change in output in the tobacco processing firms. Step 3: Estimation of the change in employment by industry Lastly, we can estimate the change in employment by using our estimates of output elasticity of labor. Let EeQ be the estimated output elasticity of labor. Then, the change in employment is: %∆Ej = EeQ,j ˙ %∆Qj Given the base number of employment, we can calculate the change in terms of number of workers. Specifically, let E1 be the number of workers before the price change. The change in the number of workers is: %∆Ej /100 ˙ E1 68 // Annex 69 The Economics of Tobacco Taxation and Employment in Indonesia