Essays on the economics of tobacco and alcohol control policy in Kenya

This thesis uses data from Kenya to contribute to the literature on tobacco and alcohol control policies in low and middle-income countries. The thesis uses the two most recent household and budget surveys (2005/6 and 2015/16), to examine some of the effects of the tobacco and alcohol control policies that were implemented in Kenya between 2005 and 2015. Chapter 2 considers the impact of consumption and taxation of tobacco and alcohol on household spending patterns. An instrumental variable approach is used in the estimation of the difference in spending patterns, between tobacco-consuming (alcohol-consuming) households and households that do not consume tobacco (alcohol). Following the precedent of some other studies, the adult sex ratio is used as an instrumental variable for the tobacco and alcohol consumption status of households. However, the adult sex ratio may not meet the exclusion restriction. In order to address this concern, I change the specification of the instrumental variable and relax the exclusion restriction. As a result, the upper and lower bounds of the difference in expenditure shares between households that consumed tobacco (alcohol) and the households that did not consume tobacco (alcohol) are estimated. A natural experiment involving tobacco and alcohol taxes occurred during the data collection period of one of the household surveys: the excise tax on tobacco and alcohol was increased during the data collection phase. A matched difference-in-differences (MDID) technique is used to estimate the implication of a tobacco (alcohol) tax increase on household spending patterns. The pseudo-panel generated from MDID also provided a new way of controlling for possible endogeneity arising from time-invariant unobservable variables. Therefore, MDID is used as a new approach, for comparing household spending patterns of tobacco-consuming (alcohol-consuming) households with those of households that do not consume tobacco (alcohol). The price and non-price tobacco-control policies that were implemented between 2005/6 and 2015/16 contributed to a decrease in household tobacco use prevalence. However, alcohol-control policies implemented over the same period did not successfully reduce the overall prevalence of alcohol consumption among Kenyan households. Tobacco- and alcohol-consuming households were found to spend less on education, energy, rent, healthcare, and food items. MDID results confirmed that tobacco and alcohol consuming households had lower expenditure shares on items necessary for human capital development. The increase in tobacco taxes did not have an impact on household spending patterns among tobacco-consuming households. However, an increase in taxes on alcohol led to further crowding out of expenditure on fruits. Chapter 3 uses the risk of child malnutrition in Kenya, to examine the effectiveness of tobacco and alcohol control policies, in reducing tobacco and alcohol consumption prevalence among vulnerable households. Past studies in this literature did not explicitly control for cluster/village level factors that may affect child nutrition. A multilevel/mixed effects logit and general equations estimation logit model are used to estimate the difference in the risk of child malnutrition, between households that consumed tobacco and alcohol and those that did not consume the two goods. The two models account for the possibility of correlation in nutritional outcomes for children living in the same cluster/village. The two methods also allowed for the inclusion of contextual effects that could inform public health policy in Kenya. In 2005/6 the odds of long-term child malnutrition were higher for children living in tobacco and alcohol consuming households in rural Kenya. The tobacco and alcohol control policies implemented between 2005/6 and 2015/16 were more effective in reducing tobacco and alcohol consumption prevalence among the poorest rural households. As a result, the decrease in child malnutrition prevalence was greater among households that consumed tobacco and alcohol. In 2015/16 the risk of child malnutrition in tobacco and alcohol consuming households was similar to that of non-consuming households. The results from chapter 3 indicate that tobacco and alcohol control policies that were implemented in Kenya over the ten-year period, contributed to the reduction in consumption of the two goods among the poorest rural households. Therefore, very poor households invested resources, which could have been used for tobacco/alcohol consumption, on human capital development. Chapter 2 and chapter 3 provide evidence on some of the opportunity costs associated with tobacco/alcohol consumption as well as potential benefits that may arise from controlling consumption of the two goods. Subsequently, Chapter 4 focuses on the price elasticity of demand for tobacco and alcohol products. Tax-induced price increases is one of the most effective policy tools for controlling the demand for the two goods. The effectiveness of price policy in controlling demand for tobacco and alcohol may be hindered by among other things, the consumption of informally produced alcohol and noncigarette tobacco products as well as other stimulants. Majority of the relatively few studies done on African countries were on South Africa and most of them estimated the price elasticity of demand for cigarettes and formally produced alcohol products. Further, I am not aware of any peer-reviewed study that has estimated the price elasticity of demand for alcohol and tobacco products in Kenya. Household survey data is used to estimate the own-price and expenditure elasticities (as proxies for income elasticities) of demand for tobacco and alcohol products in Kenya. The responsiveness of the demand for informally produced alcohol as well as non-cigarette tobacco products to changes in prices of cigarettes and formally produced beers is also estimated. This thesis also estimates the responsiveness of demand for khat to changes in the price of cigarettes and formally produced beers. Khat is a stimulant mostly consumed in Arabia and the Horn of Africa. One of the concerns about the use of taxes as a measure to control tobacco and alcohol consumption is the possible regressive nature of tobacco and alcohol taxes. Therefore, chapter 4 also examines the impact of price and non-price tobacco- and alcohol-control policies on the regressivity/progressivity of household tobacco and alcohol burdens (budget shares). Tobacco-control policies implemented between 2005/6 and 2015/16, contributed to a reduction in the regressivity of household tobacco budget shares. The changing profile of tobacco consuming households as well as economic growth over the period may have also contributed to the increase in the estimated price elasticity of demand for cigarettes. Over the ten-year period, the estimated price elasticity of demand for cigarettes increased from -0.63 to -0.42. Khat and snuff tobacco were found to be complements for cigarettes. Khat and informally produced spirits were found to be substitutes for formally produced beers. Further, the demand for formally produced beers was found to be price elastic. The alcohol-control policies that were implemented over the ten-year period, contributed to a reduction in the regressivity of overall alcohol budget shares. However, over the ten-year period, there was a rapid increase in demand for spirits in Kenya