This thesis consists of three self-contained essays on the economics of alcohol demand.
Chapter 2 examines the price elasticity of demand for alcohol across the drinking distribution, using household expenditure data to test whether heavy drinkers respond the same as light drinkers to price increases. Both conditional and unconditional quantile regression are used to compare results generated by the two different methods. The chapter finds that when price increases, heavier drinkers decrease consumption proportionately less than lighter drinkers whilst substituting more towards lower quality beverages. This is an important result since it shows that price-based policies may have little effect in reducing heavy consumption whilst creating large welfare losses for moderate drinkers.
Chapter 3 uses several different methods including the Tobit and Double-Hurdle models to estimate the mean price elasticity of demand for alcohol. In doing so, it tests how the price elasticity estimates can differ depending on model choice, even when the same data is used. Household expenditure data contains a large number of households who do not purchase any alcohol, for three distinct reasons: price reasons, non-price reasons, and infrequent purchase. A double-hurdle model is developed which can accommodate all three types of non-purchase. The results suggest that, compared to the double-hurdle model, the frequently-used Tobit model produces larger absolute estimates of the price elasticity of demand for alcohol. The double-hurdle model is the preferred specification since it incorporates all reasons for zeros in alcohol expenditure. Chapter 4 explores changes in alcohol consumption across the lifecourse using a large number of waves of a cross-sectional survey, the General Household Survey, to create synthetic cohorts. Whilst the existing literature looks at how the mean consumption differs across birth cohorts, this chapter instead looks at different quantiles of the drinking distribution to examine whether the changes are consistent across all drinkers, including abstention. This is important because it shows how the alcohol consumption distribution has changed across time, age and birth cohort. It finds that generally, alcohol consumption decreases both as age increases and in older birth cohorts. Alcohol consumption by females has particularly changed; younger birth cohorts drinking much more than their parents’ cohorts did, yet younger birth cohorts are also more likely not to drink at all.
Whilst these chapters can be considered stand-alone essays, they are also linked and show how different and cutting edge techniques, applied to the best available data, can be used to show new and interesting results which can aid policymakers and policy decisions.