Working paper 745
This paper provides an extensive analysis of the demand for cigarettes based on longitudinal data drawn from the South Africa National Income and Dynamic Study (NIDS: 2008 - 2014). We compare the results of the pooled OLS (POLS), the standard two-part model, the random and the fixed effect (RE, FE) panel regression. Like previous evidence into cigarette prices, we obtain negative price elasticity of demand for cigarettes, with the conditional elasticity (POLS and RE estimates) signicantly smaller than the total price elasticity (two-part model estimates). We find that over the same period, estimates from the fixed effect model are insignicant, properly due to the limited within variation in prices. However, it should be noted that estimates from the between variation models (RE and POLS) could be biased as they do not control for unobserved heterogeneity. Thus, with between variation models, increased tobacco taxes can, in the presence of the changing market structure, remain a desirable policy tool for reducing cigarette consumption.