Panel Paper:
The Relationship Between Cigarette Taxes and Illicit Trade in Europe
*Names in bold indicate Presenter
These claims are counterintuitive to the economics of illicit markets. Economic theory suggests that, all else equal, increasing the profit from an activity (and the profit from illicit trade is tied to the cost of the tax or regulation avoided) will induce a greater supply of that product. Applied to tax evasion, this implies that an increase in taxes will lead to an increase in tax evasion, all else equal. Higher taxes and regulations drive more of the economy underground, beyond the reach of the tax collector.
The discrepancy between the claimed empirical relationship between tobacco taxes and smuggling, and the theory (as well as empirical evidence from other industries) may be due to the well-known possibility that cross-sectional results can be biased due to missing variables or endogeneity.
As noted in the tobacco control literature, there are many variables that potentially determine the volume of the illicit tobacco trade: the cost of smuggling, availability of illicit supplies, strength of organized crime networks, law enforcement efforts, punishment, and corruption. Existing studies have not adequately controlled for such factors. As a result, these determinants are missing (or “omitted”) in simple cross-sectional comparisons of tax or price against illicit share. Omission of these variables may lead to biased results.
We use fixed effects panel data econometrics to analyze the relationship between tax rates and illicit share in a sample of European countries. This analysis produces results that are inconsistent with the claims of the cross-sectional studies but are consistent with economic theory and empirical studies of other goods and services. After considering several alternative explanations, we conclude that tobacco taxes are a significant determinant of illicit market share, and that previous finding to the contrary are attributable to the well-understood phenomenon of omitted variable bias.