The results discussed to this point show rather clearly that the elasticity of taxed cigarette sales have grown and that the growth in the elasticity appears to be correlated with the rise of the Internet in different states. The implications of something that facilitates the purchase of tax-free cigarettes are, however, quite different for smoking than they are for cigarette purchases. Consumption should be less sensitive than taxed purchases to the home state tax rate because people can avoid it by buying cigarettes in neighboring states or on the Internet. In addition, greater Internet access probably does not make consumption more sensitive to changes in tax rates. Indeed, it is likely that rising Internet usage makes smoking less sensitive to changes in the tax rate because it provides a new way to get around state taxes and because those smokers that already buy over the Internet would not be affected at all.
To investigate this issue, we turn to the BRFSS for data on cigarette consumption per day and convert it into a state level measure of annual packs-per-person so that it is comparable to our taxable sales data. We show in columns (1)-(3) of Table 4 the results of estimating many of the same regression specifications as we did previously, but with smoking rather than taxed purchases as the dependent variable. The baseline price sensitivity of smoking is clearly much smaller than the elasticity of taxable purchases. Further, the Internet shows no significant influence on that elasticity. In column (1) the point estimate of the interaction term is negative, but small in absolute value and insignificant. In columns (2) and (3), where we allow state-specific elasticities, it is positive (implying that rising Internet usage has made consumption less sensitive to tax changes), although again it is not significantly different from zero.
These results are consistent with a change in the technology of smuggling. Taxed purchases have become dramatically more elastic with respect to state tax rate changes, while the price elasticity of smoking has remained the same or perhaps gotten smaller in absolute value. In columns (4)-(6) we create an indicator of the amount of smuggled cigarettes by taking the log difference of the amount of cigarettes people say they smoked in the year and the amount they purchased in-state. This should respond positively to changes in the price caused by tax changes and in all three cases it does, although as we add more and more flexible controls, the standard error gets large. In the baseline, column (4), higher taxes lead to more smuggling and the amount of additional smuggling has grown significantly with the rise of the Internet. The magnitude in the baseline specification says that the amount of smuggling arising from a change in a state’s tax rate has almost doubled due to the rise of the Internet in this sample. Allowing each state to have a separate baseline elasticity, in column (5), still shows a significant impact of Internet growth on tax induced smuggling, here more than twice as big as the baseline. Introducing a linear time trend with the state-specific elasticities, as in (6) yields positive point estimates of the Internet interaction term, but with a large standard error.
These last three columns provide the clearest evidence that as access to tax-free Internet sales rises, smokers are not smoking less, they are merely paying less and buying their cigarettes from other sources. Such tax avoidance behavior is highly relevant for forecasting tax revenue, of course.
To investigate this issue, we turn to the BRFSS for data on cigarette consumption per day and convert it into a state level measure of annual packs-per-person so that it is comparable to our taxable sales data. We show in columns (1)-(3) of Table 4 the results of estimating many of the same regression specifications as we did previously, but with smoking rather than taxed purchases as the dependent variable. The baseline price sensitivity of smoking is clearly much smaller than the elasticity of taxable purchases. Further, the Internet shows no significant influence on that elasticity. In column (1) the point estimate of the interaction term is negative, but small in absolute value and insignificant. In columns (2) and (3), where we allow state-specific elasticities, it is positive (implying that rising Internet usage has made consumption less sensitive to tax changes), although again it is not significantly different from zero.
These results are consistent with a change in the technology of smuggling. Taxed purchases have become dramatically more elastic with respect to state tax rate changes, while the price elasticity of smoking has remained the same or perhaps gotten smaller in absolute value. In columns (4)-(6) we create an indicator of the amount of smuggled cigarettes by taking the log difference of the amount of cigarettes people say they smoked in the year and the amount they purchased in-state. This should respond positively to changes in the price caused by tax changes and in all three cases it does, although as we add more and more flexible controls, the standard error gets large. In the baseline, column (4), higher taxes lead to more smuggling and the amount of additional smuggling has grown significantly with the rise of the Internet. The magnitude in the baseline specification says that the amount of smuggling arising from a change in a state’s tax rate has almost doubled due to the rise of the Internet in this sample. Allowing each state to have a separate baseline elasticity, in column (5), still shows a significant impact of Internet growth on tax induced smuggling, here more than twice as big as the baseline. Introducing a linear time trend with the state-specific elasticities, as in (6) yields positive point estimates of the Internet interaction term, but with a large standard error.
These last three columns provide the clearest evidence that as access to tax-free Internet sales rises, smokers are not smoking less, they are merely paying less and buying their cigarettes from other sources. Such tax avoidance behavior is highly relevant for forecasting tax revenue, of course.
