At the same time, the data reviewed earlier suggests that smoking is a very negative influence in the lives of many smokers so it is plausible that there could be large effects on happiness from smoking reduction. Nevertheless, given these large magnitudes, we turn next to specification checks to demonstrate that the estimates are robust.
Comparing the results across the countries suggests that a similar level change in taxes has a much smaller effect in Canada than in the US. One possible interpretation of this could come from the higher level of base prices in Canada, relative to the U.S. Given these high taxes already in place in Canada, the remaining pool of smokers may be those with the largest selfcontrol problems. These smokers may need much larger tax changes to dissuade them from smoking. This contention is consistent with the fact that the elasticities of smoking with respect to price are similar across the two countries, despite the fact that base prices are much higher in Canada.
In Table 3, we further explore the sensitivity of our findings to concerns about omitted state variables that might be correlated with cigarette excise tax policy, focusing on the “unhappy” variables in both countries. In the first column, we show our basic results from Table 2 for comparison. In the second column, we interact the state/year unemployment rate with PREDSMOK, to capture any differential impacts of the cycle on the happiness of predicted smokers and nonsmokers; this has no impact on our estimates. natural inhalers for asthma
In the third column, we include state-specific linear time trends to capture any slow-moving trends in tax policy and happiness that might confound our results; including these trends raises the estimates in both samples. In the fourth column, we interact a time trend with the predicted smoking measure, to allow for separate trends in well-being for predicted smokers and nonsmokers; once again, there is little impact. In the fifth column, we interact each state dummy with PREDSMOK to allow for the effect of predicted smoking to vary by state; this reduces the estimate somewhat in the U.S., but raises it in Canada. Overall, our findings are reasonably robust to these controls for slow-moving trends in the data or heterogeneity in populations across states.
In Table 4, we address a different worry: that our happiness result arises through the spending financed by cigarette excise tax increases. Suppose, for example, that government spending is more redistributive than excise taxation, or at least valued more by the types of individuals who are predicted to smoke. Then our finding could reflect the happiness effects of spending, not excise taxation.
To address this point, we have gathered data on three other state or province taxes: the excise taxes on gas and alcohol, and the state or province sales tax rate. We have also gathered data on state or province real revenues per capita. If this is a spending effect, so long as cigarette excise revenues are spent in a similar fashion to other tax revenues, then we should see a similar happiness effect from these other taxes.