In Canada, however, we see a different distribution: only 5% of the people report being unhappy, 34% report being “somewhat happy” and 59% report being “very happy” (with 2% missing). These differences are consistent with the literature reviewed earlier, which discusses the sensitivity of the happiness responses across countries or types of wording. But the consistent impacts of cigarette taxation we will see in both countries below confirm that these differences do not interfere with our tests.
Both data sets collected data directly on smoking behavior, but only periodically. In the U.S. GSS, these smoking data were collected from 1977 to 1993; in the Canadian GSS, they were collected in 1986, 1991, and 1996. In those years, 35% of the U.S. sample reports themselves as smokers, which is consistent with prevalence rates over this time period; in the Canadian data 30% report smoking. Table 1 also summarizes the data sets based on whether the person reports being a smoker, a non-smoker and whether the data is missing. Smokers are somewhat less happy than average in both data sets. While this consistent with the notion that they would like to quit but cannot, it is equally consistent with heterogeneity in smoking behavior by underlying happiness.
The average real (in 1999 dollars) excise tax rate on cigarettes in the US is 31.6 cents, with a standard deviation of 15.8 cents, while in Canada it is 1.17 Canadian dollars, with a standard deviation of 39 cents. There is wide variation in excise taxes across states, over time, and within states over time; 25% of the variation in excise taxes in the United States, and 32% in Canada, is within states/provinces over time. This allows us to control for fixed state/province differences in cigarette taxes and happiness in our analysis below, as well as time trends in both. More info
Table 1 also shows the means for the key control variables used in our analysis. Some interesting features are worth noting. Smokers are less educated. For example in the United States, they have a high school dropout rate of 33% compared to 25% for non-smokers. They are also more likely to be unemployed and less likely to be out of the labor force, although this likely largely reflects the fact that the smoking rate is much higher among males. We have endeavored to use as much as possible a common set of control variables in the two data sets, but the available variables are not identical (e.g. there are no consistent labor supply measures in the Canadian GSS).
Finally, income is available only categorically in the U.S. GSS, in fine gradations until the top of the income distribution, then in larger intervals and finally a top code. In order to create a smooth income measure, we have used data from each year’s Current Population Survey to impute values to each of these larger ranges and the top-coded range. Income is measured continuously in some years of the Canadian GSS, and in categories in other years; in the latter set of years, we use the midpoint of the income ranges (or 1.25 times the top value for the upper range). In the regressions in each country we control for quartiles of the real income distribution.
In Canada, but not in the U.S., data are available on both personal and household income, so we include measures for both types of income.