The probability of households purchasing tobacco in Canada and the challenges associated with identifying price effect

Abstract

A logit specification is used to examine the Statistics Canada Survey of Household Spending cross sections collected within the 10 provinces across Canada. These cross sections are used to investigate the relationship between whether households purchase tobacco over an entire year and the price of tobacco. Unlike many published papers, and similar to some, the price parameter estimate is positive and statistically insignificant. This finding raises questions about how the likelihood of households purchasing tobacco products has been changing across provinces and calendar years. In order to examine these factors the price variable is replaced with province crossed with year indicator variables. Finally in order to examine the stability of findings, 5 geographic regions and 4 time periods are each separately examined. These 20 region-time period findings are generally the same as other findings in terms of sign and magnitude. In contrast there is statistical evidence to reject the hypothesis that parameters are the same across the 20 region-time periods. In summary of the findings, for particular provinces and years, the likelihood of purchasing tobacco increased along with the price and this holds even after statistically controlling for variation in other factors. One possible interpretation is that the role of the price of tobacco is challenging to identify. For each province the likelihood of purchasing tobacco generally drops over calendar time however there are some regions and time periods during which this time trend is offset. Across each set of estimates, higher income lowers the odds of purchasing tobacco and this change becomes smaller as income rises. Generally, when the household member that Statistics Canada surveys is married, older, or the only individual in the household then the household is less likely to purchase tobacco.