On Quality

M. J. Stewart failed to critique the correct data set we sent him. He had two data sets of ours in his possession, and has written a critique (Stewart, 1992) of the earlier data set, data already rejected by us. Stewart's ‘Critique of Laugesen and Meads 1991’, wrongly titled, critiques our 1991 paper using a 1990 data set. We discarded the 1990 data set and revised and replaced it in 1991 prior to publication (Laugesen and Meads, 1991). The 1991 data set used for our 1991 paper differs from the 1990 version with regard to tobacco price in most countries; in addition the advertising restriction score for Sweden has been updated.

What Stewart wrote and asked for, and what we sent him, which he received but did not use, was a copy of the data we used in our 1991 paper (Laugesen and Meads, 1991). If he had any doubts as to which data set he should critique, he was entitled to ask us. As he did not use our 1991 data set, Stewart failed to note the very high correlation between our 1991 data and his UK price series.

We did not send Stewart the 1990 data set. Stewart evidently obtained it after the 1990 Quebec tobacco advertising trial. We made it available to the Canadian government, but with some reluctance to the Quebec Superior Court and hence to tobacco manufacturer plaintiffs, as we intended to revise and publish.

In response to the parts of Stewart's critique, applicable despite his using an earlier data set: 1. Stewart says we should calculate price independent of consumption, but the necessary data were only fully available for nine out of 22 countries; the method we used was the only feasible method, and gave a very high degree of correlation with Stewart's method.

2. Stewart misquotes us on income: as a comparator of incomes, we use gross domestic product per capita, in preference to private consumption which does not allow for the value of free public sector services.

3. After allowing for data errors in some values for tobacco consumption for the United Kingdom, and for Portugal's female work-force data, we recalculated our equations and confirm our published findings that adbans tend to be significantly associated with falls in consumption.

4. We preferred Generalized Least Squares for regression. The pre-conditions for using Ordinary Least Squares could not be met.

5. We confirm the general pattern of coefficients, and the tendency of advertising bans to be followed by falls in tobacco consumption, if we use country dummies and Ordinary Least Squares as Stewart suggests. The large residuals noted in our 1991 analysis disappear; between-country differences are significant, consistent with historical differences in intrinsic tobacco consumption.

6. We find significant correlations between adbans and later tobacco consumption. The converse was not true; earlier consumption did not correlate with later ban levels.