Chapter 9 
The interpretation of data in Table 3 is complicated by almost certainly 
differing beliefs of smokers in the four tar categories regarding the risks of 
smoking a 20-mg tar cigarette and hence about the decrease in risk from 
any reduction in tar level. Because the belief factor is likely to be a constant 
in the two versions of this question, it is useful to examine the relative 
reduction in health risk (i.e., the difference in benefits between switching 
to the 5-mg tar alternative compared with the 16-mg alternative), shown in 
the last row of Table 3. Once again, the evidence points to a clear difference 
between smokers of cigarettes with 1 to 5 mg of tar and all other smokers. 
These very-low-tar smokers believe that it takes a substantial reduction in 
tar yields to significantly reduce health risk, whereas this belief does not 
appear to be held by a substantial number of smokers in other categories. 
Unfortunately, this belief also may support a judgment that a substantial 
reduction in tar levels may be a reasonable substitute for quitting. 
In the second approach, we examined smokers' understanding of the 
distinction between tar yield and delivery, together with their willingness to 
treat the numerical information as if it had ratio-scale properties rather than 
merely ordinal properties. Many of those supporting the dissemination of 
tar numbers have assumed that consumers would use these numbers in an 
ordinal fashion, essentially as if they were simply rank-ordered data. Ordinal 
scales do not possess the property that each numerical interval is of the same 
magnitude (i.e., the difference between 1 and 2 being precisely equal to the 
difference between 10 and 11). The FTC method may produce tar ratings 
that have this interval scale property for tar yields, but it cannot be said to 
do so for actual deliveries of tar because smokers' inhalation patterns seem 
to vary as they move lower on the scale. A ratio scale has the further property 
of having a genuine zero point so that it is proper to regard a scale score of 
10 as being twice as high as a scale score of 5. 
Respondents were asked to assume that a person switched from a 10-mg 
tar cigarette to a 1-mg tar cigarette. Then the three statements shown in 
Table 4 were read twice, and respondents were asked to decide which of these 
came closest to their opinion. Primacy and recency effects were controlled 
by rotating the order of the first and third statements. The first answer is the 
correct choice, whereas the second answer suggests some reluctance to rely 
on the absolute numerical values when thinking about such tradeoffs. 
The general conclusion to be drawn from these data is that at least one- 
quarter of smokers (i.e., those selecting the third interpretation) clearly have 
been misled about the meaning of the tar yield numbers. Interestingly, this 
increases to 44 percent for smokers of very-low-tar cigarettes, in line with 
other evidence presented here; it also increases concern about the safety 
reassurances that such very-low-tar cigarettes appear to provide. 
131 
