Example 3.20 supposedly concerns the question "Are attitudes towards shopping changing?" However, the statement of this question is followed by a description of the results of a single poll, which obviously can say nothing about whether anything is changing.
The question supposedly asked in this poll is "I like buying new clothes, but shopping is often frustrating and time-consuming". This is a strikingly bad question. It would be ridiculous to expect people to answer this literally, so that, for example, someone who did like buying new clothes, and did find this to be time-consuming, but who did not find it to be particularly frustrating, would answer "no".
I looked up the New York Times article referenced, to try to see whether any polling organization really asked this question. Apparently they did not. The question appears in the caption of a graph in the article, presumably written by the NYT, but from the text it appears that the 2500 people polled were actually first asked whether they liked to shop for clothes, and only those who said they did (44%) were asked whether they found the experience frustrating and time consuming (the exact text of the question was not given). One should note that the actual proportion who would have answered "yes" to the original question (if they could have figured out what it meant) was therefore 66% of 44%, which is 29.0%. Data from the article imply that the figure the previous year was 60% of 47%, which is 28.2%. If there was any substantial change at all, it was a decrease in the number of people who like shopping for clothes and don't find it frustrating or time-consuming (from 18.8% to 15.0%). Since this was apparently the fourth year for such surveys, however, and it is the first that found a change that is even marginally significant, one might suspect that nothing's really changing at all.