Persuasive Advertising - J. Scott Armstrong [206]
While you might ignore the limitations listed above, keep the following in mind:
Only a small percentage of relevant studies in advertising look at behavior
This is true. Fortunately, there is ample evidence that attitudes are related to behavior. A meta-analysis by Kim and Hunter (1993a) found 138 comparisons of attitude and behavior with a total sample of almost 91,000 subjects. Attitudes and behavior were closely related overall (r=0.79).
More generally, the criteria in the studies were based on intermediate measures including intentions to buy, which are predictive of behavior. In Kim and Hunter (1993b), a meta-analysis found 47 correlations between intentions and behavior with 10,203 subjects; they showed a strong relationship between intentions and behavior. This is especially true for high-involvement goods, as Morwitz (2001) showed. Wright and MacRae (2007), in their meta-analysis, found that purchase intentions studies provide unbiased predictions of behavior.
In addition, while only a small percentage of advertising studies look at behavior, that percentage amounts to a large number of studies; that is, in absolute numbers, many studies have examined behavior in advertising.
When behavior, such as sales, was studied, it was done typically only through short-term responses. This does not pose a serious limitation, because advertising seldom has a long-term effect without first having a short-term effect (Jones and Blair 1996; Lodish et al. 1995a; Riskey 1997).
Most advertising studies are one-shot studies. Few studies in marketing are replicated or extended, as only about 1 percent of marketing journal space has been dedicated to replications. Of the replications and extensions, 60 percent conflicted in some way with the original studies (Hubbard and Armstrong 1994). For example, in doing literature searches for this book, I found that some interesting findings (e.g., principles related to the Golden Section and to subliminal perception) were not supported when replicated. Other findings, such as two-sided arguments, apply only under certain conditions, as was discovered by research extensions. Unfortunately, this situation with respect to replications and extensions has deteriorated over time. The replication rate of studies in major marketing journals dropped by over half in the last decade of the 20th century, with the result that only a fraction of 1 percent of findings in marketing journals have been successfully replicated (Evanchitzky et al. 2007). The situation is not so bad, however, considering that few publications in marketing are important and thus worth replicating. I focus on important studies in this book, and many of them have been replicated and extended.
* * *
Appendix B
Data on print ads from Which Ad Pulled Best (WAPB)
* * *
Sandeep Patnaik and I, along with a team of research assistants, analyzed full-page print ads from the Which Ad Pulled Best (WAPB) using primarily editions 5 through 9 (Burton and Purvis 1987–2002), although edition 4 was used for principles with small samples Each edition contains 50 pairs of ads except for the 9th edition, which has 40 pairs. These ads were prepared by leading U.S. advertisers and were copy tested by Gallup & Robinson.
The WAPB database allows testing of the direction and effect size for the advertising principles because most conditions were identical for each ad in the pairs. The target market, product, and media were the same. Of the 240 pairs of ads in editions 5 through 9, just over half were for the same brand. Given these controls