Persuasive Advertising - J. Scott Armstrong [7]
Empirical evidence
When situations are complex, empirical research is needed to identify how conditions affect the outcomes of various actions.
The dangers of complexity can be illustrated by an example from medicine. In 1847, Ignaz Semmelweis observed that the obstetric clinic of a hospital that was served by doctors had three times the mortality rate of a “lower-class” clinic, which was served by midwives. Given the many factors involved, it required much research to determine the cause—which was that the doctors were coming into the obstetric clinic after having performed various procedures and their hands contained germs. Semmelweis concluded that they should wash their hands prior to each procedure. However, his findings were met with skepticism. Despite his scientific publications, it was years after his death that hand washing gained acceptance. Typical practice and expert opinion failed in this life-and-death situation involving hand washing. You might ask whether this would happen in advertising.
Let’s look more closely at the types of empirical evidence: non-experimental, quasi-experimental (meaning partly experimental), and experimental.
Non-experimental data
You have seen thousands of advertisements for analog watches in your lifetime. Do you know what time is typically displayed on the watches in the ads? When I have asked my people this question, approximately 10 percent give the correct answer.
Systematic observation and record keeping can greatly improve the ability to learn. Certainly it is more effective than casual observation. For example, if you record the time shown in the next few watch ads you see, you’ll quickly learn that almost all watch ads show the same time.
Non-experimental studies are most useful when studying simple situations—that is, the relationship depends upon few conditions. Sometimes even a single condition can obscure the effects of a persuasive technique.
In reasonably simple situations, where there is ample data, it is sometimes possible to use statistical procedures to control for some conditions. For example, analyses of extensive records kept in baseball and football have shown that coaches’ guidelines (e.g., go for a field goal on fourth down when inside the opponent’s 10-yard line) often conflict with those based on empirical studies of what is best. Surowiecki (2004) reports that some football coaches benefit from analyses of non-experimental data. Lewis (2003) describes how analyses of non-experimental data have aided decision-making in baseball. The 15 National Basketball teams with at least one full-time statistician on their staff won 59 percent of 962 games to date in 2009–10, while the 15 teams with no statisticians won only 41 percent of 958 games (David Biderman, Wall Street Journal, March 12, 2010).
However, in complex situations, non-experimental data can be misleading. Consider another example from medicine. In a former British colony, healers believed that a distillation of fluids extracted from the urine of horses, if dried to a powder and fed to older women, could preserve youth and ward off a variety of diseases. The preparation become popular and non-experimental studies using hundreds of thousands of older women showed that those who took the drug were healthier.
The former colony is the United States; the time was the latter part of the 20th century and on into the 21st century; the drug is a female hormone-replacement, such as Premarin. Women who were diligent about their health issues sought the best care on all types of problems. These non-experimental studies compared them with women who put less emphasis on taking care of their health. Thus, the non-experimental analysis of data misled doctors. This ended