Frequently, product tests are also of the “after only with control group” design. General Motors ran such an experiment to determine the desirability of nylon cord tires as compared to the traditional rayon cord tires. Nylon cord tires were more expensive than rayon cord tires and were alleged to whine and thump; but there was little evidence as to the importance of these defects if in fact they existed at all. Accordingly, General Motors equipped 40,000 Chevrolets with nylon cord tires and kept track of the serial numbers of the cars. Later they interviewed owners of cars with both types of tires to get their appraisals of their tires.
Sometimes the difficulty of establishing a control group that is not exposed to the experimental variable causes researchers to substitute something else as a control, as in the following example:
Procter & amble clearly set the sales growth target of 250,000 cases as a control. The student will recognize the dangers in this procedure. Indeed, the design is very close to the after only design previously designated as unsatisfactory.
Ex Post Facto Design:
One variation of the “after only” design is called the ex post facto design. This differs from the “after only” design because the experimental and control groups are selected after the experimental variable is introduced instead or before. One advantage is that the test subjects cannot be influenced, pro or con, toward the subject by knowing they are being tested, since they are exposed to the experimental variable before being selected form the sample.
Another advantage of this method is that it permits the experimenter to let the experimental variable be introduced realistically and to control only observations. This is useful in advertising tests which use commercial media. A large grocery product manufacturer ran an advertising campaign in one Midwestern city. Then, it selected an experimental group of consumer who reported they had seen the advertisements and a control group of consumers who reported they had not seen ads. The two groups were asked questions about the tendency of the manufacturer’s product to cause people to gain weight. Because the advertising campaign had emphasized that the product was not fattening, it was hypothesized that those who had seen the advertisements would report the product as being non-fattening to a greater degree than would those who had not seen the campaign.
The results supported this hypothesis: the product was reported not fattening by 63 percent of these seeing the advertisement but by only 56 percent of the control group. But members of the experimental and control groups were actually self-selected. Those who said they had seen the ads were very likely the ones on whom the ads made some impression. It is also quite likely that some of those who said they had not seen the ads actually had but did not remember having seen them.
If the experimental variable is such that exposure to it can be determined objectively on an ex post facto basis, this bias of self selection can be eliminated and the design becomes essentially the same as the “after only with control group” design. In this latter case, the ex post facto design may have a definite advantage over the other design, as the experimental variable will have exerted its influence in an entirely natural settling. An example would be a TV public service announcement designed to inform consumers about the pros and cons of nuclear energy. This could be broadcast over cable TV only, and interviewers could then determine with some objective accuracy whether a home had cable TV or not.
It should be noted that the ex post facto design is the same as the statistical, cross classification type of study discussed earlier in this article. For example, families under study might be classified into two classes – those with cable television and those without. Within each of these classes the percent having specific bits of knowledge of the nuclear energy controversy could be calculated.
Because the ex post facto design is essentially cross classification some students prefer not to consider it an experimental design.