Extraneous Variables

Extraneous variables are factors that may confound a researcher's ability to demonstrate causation. Here is a partial list of extraneous variables marketing researchers confront:

1. History: History refers to events that are external to the experiment. These events occur at the same time as the experiment. History makes it more difficult for marketing researchers to get a "clean read" from their test market because the change in the dependent variable may be due to historical events and not the study's independent variables. The longer the experiment the greater the probability that history will impact the research.

Here is an example of this phenomenon from my own career. During my first month as an advertising executive, I travelled to Green Bay, WI with my client from Clairol. Clairol was test marketing a new shampoo that was to compete directly with a shampoo from Proctor & Gamble[i]. We made this trip to visit retailers. Our goal was to check shelf placement, in-store displays, and the availability of our "saleable sample." Clairol was using this saleable sample—a 1 oz. bottle of the shampoo selling at 39¢—to gain trial; which is to say, get shampoo users to try the product. Upon visiting the first store, we immediately noticed that P&G priced a 6 oz. bottle of their brand at 39¢. As a consequence, Clairol's saleable sample tactic failed to achieve the hoped form levels of trial.

2. Maturation: Maturation refers to changes that occur to the test subjects during a test market that are not related to the test market. Maturation effects the test market. The target markets preferences may change because of maturation factors—changes in test subjects' demographics, psychographics, usage behaviors rather than the test variables. The longer the test market, the more likely it will suffer from maturation. Imagine a two-year experiment conducted among teenagers for an Acne remedy. The normal aging of test subjects is a maturation effect, which could severely limit researchers' attempt to make sound conclusions from their findings.

3. Testing Effects: Being part of an experiment changes people and could confound the results. The mere fact of being observed can cause people to change their attitudes and behavior. Advertisers frequently use "pre-post" persuasion tests to measure the effectiveness of advertising. But, such tests can have a testing effect. The fact that people are asked to discuss their purchase intent before seeing an advertisement may influence their perception of the advertisement.

4. Mortality: Mortality refers to the loss of test subjects over time. This is an especially serious problem with longitudinal tests that measure test variables over a long period of time. If a researcher is going to get a read of consumers' attitudes over several time periods, the impact of people dropping out of the study can undermine the validity of the study.

5. Selection Bias: Selection bias is an extraneous variable that undermines an experiment's validity. Selection bias occurs when the test group or control group is significantly different from the population in purports to represent. Let's go back to Clairol's test market in Green Bay. Why Green Bay? Well, when conducting field experiments, marketing researchers look for small, relatively isolated markets, to represent the United States. The goal is to find markets that are "Little USAs." Of course, there are no perfect test markets that give a 100 percent accurate portrayal of the USA.

These extraneous variables are often called confounding variables as they undermine, or confound, the market researcher's ability to draw clear conclusions from an experiment. When conducting an experiment, researchers attempt to control the influence of extraneous variables. Here are some of the techniques they use:

Randomization: Randomization refers to assigning test subjects to different treatment groups randomly. A treatment group is a group of subjects in an experimental design. There are two categories of treatment groups: Experimental groups, which receive a treatment, and control groups, which do not receive a treatment.
Matching: Matching involves balancing test subjects on a set of background variables before assigning them to treatments.
Design Control: Design control deals with the organization of the research design.
Statistical Control: Statistical control refers to the use of statistical techniques to adjust for the influence of confounding variables.

[i] At that time Clairol was own by Bristol-Myers. In 2001 Proctor & Gamble acquired Clairol from Bristol-Myers Squibb.

toc | return to top | previous page | next page