Random Assignment

Random assignment is a term that is associated with true experiments (called controlled clinical trials in medical research) in which the effects of two or more "treatments" are compared with one another. Participants (respondents, subjects, etc.) are allocated to treatment conditions in such a way that each participant has the same chance of being a member of a particular treatment group. The groups so constituted are therefore as similar as possible at the beginning of the experiment before the treatments are introduced. If they differ at the end of the experiment, there will be a statistical basis for determining whether or not, or to what extent, it is the treatments that were the cause of the difference.

Traditionally, survey research and experimental research have not had much in common with one another. However, the following examples illustrate some of the many ways that survey researchers can benefit from random assignment and experimental design.

A survey researcher would like to determine whether it would be better to use answer sheets for a particular questionnaire in which the responses to individual items are given by blackening in small ovals (the "new" way) or by blackening in the spaces between pairs of vertical dotted lines (the "old" way). In a pilot study carried out prior to the main study, two versions of the answer sheets would be prepared and participants would be assigned to one version or the other in such a way that chance, and chance alone, determines which form they receive. This illustrates simple random assignment. (The effect of type of answer sheet might be operationalized by the difference between the percentages of omitted items for the two forms, since large numbers of omitted items are not desirable in survey research.)

For stratified random assignment, better designated as blocking, the participants are divided into two or more strata and then randomly assigned to treatments within strata (blocks). For the answer sheet example, if a researcher were carrying out the experiment to test the effect of the new way (the experimental treatment) versus the old way (the control treatment) for a sample that consisted of 10 males and 20 females and wanted to be sure that there were proper proportions of males and females in each of the treatments, he or she would randomly assign 5 of the males to the experimental treatment and the other 5 males to the control treatment, and would randomly assign 10 of the females to the experimental treatment and the other 10 females to the control treatment. That would permit testing for the "main effect" of treatment, the "main effect" of sex, and the sex-by-treatment "interaction effect."

There is occasional confusion in the literature between random assignment and random sampling. The purpose of the former, as noted earlier, is pre-experimental equivalence of treatment groups so that post-treatment differences can be attributed to the effects that are caused by the treatments themselves. Thus the objective is one of internal validity. The purpose of the latter, however, is to be able to generalize from a sample to the population from which the sample has been drawn. That objective is concerned with external validity.

...