SURVEY DESIGN AND SAMPLING
In some very fundamental ways, survey research is different from the experimentally oriented methods that were described in the initial chapters of this book. Although, as in experimentation, questions of bias are raised in the evaluation of all surveys, the major critical focus is with external, rather than internal, validity. The usual question in survey contexts is, “How well do the responses of a subset of a population (the total group to which we wish our findings to generalize) actually represent the underlying perceptions or feelings of the population?” Generally, we are not concerned with issues of internal validity (i.e., is the manipulation responsible for the obtained findings?) because surveys rarely include manipulations—although survey researchers are increasingly using the technique as a vehicle for experimentation. Typically, the issues of concern for the survey researcher are different from those of the laboratory experimenter. This chapter provides an introduction to this alternate form of research. It is important to bear in mind throughout this chapter that an important goal of much survey research is to provide estimates of population values that are as accurate as possible. Most of the technical aspects of survey sampling have been developed in the service of this goal.
To draw the distinction between the experimental techniques and survey research, it is useful to emphasize the distinction between sampling and assignment. Recall that in the sections of the book devoted to experimental design (chaps. 1–5), we stressed the importance of random assignment of participants to conditions. In experiments, we are concerned primarily with ensuring that participants are randomly assigned to the various conditions of the study; often, however, we are less concerned with the characteristics of the pool from which participants were drawn. The core requirement for random assignment is that each person (or unit) in the pool has the same chance of being assigned to a specific experimental or control condition as any other person (or unit) in the pool. Random assignment is essential if the full power of experimental techniques to foster causal statements is to be realized. Most experimentalists realize that the generalizability of results is dependent on