The Accuracy of the Impressions of Survey Interviewers

Abstract

Living Research is a section of the Quarterly reserved for brief reports of research, discussions of unsolved problems, presentations of neat methodological tricks, and other items that arise out of the daily work of researchers. It provides an opportunity for discussion in print of questions and results that may not seem to warrant a full-blown article. Researchers will find in this section a place to exhibit data which are not adequate to substantiate important generalizations but seem promising in their implications or suggest lines of further theoretical exploration. Succinct case histories are welcomed, as well as hypotheses and insights which may be useful to other students of public opinion. Notes published here are not intended to rank below the regular articles in quality or significance, but are distinguished by their shorter length, greater informality, and more tentative nature.While conducting an opinion survey, interviewers can hardly avoid forming an impression of the trends of the responses of successive respondents to the questions being asked. This circumstance gives rise to a number of serious methodological questions, particularly: (1) How accurate are these impressions? (s) What interviewer characteristics are associated with accuracy or inaccuracy of impressions? and (3) Can these interviewer impressions affect the validity of the subjects' responses, through interviewer influence?For example, if an interviewer, after asking a number of people the same set of questions, arrives at the conclusion (with or without explicit verbalization to himself) that about three-fourths of the respondents say "Yes" to Question X, or that the respondents whom the interviewer likes are "strongly opposed" to Statement Y, it is easily conceivable that selective perception on the part of the interviewer could occur in subsequent interviews. Presumably, these distortions would be in the direction of confirming the impressionistic hypotheses of the interviewer. The survey results could then be affected either directly, by causing the interviewer to be something other than a neutral recording instrument, or indirectly, through the interviewer's unintended influence upon the respondent's statements.

Interviewer Effect in a Survey of Drinking Practices

“…I doubt this for three reasons: ( 1 ) On the average, each worker interviewed eleven respondents, or three families, hardly enough to discover a trend; ( 2 ) In the two years that had elapsed prior to the reinterview, interviewers had probably forgotten. ( 3 ) Olmsted (1962) experimentally found that interviewing made no difference in the accuracy of interviewers' estimates of sample characteristics.…”