Multiple Indicators in Survey Research

Abstract

When analyzing survey data, valuable evidence may be obtained by using several indicators of each of the conceptual variables to be related and observing the pattern of associations between each indicator of the independent variable(s) and each indicator of the dependent variable(s). These patterns of associations can help the researcher to decide whether the results are spuriously produced by a third factor and/or whether one or more of the indicators is invalid. Also, such information can show the need, and provide guidelines for conceptual reformulation. This procedure complements, but does not replace, other analysis techniques in survey research. Questions of evidence in social research are answered by reference to the general canons of science. As in any discipline, however , special problems of evidence arise in sociology stemming from its special methods of research. Inasmuch as survey research in particular poses difficult problems in the application of "classic" canons of evidence,2 the development of analysis techniques specifically intended for social surveys is highly desirable. Such techniques have wide applicability, since the problems of evidence presented by the survey method are fairly typical of all social science research. One existing approach to the special problems of survey research is the use of several different indicators to represent a single underlying concept.3 This paper suggests , in addition to the standard, combined use of multiple indicators, that the researcher observe individual associations between each indicator of the independent concept and each indicator of the dependent concept. Although this procedure dates back at least to Durkheim's analysis of suicide statistics, it is rarely used today, despite its important benefits. This paper outlines the specific advantages of this procedure and discusses the ways in which it provides evidence in combination with other techniques of survey research. MULTIPLE INDICATORS The use of multiple indicators is called for whenever the researcher has definite theoretical concepts which he wishes to relate , but for which he is unable to obtain or defend single, unambiguous, direct, operational definitions. Concepts such as age, sex, and group size are close to standard and agreed-upon operations, but this is not the case when studying the relationships between such variables as integration, morale, prejudice, adjustment, and so on. This situation, in which the detailed information gathered by interviewers is not in itself of theoretical importance, occurs with 'The authors wish to thank Lee Bean, H. M. Blalock, Jr., Phillip E. Hammond, and Stanley H. Udy, Jr., for their valuable criticism and advice.