Measurement Errors in the Behavioral Sciences: The Case of Personality/Attitude Research

Abstract

It has been established that construct measures often contain measurement error and method effects. However, the extent of these measurement problems and their effects on personality constructs has not been fully examined. In this study, an analysis of published data sets affords estimates of the amount of trait, method, and error variance of measures used in the social/behavioral sciences. It was found that traits accounted for less than 50 percent of the variance in construct measures among the entire group of data sets included in the study. Measures of attitudes and personality variables contain considerably less trait variance. Implications for theory testing are discussed.