Measurement Error in Self-Reported Health Variables

Abstract

Measurement error may be an important source of bias in studies using self-reported health indicators to explain work behavior. As a test of measurement error, the tetrachoric correlation coefficient is used to examine the relationship between two alternative measures of arthritis, a standard self-reported measure and a simulated clinical measure. While the two measures are highly correlated, measurement error is found. Regression analysis demonstrates that it varies systematically across different socioeconomic groups. In particular, individuals who are not working tend to report their health incorrectly, perhaps owing to social pressure to justify not having a job. Coauthors are Richard V. Burkhauser, Jean M. Mitchell, and Theodore P. Pincus. Copyright 1987 by MIT Press.