What is coefficient alpha? An examination of theory and applications.

TLDR

Coefficient alpha, introduced by Cronbach in 1951, is a widely used reliability statistic across many disciplines, yet its meaning and interpretation remain unclear. This article seeks to clarify alpha by addressing the confusion through theoretical and practical analyses. The authors first provide a theoretical review integrating various statements about alpha, then conduct a practical examination of how alpha changes with item quantity and inter‑item correlations.

Abstract

Coefficient alpha (Cronbach, 1951) is certainly one of the most important and pervasive statistics in research involving test construction and use. A review of the Social Sciences Citations Index for the literature from 1966 to 1990 revealed that Cronbach's (1951) article had been cited approximately 60 times per year and in a total of 278 different journals. In addition to the areas of psychology in which one may expect to see alpha used, such as educational, industrial, social, clinical, child, community, and abnormal psychology, this list of journals included representatives from experimental psychology, sociology, statistics, medicine, counseling, nursing, economics, political science, criminology, gerontology, broadcasting, anthropology, and accounting. In spite of its widespread use, however, there is some confusion as to the true meaning and proper interpretation of the statistic. In this article I address this confusion in two ways. First, a theoretical discussion of alpha is presented. This includes some of the many statements that have been made about alpha and an attempt to integrate these statements. Second, I take a more practical approach in which the interpretation of alpha is examined by observing the changes in alpha as the number of items and interitem correlations are manipulated.