Meta‐Analysis of Coefficient Alpha: A Reliability Generalization Study

• DOI: 10.1111/joms.12328
• Publication Date: 2017-11-15T11:22:39Z
• Authors: Lindsey M. Greco; Ernest H. O'Boyle; Bethany S. Cockburn; Zhenyu Yuan
• Keywords: Alpha (finance); Benchmark (surveying)

Abstract Increasing precision of measurement is a goal scientific advancement, but Nunnally's (1978) .70 benchmark for coefficient alpha (alpha) has remained the omnibus test reliability nearly 40 years. This likely arises due to there only being scattered empirical evidence degree which field met or surpassed this standard. Using meta‐analytic techniques known as generalization (RG), we cumulate alphas across 36 commonly used individual differences, attitudes, and behaviours from 1675 independent samples ( N = 991,588). Our primary finding that almost always exceed generally fall above .80. In addition, identified factors moderate including specific measure used, number scale items, rater. The study provides baseline can be research planning design; it also offers best practices RG notes benefits understanding systematic fluctuations in reliability.