Key points are not available for this paper at this time.
Abstract Increasing precision of measurement is a goal of scientific advancement, but Nunnally's (1978) .70 benchmark for coefficient alpha (alpha) has remained the omnibus test for reliability for nearly 40 years. This likely arises due to there only being scattered empirical evidence of the degree to which the field has met or surpassed this standard. Using meta‐analytic techniques known as reliability generalization (RG), we cumulate alphas across 36 commonly used individual differences, attitudes, and behaviours from 1675 independent samples ( N = 991,588). Our primary finding is that alphas almost always exceed .70 and generally fall above .80. In addition, we identified factors that moderate alpha including the specific measure used, the number of scale items, and the rater. The study provides baseline alphas that can be used for research planning and design; it also offers best practices for RG and notes the benefits of RG for understanding systematic fluctuations in reliability.
Greco et al. (Wed,) studied this question.