Cronbach’s alpha - what makes it really good? Some advice for planning and criticizing psychological questionnaires

Abstract

Whatever Cronbach’s alpha measures – it’s not internal consistency, commonly misunderstood in psychology as the average strength of relationships within questionnaire items. In this article, we explore the reasons why the understanding of alpha as internal consistency is particularly flawed, and focus on how alpha inflation works in a practical way. Using the simulation method, we determine the precise (common) influence of the number of respondents, the range of measurement (Likert) scales, the number of questions in the questionnaire and the average correlation of items on the alpha level. The results confirm alpha-level inflation due to a greater number of questions: alpha gets a satisfactory level even with minimal internal consistency if there are many questions in the questionnaire. We suggest that the reliability of weak psychological tools is overestimated because of presented rapid alpha inflation. Number of subjects and the range of the scale had no influence on alpha.