Estimating Reliability Coefficients with Heterogeneous Item Weightings Using Stata: A Factor Based Approach

Abstract

We show how to estimate a Cronbach's alpha reliability coefficient in Stata after running a principal component or factor analysis. Alpha evaluates to what extent items measure the same underlying content when the items are combined into a scale or used for latent variable. Stata allows for testing the reliability coefficient (alpha) of a scale only when all items receive homogenous weights. We present a user-written program that computes reliability coefficients when implementation of principal component or factor analysis shows heterogeneous item loadings. We use data on management practices from Bloom and Van Reenen (2010, JEL) to explain how to implement and interpret the adjusted internal consistency measure using the user-command afa.

Keywords: afa, factor analysis, Cronbach's alpha, reliability, heterogeneous scale construction, latent variable

JEL Classification: C49, C87, M21