A Cautionary Note on Using Univariate Methods for Meta-Analytic Structural Equation Modeling

Abstract

Meta-analytic structural equation modeling (MASEM) is an increasingly popular technique in psychology, especially in management and organizational psychology. MASEM refers to fitting structural equation models (SEMs), such as path models or factor models, to meta-analytic data. The meta-analytic data, obtained from multiple primary studies, generally consist of correlations across the variables in the path or factor model. In this study, we contrast the method that is most often applied in management and organizational psychology (the univariate-r method) to several multivariate methods. “Univariate-r” refers to performing multiple univariate meta-analyses to obtain a synthesized correlation matrix as input in an SEM program. In multivariate MASEM, a multivariate meta-analysis is used to synthesize correlation matrices across studies (e.g., generalized least squares, two-stage SEM, one-stage MASEM). We conducted a systematic search on applications of MASEM in the field of management and organizational psychology and showed that reanalysis of the four available data sets using multivariate MASEM can lead to different conclusions than applying univariate-r. In two simulation studies, we show that the univariate-r method leads to biased standard errors of path coefficients and incorrect fit statistics, whereas the multivariate methods generally perform adequately. In the article, we also discuss some issues that possibly hinder researchers from applying multivariate methods in MASEM.