A Critique of the Structural Equation Modeling Technique

The paper is non-technical critique of structural equation modelling (SEM). The SEM technique is discussed as an instrument to address some of the shortcomings of traditional statistical approaches. Traditional statistical techniques utilised in data analysis specify default models, assume measurement occurs without error, and are generally inflexible. In contrast, SEM requires specification of a model based on theory and research and is a multivariate technique with the ability to analyse both observed variables and latent (unobserved) constructs. The objective of this paper is to review the celebrated virtues of SEM vis-a-vis the traditional statistical approaches and to critically discuss the technique's limitations and misinterpretations in practice.