Summary: Structural equation modeling has developed into a standard application in marketing. However, researchers are divided into two camps: proponents of a "hard", parametric approach exemplified by LISREL, and proponents of a "soft", non-parametric approach exemplified by PLS. Hardly any up-to-date literature exists that compares the statistical properties of the two approaches. The present paper aims to fill this gap. First, the development of both approaches is traced back to their roots in psychometrics and econometrics. Then, the nature of the mo-dels, the statistical properties of their estimators, and their key differences are discussed in detail. We conclude with recommendations in which cases one approach is to be favored over the other.