Modelling Formative Second Order Constructs in PLS

PLS-SEM has become a popular technique in the past years with an increasing number of articles. Using PLS-SEM, researchers have created more sophisticated models, with second order constructs (also known as Higher Order constructs, multidimensional constructs or hierarchical latent variable models). Second order constructs are becoming more and more popular in the PLS-SEM literature. However, research discussing the use of second order constructs has mainly focused on second order constructs with a reflective measurement, with limited research with guidelines on how to estimate formative second order constructs. Therefore, this paper strives to help researchers by providing a comparison of the different approaches typically used to estimate a formative second order construct (repeated indicators, two stage and hybrid approaches) with an empirical example. The findings from this study provide useful insights for academics and researchers deciding how to estimate a second order formative construct.