The examination of between-group differences in theoretical relationships of interest is central to the conduct of research in the organizational and behavioral sciences, including Information Systems research. One such approach for examining these differences relies on the conduct of multi-group PLS analyses, where each group of interest is modeled separately, and the structural results from each analysis are then compared. Though this approach has been employed in the empirical literature, the ability of separate analyses to obtain models that are equivalent from a measurement perspective - which is a pre-requisite to making any comparisons between them - has not been carefully studied. In this research we perform three simulation studies under varying invariant conditions that highlight the performance of PLS in this regard. Our results indicate that sampling variability plays a major role in whether equivalent results can be obtained, and showcase the conditions in which the technique performs best.
