Estimating hierarchical constructs using consistent partial least squares: The case of second-order composites of common factors

Purpose
Many important constructs of business and social sciences are conceptualized as composites of common factors, i.e. as second-order constructs composed of reflectively measured first-order constructs. Current approaches to model this type of second-order construct provide inconsistent estimates and lack a model test that helps assess the existence and/or usefulness of a second-order construct. The purpose of this paper is to present a novel three-stage approach to model, estimate, and test second-order constructs composed of reflectively measured first-order constructs.

Design/methodology/approach
The authors compare the efficacy of the proposed three-stage approach with that of the dominant extant approaches, i.e. the repeated indicator approach, the two-stage approach, and the hybrid approach by means of simulated data whose underlying population model is known. Moreover, the authors apply the three-stage approach to a real research setting in business research.

Findings
The study based on simulated data illustrates that the three-stage approach is Fisher-consistent, whereas the dominant extant approaches are not. The study based on real data shows that the three-stage approach is meaningfully applicable in typical research settings of business research. Its results can differ substantially from those of the extant approaches.

Research limitations/implications
Analysts aiming at modeling composites of common factors should apply the proposed procedure in order to test the existence and/or usefulness of a second-order construct and to obtain consistent estimates.

Originality/value
The three-stage approach is the only consistent approach for modeling, estimating, and testing composite second-order constructs made up of reflectively measured first-order constructs.