Working with Second-order Construct in Measurement Model: An Illustration Using Empirical Data

Abstract

With human resource (HR) roles evolving to encompass wider responsibilities, HR decision-making in organizations has become more complex than ever. This has compelled researchers in the area to move beyond simplistic models to testing models that involve studying the relationship between multiple independent and dependent variables in the presence of moderators and mediators, in order to make relevant contribution to managerial decision-making. Thus, research in the field is heavily dependent on multivariate techniques that can run several regressions simultaneously and can study the influence of one variable on the other, in presence of the other variables in the model. Structural equation modeling is the most widely used multivariate technique and involves two phases – measurement model to test reliability and validity of study constructs and structural model that involves path diagrams to test the causal relationships between these constructs. At times, however, the researcher might run into trouble with validity issues of constructs in the measurement model; especially when dimensions of a larger construct are used as independent constructs in the study. Introducing a second-order construct in such a case could be the solution to proceed further. Using empirical data, this chater illustrates the case of such a problematic measurement model and details the research methodology of introducing and working with a second-order construct in a step-wise manner, starting with an exploratory factor analysis and subsequently, moving toward a confirmatory factor analysis, highlighting the best practices to be followed while using these statistical techniques.