Methods for modelling reflective-formative second order constructs in PLS
Purpose This paper aims to discuss the estimation of PLS models with second-order formative constructs as existing research has mainly focused on second-order constructs with a reflective measurement. Design/methodology/approach Using a model grounded on Roger’s innovation diffusion theory applied to online travel shopping, an empirical application is used to assess and compare the different approaches used to estimate a formative second-order construct. The proposed model examines the innovations characteristics that have an impact on intentions to purchase travel online, using data from a convenience sample of 1,732 responses. Findings The findings show that all approaches produce similar results regarding the path coefficients, the predictive relevance of the model and the explained variance. The main differences between the approaches are related to the weights of the first-order constructs on the second-order construct and the significance of those weights. Several recommendations are made for researchers on which approach to use. Originality/value Since most research has focused on second-order constructs with a reflective measurement and there is limited research with formative second-order constructs, this paper provides a comparison of the different approaches typically used to estimate a formative second-order construct and present useful guidelines for researchers to decide the method to analyse a model with second-order constructs and how to assess formative second-order constructs.