Partial least squares structural equation modeling: recent advances in banking and finance.

Abstract

We conceived this book because of the absence of partial least squares structural equation modelling (PLS-SEM) in banking and finance disciplines. Yet, the PLS-SEM method has been broadly accepted and used in disciplines such as accounting, health care, hospitality management, management information systems, marketing, operations management, strategic management, supply chain management, and tourism. Besides, the method enjoys an increasing application in a wide range of additional disciplines such as economics, engineering, environmental sciences, medicine, political sciences, and psychology (Richter et al. 2016). Against this background, we also expect an adoption of PLS-SEM in the banking and finance disciplines. As a causal-predictive method, PLS-SEM has a wide spectrum of practical applications to managerial challenges. Unfortunately, secondary data frequently found in business databases are unlikely to satisfy such constraints as homogeneity in the population, and measurement errors being uncorrelated. With the ever-increasing availability of secondary data, PLS-SEM’s soft modelling approach fits exploratory research, where theory has not been fully developed. Using the PLS-SEM approach is recommended when (a) the objective is explaining and predicting target constructs and/or detecting important driver constructs, (b) the structural model has formatively measured constructs, (c) the model is complex (with many constructs and indicators), (d) the researcher is working with a small sample size (due to a small population size), and (e) the researcher intends to use latent variable scores in follow-up studies (Sarstedt et al. 2017). PLS-SEM is relatively robust with non-normal data. However, researchers should not use the latter characteristic and/or small sample sizes as the sole argument for selecting PLS-SEM but focus on the goal of their empirical analysis (Rigdon 2016). This is important to proactively attend to potential criticism that has been put forward with regard to PLS-SEM (for further details on this debate, see Sarstedt et al. 2016) The applications in this handbook further pioneer PLS-SEM adoptions in the banking and finance disciplines. New PLS-SEM developments will further expand the method’s usefulness to banking and finance studies. These advances primarily address both the method’s explanatory and predictive capabilities. Examples of recent enhancements include methods for uncovering unobserved heterogeneity, different multi-group analysis approaches, testing measurement invariance of composites, overall goodness-of-fit measures, and novel approaches of prediction-oriented results evaluations (Richter et al. 2016; Sarstedt et al. 2017). We also expect that the PLS-SEM method will experience extensions in the direction of longitudinal data analysis and multilevel modelling, which will become particularly beneficial for the banking and finance disciplines given the characteristics of data usually used in such studies.