Journal of Global Marketing

Multimethod Approaches to Structural Equation Modeling (SEM) in Global Marketing Research

Selecting an appropriate data analysis method is difficult in every research project. Even when researchers identify a certain method type, an estimator's choice and application involve numerous decisions. This also holds for structural equation modeling (SEM) applications, which facilitate the estimation of complex interrelationships between constructs and their indicators (Baumgartner & Weijters, 2020; Sarstedt et al., 2021). Addressing this concern, researchers have compared different SEM types—factor-based and composite-based—on conceptual and empirical grounds (e.g., Fornell & Bookstein, 1982; Rigdon et al., 2017; Sarstedt et al., 2016) and alternative estimators for each type (e.g., Boomsma & Hoogland, 2001; Hair et al., 2017). Moreover, review studies on the use of factor-based (e.g., Baumgartner & Homburg, 1996) and composite-based SEM methods, most notably partial least squares (e.g., Sarstedt et al., 2022), have revealed considerable variability in researchers’ applications of these methods.

More recently, research by Sarstedt et al. (2024) has highlighted the role of subjectivity in the scientific process when working with SEM methods. Despite following rigorous scientific methods, research findings can vary due to the complexity of data analysis and the choices researchers make in the course of their analysis (Buchanan et al., 1998; Schweinsberg et al., 2021; Silberzahn et al., 2018). Sarstedt et al. (2024) demonstrated how different analytical decisions in applying different SEM estimators led to diverse outcomes, even when analyzing the same dataset. Their findings underscore the need for transparency in the research process, a responsibility that lies with all of us as researchers. They suggest that we consider alternative workflows to account for uncertainty and variability in results. At the same time, this finding calls for a multimethod approach to SEM—particularly in global and international marketing research, which has a long tradition in SEM use (Richter et al., 2016).

Using multiple SEM estimators helps overcome the narrow perspective that one method is universally superior to others (e.g., Rönkkö et al., 2016; Sarstedt et al., 2016). Instead, such an approach helps establish a complementary view that appreciates and considers each method's specific goals, advantages, and limitations (Cook & Forzani, 2023; Jöreskog & Wold, 1982; Rigdon et al., 2017). This will enable researchers to better focus on advancing the various SEM estimators, especially considering the manifold challenges of big data, machine learning, and artificial intelligence.

In addition, a multimethod approach to SEM helps ensure robust results are presented. For example, coefficients may vary slightly between methods, but the conclusions regarding the significance and relevance of the coefficients should be consistent. When this is the case, a multimethod approach to SEM supports the robustness of results and the derivation of reliable managerial implications. Otherwise, the divergence of results across methods helps researchers identify potential problems with their theoretically established model, the dataset used, the type of statistical methods, or the technical requirements and pitfalls associated with each method.

Finally, a multimethod approach to SEM enables researchers to incorporate methods with different analytical objectives into their projects (Sharma et al., 2024). For example, while factor-based SEM emphasizes an explanatory perspective (e.g., based on model fit), partial least squares implement a causal-predictive perspective (e.g., based on the model’s predictive power) (Hair & Sarstedt, 2021). As a result, researchers can combine two sides of the same coin in their research, leading to a stronger consideration of prediction in combination with explanation since researchers have long focused on an explanation-oriented approach in social science disciplines (Hofman et al., 2017; Hofman et al., 2021; Sarstedt & Danks, 2022).

Taken together, scholarly attention is needed to comprehensively examine how a multimethod approach to SEM use will strengthen global marketing research by providing richer and more robust findings, implications, and conclusions. This special issue aims to shed light on this research question from both conceptual and methodological perspectives and in the context of applications in global marketing research. More specifically, the special issue editors are interested in contributions within the following broad themes in multimethod SEM, including examples of potential topics (but not limited to the themes listed below):

• Conceptual frameworks combine alternative SEM methods for different purposes, especially when considering explanation and prediction.
• Examples of how different subjective choices can lead to different SEM results and recommendations on addressing potential divergencies.
• Methodological contributions further strengthen the joint use of alternative SEM methods (e.g., model specification searches that aim for high model fit and predictive power).
• Global marketing studies that combine multiple SEM methods in one research project. For example, such studies could contrast results from factor-based SEM with different estimators, composite-based SEM methods (e.g., generalized structured component analysis, partial least squares), and sum scores regression.
• Methodological uncertainty: Developing explanations and recommendations when results vary across methods. This topic may include conceptual contributions, applications, or both.
• Replication studies of global marketing research demonstrate how a multimethod approach to SEM supports the robustness of results and extends findings and conclusions based on additional analytical results.
• Studies demonstrating the variability of results and thus the robustness methods for data generation processes that are prototypical for marketing studies using SEM.
• The combination of qualitative research and necessary conditions with quantitative research using various SEM methods.

The editors of this special issue invite studies that make theoretical, conceptual, methodological, or empirical advances. While this special issue is primarily open to quantitative studies, we also welcome conceptual articles and systematic literature reviews. Studies that take an interdisciplinary approach are also interesting for this special issue.

(for list of references, please contact the Guest Editors)