CB-SEM vs PLS-SEM methods for research in social sciences and technology forecasting

Abstract

This study compares the two widely used methods of Structural Equation Modeling (SEM): Covariance based Structural Equation Modeling (CB-SEM) and Partial Least Squares based Structural Equation Modeling (PLS-SEM). The first approach is based on covariance, and the second one is based on variance (partial least squares). It further assesses the difference between PLS and Consistent PLS algorithms. To assess the same, empirical data is used. Four hundred sixty-six respondents from India, Saudi Arabia, South Africa, the USA, and few other countries are considered. The structural model is tested with the help of both approaches. Findings indicate that the item loadings are usually higher in PLS-SEM than CB-SEM. The structural relationship is closer to CB-SEM if a consistent PLS algorithm is undertaken in PLS-SEM. It is also found that average variance extracted (AVE) and composite reliability (CR) values are higher in the PLS-SEM method, indicating better construct reliability and validity. CB-SEM is better in providing model fit indices, whereas PLS-SEM fit indices are still evolving. CB-SEM models are better for factor-based models like ours, whereas composite-based models provide excellent outcomes in PLS-SEM. This study contributes to the existing literature significantly by providing an empirical comparison of all the three methods for predictive research domains. The multi-national context makes the study relevant and replicable universally. We call for researchers to revisit the widely used SEM approaches, especially using appropriate SEM methods for factor-based and composite-based models.

Cite this

Ganesh Dash, Justin Paul (2021). CB-SEM vs PLS-SEM methods for research in social sciences and technology forecasting. Technological Forecasting and Social Change. https://doi.org/10.1016/j.techfore.2021.121092

@article{ganesh2021cbsem,
  title  = {CB-SEM vs PLS-SEM methods for research in social sciences and technology forecasting},
  author = {Dash, Ganesh and Paul, Justin},
  year   = {2021},
  journal = {Technological Forecasting and Social Change},
  doi    = {10.1016/j.techfore.2021.121092}
}

Citations in this library

Stored only when both papers are in the corpus — 2,482 citations exist upstream.

References

Cited by