An Empirical Comparison of the Efficacy of Covariance-Based and Variance-Based SEM

An Empirical Comparison of the Efficacy of Covariance-Based and Variance-Based SEM

2009 | Reinartz, W.; Haenlein, M.; Henseler, J.
An empirical comparison of the efficacy of covariance-based and variance-based SEM (CBSEM vs. PLS) is conducted through a large-scale Monte Carlo simulation. The study investigates the relative performance of these two approaches under various conditions, including the number of indicators per construct, sample size, distribution, and indicator loadings. The results show that CBSEM generally outperforms PLS in terms of parameter consistency and accuracy, especially when the sample size exceeds 250 observations. However, PLS is preferred when the focus is on prediction and theory development, as it tends to have higher statistical power. PLS is also more suitable for small sample sizes and when dealing with formative indicators. The study highlights that CBSEM is more robust to violations of distributional assumptions, while PLS does not require such assumptions. The findings suggest that researchers should choose CBSEM for confirmatory studies and PLS for exploratory or predictive research. The study provides quantitative guidelines for selecting between the two methods based on the specific research context and requirements.An empirical comparison of the efficacy of covariance-based and variance-based SEM (CBSEM vs. PLS) is conducted through a large-scale Monte Carlo simulation. The study investigates the relative performance of these two approaches under various conditions, including the number of indicators per construct, sample size, distribution, and indicator loadings. The results show that CBSEM generally outperforms PLS in terms of parameter consistency and accuracy, especially when the sample size exceeds 250 observations. However, PLS is preferred when the focus is on prediction and theory development, as it tends to have higher statistical power. PLS is also more suitable for small sample sizes and when dealing with formative indicators. The study highlights that CBSEM is more robust to violations of distributional assumptions, while PLS does not require such assumptions. The findings suggest that researchers should choose CBSEM for confirmatory studies and PLS for exploratory or predictive research. The study provides quantitative guidelines for selecting between the two methods based on the specific research context and requirements.
Reach us at info@futurestudyspace.com