SEM is more powerful than OLS path analysis because it avoids several OLS problems, notably difficulties in meeting assumptions. Unlike OLS, SEM can accommodate measurement errors, correlated residuals, and nonrecursive models that allow for reciprocal causation, and it can analyze causal models involving latent variables.

According to the text, structural equation modeling (SEM) using maximum likelihood estimation is a more powerful approach to path analysis that avoids several problems in ordinary least-squares (OLS) estimation, especially difficulties in meeting assumptions. OLS path analysis requires more restrictive assumptions, whereas SEM relaxes these. Specifically, SEM can accommodate measurement errors, correlated residuals, and nonrecursive models that allow for reciprocal causation. Another key advantage is that SEM can be used to analyze causal models involving latent variables, which are constructs not measured directly but captured through two or more measured indicator variables. In practice, SEM proceeds in two phases: first testing a measurement model via confirmatory factor analysis, and then testing the theoretical causal model. SEM also yields information on the significance of individual paths and overall model fit using statistics such as the goodness-of-fit index.

Key points

SEM avoids several problems in OLS estimation, notably difficulties in meeting assumptions.
SEM can accommodate measurement errors, which OLS path analysis cannot.
SEM can include correlated residuals, addressing a limitation of OLS.
SEM supports nonrecursive models with reciprocal causation.
SEM can analyze causal models involving latent variables, captured by multiple measured indicators.
SEM offers overall model fit statistics, such as the goodness-of-fit index (GFI).

What are the advantages of structural equation modeling (SEM) over ordinary least-squares (OLS) path analysis?