Cross-Lagged Panel Analysis Using Structural Equation Modeling: Trends and Suggestions Based on a Literature Review of 111 Studies

Abstract

Cross-lagged panel analysis based on structural equation modeling (SEM) has been widely used in the behavioral sciences. The present study provides a literature review based on 111 studies published in major psychology journals that conducted cross-lagged panel analysis. Through this review, we offer an overview of current trends in study design, analytical methods (e.g., estimation method, missing data handling), and model selection, and discuss practical challenges and implications for future applications. A major finding is that the use of the random intercept cross-lagged panel model (RI-CLPM) has increased substantially (41% of the studies) over the past few years. At the same time, most studies rely on longitudinal data with a small number of measurement occasions (e.g., T = 3 − 4). Studies that employed the RI-CLPM on the grounds that it enables inference about within-person relations tended to conduct only limited comparisons with potential alternative models, and only 3% of the studies evaluated model fit from the perspective of local fit (i.e., residual correlations). Although many studies using psychological scales relied on sum scores, only 40% of the studies examined longitudinal measurement invariance, suggesting that there isstill room for improvement.