Temporal Asymmetry in Cross-Lagged Panel Models: Improved Causal Inference from Longitudinal Data

Inferring causal direction from longitudinal data remains challenging in social sciences, especially without experimental control. Traditional methods like Granger causality and cross-lagged panel models (CLPMs) assess predictive relationships but cannot distinguish directional causal effects from reciprocal associations due to autoregression, latent confounding, or measurement error. We propose a structured, hypothesis-driven CLPM extension that tests temporal asymmetry through nested structural equation models. By formally comparing forward and reverse cross-lagged effects under constrained models, this method identifies asymmetric predictive influences, strengthening causal inference. Demonstrated through theoretical analysis, simulations, and real-world data, our approach maintains CLPM accessibility while providing a principled, statistically robust test for causal direction in observational time-series