A Cautionary Note on Avoiding the Initial Conditions Problem in the CLPM-RI

The cross-lagged panel model with random intercepts (CLPM-RI) is often used to analyze reciprocal relationships, because it separates within-unit dynamics and between-unit differences by modeling unit-specific heterogeneity as a latent variable. Through a simulation study, it is shown that common specifications of the CLPM-RI often produce biased estimates of autoregressive (AR) and cross-lagged (CL) effects due to the initial conditions problem. These biases are often substantial enough to produce inaccurate inferences about the long-term stability of the observed system. An alternative approach suggested by Bollen and Brand (2010) resolves the initial conditions issue by conditioning unit-specific heterogeneity on the first observation and provides more accurate estimates and reliable inferences about system stability compared to the CLPM-RI.