Experimentally established expected effects in random-intercept cross-lagged panel models when causality holds

There are indications that findings from random-intercept cross-lagged panel models (RI-CLPM) may be biased and spurious, similarly to findings from traditional cross-lagged panel models. The objective of the present study was to establish what effects to expect in analyses with RI-CLPM when causality holds. We generated experimental data with truly causal effects by estimating the weight of a container after adding or removing stones to/from the container. Data were analyzed with three complementary RI-CLPMs as well as three corresponding multilevel model (MLM) analyses of person-mean centered scores. The results suggested that if a prior level on a predictor X has a within-individual causal increasing effect on an outcome Y, we should expect analyses with RI-CLPM to show that prior within-individual X has: (1) A positive effect on subsequent within-individual level of Y when adjusting for prior within-individual level of Y; (2) A negative effect on prior within-individual Y when adjusting for subsequent within-individual Y; (3) A positive effect on the subsequent within-individual Y - prior within-individual Y difference. Corresponding analyses of person-mean centered scores with MLM should yield similar effects. In the case of causal decreasing effects, we should expect effects with the opposite signs (i.e., negative, positive, negative, respectively). These experimentally established expected effects when causality holds can help researchers triangulate and scrutinize findings from RI-CLPM in order to establish if observed effects might be truly causal or if they appear to be spurious.