ABSTRACT
Herbert L. Costner had demonstrated earlier that with three indicators for each substantive variable in a recursive cross-sectional model the analyst can both test measurement models and estimate causal parameters. While the panel design has been used primarily to resolve difficulties in causal analysis, there has been an emphasis in the causal-models approach to measurement error on using panel observations to eliminate complications suggested by the acknowledgment of measurement imperfection. Work on measurement error in panel models has focused on only the very simplest cases. In attempting analysis, the researcher must deal simultaneously with the inference problems of linear panel models and with those arising from the existence of measurement error. David E. Wiley and James A. Wiley have demonstrated that the assumption of stable standardized coefficients requires that both the true population variances and the measurement-error variances be stable over the waves of observations.
