This paper explores various procedures of panel data in the estimation of causal models. The reported analyses are from the Monitoring the Future study, a nationwide questionnaire survey of 16,000 to 17,000 high school seniors conducted annually since 1975. First, the parameters of causal models are estimated in which the dependent variables are two reports on the overall quality of life. The exercise is then repeated using two measures of drug abuse as the dependent variables. Multiple regression analyses were performed and results are presented briefly in tabular form. Change scores offered a clear advantage over the use of static scores by reducing the potential for misspecified models. In general, the standard errors of estimates based on change scores are greater than those based on static scores. However, present results indicate that for reasonably large sample sizes, and for a wide range of stabilities of the true values and measurement errors in the predictor variables, the ratio of sampling errors for estimates based on change scores to those based on static scores is not excessively large. Panel data as estimates of dynamic causal models are also discussed. (PN)