The authors describe methodological issues of panel analysis designs; in particular, autocorrelation of errors across waves. They recommend constructing a “pooled” model of data from all waves to deal with design and estimation problems, and present results from a simulation of the behavior of alternate estimators for pooled models.