This paper reports on the results of some sampling experiments on six structural equation estimators: direct least squares, two-stage least squares, Nagar's unbaised k-class estimator, limited-information maximum likelihood, three-stage least squares and full-information maximum likelihood. Sensitivity of the relative performances of the estimators to the parameter values chosen in the models and to the values of the exogenous variables used is studied. It is found that differences among the estimators are not great and that they are sensitive to the parameters used.