For a suitably chosen ridge penalty parameter, the ridge regression estimator uniformly dominates the maximum likelihood regression estimator in terms of the mean squared error. Analogous results for the ridge maximum likelihood estimators of covariance and precision matrix are presented.