The variance-covariance matrix for the seemingly unrelated regression estimator is expressed as an ordinary-least-squares variance-covariance matrix.This permits new insights into the efficiency of seemingly unrelated regression, especially the role of correlations among variables.