For linear regression models with cross-section or panel data, it is natural to assume that the
disturbances are clustered in two dimensions. However, the finite-sample properties of two …