Performance of shiller lag estimators: some additional evidence