Improving the efficiency of the shifted penalty method