Estimating willingness to accept using paired comparison choice experiments: tests of robustness