Multiple imputation (MI) is often presented as an improvement over listwise deletion (LWD) for regression estimation in the presence of missing data. Against a common view, we demonstrate anew that ...