Estimators of average treatment effects under unconfounded treatment assignment are known to become rather imprecise if there is limited overlap in the covariate distributions between the treatment groups. But such limited overlap can also have a detrimental effect on inference, and lead for...