Should observational studies be designed to allow lack of balance in covariate distributions across treatment groups?

Rubin, Donald B. · Statistics in Medicine · 2009