A specific type of bias that occurs when unobserved or
unobservable characteristics of units in an intervention or program differ from the
characteristics of the control or comparison group, and those unobservable
characteristics are correlated with outcomes of interest