Selection Bias

Selection bias relates to unobservables that may bias outcomes (for example, individual ability, preexisting conditions). Randomized experiments solve the problem of selection bias by generating an experimental control group of people who would have participated in a program but who were randomly denied access to the program or treatment. The random assignment does not remove selection bias but instead balances the bias between the participant and nonparticipant samples. In quasi-experimental designs, statistical models (for example, matching, double differences, nstrumental variables) approach this by modeling the selection processes to arrive at an unbiased estimate using nonexperimental data. The general idea is to compare program participants and nonparticipants holding selection processes constant. The validity of this model depends on how well the model is specified.

Source: Baker, J. L. (2000): Evaluating the Impact of Development Projects on Poverty. A Handbook for Practioners. The World Bank, Washington, D.C.