Propensity Scores

A propensity score is the probability that an individual received a particular treatment, given a set of researcher-identified variables related to self-selected treatment participation. The use of propensity scores is grounded in the logic of the counterfactual: To know the true effect of a treatment on the outcome, a researcher must also know what the outcome would have been had participants not received treatment. However, this is not something researchers will ever know. Accounting for the propensity for treatment, given a set of variables related to self-selection into that treatment, can increase precision in estimating the effect of the treatment on the outcome.

In the context of educational and social sciences research, treatment typically refers to a course, intervention, or program. Propensity scores are used in observational or quasi-experimental studies, in which researchers are unable to randomly assign participants to conditions, making accurate evidence-based causal claims challenging. Instead, students typically self-select or choose to participate in educational programs. Propensity scores provide a means of accounting for variables related to self-selection bias, allowing the researcher to create matched, or balanced, treatment and comparison groups. Propensity scores can also be used to adjust, or assign weight to, outcome analyses to account for self-selection bias, thereby mimicking a randomized controlled study.