Evidence and evaluation
Program and policy evaluation
Most programs are judged on activity: people served, dollars spent. Funders and boards increasingly want to know something harder, which is what changed because of the program, compared with what would have happened anyway.
Questions this answers
- Did the program reduce the outcome it was designed to reduce, and by how much?
- Did the effect differ across regions, income groups or types of participant?
- Is the effect large enough to justify continuing or expanding the program?
- What can the data we already hold tell us, and what would a stronger evaluation need?
Methods
Difference-in-differences and event-study designs, matching and weighting, regression discontinuity, interrupted time series, target-trial emulation on administrative data, and randomized designs where they are feasible. We specify the estimand, the precise quantity being measured, before any analysis starts.