Timing alone does not establish causation. If unemployment falls after a job-training program begins, the program may have helped, but a broader economic recovery, migration, or changes in who enrolled could also explain the decline. Strong evaluation uses a relevant baseline or comparison group, examines conditions before and after the intervention, and states uncertainty honestly. Measurement choices matter too: an average improvement can hide worse outcomes for one region, while counting only participants may omit eligible people who were unable to navigate the application process.

Evaluation should distinguish a failure of execution from a failure of theory. If an agency never delivered the promised service, the implementation chain is the first place to investigate. If the service reached its target population exactly as designed but the social outcome did not improve, the policy's causal logic may be weak. In the water example, replacing every identified pipe is an output; lower contaminant levels and fewer illnesses are outcomes. Both matter, but they are not interchangeable evidence.

Findings do not mechanically dictate a political decision. Officials may disagree about goals, acceptable costs, fairness, and how much evidence is enough. Evaluation can lead to continuation, expansion, redesign, termination, or symbolic response. It can also return an issue to the agenda when journalists, auditors, courts, or affected communities publicize an implementation gap. On an exam, look for comparisons with goals or baselines rather than mere descriptions of activity.

Video lesson: Introduction to the public policy process | US government and civics | Khan Academy