Joshua Baer Gilbert

Examines how teacher and school “value-added” ratings—measures based on student test scores—can change depending on which test questions are included on an exam. Finds that these ratings are often less reliable and less stable than they appear, suggesting that schools and teachers may be judged too confidently based on tests that capture only part of what students know.

Examines how researchers’ choices about measuring outcomes—such as how they score surveys or tests—can affect conclusions about whether a policy or treatment worked. Finds that simple measurement error usually matters more than the specific scoring method, meaning weak or noisy measures can make real effects look smaller or disappear altogether.

Explores a new statistical method for figuring out which students or groups benefit most from a policy or educational program by analyzing responses to individual test questions instead of only overall scores. Argues that looking at item-level data can reveal more detailed and accurate patterns of who is helped—or not helped—by an intervention, which could improve how policies and programs are evaluated.