She studies how AI changes the work of economists and policy teams, combining causal inference, public data, and software testing to keep every step of a faster workflow checkable.
Too Early To Say is an independent research project. The site publishes policy findings, known-truth econometric cases, open methods, and a public queue of contributions still under validation.
Does AI move the real cost of policy research from running the analysis to specifying it, checking it, and judging what it means?
Three fields meet in the work: policy applications set the stakes, econometrics brings the identification discipline, and software engineering adds the tests, assertions, versioned artifacts, and failure gates that probabilistic tools need.
Evidence on productivity, error, review, and skill.
Difference-in-differences in Python plants an effect of 1.60 and catches a clean regression that answers the wrong comparison.
Agents: implement and rerun. Researcher: defines the estimand, answer key, and failure rule.
Findings with traceable data, methods, and limits.
The Food Desert Myth records a published 4.7-fold comparison now under reconciliation across the article, data, and public materials.
Agents: assemble and reconcile. Researcher: defends the comparison and interprets the policy claim.
Reusable tests for econometric and workflow claims.
Instrumental Variables in Python pairs an F-statistic of 2,051 with a causally invalid estimate when exclusion fails.
Agents: execute diagnostics. Researcher: defines the data-generating process and release standard.
The work spans academic research, public-health practice, applied econometrics, and open research infrastructure.
Senior Health Economist at the Santa Clara County Department of Public Health and visiting faculty at the University of California, Berkeley, School of Public Health.
Doctoral work spanning health, education, public finance, and social policy.
Co-founder of the California Association of Public Health Economists and recipient of the APPAM 40 for 40 Fellowship.
Research partners can contribute policy cases, external replications, domain adjudication, or institutional settings for measuring what AI-assisted evidence production costs.
The analyses on Too Early To Say represent personal research interests and do not reflect any employer or affiliated institution.