Abstract:
|
A good data steward produces and disseminates relevant and timely statistical information; conducts objective, credible and accurate statistical activities; and protects the trust of information providers by ensuring the confidentiality and exclusive statistical use of their responses. If you recognize that language, then you know it is from Title III of the Foundations for Evidence-Based Policymaking Act of 2018. It is tempting to say that we should do as much as possible for all three objectives. As statistical agencies and Internet giants are both realizing, the second and third objectives are competing uses of the same resources. The ability to deliver the first objective depends critically on the way that competition is resolved. Formal privacy—protecting confidentiality with provable mathematical models—provides the technology for understanding the feasible trade-offs between accuracy and privacy loss. Delivering timely, relevant, objective, credible, and accurate statistics within the scope of this technology requires a much deeper understanding of actual uses, and limitations, of the statistical information produced by agencies and private companies. My talk will look at some important current applications, including the 2020 Census.
|