Analyzing Administrative Data with Privacy Protection in Place
summary
Administrative wage and employment records (nearly complete since 1972) require heavy episode-splitting and cross-source linkage (linkage handled by ~10 staff) before analysis, because fields are sparse and events overlap; secure options exist (IZA-style remote-execution software, New York ADRF) but current review is manual and brittle — about 1,000 outputs/month get checked and median requires 40 observations under rule-based filters; exploratory SQL-heavy feature engineering drives privacy questions (person vs. spell vs. event) and breaks straightforward differential privacy.
AI-written summary · about
chapters (11)
- 0:00Admin data analysis intro
- 1:03Need for raw data access
- 7:07Challenges with admin data
- 15:13Data linkage and sampling
- 21:12Privacy methods and gaps
- 25:52Research workload diversity
- 31:09Output review processes
- 33:56Direct vs result protection
- 36:25Standardizing data prep
- 40:01Differential privacy for outputs
- 41:47Privacy versus research utility
related talks
55mPANEL: Where Are Statistical Agencies Going?
01h 00mStepping-up: The Census Bureau Sets an Example of How to Be a Good Data Steward in the 21st Century
19mCreating Global Economic Opportunity with Responsible Data Science
12mPer-Record Privacy and Its Application to Heavy-Tailed Economic Data
22mEnhancing Privacy through an Interactive On-demand Incremental Information
40mApplied Privacy Engineering Study
17mNegotiating Privacy/Utility Trade-Offs under Differential Privacy
01h 43mPerspective