Sally Obenski and Jim FarberU.S. Census Bureau

CNSTAT Panel January 26, 2007

Expanding the Use of Administrative Records: Methods, Applications, and

Challenges

2

Overview

Overview and History Key technical breakthroughs Decennial and Survey Applications Medicaid Undercount Study Value of Integrated Data Sets Operational and Technical Constraints Policy Challenges Conclusions

3

Mandate for Administrative Records Use

Title 13, Section 6: Use administrative records information as

extensively as possible in lieu of conducting direct inquires  

Census Bureau Strategic Plan:Reduce reporting burden and minimize cost to

taxpayer by acquiring and developing high-quality data from sources maintained by other government and commercial entities

4

Legal Guidance and Protections

Title 13, U.S.C., Section 6, 9, and 214

Title 26, U.S.C., Section 6103(j)

Privacy Act of 1974

Paperwork Reduction Act

Government Information Security Reform Act

(GISRA)

E-Government Act of 2002, including Federal Information Security Management Act (FISMA)

5

Safeguarding Administrative Records

Consistent Application of Policies To ensure that projects have the appropriate legal

authorization, comply with existing data agreements, and provide adequate controls to protect confidentiality and privacy

Administrative Controls Numerous levels of approval Need-to-know access Removal of identifiable information Administrative Records Tracking System Security and confidentiality training

6

AR Program Evolution

Today

Program begins

mid 1990s

July 1993 1996

AR Research staff created

Survey launched to gather info on potential AR files

Early 1990s

Statistical uses of AR

conference held

July 1993

1999/2000

Projects included AREX 2000 and the

1999 StARS prototype

1999

Centralized program emerges

Data Stewardship

program begins

2001

AR Test for 2000

Census

Race Model Addresses

Quality Concerns

PVS Increases Linking

Capacity

Infrastructure investments

allow new interagency

collaborations

2004

2002 2005

7

StARS Provides Technical Infrastructure

Person & Address Databases consist of 7 national files:

IRS 1040 HUD TRACS IHS SSS

IRS 1099 HUD PIC Medicare

CY2004 records Persons Addresses

Raw input 894 million 767 million

StARS 308 million 152 million

8

Administrative Records Experiment Validated StARS

Local test of AR census models conducted in 5 counties

Coverage issues similar to Census 2000

Validated conformance of StARS to Census 2000 addresses & persons

Improvements to StARS continue, including move to more real-time redesign (E-StARS)

9

NUMIDENT Provides National Reference File

Social Security Administration (SSA) Numerical Identification (Numident) Transaction file with 803 million records

Collapse to 431 million unique SSN records

Usages:Look-up file that provides demographic dataSocial Security Numbers (SSNs) verification/

validation

10

Race and Hispanic Origin Model Rectified Quality Concerns

Initial weakness was dependence on race data from SSA’s SSN transaction file

Census 2000 records matched to SSN transaction file

Model completed missing linkages

11

Person Validation System (PVS) Increases Linking Capacity

Use master file of SSN/name/DOB as reference file

Link addresses with SSN reference file Match incoming census or survey record using

name, address, DOB Search within address first (high quality match) Search by name/DOB nationally if address

search not successful Replaces SSN with unique identifier (PIK)

12

Person Validation System (PVS) Increases Linking Capacity

Prepare NumidentPrepare Incoming DataRun Verification PhaseRun Geokey SearchRun Name SearchApply PIKs

Incoming Incoming DataData

Incoming Incoming DataDataNumidentNumidentNumidentNumident

VerificationVerificationVerificationVerification

Geokey Geokey SearchSearch

Geokey Geokey SearchSearch

NameNameSearchSearch

NameNameSearchSearch

FinalFinalFinalFinal

13

Implementing the ACS Provides Current Long Form Data

Designed to ameliorate constraints of decennial long form data collection

Provides means for timely analyses and estimates at small geographic areas

Provides means to push models based on less granular surveys to smaller geographies

14

AR Integral to Census Bureau Programs

Internal Revenue Service (IRS) 1040 Intercensal EstimatesSmall Area Income and Poverty Estimates

CMS Medicare and MedicaidNational Longitudinal Mortality StudySmall Area Income and Poverty EstimatesSmall Area Health Insurance Estimates

State Unemployment Insurance FilesLongitudinal Employer-Household Dynamics

15

Current Decennial Census Research

Using AR to “assign” age, race, sex, Hispanic Origin, when a record can be matched

Use AR to identify households with coverage problems

Determine if commercially available & other lists can improve & help build GQ frame

16

Emerging Survey Improvement (1)

Reducing ACS small area varianceUse AR as controls to adjust survey weights right

after nonresponse adjustmentPreliminary research highly promising

Obtaining characteristics on nonrespondentsCompared StARS persons to CPS responders to

ensure consistencyUsed StARS to obtain characteristics of

nonresponding households

17

Emerging Survey Improvement (2)

Reacting to disaster and other near-real time requirementsKatrina’s effect on the federal statistical system

and our lack of current response data highlighted need

Acquired the USPS National Change of Address File and FEMA’s emergency management and flood insurance files

Developing next generation StARS – near real-time measurements

18

Medicaid Undercount Study (1)

Survey estimates are important to policy research

Examining the large discrepancy between survey estimates and Medicaid enrollment figures

Multi-phased, interagency study, including academia

19

Medicaid Undercount Study (2)

Phase I: Examines quality and characteristics of Medicaid and Medicare files (2000-2002)

Phase II: Conducts national match of Medicaid files to Current Population Survey ASEC (2001-2002)

Phase III: Conducts selected state matches of Medicaid files to states in CPS ASEC

Phase IV: Conducts national match of Medicaid files to the National Health Interview Survey and compares results

20

Value of Integrated Data Sets

Provides more robust and accurate picture

Builds on strengths of both views while controlling for their weaknesses

Provides better statistics for input into simulations for predictions and funds distribution

As the demand for data increases and budgets decrease data re-use many be the only cost-effective option

21

Operational Constraints

File Acquisition ComplexitiesComplex Memoranda of UnderstandingState by state negotiationDifferences in content definition, quality, and

program rules over time

File lag timeMSIS (Federal Medicaid) lags by about 4 yearsMost lag for about a yearMany applications require more near real-time

response

22

Illustrative Examples

Federal Files—yearly acquisitions IRS 1040, IRS 1099, CMS Medicare, Medicaid

Federal Files—quarterly or monthly updatesSSA Numident, USPS NCOA

State Files—on as needed basisMD Food Stamps, MD TANF, MD Child Care

Subsidy State Files—on a quarterly basis

UI Wage and ES 202 (currently from 40 states) Commercial Files—INFOUSA

23

Technical Constraints

Obtaining the right data in the right format

Varying rates of validation (e.g., Medicare 99%, Medicaid 91%)

Coarseness of administrative data compared to nuances of surveys

Measuring error

24

Policy Challenges

Communicating the benefits vs. privacy concerns

Need for interagency teams to ensure accurate results

Interagency agreements and mission

“Ownership” of the integrated data sets

Growth of possible disclosure risks

Need for longitudinal data bases in order to find an anonomyzed person at an address at a point in time

25

Overcoming the Constraints

Resolving file acquisition issues may require OMB or Congressional assistance

Lag time for general demographics addressed by National Change of Address file—planning move to Enhanced StARS for more near real-time response

Standardized and centralized file acquisition

New files in address search phase and SAS-based matcher increased validation rates

Data Quality Standards team addressing measuring error in integrated data sets

Increasing inter-agency efforts for generating integrated data sets

26

Conclusions

New files and innovations leading to expansion of AR uses

New challenges continue to arise Regular updating of billions of records to have a near real-

time response system Effectively acquiring state-based records Understanding integrated data sets

At incipience of a new generation of products, services, and inter-agency opportunities

27

Contact Information

Sally M. ObenskiAssistant Division Chief for Administrative

Records ApplicationsData Integration DivisionU.S. Census BureauWashington D.C. 20233-8100

Email: Sally.M.Obenski@Census.GovPhone: (301) 763-4044Cell: (301) 873-7615