430 likes | 578 Views
Vital Statistics Enhancement Team. Amy Symens Smith U.S. Census Bureau Linda Gage , California Gregg Williams , Alaska FSCPE, Research & Methodology Subcommittee FSCPE Mid-year Meeting March 28-29, 2005 Philadelphia, PA. Vital Statistics Enhancement Team.
E N D
Vital Statistics Enhancement Team Amy Symens Smith U.S. Census Bureau Linda Gage, California Gregg Williams, Alaska FSCPE, Research & Methodology Subcommittee FSCPE Mid-year Meeting March 28-29, 2005 Philadelphia, PA
Vital Statistics Enhancement Team • Census Bureau and FSCPE collaboration • 1st research project of this type. • Project designed to address FSCPE • recommendations and Census Bureau response • concerning the current processing of vital • statistics input data used to produce subnational • population estimates.
FSCPE Recommendations • FSCPE vital statistics inputs should be used in the estimates. • More current vital statistics inputs than NCHS • More accurate inputs by race and Hispanic-origin • Actual July-June annual period • PEP should • Pursue a data sharing agreement with NCHS for the FSCPE • Enhance annual vital statistics reports to learn more about effects of cut-off dates and timing of final data. • PEP and FSCPE Research and Methods Subcommittee should conduct joint research on NCHS and FSCPE vital statistics data and make further recommendations.
Census Bureau Response • Overall agreement with all recommendations and enthusiastic about a joint project. • Concerns raised: • Recommendation to use more current data that have not been previously requested. • Because FSCPEs may differ in their ability to provide these data we may have to design a new system to accommodate missing data. • While designing a new system is not insurmountable, it likely will require additional time in the processing.
Vital Statistics Enhancement Research • Objective: Design a process and methodology to efficiently use all the vital statistics data available from the National Center for Health Statistics (NCHS) and the FSCPEs in the production of subnational population estimates. • Goals: • Streamline requests for data from FSCPE. • Examine NCHS/FSCPE differences and design a method to combine data drawing on the strengths of each data source. • Examine characteristics data. • Examine data for more current years.
Research Agenda & Schedule • 1. Review of NCHS & FSCPE raw birth & death data. Comparison to • subnational population estimates processing requirements. • End date: 2/3/05 • 2. Review of current method used to reconcile differences in the • NCHS & FSCPE data to create population estimates input data. • End date 2/24/05 • 3. Comparison of NCHS & FSCPE data. End date: 3/10/05-6/2/05 • 4. Review of the Data Collection Process Survey to gain insights into • cut-off dates & additional data availability. End date: 3/3/05 • 5. Testing, if changes are recommended. End date: 6/24/05 • 6. Preparation of report/recommendation. End date: 7/1/05
Comparison of NCHS/FSCPE Data • Evaluation tools: • National maps of outliers by year for births and deaths • Ranking of outliers: 2000 births • Regional maps based on ranking of 2000 births outliers • Census 2000 data for population under age 1 • Index of Dissimilarity: 2000 births • ID = ½ * {ABS(NCHS or FSCPE – Census)}
Ranking of Outliers: 2000 Births • Focus on 4 areas: • New Jersey and New York • DC Metro: DC, VA and MD • State of Georgia • Nevada, California and Arizona
Index of Dissimilarity: 2000 Births • Population estimates must reflect change in the population as if a census was taken. • To use birth certificate data it must capture the same universe as that captured by the census. • To measure differences use the Index of Dissimilarity to quantify the differences between each data source and the census age zero population.
Index of Dissimilarity: 2000 Births • In 22 states the NCHS index was smaller. • In 16 states the FSCPE index was smaller. • In 9 states both indexes were the same. • NCHS index larger than national index of 2.54 in: VA, IA, AR, GA, MT, WY, CO and AK • FSCPE index larger than national index of 2.59 in: WV, AR, IA, WI, WY, MT, CO, HI, and AK
Current Reconciliation Method • Goal is to draw on the strengths of each data source. • Current Reconciliation Method: • Remove counties with zero differences or counties where no FSCPE data was supplied so NCHS data is used. Sum counties • Rake FSCPE counties to the sum of counties in #1 • Rake NCHS data with characteristics to new county totals in #2. • Method draws on the strengths of each data source: • FSCPE st/co distributions NCHS characteristics data
Current Reconciliation Method In both 2001 and 2002 the majority of counties (54.1% and 50.6%) were removed from the rake process and in 2000 nearly forty-nine percent were removed. Of those counties that were raked, the rake factors for each year were very small, ranging from 1.00200 in 2002 to 1.00324 in 2000.
FSCPE or NCHS Input Births Compared to Census Bureau Output Births
NCHS/FSCPE Discrepancies in Births(California Counties adjacent to Nevada)
Census Bureau/FSCPE Joint Research • Found great consistency between NCHS and FSCPE vital statistics. • Identified and continue to analyze areas with large differences in NCHS/FSCPE vital statistics. • Identified processing methods that distort the vital statistics input data.
Future Steps • Report/Recommendation by July 1, 2005 • Likely will not be included in 2005 vintage, but will be considered for 2006 vintage • Additional research needed on assignment of race/Hispanic origin in subnational estimates.