Comparing Alternative Models for Using Decennial Census Data in SAIPE State Poverty Estimates

• Activity Number: 386
• Type: Contributed
• Date/Time: Thursday, August 15, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods*
• Author(s): Elizabeth Huang*+ and William Bell
• Affiliation(s): U.S. Census Bureau and U.S. Census Bureau
• Keywords: Small area estimation ; Fay-Herriot model ; Akaike's AIC

Abstract

The Census Bureau's Small Area Income and Poverty Estimates (SAIPE) Program produces poverty estimates for various age groups for states, counties, and school districts. The state estimates come from a Fay-Herriot model developed by Fay and Train (1997) for direct state estimates from the March Supplement to the Current Population Survey (CPS). This model uses predictor variables constructed from IRS tax file data, other administrative sources, and state poverty estimates for income year (IY) 1989 from the 1990 census. The model allows for sampling error in the CPS estimates. To date, SAIPE state production estimates have been released for IYs 1993 and 1995-98. In developing the IY 1999 SAIPE state estimates, we face some issues about how best to use the 2000 census results. One extreme option is to simply use the 2000 census estimates (which are for IY 1999) as the SAIPE state poverty estimates; another is to use the 2000 census estimates instead of the 1990 census estimates in the IY 1999 SAIPE state models. We represent these and various options in between as alternative models, and use statistical comparisons to determine which model fits the CPS data best.