40 – Survey Weighting, Imputation, and Estimation
Using Administrative Data to Improve Nonresponse Weighting Procedures in the Consumer Expenditure Survey
Barry Steinberg
Bureau of Labor Statistics
The Census Bureau has begun research in collaboration with survey sponsors to improve the weighting procedures used in Title 13 household surveys by linking survey data to administrative data collected by other government agencies. The Consumer Expenditure Survey (CE) is one such survey where this research is ongoing, looking into the use of household-level income data collected by the Internal Revenue Service (IRS) to improve its nonresponse adjustment factors. Research shows that income is correlated with both response rates and expenditure reports. Currently, CE uses publicly available zip code level data from the IRS to stratify its sample of households, but CE is looking into the use of non-public household-level data from the IRS to improve the stratification process. The IRS’s publicly available data works fairly well. However, income varies within zip codes, and using IRS’s non-public household-level data may improve CE’s stratification process by putting the sample households into more homogeneous groups. This paper examines the impact of this linkage on CE’s household weights and expenditure estimates.