Document Type
Article
Publication Date
7-8-2021
Department
Department of Social Sciences
Abstract
Each decade since the 1950s, demographers have generated high-quality net migration estimates by age, sex, and race for US counties using decennial census data as starting and ending populations. The estimates have been downloaded tens of thousands of times and widely used for planning, diverse applications, and research. Census 2020 should allow the series to extend through the 2010–2020 decade. The accuracy of new estimates, however, could be challenged by differentially private (DP) disclosure avoidance techniques in Census 2020 data products. This research brief estimates the impact of DP implementation on the accuracy of county-level net migration estimates. Using differentially private Census 2010 demonstration data, we construct a hypothetical set of DP migration estimates for 2000–2010 and compare them to published estimates, using common accuracy metrics and spatial analysis. Findings show that based on demonstration data released in 2020, net migration estimates by five-year age groups would only be accurate enough for use in about half of counties. Inaccuracies are larger in counties with populations less than 50,000, among age groups 65 and over, and among Hispanics. These problems are not fully resolved by grouping into broader age groups. Moreover, errors tend to cluster spatially in some regions of the country. Ultimately, the ability to generate accurate net migration estimates at the same level of detail as in the past will depend on the Census Bureau’s allocation of the privacy loss budget.
Publication Title
Population Research and Policy Review
Recommended Citation
Winkler, R.,
Butler, J.,
Curtis, K.,
&
Egan-Robertson, D.
(2021).
Differential Privacy and the Accuracy of County-Level Net Migration Estimates.
Population Research and Policy Review.
http://doi.org/10.1007/s11113-021-09664-5
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p/15133
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Version
Publisher's PDF
Publisher's Statement
© The Author(s) 2021. Publisher’s version of record: https://doi.org/10.1007/s11113-021-09664-5