In 2020, the U.S. Census Bureau began injecting noise into census counts using a differential privacy technique. See here for a popular press description and here for some official literature.
The introduction of this noise makes counts from individual blocks unreliable, but the counts should become increasingly accurate as blocks are aggregated into larger geographies.
My understanding is that we should be able to calculate the confidence interval stemming from the uncertainty created by the privacy algorithm. I assume that the particulars of this calculation depend on the parameters of the algorithm. Further complicating things, the noise must not be truly normally distributed around the mean, because it has to be truncated to avoid negative numbers.
I haven't been able to find any official or unofficial guidance or demonstrations of how to calculate confidence intervals for 2020 census data. Does anyone know of anything?
