Fast-tracked census IT systems receive more scrutiny from GAO

The information technology systems processing 2020 census data for congressional apportionment face an increased risk of defects because testing was fast-tracked to end in October, rather than January as originally scheduled, a new report says.

And those defects may go unaddressed because the time to complete data processing was shortened from 153 days to 77 days, according to a Government Accountability Office report released Wednesday.

The Census Bureau made numerous adjustments to its processes due to coronavirus-related delays and pressure from the Trump administration to end nonresponse follow-ups (NRFUs) early that have left it with less time to ensure data quality.

“To complete response processing in fewer days, the bureau made changes to its process, including locking down its Master Address File prior to the end of data collection and shortening the amount of time for reviews by subject matter experts in the Bureau’s statistics divisions,” reads the report. “The bureau is also prioritizing tasks needed to produce apportionment counts rather than simultaneously preparing redistricting data, which involves more data elements.”

Even so, the bureau isn’t guaranteeing it will meet its statutory deadline of Dec. 31 for congressional apportionment data.

The IT system used to optimize the assignment and routing of cases for census enumerators has also been criticized, with less than a quarter of area census office managers reporting satisfaction in an October survey. Managers reported performance issues, slowness and inefficiencies, while bureau officials praised the system’s performance given the higher number of cases completed per hour, according to GAO.

“However, they also noted that dissatisfaction with the system may have been due to, among other things, ACO managers and enumerators not fully understanding how the system worked, which led to expectations that were not met,” reads the report.

While 99.98% of housing units were counted through self-reporting and follow-up outreach, gaps in data were often filled with less reliable sources.

Proxy responses from people like landlords or neighbors accounted for 7.4 million people or 24.1%, compared to 23.8% in 2010.

The bureau hasn’t calculated the number of partial responses —where only the most important information like whether a housing unit is occupied and by how many people — it received.

Administrative records from sources like the IRS or a previous census accounted for 8.4 million households, or 14%, in the NRFU workload, which was less than planned. But some of those records lacked corroboration.