DEC. 31, 2020 — The 2020 Census data collection contained numerous quality checks, including analyzing data about census takers’ work and reinterviewing cases where the data suggested a potential problem. Every completed case and every census taker’s work went through a battery of checks designed to spot unusual patterns of work, called analytic selection. If the analytical process indicated a census taker had zero cases selected for review, it was an indication that nothing was out of the ordinary about their work characteristics. The analytic selection techniques were sufficient for detecting unusual patterns of work, alleviating the need to check every census taker and tens of millions of results through reinterviews.
During part of the Nonresponse Followup operation (from August 3 to September 6), the Census Bureau reduced the period of attempted reinterview contacts with households from six days to three days. This change allowed the Census Bureau to focus on collecting data from the remaining nonresponding households. Experience had shown that about 90% of completed reinterviews that allowed up to six attempt-days were completed within three attempt-days, so the Census Bureau did not expect the additional attempt-days to generate a significant number of additional reinterviews. After September 6, the Census Bureau increased the visits for attempting contact with households back to six days.
We also expected to find unresolved reinterview cases. As in numerous previous censuses, we incorporated special strategies for completing the very last Nonresponse Followup cases. The “closeout” operation, as outlined in the 2020 Census Operational Plan, directs census takers to resolve these cases (which have received numerous contacts and opportunities to self-respond) with at least a population count. Our analytic selection criteria did not change during this closeout phase, and as a result the number of census takers flagged as outliers was high during this stage.
Preliminary metrics on census takers’ work reveal that far below 1% (i.e., 0.4%) of census takers may have falsified their work or performed work unsuccessfully to the extent that it needed to be reworked. According to preliminary metrics, about 13% of those completed rework cases for the census takers were found to be either a hard fail or a performance fail (meaning that the cases needed to be reworked). The other 87% of rework cases examined passed the reinterview. Therefore, enumerators who may have falsified data or performed poor quality work were very rare.
The Census Bureau is committed to data quality and completed almost 1.6 million reinterviews for this purpose during a national pandemic and severe weather challenges. The Census Bureau will continue to assess 2020 Census operations and their impact on data quality, and will publicly share the information.