On Thursday, May 26, NGA hosted a briefing for Governors’ offices with the U.S. Census Bureau to review the Post-Enumeration Survey (PES), which measures the accuracy of the census by independently surveying a sample of the population.
The Census Bureau has used PES with dual system estimation to measure coverage in the Decennial Censuses of Population and Housing since 1980. This approach involves case-by-case matching of persons in an independent survey with persons in the census to determine who was missed or counted in error.
The prior week, the Census Bureau released a report, tables, and an American Counts story showing additional results on the coverage of the 2020 Census. The Census Bureau uses the PES to estimate the number of people in the country. Then Census officials compare the census counts to the independent PES estimate to see if the census counts may have been too high or too low. The PES also provides information about how many people were correctly counted in the census, missed, or erroneously enumerated. Consistent with prior practice, the Census Bureau will not be adjusting the census counts for Apportionment or Redistricting.
The Post-Enumeration Survey is a probability survey of about 161,000 housing units in about 10,000 blocks across the country that is independent of the census operations. After creating independent lists of housing units and people in these 10,000 blocks, the Census Bureau looked for these people in the census to determine who was missed from the census or counted in error.
The Census Bureau also measures census coverage using Demographic Analysis. Both Demographic Analysis and the PES provide independent estimates of the population size. One way they differ is in who is considered in scope. The Post-Enumeration Survey excludes people living in Group Quarters and Remote Alaska areas, while Demographic Analysis includes both groups. In this slide, we see the census counts for the Demographic Analysis and the Post-Enumeration Survey. For the Demographic Analysis universe, which includes group quarters and remote Alaska, the census counted 331.4 million people and the demographic analysis independently estimated 332.6 million people. For the PES universe, which excludes people in group quarters and people in remote Alaska areas, the census count was 323.2 million people, and the PES estimate was 323.9 million people. It is important to keep in mind that the PES estimates for each state apply only to the household population in each state. The PES does not estimate census coverage for people who are living in group quarters in each state.
No survey is without challenges or errors. Most of the PES operations were conducted during the COVID-19 pandemic. Some modifications to the PES were made in response to the pandemic. Probably the most visible modification was delaying the fieldwork. Many surveys doing in-person interviewing in the summer and fall of 2020 suffered from large amounts of nonresponse. Delaying field work and extending deadlines probably contributed to the PES having higher response rates than many other surveys. Nevertheless, these delays also increased the time between the census and PES interviews.
Many colleges and universities either closed or pivoted to virtual learning in 2020. This contributed to a major migration of young adults, often back home and into the household population. This migration made it challenging to determine who should be included in the PES and who was out of scope because they should have been counted in college dorms or other group quarters.
Another challenge was related to increased levels of people not answering specific questions, called item nonresponse, in the Census or PES. It is hard to conduct the matching and follow-up work required for dual-system estimation when characteristics are missing.
Overall, the Census Bureau overcame many challenges over the past two years. Staff adapted their operations given the changing pandemic environment and continued to implement quality safeguards into all the field and clerical matching activities.
