Counting to 99.9%?
The Census Bureau says that it is well on its way to "enumerate" 99.9% of the nation's housing units. The Bureau may attain that percentage if the courts require the 2020 Census field operations to continue through Oct. 31. But we don't know yet what the 99.9% number tells us about the accuracy or quality of the actual census, which is a count of the population (not just housing units) and some basic but important characteristics of the population: gender, age distribution, race/Hispanic origin, tenure (owners or renters), and household structure.
The Census Counts coalition has produced a helpful explainer [PDF] that discusses the Bureau's goal of enumerating almost all housing units and what that does & does not tell us about the population count's accuracy & quality, especially if the Bureau cuts short its field operations.
In this email newsletter we highlight some additional concerns based on feedback and questions we've received on the topic, and to highlight how to use our online HTC/Response Rate map to help gauge progress toward a fair and accurate 2020 Census.
Different metrics
The Census Bureau has published several metrics regarding responses to the 2020 Census. These are listed here, and we elaborate on each one below the list:
- self-response rates;
- share of housing units "enumerated" through the door-knocking, nonresponse follow-up (NRFU) operation (but a more accurate term would be "accounted for");
- the combination of these first two metrics, i.e., total response rate; and
- the NRFU workload "completion" rate.
Self-response rates
The U.S. self-response rate is now 66.7%. It is higher in many states and local communities, and it continues to rise. Our prior email newsletter highlighted the importance of achieving, and now surpassing, the nation's 2010 self-response rate.
We noted in that newsletter that meeting the 2010 rate may not mean that the 2020 count will avoid similar (or worse) undercounts as in 2010 (for example, the significant disruptions to the current census from COVID-19and the Bureau's recent efforts to rush field operations may be greater factors in determining accuracy and any differential undercounts). But ongoing increases in self-response, even as the NRFU operation continues, will help in terms of a fairer and more accurate 2020 count. For example:
- the census tracts with self-response rates in the lowest fifth nationwide (hardest to count) now have a Hispanic population that is smaller by almost 1 million people (from 13 million to just over 12 million; a 7% reduction) than the tracts with the lowest rates in early August (just before NRFU began nationwide);
- census tracts with the lowest self-response rates now have a foreign-born population smaller by more than 800,000 people (from 8 million to 7.2 million; a 10% reduction) compared with tracts with the lowest rates in early August; and
- the number of renter households in census tracts with the lowest self-response rates is now smaller by 350,000 (from 9.2 million to almost 8.9 million; a 4% reduction) than in the lowest-rate tracts in early August.
These statistics indicate that the communities with the greatest need for door-to-door enumeration — i.e., the hardest to count areas — are changing. On a nationwide basis the overall population in tracts with the lowest self-response rates still tends to be disproportionately people of color, immigrants, and people with lower incomes. But in cities such as Chicago, for example, our map shows that self-response rates have increased notably in areas that are predominantly Hispanic. Or in NYC, self-response rates have risen substantially in diverse neighborhoods across Queens County.
If the opportunity for self-response continues throughout October, combined with more thorough door-to-door enumeration, this could help address concerns about accuracy of 2020 census data in these traditionally hard-to-count communities.
NRFU "enumeration" rates
When housing units are "enumerated" via door-to-door visits, it doesn't necessarily mean that people were counted or were counted accurately. A more "plain English" term would be that housing units visited during NRFU are "accounted for" or "resolved." In other words, a census enumerator visiting a housing unit hopefully would be able to talk with a household representative and count everyone in the household during the enumerator's visit. But a "successful" enumerator visit could also mean:
- the enumerators weren't able to talk with anyone in the household itself, but instead relied on someone else for data about people in the household (enumeration by "proxy"); or
- the number of people in a household and/or their characteristics (gender, age, race/Hispanic origin) needed to be counted via administrative records (this approach can systematically omit groups already likely to be missed, such as children or young adult men of color); or
- the housing unit was designated as vacant (perhaps without confirming whether people lived there on April 1, i.e., Census Day).
Without more detailed metrics from the Census Bureau about the share of housing units "accounted for" during NRFU in these various ways, we cannot evaluate the accuracy or thoroughness of the NRFU enumeration process. Demographers and census stakeholders across the country have requested this data from the Census Bureau — preferably at the local level but even on a state-by-state basis — but the Bureau has not provided this information.
Total response rates
The Census Bureau combines the percent of housing units accounted for during NRFU with the share of households that have self-responded in order to calculate a "total" response rate to measure progress toward the goal of "enumerating" 99.9% of the nation's housing units.
Census Bureau research shows that self-response produces the best quality data; lower self-response at the state level is correlated with lower decennial census data quality. At this point without better metrics about the distribution locally and nationwide of the methods used to "successfully" enumerate housing units during NRFU, stakeholders cannot evaluate the quality of data collected during NRFU. Therefore, the "total" response rate — which the Census Bureau is using to measure progress toward a successful 2020 Census — is a metric with limited value for those of us hoping for a fair and accurate count.
NRFU "completion" rates
At the request of census stakeholders, the Bureau is publishing NRFU "completion" rates using geographic areas called "Area Census Offices" (there are 248 of these stateside and in Puerto Rico). Completion rates measure the actual work being done on the ground to visit housing units and determine their status. This is separate from NRFU enumeration rates; it measures something different. Completion rates reflect the share of the door-to-door "workload," which includes the number of housing units that need to be visited but also includes revisits to these housing units and other quality-checking work.
For example, the total response rate as of Oct. 6 for Montana was 98.8%. But the NRFU "completion" rate for the Billings Area Census Office (coterminous with the state) on Oct. 6 was 96.8%. The total rate of 98.8% reflects the share of the state's housing units that filled out the census form on their own plus the units that were resolved during NRFU (regardless of how the units were resolved during NRFU). The completion rate — two percentage points lower — reflects the share of the door-to-door workload that is complete.
Not only do these two metrics measure different things, but according to the Census Bureau the total response rate will always go up but the NRFU completion rate can go up or down depending on how the workload is changing.
On Oct. 6 we noticed that the completion rates decreased in 27 Area Census Offices (ACOs) across 13 states; our HTC/Response Rate map highlights the completion rate changes:
The decreases were in the 0.1 to 0.4 percentage point range. It indicates that the NRFU “workload” in these areas has gone up because the Bureau is sending enumerators back out to be more thorough in their efforts, in response to court orders. The Bureau described this approach in a footnote (since removed) at its NRFU webpage. The footnote stated that a Temporary Restraining Order issued by a federal court on Sept. 5 "is currently preventing the enumeration of certain households using administrative records that the Census Bureau has deemed to be a valid and more cost efficient way to enumerate and the preferred method of enumeration." Using administrative records tends to be less accurate than in-person enumeration, so aside from cost efficiencies, the TRO is requiring the Bureau to rely on more accurate methods.
This is not the first time we have noticed completion rate decreases. But the extent of the decreases across 27 ACOs was greatest so far on Oct. 6. We will continue to monitor and map this trend if it continues.
Links to earlier updates
Make sure to follow us on Twitter at @Census2020Map !
The HTC 2020 map is a work in progress. Other recent updates and enhancements are described here:
- September 29, 2020: U.S. achieves 2010 self-response rate goal as door-knocking continues; HTC Map shows self-response increases during NRFU.
- September 4, 2020: Census Bureau releases door-knocking completion rates for each "Area Census Office" across the U.S.
- August 27, 2020: 2020 Census Total Response Rates added to the map; tracts receiving a 7th mailing also highlighted.
- August 11, 2020: Census self-response analysis leading into door-knocking follow-up; 8 million more homes need to be visited compared with 2010 in a shorter timeframe.
- August 5, 2020: Online maps pinpoint areas at greatest risk of a rushed 2020 Census.
- July 24, 2020: Door-knocking enumeration begins in selected areas; our map shows where and what to expect. Also updates on latest response rate analysis & recent online census self-response highlights.
- July 15, 2020: New "Data Trends" feature added to the mapping site, with dynamic trendlines & animated map of response rate trajectories. Estimated response rates by state legislative district also added to the map.
- June 25, 2020: Dynamic new metric to tract response rates at the tract, city, & county levels. Examples of other resources analyzing the latest 2020 Census self-response rates, and our latest analysis of the nation's response rate trends.
- May 27, 2020: Update/Leave operations have resumed in 42 states & Puerto Rico; rates are increasing slowly nationwide but bright spots highlight importance of census outreach to continue to boost response; tracts with lowest response rates are very different than tracts with highest rates.
- May 14, 2020: New map search feature for Area Census Offices (where the Update/Leave operation is resuming), our latest self-response rate analysis (Week7), & links to other projects analyzing self-response rates.
- May 6, 2020: News about where the Update/Leave operation is resuming, and our latest self-response rate analyses (Weeks 4, 5, & 6).
- April 13, 2020: Week 3 Response Rate Analysis; Trendlines Added to the Map.
- April 7, 2020: Week 2 Response Rate analysis; Tribal Lands added to the map.
- April 3, 2020: Census 2020 HTC map news: "Census Day" Self-Response Bump; Data Q&A.
- March 31, 2020: Week 1 Response Rate analysis.
- March 23, 2020: Update on mapping self-response rates, with emphasis on the 2020 progress bar that fills in daily after the latest rates are published, easy share/embed options for your map, and some notes on the data.
- March 19 2020: In a joint statement with our colleagues at the Leadership Conference on Civil and Human Rights (LCCHR), we highlighted the importance of a fair and accurate 2020 Census as the coronavirus challenges grow, and lifted up the resources available at our HTC 2020 map to help inspire Americans to fill out the 2020 Census form on their own. Self-responding helps ensure that the Census Bureau collects reliable data about the nation's population. And self-responding is a way to practice social distancing because it avoids a knock at your door later on from a census taker.
- March 2020: The HTC map is now focused on census self-response rates. All the former info at the map is still there. But now that census mailings are going out, the map has been updated to reflect 2010 response rates in anticipation of integrating 2020 real-time rates after March 20. NB: the map not only displays response rates for each state, county, and tract, but also include a bar chart in the map's left-hand panel that shows the 2010 rate (and 2000, for historical context) for now. The 2020 column in the bar chart is empty, but it will start to go up after March 20. How quickly and how far it rises depends on local census stakeholders!
- February 2020: New data on the risk of undercounting young children, in collaboration with the Population Reference Bureau. For more info, visit PRB's website.
- January 2020: To help promote the official start of the 2020 Census in Alaska, we added a special "It Starts Here" (in Toksook Bay, AK) graphic on the map. Updates in January also included new 2014-2018 population estimates for tracts, counties, states, legislative districts, and more.
- December 2019: New advanced tract search feature, statewide maps of Census Bureau contact strategies, and more.
- November 2019: Comprehensive information for all 2020 Census contact techniques combined in one place at the HTC 2020 map, so census stakeholders can more easily inform local residents about what to expect when the 2020 decennial census takes place. Also see the CUNY Center for Urban Research website for a state-by-state analysis.
- October 2019: Updates to TEA designations; the latest examples of groups using the HTC map across the country; enhancing the HTC metrics with the Census Bureau's "low response score", the Urban Institute's projections of undercount by state; & more.
- August 2019: In-Field Address Canvassing areas & stats on the map; organizations that are using the HTC map for local grant assistance; new examples of linking to and/or embedding the HTC map.
- July 2019: new feature to highlight tracts based on share of households without internet access; a list of other census maps nationwide, and more.
- June 2019: Census contacts by state/county; census tract search feature.
- April 2019: customized printing, data downloads, and more.
- March 2019: mapping Type of Enumeration Areas (TEA) and Area Census Offices (ACOs)
- January 2019: new ACS data for the 2013-17 period (including internet access), new legislative info, public library locations, and tribal lands added to the map.
If you haven't signed up for our HTC 2020 map updates, please do so here.
We look forward to hearing your suggestions for improving the map. Please contact the Mapping Service at the CUNY Graduate Center with your feedback.
Thanks!
If you need a bit of inspiration, you can see what other MailChimp users are doing, or learn about email design and blaze your own trail.
|
