Show The Graduate Center Menu

Census 2020 Response Rate Analysis: Week 5

Mapping "Self-Response" for a Fair and Accurate 2020 Census (Week 5)

Prepared April 24, 2020

HTC 2020 map news

Links to analyses from prior weeks:

WEEK 5 (April 17 through April 23)

Summary materials:

National Response Rate Trends

The nationwide 2020 Census self-response rate as of Thursday, April 23, was 51.8%. There was a modest boost to the day-over-day increase on Wednesday April 22 (a 0.6 percentage point increase), but that dropped to a 0.2 percentage point increase the following day (Thursday April 23).

Earlier this month the Census Bureau indicated that the "current national response rate is on track for what was expected," per an Associated Press report. However, if the national daily response rates increase at the same pace as the last two weeks (roughly a 0.3 percentage point average daily increase), the nation will not achieve the Census Bureau's projected 2020 final self-response rate of 60.5% until May 22, 2020. If the daily rate increases continue at the same pace, the nation will not match its 2010 final self-response rate of 66.5% until June 11.

The response rate trend is shown in the chart in Figure 1 below. The chart is annotated to show that as of April 8, the Census Bureau has been mailing paper questionnaires and reminder letters to non-responding households in mail-out areas. Due to the COVID-19 crisis, the Census Bureau has extended the timeframe of these mailings: paper questionnaires (and related reminder letters) will continue to be mailed through April 30, and reminder postcards will be mailed from April 27 to May 9. In addition, the Bureau sent previously-planned reminder letters bearing the unique ID numbers and telephone assistance information to “mailable addresses” in Update/Leave areas, despite most households not yet receiving their initial census packets due to COVID-19 related suspension of field operations.

The chart visualizes the day-to-day response rates in 2010 (the long red line) compared with the rates so far in 2020 (the shorter red line noted with the arrow). The other lines represent state-by-state response rates in 2020. (The Census Bureau did not publish daily response rates from 2010 at the state level.) Click to view larger image.


NB: Keep in mind the differences in census operations between 2010 and now when comparing response rate trends. Census operations were different in 2010, which may impact the difference in the national trend. For example, the 2010 Census was largely paper-based (the Bureau didn’t advertise the online option, which was used minimally); there were only two initial mailings (rather than staggered waves of initial mailings and multiple follow up mailings in 2020); there was a targeted replacement mailing in 2010; and of course, the 2010 Census did not face a pandemic that delayed key census operations and upended Get Out the Count (GOTC) campaigns that were based significantly on in-person outreach. Nonetheless, stakeholders will measure the success of the 2020 self-response operation by whether the nation, states, and localities meet and surpass their 2010 rates. Our map allows you to compare not only the 2020 rates with final 2010 “goalpost” rates, but also assess how your area is doing compared with nearby communities or similar areas across the state, and for larger regions such as counties, the state, and nation as a whole.

The following graphs show the latest 2020 response rates by state and the percentage point change by state from March 20 (when the Census Bureau first reported response rates) until now.

FIGURE 2 States ranked by April 23 census response rate (the final 2010 U.S. rate and the current U.S. 2020 rate are included for comparison). Click to view larger image.

FIGURE 3 States ordered left-to-right by lowest response rate to highest on April 23 (including the current national rate and final 2010 response rate for comparison). Brown bars represent the beginning response rate (March 20); the blue bars represent the increase between 3/20 and 4/23. Click to view larger image.

The majority of responses continue to be submitted via the online portal. According to the response rate data published on April 23, 44.8% of responses the day before were submitted via the internet, while only 7% of responses were submitted by mail or phone, for an overall response rate reported for April 23 of 51.8%.

On a daily basis, the increase in response also has been driven by internet responses until recently, as shown in Figure 4 below. The blue line in the chart shows the daily change in response rate from internet responses, and the orange line represents daily change from mail/phone responses. From March 21 through April 11, the daily change from internet responses was greater than the daily change from mail/phone. (On April 2, reflecting the responses from "Census Day" on April 1, the daily increase from internet responses was much greater: a 2.7 percentage point increase from the internet, compared with only 0.2 percentage point increase from mail/phone.)

On two separate days (March 24 and April 13), the daily change in response rates from mail/phone was negative, as shown in the chart where the orange line drops below the horizontal gray line that indicates zero change from the day before. In other words, on March 24 and April 13 there were fewer mail/phone responses than the day before. On April 19, the daily increase from internet responses fell below the daily increase from mail/phone (i.e., there were fewer internet responses that day than mail/phone responses).

On April 22, the daily increase from mail/phone rose above internet responses; on that day there was a 0.6 percentage point increase in daily response, most of which (0.4 percentage points) was due to mail/phone response. The following day (April 23) the mail/phone increase fell back to 0 (there were the same number of mail/phone responses as the day before). Nonetheless, we have noticed in several tracts across the country a growing increase in mail/phone responses compared to internet responses, and we will continue to monitor this pattern in the coming week to detect if the Census Bureau's ongoing mailings of paper questionnaires and reminders will have helped to boost mail response.

FIGURE 4 (click to view larger image)

Key areas of concern analyzed for Week 5

In addition to the nationwide response rate trends, we focus our Week 5 analysis on the following issues:

  1. City/metropolitan area/suburb/rural trends
    • State-level rates can mask substantial regional/local differences
    • Two examples: Detroit and Houston
  2. Demographic trends for tracts with lowest rates
    • Can these trends be changed by the time NRFU begins?
  3. Response rate gap between Internet Choice and Internet First tracts
    • What types of communities are responding more slowly?


At the national level and state-by-state, the 2020 Census self-response rates are improving daily. Figure 5 below from the CUNY HTC 2020/Response Rate map shows that as of April 1, each state's response rate was below 50% (and the national rate was only 38.4%), while Figure 6 from the April 23 view of the map shows that the U.S. rate is above 50% and more than half the states each have rates above 50%.

FIGURE 5 (click to view larger image)

FIGURE 6 (click to view larger image)

But this national and state-level view masks local differences in response rates. Below we examine the differences in average response rates by tract for cities (population 50,000 or more) in metropolitan areas, suburban areas outside these cities, areas outside metropolitan regions, and tracts covered by Update/Leave to illustrate the uneven rates across the country. In particular, we focus on two examples: Michigan (with an emphasis on Detroit), and Texas (with an emphasis on Houston).

Michigan / Detroit

Michigan's response rate on April 23 was 58.2% — 6.4 points above the U.S. rate. The average rate for tracts in cities with at least 50,000 people in metropolitan areas in Michigan was lower: 55.2%. In tracts in metropolitan areas outside cities (suburbs, generally, and including incorporated places of fewer than 50,000 people), however, the rate was 64% — almost 6 points above the statewide rate and 12 points above the U.S. rate.

Outside metropolitan areas in Michigan, we examined average response rates for tracts with fewer than 10% of their units covered by Update/Leave. The average response rate in these tracts was 52.9% — below the statewide rate and below the average rate in metropolitan areas. Separately we examined response rates for "Update/Leave" tracts, almost all of which in Michigan are outside metropolitan areas. The average response rate for these tracts was very low: only 32.1%.

Figures 7 and 8 below display metropolitan areas across Michigan with red outlines that represent boundaries of metropolitan areas. In Figure 7, the map shows county-level response rates as of April 23. Figure 8 shows the patterns of the Census Bureau's 2020 contact strategies: the map highlights in yellow where housing units are covered by Update/Leave (in most areas across Michigan's upper peninsula, as well as some parts of northern Michigan).

FIGURE 7 (click to view larger image)

FIGURE 8 (click to view larger image)

Focusing on one of Michigan's city's, however, reveals that response rates in specific cities are often lower than the metropolitan average. In the Detroit-Warren-Dearborn metropolitan area, the average tract-level response rate is 55.2% (the Census Bureau does not publish response rates for metropolitan areas, so we identified tracts within metropolitan Core-Based Statistical Areas/CBSAs and averaged the response rates across all these tracts by metro CBSA). This average metropolitan rate is 3.4 points above the U.S. rate, and on par with response rates for Michigan's urban areas in general.

But within the metropolitan area, Wayne County's response rate is lower (53.9%), and Detroit's rate is substantially lower: 41.2%. (The city's rate is 10.6 points below the U.S. rate, and 17 points below Michigan's statewide rate.)

Detroit's population is predominantly Black, its households are mostly renters, and a third of the population has incomes below the poverty line — each of which are historically undercounted or low self-response populations. In addition, as shown in Figure 9 below, homes in most tracts in Detroit received the Census Bureau's "Internet Choice" mailing which included the paper questionnaire in the mailing as well as information on responding online or by phone. As our analysis has shown in prior weeks (and below for this week), Internet Choice tracts tend to have much lower response rates than Internet First tracts, where homes received a mailing with information only for responding online or by phone.

FIGURE 9 (click to view online map)

Texas / Houston

Our second example regarding metropolitan/suburban/rural response rate differences is Texas, with an emphasis on Houston. Response rates in Texas are lower than in Michigan, al they exhibit similar patterns across these regions.

The statewide response rate in Texas for April 23 was 47.3%. This is 4.5 points below the U.S. rate. The average rate for tracts in cities of 50,000 people or more in metropolitan areas in Texas is 48.5%, slightly higher than the statewide rate. In tracts in metropolitan areas outside cities (generally these are suburban tracts, but also include incorporated places of fewer than 50,000 people), the average response rate rises to 50.2%.

The average rate for "non-Update/Leave" tracts outside metropolitan areas is lower, at 42.4%. The average response rate for "Update/Leave" tracts (with 10% or more homes covered by Update/Leave) is the lowest across all these typologies: 26.7%.

Figures 10 and 11 below display metropolitan areas across Texas with red outlines representing metropolitan areas. In Figure 10, the map shows county-level response rates as of April 23. Figure 11 shows the patterns of the Census Bureau's 2020 contact strategies: the map highlights in yellow where housing units are covered by Update/Leave (only 3.5% of housing units in Texas are covered by Update/Leave, but these areas extend across much of the state).

FIGURE 10 (click to view larger image)

FIGURE 11 (click to view larger image)

In Houston, however, there are no Update/Leave areas. The "Greater Houston" metropolitan area has an average tract-level response rate of 45.4% (below the U.S. rate and below the Texas urban area response rate). Harris County within this metro area has a slightly higher response rate of 47%. Houston itself has a lower response rate: 44% (roughly 8 points below the U.S. rate and about 3 points below the Texas statewide rate).

As Figure 12 below shows, the tract-level response rates in and around Houston illustrate a pattern similar to many other urban areas nationwide: relatively low response rates within the city, higher rates in the immediately adjacent suburban tracts, but lower (sometimes lowest) response rates outside the metropolitan area.

FIGURE 12 (click to view online map)


Who will be in the NRFU universe by August, and how will that have changed during the extended 2020 period for self-response?

This analysis expands on our research from Week 4, focusing on the “bottom 20%” of tracts based on the latest 2020 Census response rates. This enables us to understand the geographic patterns and concentrations of tracts that will eventually require the greatest share of NRFU effort.

This week, we slightly modified our methodology to determine the bottom 20% of tracts. Instead of omitting tracts that have any housing units covered by the Update/Leave operation, we only omitted tracts that have 10% or more housing units in Update/Leave. This enables us to examine more tracts, but still separate the tracts where suspension of the Update/Leave operation is causing very low response rates.

We use the latest data from the Census Bureau’s American Community Survey (ACS) (2014-2018 “5-year estimates”) to determine the characteristics of the overall population in these “bottom 20%” census tracts. This shows not only if historically undercounted population groups are in these tracts, but if the tracts that eventually will require the greatest Nonresponse Follow-up effort are home to disproportionately high numbers of these groups.

Of course, this approach does not identify who is being counted as the self-response operation continues. It only identifies the demographic characteristics of census tracts overall. But it provides an indication of which population groups (and which communities) are at risk of being undercounted, and whether that risk is increasing or decreasing for historically undercounted groups if they continue to represent a large share of the population of the bottom 20% of tracts.

Comparisons from Week 1 to Week 4, and then Week 4 to Week 5

On March 20, when the Census Bureau began reporting response rates, tracts with response rates in the lowest fifth had a response rate of less than 8.6% (approximately 15,000 census tracts). Based on ACS estimates for the 2014-2018 period (using our modified approach regarding U/L tracts), there are approximately 59 million people in these tracts, approximately 23.6 million total housing units, and approximately 20.8 million occupied housing units (i.e., households).

Some population characteristics of these tracts are as follows:

  • 8.7 million people (~15% of population in bottom 20% of tracts) had incomes below the poverty line
  • Language challenges: 1.3 million households (6.4%) were "limited English proficiency" (LEP)
  • Non-Hispanic White population: 31.4 million (53.2%)
  • Non-Hispanic Black population: 8.8 million (15%)
  • Hispanic population: 13.8 million (23.4%)
  • Households with no internet: 3.1 million (15%)

On April 16, tracts with response rates in the lowest fifth had rates less than 41.4%. This also represents approximately 15,000 tracts, but they are a different set of tracts than on March 20. Based on ACS estimates for the 2014-2018 period, the universe of population and housing units in these tracts had changed:

  • overall population decreased about 14% to 50.7 million people;
  • the number of housing units decreased to 22.4 million; and
  • the number of occupied housing units (households) decreased to 18.2 million.

However, the population groups listed above increased their concentration in the bottom 20% of census tracts between March 20 and April 16:

  • Population in poverty increased to 12.3 million (from 15% to 25% of population in bottom 20% of tracts)
  • Language challenges: LEP households increased to 1.8 million households (9.9% of the population in the bottom 20% of tracts as of April 16)
  • Non-Hispanic White population: fell to 19 million (37.6%)
  • Non-Hispanic Black population: grew to 11.2 million (22%)
  • Hispanic population: grew to 16.2 million (32%)
  • Households with no internet: grew to 4.2 million (23.4%)

By April 23, because response rates overall had increased, tracts with response rates in the lowest fifth had rates less than 43.1%. Based on ACS estimates for the 2014-2018 period, the universe of population and housing units in these tracts had changed, but not substantially since the prior week. Overall population increased slightly to 50.9 million people, the number of housing units increased slightly to 22.6 million, and the number of occupied housing units (households) increased to 18.3 million.

The population groups listed above in the bottom 20% of census tracts also had only modest changes between April 16 and April 23:

  • Population in poverty: slight increase (12.4 million households, or 25.4%)
  • Language challenges: slight decrease in LEP households to 1.73 million (9.5%)
  • Non-Hispanic White population: grew slightly to 19.3 million (37.8%)
  • Non-Hispanic Black population: grew to 11.9 million (23%)
  • Hispanic population: decreased to 15.7 million (31%)
  • Households with no internet: increased slightly to 4.4 million (23.9%)

Week 5 comparisons between the Bottom 20% and Top 20% of census tracts

In contrast to the population in the bottom 20% of census tracts on April 23, the population in the top 20% of tracts (those with the highest response rates) had a smaller share of historically undercounted population groups. Tracts in the top 20% had response rates of 64.1% or more. Based on the 2014-18 ACS estimates, these tracts represent 66.5 million people (30% more than the bottom 20%), 25.8 million housing units, and 24.5 million households.

The population groups listed above in the top 20% of census tracts for April 23 are as follows:

  • Population in poverty: 4 million people (8.5 million fewer than bottom 20%)
  • Language challenges: 500,000 LEP households (70% fewer than the bottom 20)
  • Non-Hispanic White population: 50.7 million (more than 1.5x as many as in the bottom 20)
  • Non-Hispanic Black population: 3.1 million (74% fewer)
  • Hispanic population: 5.6 million (65% fewer)
  • Households with no internet: 2 million (53% fewer)

Analysis: response rates for Internet Choice tracts are relatively low. What population groups are in these tracts?

About 80% of homes that received a mailed invitation from the Census Bureau for the 2020 Census (95% of all addresses) received Internet First packets (a letter of invitation to respond on-line, which also included toll-free phone numbers to call in the responses). The remaining 20% of housing units across the country received Internet Choice packets (these included the same materials as the Internet First mailing plus a paper questionnaire).

Response rates for Internet Choice tracts are increasing slowly compared with Internet First tracts. As of April 23, the response rate gap between the two had grown to more than 10 points. This week, the average response rate for Internet Choice tracts is more than 6 points below the U.S. rate; Internet First tracts have an average rate 4 points above the U.S. rate.

More people live in Internet First tracts (an estimated 244 million) compared with Internet Choice tracts (~59 million). But, “Choice” tracts tend to have:

  • a greater share of population in poverty (24% vs 11%);
  • a greater share of people of color. For example, non-Hispanic Black population is 26% of “Choice” tracts vs 9.3% of “First” tracts, and Hispanic population comprises 25% of “Choice” tracts vs 16.6% of “First” tracts; and
  • a smaller share of non-Hispanic White population (44.4% in “Choice” tracts vs 64.3% in “First” tracts).

Internet First tracts also have more households (almost 91 million compared with 21.6 million). But Internet Choice tracts tend to have:

  • a greater share of renter households (44% of “Choice” households vs 35% of “First” households);
  • a greater share of “limited English proficiency” households (6.4% vs 4.1%); and
  • a greater share of households with no home internet subscriptions (28% compared with 12.3%).

For more information, contact:

Center for Urban Research

at the Graduate Center, City University of New York

All work and materials are supported by a grant from the 2020 Census Project

and developed in partnership with the Leadership Conference on Civil and Human Rights.