The Census, Hard-to-Count Populations, and Voter Suppression?

The first census post discussed what is called “prison gerrymandering.” This post focuses on the census’s differential undercount. The term “differential undercount” refers to the long-standing bias in the decennial census that undercounts what are called hard-to-count populations (HTC). I discuss this now before returning to the 2020 Census and the citizenship question because this next post continues the HTC discussion.

Danahy and Lang start with the premise that the census count is a voting rights issue. Voting rights can be understood in several distinct ways. One can consider voting as participation, which is the right to cast a ballot that is counted. The census does not affect whether an individual can cast a ballot. However, in two other aspects of voting census data play a role.

Voting can also be discussed in terms of voting as aggregation. Most voting ballots are cast in an aggregated fashion. A legislative body or some other body specifically tasked to do so, creates districts for the purposes of electing representatives. For example, nearly all states are subdivided into two or more districts to elect members of the U.S. House of Representatives. All states subdivide themselves into multiple districts to elect members of the state legislature. Many cities and counties sub-divide their jurisdictions to elect representatives to their legislative bodies.

Finally, one can consider voting as governance. In this perspective, voting rights relate to how much representation communities receive in their governing bodies. This aspect of voting is especially important in sub-state districts.

Voting aggregation

In voting aggregation, people are combined to produce a collective decision about the selection of a representative. When people are undercounted, some districts represent more people than others. For example, if the Census undercounts people with green eyes, and district X has many green-eyed people, the representative elected from district X represents more people than a representative from district Y that has no green-eyed people. Jurisdictions, following the one-person, one-vote principle, usually ensure that each jurisdiction represents a relatively equal number of persons or persons of voting age. If some green-eyed people are missed, then non-green-eyed people make up the balance. Voting district boundaries can be drawn that pack in green-eyed people in certain districts.

As expressed by Danahy and Lang, “the differential undercount may limit minority communities’ ability to aggregate their votes to elect candidates by packing minority communities into artificially overpopulated districts.”

Intersection with Voting Rights Act

The differential undercount relates to the Voting Rights Act. In Thornburg v. Gingles, the Supreme Court determined in 1986 that a minority group can challenge an at-large election [where there are no subdivided districts for electing representatives] only if it meets three standards. One is a ”’sufficiently large and geographically compact’ minority population to constitute a majority in a single-member district.” A second is “cohesive voting patterns among the minority community.” The final standard is “a cohesive majority voting bloc that blocks the election of minority candidates of choice.”

This situation changed when in 2009 the Court held in Bartlett v. Strickland, as summarized by Danahy and Lang, “that voters can only succeed in Voting Rights Act claims if they can prove their group constitutes at least 50% plus one of a proposed single-member district …In other words, the Court held that minority voters cannot invoke the Voting Rights Act to create ‘crossover’ districts. [Previously it was enough in proposed single-member districts “for minority voters to show that they can create a coalition of minority and white voters to elect their candidate of choice.”]

Danahy and Lang conclude that “the Supreme Court has put minority voters to strict proof regarding their numerosity to gain representation under the Voting Rights Act even though available data are known to systematically undercount these communities.”

Voting as governance

The “one-person, one-vote” principle is about voting as governance because it is about both direct and virtual representation. If certain localities are systematically undercounted while other localities are overcounted (because of prison gerrymandering), widespread violation of the principle may occur.

As shown below, the (HTC) are consistently the same. The authors remark, “In 2016, the Supreme Court [Evenwel v. Abbott, 136 S. Ct. 1120, 1132] held that ‘ensuring that each representative is subject to requests and suggestions from the same number of constituents, total-population apportionment promotes equitable and effective representation.’ If minority groups are systematically undercounted, the result is less equitable direct and virtual representation for those communities.”

The systematic differential undercount

• The Census Bureau has well-identified persons who are hard to count [counting everyone once]. At the most general level, these include
• Hard to locate – housing units not in the Census’ frame and/or persons wanting to remain hidden
• Hard to contact – highly mobile people, people experiencing homelessness, physical access barriers such as gated communities
• Hard to persuade – suspicious of government, low levels of civic engagement
• Hard to interview – participation hindered by language barriers, low literacy, lack of internet access

More specifically, the following are hard to count populations:

• Young children
• Highly mobile persons
• Racial and ethnic minorities
• Non-English speakers
• Low-income persons
• Persons experiencing homelessness
• Undocumented immigrants
• Persons who distrust the government
• LGBTQ persons
• Persons with mental or physical disabilities
• Persons who do not live in traditional housing

The Census Bureau estimated the 2010 census (1) overcounted the non-Hispanic white population by 0.8%, (2) undercounted the Black population by 2.1%, (3) undercounted the Hispanic population by 1.5%, and (4) undercounted Native Americans living on reservations by 4.9%.

Consequently, the differential undercount between whites was (1) 2.9% for Blacks, (2) 2.3% for Latinos, and (3) 5.7% for Native Americans living on reservations.

More specifically the Census Bureau reported that the two sub-groups with the most undercounts in the 2010 Census were

• Black men ages 30-49 – -10.1%
• Children ages 0 to 4 – -4.6%

As Danahy and Lang show, the census differential is long-standing. The Black percentage differential (compared to whites) was 3.4 for 1940; 3.7 for 1950; 3.9 for 1960; 4.3 for 1970; 3.7 for 1980; and 4.4 for 1990.

The hard-to-count Blacks living in hard-to-count census tracts*

The Census Bureau estimates that nation-wide about 15.4 million Blacks lived in HTC census tracts. The 5 states with the highest number of Blacks living in HTC census tracts were the following:

• New York – 2.2 million
• Texas – 1.8 million
• Florida – 1.1 million
• California – 1 million
• Georgia – 1 million

Combined, these 5 states have about 42% of the nation’s Black population living in HTC census tracts. Arguably, these states may house a similar proportion of the Black population not counted by the census. Again arguably, these 5 states bear the brunt of the census’s undercount of the Black population. The Census Bureau estimated that about 800,000 blacks were not counted in the 2010 Census.

The 5 states with the lowest number of blacks living in HTC census tracts include Iowa, Utah, New Hampshire, Maine, and North Dakota. There is just about no chance for any Black undercounts in these states.

For Hispanics, the 5 states with the highest number of Hispanics living in HTC census tracts are

• California – 5.1 million
• Texas – 3.4 million
• New York – 2 million
• Arizona – 896,000
• Florida – 734,513

These 5 states have about 68% of the nation’s Hispanic population residing in HTC census tracts.

The 5 states with the lowest number of Hispanics in HTC census tracts include New Hampshire, Idaho, West Virginia, Wyoming, and Montana. These five states are apt to experience no meaningful Hispanic undercounts.

In sum, relative to census undercounts of black and Hispanic populations, California, New York, and Texas are arguably most at risk of losing or not gaining a U.S. House seat in apportionment due to census undercounts. Focusing solely on Hispanics, the likely source of undercounts due to the citizenship question controversy, California, Texas, and New York have the most to lose.

How the Bureau of Census responds to HTC populations

Over time the Census Bureau has employed several strategies to deal with HTC populations. This section reviews the key elements of its approach.

Prior to undertaking census enumeration and beginning with the 2000 Census, the Bureau developed public advertising campaigns to reach HTC immigrant communities and other HTC populations. These targeted marketing and outreach campaigns use paid media in over a dozen different languages to improve responsiveness. The Bureau undertakes these efforts in its Partnership and Communications Program. The Bureau works with community organizations, local businesses, faith-based groups, elected officials to improve count accuracy.

The Bureau’s key strategy entails its Non-Response Follow-Up (NRFU) operations. In the 2010 Census, about 27 percent of the persons enumerated came through the NFRU workload. The NFRU activities include (1) in-person follow-up enumeration, (2) proxy enumeration, (3) administrative record enumeration, and (4) imputation by other means.

In-person follow-up enumeration. Census experience suggests the following trends since the 2010 Census: (1) in-person follow-up enumeration has been less effective over time in all census tracts, (2) in-person follow-up enumeration has been differentially less effective in census tracts with a higher proportion of households containing a noncitizen, and (3) the differential between census tracts with a higher proportion of households containing a noncitizen and census tracts with a lower proportion of households containing a noncitizen has grown over time.

Additionally, in a December 2018 report, the U.S. Government Accountability Office found problems in the Census Bureau’s NRUF test site in Rhode Island. It made four recommendations to improve NRFU procedures. Among the deficiencies noted by the GAO were these (1) the Bureau’s field workforce was not fully prepared to face all of the enumeration challenges that arose during the test and (2) the Bureau lacked any standardized form of mid-operation training or guidance as new procedures were implemented.

An earlier July 2018 GAO report also noted potential problems the Census Bureau may face in counting HTC populations.

Proxy-enumeration. Locating a proxy respondent (a neighbor, landlord, postal worker, or other knowledgeable person) who will provide information about another household is not easy. Even if proxies are found they generally provide lower quality enumeration data than self-responses. For example, in the 2010 Census, 97.3 percent of self-responses resulted in a correct enumeration, but the correct enumeration rate for proxy responses was 70.2 percent. Proxy responses are particularly inaccurate for persons in tenuous residential arrangements (such as people living in converted garages) because proxy responses may not know how many people live in these tenuous residential arrangements.

Administrative record enumeration. The quality of administrative records varies depending on the subpopulations, especially Hispanics and noncitizens. Generally, administrative record enumeration will not remediate the differential decline in self-response rates.

Imputation. When the above tactics do not lead to full enumeration the Census Bureau computes or models, the number of persons in the household and their characteristics. In the decennial census, the Bureau uses count imputation to compute the size of the household and whole-person imputation to impute both the size of the household and the characteristics of the people in the household.
Overall, the Census Bureau makes a concerted and expensive effort to enumerate households who do not respond to the mail questionnaire.

Nonetheless, differential undercounts still occur. What makes this matter more significant is that HTC populations have listened since 2015 to candidate Trump’s and then President Trump’s derision and scorn for the very populations that are HTC – blacks, Latinos, and, especially, noncitizens. This context, as shown next post, makes for growing concern about the enumeration effectiveness for the 2020 Census.

Comment

The Census Bureau works diligently to try to minimize undercounts given that the Constitution’s only demand on the decennial census is that it enumerates all residents of the United States. Both the recent presidential rhetoric against immigrants and the administration’s actions against immigrants and the vacancies within the Census Bureau make its task more difficult.

In addition to these challenges, the Census Bureau is undertaking a new information technology system [see GAO report]. The new system will enable it for the first time in 2020 to urge most households to submit their responses online. All households will have the option of using email, or their phone, or to complete the mail-in questionnaire. The Bureau is aware that this move may be problematic because of the lack of good internet access or experience in using the internet. For example, the Bureau estimates that 25.8% of West Virginia households during 2013-2017 had either no home internet subscription or dial-up only; 6.1% had a cellular data plan only.  For Maryland, the estimate is 14.0% with either no home internet subscription or dial-up only and 6.1% with only a cellular data plan.

Given some of these difficulties, the GAO recently placed the 2020 Census on its high-risk list.

As will be discussed in the next census post, the prolonged controversy over the citizenship question is likely to increase immigrant undercount. The Supreme Court’s opinions will be reviewed in regard to the Constitution’s charge for an enumeration of the population every 10 years.