The Rural Undercount

This blog was originally posted by the CBPP on March 1.

. . .

By Arloc Sherman, Senior Fellow, Center on Budget and Policy Priorities

President Trump’s latest requests for 2018 and 2019 Census Bureau funding are too low to ensure an accurate 2020 census, as I wrote recently. One point deserves further attention: an underfunded, poorly executed census could hurt a diverse range of communities, leaving them underrepresented in Congress and cutting their federal Medicaid, economic development, and child care funds.

HighRiskUndercountforPoC-Mar2018Concerns about the accuracy of the upcoming census often focus on people of color, who face a higher risk that they will be undercounted. Less well known is that disadvantaged rural Americans and other low-income and marginalized groups are also at high risk. For any community, being undercounted not only would deprive them of fair representation in Congress and state capitols and their fair share of federal dollars, it also would distort the information that a business uses to decide where to locate or a county uses to decide whether to close a school.

We don’t know how an underfunded census would play out in 2020. But a look at past censuses with a high undercount provides important clues. The 1990 census had the biggest overall undercount of any in the last 40 years, and it’s the only one that was ever measurably less accurate than the one before it. Undercount rates in 1990 were far higher for the Hispanic (5.0 percent) and black populations (4.6 percent) than for the white population overall (0.7 percent), one Census Bureau study found, but the rate for white rural renters was similarly high (5.3 percent). (See chart.) It was even higher for American Indians on reservations (12.2 percent) and Hispanic rural renters (15.8 percent).

While non-metropolitan households in general have higher response rates without follow-up by the Census Bureau, a recent study found, intensely rural areas like remote Appalachia and Indian reservations are among the hardest to count. Homes in such areas are spread farther apart and often hidden from the main road, may not have city-style addresses, and sometimes are non-traditional living quarters, such as sheds and campers. If the Census Bureau has insufficient funding for in-person visits by census workers, more rural homes could fall through the cracks.

A newer concern for rural areas is the digital divide. The 2020 census will be the first to let households respond online, which should increase participation and limit costs, but it could also widen the gap in response rates between metropolitan and rural areas because of differences in Internet access. In 2016, 26 percent of households in rural areas lacked an Internet subscription, compared with 17 percent in metro areas, according to the Census Bureau. Bad Internet connections may also slow the work of census takers in rural areas by making it harder for them to use electronic devices for address updating and collecting information from households that don’t initially respond by themselves.

Many minority groups have low Internet access rates as well. A much higher share of black (22 percent), Hispanic (19 percent), and American Indian and Alaska Native (28 percent) households don’t have an Internet subscription compared to non-Hispanic white (12 percent) and Asian (7 percent) households. Low-income and elderly individuals also tend to lack access. The Census Bureau needs adequate funding of non-Internet-based efforts such as initial outreach, telephone-based response options, and in-person visits to bridge the digital divide for all of these populations.

Compounding these concerns, funding problems forced the Census Bureau to cancel its main field tests of census methods in rural and suburban areas in 2017 and 2018. That leaves the bureau with little chance to detect and correct flaws in its many procedures and systems. For example, the Census Bureau plans to use administrative data, including from the U.S. Postal Service, to remove vacant addresses from the list of unresponsive homes that census takers will visit. But failure to fully test and fine-tune this new method for rural, suburban, and urban differences might lead the Census Bureau to misidentify some buildings as vacant and neglect to follow up with them.

All Americans have a stake in an accurate census, and the President and Congress need to provide the resources that will ensure that the 2020 census doesn’t leave any communities behind.

Providence Field Test Begins

Traditionally, the U.S. Census Bureau in the 8th year of its decennial planning cycle, tests new counting techniques with a large-scale field test. Last year, due to funding shortages, the Bureau announced that it was drastically cutting back on the 2018 End-to-End field test.

Originally, the test was scheduled for about 700,000 targeted housing units in rural West Virginia, suburban Washington State and urban Providence, Rhode Island. With funding shortfalls, the Bureau will only test 275,00 housing units in Rhode Island using new IT methods it plans to employ for the 2020 Census. The Bureau’s single test will not test the advertising/communications components of the original test plan focused on Hard-to-Count populations also due to funding shortages.

The implementation of the Providence test is underway with the establishment of local census offices and the recruitment of census takers.

On March 16, the Bureau will send two separate letters to the test residences asking for either an internet first reply (188,000 residences), or a letter plus a paper questionnaire indicating an internet choice (87,000 residences). Five separate mailings will be done by April 23. More than 1,000 census enumerators will then swing into action to complete the test among those who have not responded either by the internet or via a paper questionnaire. Peak operations will be April 1, exactly two years from the actual 2020 Census.

This Chart Says It All – Again

This blog was originally posted by the CBPP on February 16.

. . .

By Arloc Sherman, Senior Fellow, Center on Budget and Policy Priorities

With President Trump and Congress boosting overall non-defense discretionary funding for 2018 and 2019, they now must use some of these funds for the fast-approaching 2020 census. The President’s latest request for 2018 funding for the census moves in the right direction but remains below what outside experts believe is needed to ensure an accurate census count, and it omits funds for contingencies that Commerce Secretary Wilbur Ross has recommended. The President’s 2019 request also falls well below what the Commerce Department projects is needed.

The decennial census count, mandated in the Constitution, supports our electoral system and much public- and private-sector decision-making. The 300-million-plus Americans must be counted accurately to ensure that each state has fair representation in Congress, enable fair redistricting within states, and guide the allocation of federal funding for programs from Medicaid to economic development and child care. Moreover, the census forms the statistical foundation for many annual and monthly surveys, including crucial surveys on unemployment, education, health, and commuting. Businesses use this information to understand markets and plan their locations, and governments use it to plan everything from locating schools and roads to funding disaster relief and setting interest rates.

Yet funding problems over the last several years — due in part to tight annual caps on overall non-defense discretionary funding — have hamstrung census preparations. Low and delayed funding forced the Census Bureau to cancel key field tests in 2017 and 2018 and slowed progress on basic communications and outreach strategies to encourage the public to fill out the census.

To ensure a successful census, final Census Bureau 2018 funding should be well above what the House ($1.507 billion) and Senate appropriators ($1.521 billion) approved earlier this year. The President, who originally requested slightly less than those amounts, subsequently increased his official request by $187 million to $1.684 billion. Even this higher level, however, represents just a 23 percent increase over 2016 funding. At this point in the decennial census cycle — just two years before the census is conducted, when critical field preparations and a final “dress rehearsal” must be completed — funding should be ramping up significantly more. The President and Congress have traditionally provided far larger increases in funding at a similar stage in past decades, including increases of 79 percent from 2006 to 2008, 143 percent from 1996 to 1998, and 135 percent from 1986 to 1988. (See figure.)

While the President’s revised request reflected a recognition that his initial request was far too low, outside experts estimate that the Census Bureau needs at least $1.848 billion for 2018 to be ready for 2020 and avoid cuts in other important areas of the bureau’s budget. That would include funding to complete planned communications work and hire 200 partnership staff to engage with local faith groups, local officials, and others who can help build trust in the census, as well as $50 million that Secretary Ross recommended last October for the first installment of a contingency fund to address natural disasters such as wildfires and other concerns.The President’s request for 2019, $3.8 billion, is also low. It falls well short of giving the Census Bureau the strong ramp-up in funds that a December Census Bureau cost estimate said the 2020 census program would need in 2019.

The 2020 census faces particular challenges that make strong preparation even more critical. Among them is the plan to offer an online response option for the first time ever, which the Census Bureau hopes more than 2 in 5 households will use to complete the census. While that should make participating easier, improve response rates, and lower Census Bureau costs for households with online access, it will require careful system testing in advance. The online option also risks disproportionately undercounting those without Internet access. Avoiding a “digital divide” in census undercount rates will require significant outreach and follow-up to rural, low-income, and other populations that are less likely to have reliable broadband or Internet access, making the final two years of research and deployment of advertising and outreach strategies vital.

A poorly funded, poorly executed census would likely depress the overall count and leave some places and population groups more undercounted than others. The 1990 census is a case study of a relatively large undercount; 12.2 percent of American Indians on reservations were missed, the highest undercount rate for any racial group, followed by high rates for the non-black Hispanic population (5.0 percent) and black population (4.6 percent). Whites experienced far less undercounting overall, but rural white renters were also undercounted at a high rate (5.3 percent). While the 2000 and 2010 census both lowered the overall net undercount and generally narrowed the gaps between groups, underfunding for 2020 could erase these improvements — potentially leaving undercounted groups with less representation in Congress and statehouses, lower federal resources, and less attention to their needs.

In addition, policymakers must ensure that the Census Bureau doesn’t add a citizenship status question to the 2020 census. The Justice Department asked the Census Bureau in December to add such a question — an alarming, last-minute move that would likely depress immigrant participation and lead to “bad Census data,” according to four past Census Bureau directors who served in Republican and Democratic administrations.

Administration Funding Proposal for FY 2019

The Trump administration released its FY 2019 Census Budget request which the Census Project labeled as only “keeping the 2020 count on life support.”

A previous blog reported that census stakeholders will still be seeking additional monies for the FY 2018 Census Bureau budget beyond the $182 million increase already approved by policymakers as part of the emerging budget deal increasing overall domestic spending.

Meanwhile, several articles explain the administration request.

What a Citizenship Question on the Census Would Mean for Cities

This was originally posted on the NLC’s CitiesSpeak blog on February 9.

. . .

By Brian Egan, Principal Associate for Finance, Administration and Intergovernmental Relations, National League of Cities

There is no question that America’s city leaders share Commerce Secretary Wilbur Ross’s commitment to a full and fair 2020 Census. Census data is vital to cities for uses including regional planning, economic research, public health initiatives, and allocating more than $600 billion in federal funding to state and local governments.

But because city leaders understand the importance of an accurate count, they also recognize that the upcoming 2020 Census faces unique challenges — and that the stakes are high.

For starters, Congress has required the Census Bureau to spend no more on the 2020 Census than it did on the 2010 Census. Facing this challenge, the Bureau is attempting to adapt through a greater reliance on technology and state and local assistance. Even under these intense budget constraints, however, we have yet to see the increase to census funding that typically arrives two years out from a decennial census — stoking concerns about underfunding.

Still the Bureau has managed to extensively prepare for these cost saving changes. In addition, each question on the form and change made to the process has undergone extensive testing to mitigate any effects on the rate of census self-response. Non-response follow up work (NRFU), which includes the Bureau’s efforts to collect responses from residents who do not provide them voluntarily, accounts for the most costly census operation. As a result of this fact and the pressures the Bureau is facing, it cannot afford to risk inflating the need for NRFU in 2020.

Late in December, the Department of Justice (DOJ) sent a request to have a last minute question, on citizenship, added to the 2020 Census. DOJ argued that the additional data was necessary for executing the Voting Rights Act of 1965.

The Census Bureau must submit the proposed questions for congressional review by March 31 of this year. The one and only end-to-end census test begins in Providence County, Rhode Island, next month, and will not include such a question. Given this timeframe, there is no conceivable way in which this addition could be vetted with the thoroughness demanded of such an important consideration.

Even more alarming is that there is tremendous bipartisan agreement that the addition of an intrusive and untested citizenship question this late in the game would lower self-response rates among both noncitizens AND citizens. Four previous Census Bureau directors, who have served under both Republican and Democratic administrations, have expressed their concerns that such an addition would ultimately lead to worse data.

Finally, since the passage of the Voting Rights Act of 1965, the DOJ has relied exclusively on citizenship data collected through robust but much smaller surveys — most recently the American Community Survey (ACS). For the past 53 years, the DOJ has reliably used data from other Census Bureau surveys to fully enforce the Voting Rights Act, leading us to seriously question the necessity for drastic changes made to the only constitutionally-mandated count of the nation’s population.

An accurate count is of the utmost importance to city governments when it comes to planning, research, federal funding and so much more. Last week, NLC sent a letter to Secretary Ross, who has ultimate say in whether or not the question is added, asking him to reject a proposal to add any untested questions to the census. Over the next two years, NLC will continue to work with the Bureau and the Department of Commerce to ensure the 24th Decennial Census is a yet another success.

A Vital First Step on FY 2018 Decennial Funding

Today the President signed another Continuing Resolution (CR) through mid-March to keep the government running. Included in the CR was an acknowledgement of the dire need for more FY 2018 census funding as Census Project stakeholders have been insisting.

The policymakers included authorization for the Census Bureau to spend $182 million in additional funds in an anomaly.

Of course, the Census Project pointed out that this is only “an important step” towards full funding for the 2020 Census.

Earlier, the Project said it was supporting calls for a final FY 2018 funding level of $1.84 billion for the Bureau, an additional $140 million for 2020 Census preparations beyond the Administration’s updated request.

Undercount Could Cost the States

With the help of 2020 Census advocates across the country, using background materials from both the Census Project and the Leadership Conference on Civil and Human Rights, allies have written op-ed pieces and contributed to stories and editorials in their states. Below is a short listing of some of the op-eds, editorials and articles pinpointing how an inaccurate, unfair and underfunded decennial census would affect various states.

Alabama – (The Andalusia Star-News, 12/16)
California – (San Jose, The Mercury News, 12/7; Politico, 1/16)
Illinois – (Chicago Tribune, 1/19)
Kentucky – (Lexington Herald Leader, 12/13)
New York – (Daily News, 10/12)
North Carolina – (Durham, The Herald-Sun, 12/9)
Mississippi – (Jackson, Clarion-Ledger, 1/13)
Montana – (Billings Gazette, 1/8)
Utah – (The Salt Lake Tribune, 1/26)

Please feel free to utilize any of the facts and figures in the posted materials for your own outreach to the media. Or, visit the Census Project website ( for more information.