Chapter 4:
Government Data Collection and Use

The Census: Big Data for Civic Welfare

The United States Census is the government’s original big data. Since 1790, the Census Bureau has conducted surveys to enumerate the United States population. Required by the Constitution to apportion representation and taxation among the states, the Census Bureau today maintains the most complete and authoritative records of the American population. These records are used for allocating resources and planning for a variety of social and economic programs. The decennial census and the more detailed American Community Survey (the modern version of the census “long form”) have become essential for the functioning of government, civil society, and the private sector. Civil rights groups have a strong stake in the accuracy, completeness, and availability of Census Bureau data.

The decennial census is conducted every ten years and counts every individual in the United States. In 2010, the Bureau collected demographic information on each member of the household.^[120] Between 1940 and 2000, households were sampled during the decennial census to collect more detailed information, in what was known as the census “long form.”^[121] In 2005, these more detailed records were separated into the American Community Survey (ACS), which samples households on a continuous basis to provide more up-to-date data for areas as small as census tracts (average population: 4,000). The ACS samples roughly 2.5 percent of U.S. households each year.^[122]

In addition to determining Congressional representation, census data is used to allocate resources for social programs, including Medicaid, maternal and child health programs, transit programs, public housing assistance, Community Development Block Grants, Head Start, Title I education funds and grants for special and vocational education.^[123] It is also used for planning for hospitals, nursing homes, clinics, and the location of other health services, and drawing school district boundaries.^[124]

For Census 2020, the Bureau is researching ways to incorporate new technology to make the census more efficient and less costly.

For Census 2020, the Bureau is researching ways to incorporate new technology to make the census more efficient and less costly. This is, in part, a necessity: the Bureau has been tasked with running the 2020 Census at the same cost as the 2010 Census, after decades of rapid growth in the cost of census-taking.^[125] It is considering using governmental administrative records (ARs)—pre-existing government databases from sources such as Internal Revenue Service, Social Security, TANF, SNAP, and Medicare/Medicaid,^[126] as well as from state and local governments—to pin down areas of housing change. This will help to eliminate the need to physically canvass every street and address prior to the start of the population count, identify vacant housing units before sending a census taker to knock on doors, and add people to the count who do not self-respond to the census. The Bureau is also looking at new ways to strategically route census takers during non-response follow-up, optimizing their productivity. And it is researching an Internet response option to reduce costs during the self-response phase.

Unfortunately, ARs can both help and hurt the accuracy of the census. Data collected from ARs may lack detail (such as race and ethnicity, especially for subgroups, and household relationships), may be out of date, or may underrepresent the hardest-to-count populations. Yet, ARs may be suitable as a last resort, when people would otherwise have been missed even after in-person follow-up interviews. In the 1990 Census, for example, the Bureau used parolee and probationer records to add “between 400,000 and 500,000 persons to the final census counts.”^[127] Many of those added were young, black men, who had been missed. This practice was abandoned in later years because “about half of [people added using parolee and probationer records] were later estimated to have been enumerated erroneously.” But technological improvements might increase the utility of ARs in the future.

A secure Internet response option could significantly improve efficiency of the census. Recent testing of the Internet response option suggests it is likely to only marginally increase the total number of self-responders, meaning that most of those who would use the Internet to respond would otherwise have voluntarily responded by mail. But even a small shift will provide the benefit of reducing the cost of mailing out packets to tens of millions of households, which will free up monetary resources that can better be focused elsewhere—like physically canvassing neighborhoods with low Internet adoption rates. However, an Internet response option must be mobile-friendly in order to maximize its reach to certain minorities who over-index on smartphone adoption, such as Hispanics and Asian Americans.

The Census Bureau faces a difficult challenge ahead: saving money while retaining quality and accuracy. Technological advances and new data sources can help, but cost constraints still threaten to make the 2020 Census less complete than its predecessors.