For demographers like me, the census is kind of like our Super Bowl, if the Super Bowl only took place once a decade — and if the foundation of your representative democracy hinged on the winner of the game.
It’s why I’m glad that the Supreme Court, in a surprise ruling Thursday, sent a decision about adding the citizenship question to the 2020 census back to the lower courts. This Trump administration addition to the census had demographers like me really concerned: We know that any change to the census that threatens to impact response rates should be concerning, so the Court’s ruling is a smart one.
But there’s so much more we are worried about: lots of expected changes in 2020. Chronic underfunding throughout the decade has limited the Census Bureau to fully research and test ways to make the census better, which may force the bureau to rely on older and costlier methods in 2020.
Counting all Americans is mandated by the Constitution, but counting everyone who lives in the United States — and counting them accurately — is hard to do. One of the biggest challenges in obtaining a count of all residents is non-response, or when households do not complete their census form by mail. When households do not respond to the initial census questionnaire, the bureau sends additional forms. If this does not work, they may send individual enumerators door to door to try to obtain the required information. This gets expensive: In 2010, non-response follow-up efforts cost $1.6 billion.
In addition to being expensive, high rates of non-response also increase the risk of undercounting. In 2010, the overall census count was highly accurate, but certain populations were missed in the census totals, including renters, black men, American Indians living on reservations, and Hispanics. But the highest rate of undercount was for children ages 0 to 4. This matters: Census data is used to plan roads, schools, hospitals, senior centers, and emergency services in our communities. Inaccurate counts paint a distorted picture of the makeup of our communities and result in a misallocation of resources vital to families.
Non-response in 2020 may be even higher: The US population is larger and harder to count than ever before; Americans are increasingly distrustful of government data collection efforts; and overall response rates to statistical surveys have been declining steadily for the past few decades. Meanwhile, the citizenship question may create new risks of undercount for the nearly 45 million Americans living in a household with at least one noncitizen.
The census is meant to count all individuals regardless of citizenship status, but many Americans do not fully understand how the census is used. In a recent survey, nearly one in six respondents said the census did not count both citizens and noncitizens, and one in 10 thought it was used to locate individuals living in the US without documentation, which is not true. New research suggests the citizenship question will increase the non-response rate by 2.2 percent, raising overall costs of conducting the census and reducing the accuracy of the data.
At the same time, the bureau was asked by Congress to limit the total life cycle cost of the census. If the 2020 census were conducted using a design like the one used in 2010, it would cost an estimated $17.5 billion. Instead, the Census Bureau is piloting many new initiatives to reduce overall costs to $15.6 billion. These include using the internet to increase self-response; using administrative records (data that already exists in the government) to reduce the need for non-response follow-up efforts; using satellite and aerial imagery to identify where to count; and using automation to reduce staffing and office needs.
These changes, if they are not fully tested, make non-response and undercount more likely, and chronic underfunding throughout the decade has reduced the Census Bureau’s capacity to research and test planned innovations. On April 1 last year, the bureau conducted an end-to-end test, which is like a dress rehearsal for the 2020 Census. The bureau initially planned to conduct this test in three locations: Bluefield-Beckley-Oak Hill, West Virginia; Pierce County, Washington; and Providence, Rhode Island. The only complete test occurred in Providence, and the Office of Inspector General reported that in-office address canvassing — the use of satellite and aerial imagery to identify where to count —did not fully identify where to count. The cause of these errors is unknown, and the bureau does not currently have a process for correcting them.
The census is vitally important. It’s the backbone of virtually every data product researchers, governments, and businesses use to understand who we are, how we’ve changed, and what this might mean for the future. It’s also the most democratic and inclusive activity we do as a country. This once-a-decade count is the only source of basic demographic data on all individuals living in the United States.
We have one shot to nail it for the next decade. It’s less than a year away, and we must ensure that the Census Bureau has the resources it needs to conduct a complete and accurate count of all Americans. This data is critical for decision-making and research in so many sectors. We’ve got to get it right.
Rebecca Tippett, PhD, leads Carolina Demography, located in the Carolina Population Center at UNC Chapel Hill. Her most recent work looks at areas in North Carolina that are likely to be undercounted in the 2020 census and the high school to postsecondary education pipeline in North Carolina.